Protein Crowding Mediates Membrane Remodeling in Upstream ESCRT

Protein crowding mediates membrane remodeling in upstream ESCRT-induced formation of intraluminal vesicles Susanne Liesea,1, Eva Maria Wenzelb,c,1 , Ingrid Kjosb,c , Rossana Rojas Molinaa, Sebastian W. Schultzb,c , Andreas Brechb,c, Harald Stenmarkb,c , Camilla Raiborgb,c, and Andreas Carlsona,2

aDepartment of Mathematics, Mechanics Division, University of Oslo, N-0851 Oslo, Norway; bDepartment of Molecular Cell Biology, Institute for Cancer Research, Oslo University Hospital, Montebello, N-0379 Oslo, Norway; and cCentre for Cancer Cell Reprogramming, Institute of Clinical Medicine, Faculty of Medicine, University of Oslo, Montebello, N-0379 Oslo, Norway

Edited by David A. Weitz, Harvard University, Cambridge, MA, and approved September 21, 2020 (received for review July 7, 2020)

As part of the lysosomal degradation pathway, the endosomal ESCRT-0 to sequester cargo material into patches, so-called sorting complexes required for transport (ESCRT-0 to -III/VPS4) microdomains, on the endosome membrane (14). Interestingly, sequester receptors at the endosome and simultaneously deform the role clathrin plays in ILV formation differs significantly the membrane to generate intraluminal vesicles (ILVs). Whereas from its role in endocytosis, as it promotes ILV formation (15), ESCRT-III/VPS4 have an established function in ILV formation, the but does not form a basket-like scaffold. Instead, a rather flat role of upstream ESCRTs (0 to II) in membrane shape remodel- clathrin coat is bound to the ESCRT microdomain (14, 16– ing is not understood. Combining experimental measurements 18). ESCRT-0 recruits ESCRT-I that leads to the recruitment and electron microscopy analysis of ESCRT-III–depleted cells with of the complete ESCRT machinery. While ESCRT-0, -I, and a mathematical model, we show that upstream ESCRT-induced -II sequester transmembrane cargo proteins and facilitate ILV alteration of the Gaussian bending rigidity and their crowding in formation, ESCRT-III and the ATPase VPS4 enable constric- concert with the transmembrane cargo on the membrane induce tion of the membrane neck leading to the formation of an membrane deformation and facilitate ILV formation: Upstream ILV (19). Notably, the only energy-consuming step in the mem- ESCRT-driven budding does not require ATP consumption as only brane remodeling process is the membrane scission, involving the a small energy barrier needs to be overcome. Our model predicts ATPase VPS4 (20, 21). This is especially remarkable, since the that ESCRTs do not become part of the ILV, but localize with a energy required for a flat lipid bilayer to form a spherical vesicle high density at the membrane neck, where the steep decline in in absence of a protein coat is orders of magnitude larger than the Gaussian curvature likely triggers ESCRT-III/VPS4 assembly to the thermal energy, thus creating an energy barrier inhibiting enable neck constriction and scission. vesicle formation (22, 23). Experiments have shown that the recruitment of ESCRTs to ILV formation | upstream ESCRT | protein crowding the endosome occurs in a time periodic fashion (15, 24), where each recruitment cycle is associated with the formation of a single ILV. Transmission electron microscopy (TEM) images orting and compartmentalization of biomaterials lie at the and electron tomography have also provided high-resolution Sheart of cellular processes and play a fundamental role in the lysosomal degradation pathway to regulate cellular activi- ties. Transmembrane proteins in the plasma membrane, such Significance as growth factor receptors, are internalized by endocytosis and degraded in lysosomes (1). As a part of this pathway, intralumi- Intraluminal vesicle (ILV) formation plays a crucial role in the nal vesicles (ILVs) with a typical diameter in the order of 50 nm attenuation of growth factor receptor signaling. The endo- are formed inside endosomes (2) (SI Appendix, Fig. S1). The somal sorting complex required for transport (ESCRT-0 to ILV formation starts with a small deformation of the endosome -III/VPS4) mediates this process. The general dogma has been membrane, which grows over time and finally leads to a cargo- that upstream ESCRTs (0 to II) sequester receptors at the sur- containing vesicle within the endosome lumen. ILV generation face of endosomes and the downstream ESCRTs (III/VPS4) is an essential part of the endocytic down-regulation of activated remodel the endosome membrane leading to the abscission receptors to ensure signal attenuation, failure of which can result and formation of receptor-containing ILVs. We now show in tumorigenesis (2–5). In addition, ILVs can reach the extra- that upstream ESCRTs not only sequester cargo, but in addi- cellular environment as exosomes, where they become signaling tion play a crucial role for the initiation of membrane shape entities enabling intercellular communication. Altered exosomes remodeling in ILV budding. Through a combination of math- can serve as tumor biomarkers (6–8). Despite the fundamen- ematical modeling and experimental measurements we show tal role ILVs play in the endocytic pathway, surprisingly little is that upstream ESCRTs facilitate ILV budding by crowding with known about the mechanochemical crosstalk that regulates their a high density in the membrane neck region. formation. The biophysical process that leads to the formation of an ILV Author contributions: S.L., E.M.W., C.R., and A.C. designed research; S.L., E.M.W., and I.K. is in many ways inverse to clathrin-mediated endocytosis at the performed research; S.L., E.M.W., I.K., R.R.M., S.W.S., A.B., H.S., C.R., and A.C. analyzed plasma membrane, as the membrane bud protrudes away from data; and S.L., E.M.W., C.R., and A.C. wrote the paper.y the cytosol (Fig. 1A) and the vesicle shape is not dictated by The authors declare no competing interest.y a protein scaffold (3). Cargo sorting and ILV formation are This article is a PNAS Direct Submission.y mediated by the endosomal sorting complex required for trans- Published under the PNAS license.y port (ESCRT) (4, 9–13), which consists of four subcomplexes, 1 S.L. and E.M.W. contributed equally to this work.y ESCRT-0, -I, -II, and -III, and the accessory VPS4 complex. 2 To whom correspondence may be addressed. Email: [email protected] Each of the subcomplexes has different, yet complementary func- This article contains supporting information online at https://www.pnas.org/lookup/suppl/ tions (4, 10). ESCRT-0 contains binding domains for the endo- doi:10.1073/pnas.2014228117/-/DCSupplemental.y some membrane, ubiquinated cargo, and clathrin, which enables First published November 2, 2020.

28614–28624 | PNAS | November 17, 2020 | vol. 117 | no. 46 www.pnas.org/cgi/doi/10.1073/pnas.2014228117 Downloaded by guest on September 28, 2021 Downloaded by guest on September 28, 2021 hog nemdaemmrn hps lutae yteTMmicro- TEM is bin the each by in illustrated (from points classified data center shapes, and transitions graphs the of membrane membrane and number endosome intermediate nm, the the 20 formation through that ILV of such During width chosen (D) a maximized. is with endo- bins The bins diameters. the into ILV of The 10 sorted least images. is the at TEM over diameter of calculated in some are independent measured SD size the was and ILV diameter average ILV of The distribution denoted diameter. narrow is endosome a events exhibit formation endosomes ILV consecutive time two dwell by the sig- between during fluorescence time decreases The waiting and (15). measured increases ESCRT-III the intensity and from nal ESCRT-0 timescales of machinery characteristic ESCRT intensity the the illustrate fluorescence by We sorted (B) are microdomain ILVs. proteins a into cargo form where endosome, clathrin the and at ESCRTs, proteins, cargo Transmembrane 1. Fig. is tal. et Liese under licensed is S1). which (Movie 15, (15) ref. tomograms from TEM from (E shapes, brane (15). vesicle abscised and E D C B A τ wt (C . ceai lutaino h omto fa SR-reILV. ESCRT-free an of formation the of illustration Schematic (A) xeietlmaueet f47IVdaeesi eacell HeLa in diameters ILV 477 of measurements Experimental ) Left ercntuttetredmninlmem- three-dimensional the reconstruct We ) to Right CB 4.0. BY CC noptsae shape, U shape, pit into ) D and E τ r modified are d h mean The . Ω shape, xiisajm nisfloecneitniyoe utafew a just over intensity fluorescence sig- its ESCRT-III ESCRT-0 intensity, in the fluorescence Once jump in I. a maximum from to exhibits 0 a different subunits reaches distinctly ESCRT nal are the The ESCRT-III of ESCRT-0. those of of to evolution features starts temporal therefore the dynamic abruptly We consider ESCRT-0. it to to ESCRT- only dynamics before that need similar shown min, have was clathrin it 3 and addition, In I about min. 2 of about for span decrease increases time continuously recruitment a ESCRT-0 each of over In signal 1B. fluorescent Fig. the in and cycle, illustrated increase as membrane (15), limiting oscillatory endosomes the of on an concentration ESCRT by that the show of accompanied decrease which is (HeLa), formation cells cancer live- ILV human by of performed microscopy were dynamics cell protein ESCRT Measurements. of surements Experimental Underpinning Results cells in ESCRTs. profiles downstream budding the- of our endosomal depleted verify analyzing rigidity we by Finally, bending model membrane. Gaussian oretical the on the crowding alter their to upstream and the dynam- ability membrane by facilitated proteins’ the endosome is formation ESCRT the do ILV to that How find couple 3) We shape? recruitment vesicle? ESCRT endosome ESCRT-free of the an at ics external form organize of proteins source later a ESCRT to and or do form force How active to 2) What an energy? ILVs 1) require allows process questions: that this following does mechanism biomechanical the the answer theo- is with we measurements modeling experimental membrane-remodeling retical this combining in By role formation, their ILV process. investigate mediate to ESCRTs out set upstream we the which by nisms formation the understand to ILV. an essential of is bending highlight the Gaussian models of that These structure supercomplex. the ESCRT-II a and of by ESCRT-I influence caused the curvature considered Gaussian (44) Gaus- the spontaneous al. negative on et to Mercker coat binding curvature. ESCRT sian preferred uniform forma- a exhibits ILV assumed which (43) observe membrane, al. To et (39–42). Rozycki forces surface tion elastic between resistive (38), the link effects and stochastic the (37), proteins transmembrane about from pro- forces information has modeling invaluable theoretical vided where with shapes, membrane, membrane plasma studies description similar the of number at a large endocytosis the of clathrin-mediated with on lack deforma- contrasts endosome The membrane the generating at 36). ESCRTs tions mechanisms (15, crit- biophysical upstream formation as the the ILV an of play in how they role about though ical even known remodeling, is neck membrane membrane affect less by the exerted of Much closure tension cause (35). scaf- the to neck with suggested combination is and ESCRT-III in (33) which buckling (34), filaments membrane folding ESCRT-III promotes of stud- spirals polymerization experimental the into that and shown Theoretical have 25–32). and ies cellu- cytokinesis (19, of e.g., budding range scission, broad virus theoretical membrane a involving and in processes role experimental fundamental lar their both to in due studies, attention most received portion main the coat (9). ESCRTs not vesicle do that the but of neck, demonstrate vesicle experiments the in mem- vitro enriched to their are in system at The model vesicles (9). vitro micrometer-sized brane in of an formation as the used study been unilaminar have giant (GUVs) membrane vesicles endosome the of on mem- understanding detailed assembly the a ESCRT and gain markers To simultaneously. fluorescent shape of to brane able evolution not pro- time are budding the techniques measurement record the vivo during in Today, shapes (15). cess membrane the on information ie u iie nesadn ftebohsclmecha- biophysical the of understanding limited our Given have VPS4 and ESCRT-III subcomplexes, ESCRT the Among PNAS | oebr1,2020 17, November | o.117 vol. loecnemea- Fluorescence | o 46 no. | 28615

CELL BIOLOGY BIOPHYSICS AND COMPUTATIONAL BIOLOGY seconds, before it decreases with a decay time similar to that of ESCRT-0. The dynamics of ESCRT assembly are described by two timescales: the dwell time of ESCRT-0 τd = (161 ± 94) s and the periodicity of the ESCRT-recruitment cycle or equivalently the mean waiting time τwt = (203 ± 47) s (15). TEM imaging of HeLa cancer cells reveals that ILVs have an average diameter of ≈46 nm and that there is no apparent trend with respect to the endosome size (Fig. 1C). A description of the membrane-remodeling process that leads to ILV formation must therefore include a mechanism that is robust in terms of setting the vesicle diameter. TEM imaging also allows us to describe the membrane shape at different stages, where we cat- egorize the membrane profiles into three specific shapes: pit, U, and Ω shapes (Fig. 1 D and E). As there is no overweight of samples in either of these categories (15), despite that these images are taken from different cells and at random time points within the ILV formation cycle, it suggests that the membrane deforma- Fig. 2. A part of the endosome membrane reconstructed from TEM micro- tion takes place in a continuous rather than a jump-like fashion graphs (in green) is shown together with the parameterization of the over time. membrane in the mathematical model. Membrane-bound coat proteins (red blocks), which include ESCRT-0, -I, -II, and clathrin, bind to the trans- Energy Barrier. Next, we use the fluorescent signal data from the membrane proteins (Y-shaped orange markers). The endosome membrane experiments to extract information about the energy barrier that deforms into an ESCRT-free spherical vesicle with a curvature Cg, surrounded by a neck region with an elevated ESCRT density. The extent of the coat- has to be overcome as an ILV forms. The magnitude of the free area is quantified by the opening angle α. The membrane shape is energy barrier is crucial to classify whether ILV budding hap- defined by the arc length S and the azimuthal angle ψ, where we treat the pens passively, i.e., initiated by thermal fluctuations, or as an membrane as being axially symmetric around the Z axis. active process that requires energy consumption. To determine the energy barrier that is associated with the formation of a single ILV, we use a theory derived by Kim and Netz (45) that relates ILV formation involves a change in both mean and Gaus- the height of an energy barrier in a diffusive process to the ratio sian curvature (SI Appendix, Fig. S3). A common feature among of two characteristic timescales τd/τwt. We deploy this theory proteins that alter the Gaussian bending rigidity and promote for an energy landscape with the shape of a harmonic potential. membrane remodeling is α-helix motives in their secondary In other words, we assume that the system has to overcome a structure (46, 47). The structural similarity between these pro- single energy barrier with a magnitude ∆EB to form an ILV. teins and the ESCRTs (48–51) prompts us to model the Gaussian In a diffusive process the time to reach and cross this energy bending rigidity as dependent on the concentration of ESCRT barrier once, i.e., the dwell time τd in the experimental mea- proteins. The simplest mathematical description of a protein- surements, scales with ∆EB as τd ∼ (4/3 − 16/45β∆EB)β∆EB, induced Gaussian bending rigidity is to assume a linear response while the mean time that passes before the energy barrier is with respect to the ESCRT density ρ. The Gaussian bending crossed a second time, i.e., the waiting time τwt, scales with energy ∆Eg then reads √ √ −1 ∆EB as τwt ∼ πerf( β∆EB)erfi( β∆EB), with β the ther- Z ∞ mal energy. The ratio of the two experimental timescales, τd/τwt, ∆Eg = 2π ργgC1C2RdS, [1] provides an upper limit for the magnitude of the energy bar- 0 rier ∆EB / 0.6 kBT(SI Appendix, Fig. S2). Since ∆EB is of such low magnitude, it can be overcome by thermal fluctua- with the proportionality factor γg > 0. The two principle curva- tions, giving us a first suggestion that ILV budding generated tures of the membrane are denoted as C1 = sin ψ/R and C2 = by the upstream ESCRTs is a passive process as we further dψ/dS. In qualitative terms, a nonhomogenous Gaussian bend- demonstrate below. ing rigidity describes the tendency of the membrane to deform into a neck-like shape, where a homogenous protein distribution Mathematical Model. A mathematical model that describes ILV (ρ = const.) would lead to a vanishing energy contribution, since formation needs to incorporate how ESCRT proteins influence the azimuthal angle is ψ(S = 0) = 0 and ψ(S → ∞) = 0 at the the shape of the endosome membrane. We consider ESCRT-0, inner and outer boundary and hence ∆Eg = 0 according to the -I, -II, and clathrin as one effective complex that coats the endo- Gauss–Bonnet theorem (52). some membrane (Fig. 2), and the membrane together with the In addition to ∆Eg, the total membrane energy has to account embedded cargo proteins is treated as a homogenous elastic sur- for membrane bending and stretching as well as protein crowd- face. Since the ILV size is much smaller than both the endosome ing. By following the Helfrich model for lipid bilayers (22), we (Fig. 1C and SI Appendix, Fig. S1) and the ESCRT microdomain describe the membrane together with the embedded cargo pro- (several hundred nanometers in diameter) (15), we simplify the teins as a thin elastic sheet. The bending energy ∆Eκ is then model by considering only a part of the endosome membrane obtained as an integral of the squared mean curvature over the that is approximately planar (Fig. 2), where cargo proteins are entire surface evenly distributed within the domain. As part of the budding Z ∞ process a region within the ESCRT microdomain forms that is κ 2 ∆Eκ = 2π R(C1 + C2) dS, [2] not coated by ESCRTs (9), which we define as the ESCRT-free 0 2 region (Fig. 2). TEM tomography experiments show that ILVs are very close to being rotationally symmetric (Movies S1 and with the bending rigidity κ. S2), which we adopt in the mathematical model, where the mem- In the limit of an endosome that is much larger than the ILV, brane is parameterized by the arc length S and the azimuthal as is the case in this system, the pressure that acts across the angle ψ (Fig. 2). The Z coordinate and the radial coordinate lipid bilayer causes an effective far-field tension, with σ the sur- R of the membrane contour are related to S and ψ through face tension coefficient (53). This means that remodeling the dZ dR dS = sin ψ, dS = cos ψ. membrane away from a flat shape requires a surface energy ∆Eσ:

28616 | www.pnas.org/cgi/doi/10.1073/pnas.2014228117 Liese et al. Downloaded by guest on September 28, 2021 Downloaded by guest on September 28, 2021 iesols variables, dimensionless scale rigidity bending length Gaussian the of factor ality niainta hs rtisalidc iia spontaneous similar large a biophysical a induce no all for (58). remod- is proteins curvature compartment shape there these sorting and that membrane a proteins indication In for transmembrane as (9). of cause serve proteins variety ILVs a ILVs cargo coat since as by eling, not disregarded induced using do be curvature experiments can ESCRTs a spontaneous vitro play that in a to as addition, shown expected budding have not cur- ILV is spontaneous GUVs in scaffold A (SI protein role here. energy a significant membrane neglected by the is induced contri- to and vature tension correction S6 ) smear small line Fig. will very this Appendix, a membrane However, is the profile. bution density along the protein at diffusion the tension Sim- protein the region. line coated also in effective the ilarly, and an gradients ESCRT-free to the large between an leads boundary penalizes to which lead that profile, will density term proteins energy ESCRT of additional distribution nonuniform A ntepoesgvnb h u ftefieeeg contributions energy five the of sum stage the each by at given is process membrane the budding in a toward progress and brane proteins, cargo of supercomplex the clathrin. from and ESCRTs, stem facilitates that that effects energy ing binding mechanism the a Together, (Eq. 57). be (56, theoret- to deformation shown membrane experimentally been has and crowding ically Protein membrane. coated Eq. in term second The as written is eapoiaeu oscn re in which order exclusion, second volume to coefficient by up generated approximate primarily we repul- 56), steric effective (55, an sion experience they as limited, is membrane density ESCRT uniform a with membrane Eq. with h ebaeadteca spootoa otelclESCRT local the to proportional is coat density the the and from membrane energy dissociate the binding ESCRT The and 54). individual to (20, membrane i.e., recruited endosome coat; continuously dynamic are a proteins form proteins ESCRT ftemmrn-eoeigprocess. membrane-remodeling the of numbers nondimensional the introduce we of ratio The ILV. coefficient ESCRT-free virial the that of below ture show We SR est ntefltmembrane flat the on density ESCRT is tal. et Liese = h ai fteseicbnigenergy binding specific the of ratio The h oa hnei energy in change total The h muto SR rtista a idt h endosome the to bind can that proteins ESCRT of amount The ρ 4 0 µ n h trcrplin(Eq. repulsion steric the and 4) γ utat h idn nryo a otnosycoated continuously flat a of energy binding the subtracts g ρ h idn nryprui ra h eodtr in term second The area. unit per energy binding the /κ n h otarea coat the and ∆ = ν 2 C E µγ h orsodn trcrplinenergy repulsion steric corresponding The . ∆E g ∆ ∆E = ∆ htw s oepestemmrn nryin energy membrane the express to use we that g ν E /(2ν E µ 2 σ s = ensadensity a defines 2 = 2 = g −2π 2 ∆E + hc itt h nrylandscape energy the dictate which κ), 5 π π C C utat h nryo a,uniformly flat, a of energy the subtracts Z Z g Z g 0 0 0 ∞ ∞ = κ orsod otema curva- mean the to corresponds ∞ ∆E + ∆ p ν σ µ(ρS 2 R E (ρS µ/γ (1 sw tr rmafltmem- flat a from start we as σ − g ) ) ∆E + . ρ 2 cos − ersn h crowd- the represent 5) 0 − ρ hc stebaseline the is which , ρ ρ σ e 0 µ ρ ψ 0 ρ = γ ytescn virial second the by )RdS o 2 µ 0 n h proportion- the and )dS ecie h ratio the describes g )R . ∆E + σ e 2ν µ ensa inverse an defines µ 2 = 2 . . n h second the and Furthermore, . , ∆ σ/ s E . (µρ µ between 0 ) ∆ and [6] [5] [4] [3] [7] E s nooemmrn rfralwECTdniyo h non- significantly, the the quite on vary and tension density can ESCRT proteins The low ESCRT membrane. a deformed the small for or between a membrane interaction for endosome increases and weak which tension a energy, surface for high binding a and for tension surface between esefo Eq. ρ from see We Eq. minimizing density ρ ESCRT By the energy. to overall respect membrane with ESCRTs the the the in minimize changes that to to suggests density curvature local which their 54), adapt (15, quickly (20, can seconds molecules few exchanged ESCRT a rapidly which are and within proteins clathrin, ESCRT cargo by individual However, of stabilized These 54). diffusion minutes. be lateral several to inhibits for shown endosome were the microdomains at microdomain ble Endosome. the on Density magnitude ESCRT of order same the expect we in Gaussian and typically the (64), and are mean which the of with curvature, range associated physiological rigidities bending expected of the number us dimensionless leads gives the turn in of data which density these 63), ESCRT 51, bining the (50, for nm estimate 10 an of to order the of in is range proteins the proteins in and typically membranes is biological between energy interaction a ebaesae hr ohpicplcraue r given are curvatures principal both by where shape, Eq. membrane vesicle. cal with the of bud, budding region membrane the neck ESCRT-free the of in center i.e., membrane the ture, the at i.e., where while curvature, vesicle, reduced positive is a density exhibits protein the contrast, (Eq. hne sti einicess emnmz h oa energy total the minimize energy we membrane increases, the region how this understand as To changes 2. Fig. in illustrated the at cos vature two, just to model number our the in parameters of free reduction of to a equivalent enables is measurements which experimental nm, 46 curvature of diameter a ILV average an mine curvature the numbers on dimensionless only depend energy and shape Profiles. brane Density ESCRT and Shapes Membrane the Endosome determines that rigidity ILV. ESCRT-free bind- bending the between of Gaussian balance curvature the and is It energy energy. ing binding penalty reduced energetic a an pays to system Gaussian due the negative time, same a the exhibits bud At membrane curvature. a membrane the of where ESCRT-free region, form outer an the in if since forms, energy energy, gains bending system Gaussian the ESCRTs, energy upstream binding contribu- energy the negative energy: a bending generate the can to that Eqs. terms tion energy two membrane are con- the There we to 2–6. curvature, contributions vesicle individual understanding the the deeper determine sider a that gain parameters the To of rigidity. bending Gaussian the = n h rnia curvatures principal the and ftemmrn dpsashrclsaewt encur- mean a with shape spherical a adopts membrane the If C with α), ρ 1 sdtrie yablnebtenbnigeeg and energy binding between balance a by determined is 7) 0 = C when Z C g h SR-reae sgvnby given is area ESCRT-free the , 2 xsadtesato h SR-otdrgo as region ESCRT-coated the of start the and axis = ∆ C α ρ C C g PNAS E 1 nrae nrgoso eaieGusa curva- Gaussian negative of regions in increases ≈ g h nl omdbtentetpo h bud the of tip the between formed angle the g h curvature The . ∆ 0 = 0 = 0.04 E 8 | g = htteECTdniyi nfr with uniform is density ESCRT the that nteca-rergo and region coat-free the in While . or ρ oebr1,2020 17, November O nm = C and (1). ρ −1 2 µ 0 σ e σ 0 = ≈ .1t 10. to 0.01 = enn h curvature the Defining . σ e 1 ∩ ∆ nFg 1C Fig. In . h SR rtisfr sta- a form proteins ESCRT The − k 1 10 .. nafltmmrn.In membrane. flat a on i.e., , E C C −6 g B ρ g ρ C 1 0 = 6) h ieo ESCRT of size The (62). T σ 0 = C g ie sarlto between relation a us gives fteECTfe region ESCRT-free the of 2 to 2 nbooia membranes biological in | 10 ∆E olw rmaspheri- a from follows , o nfr otof coat uniform a for o.117 vol. ρ . 0 −3 n e.1 edeter- we 15 ref. and µ ≈ µ ecie h ratio the describes and / 5–1.The (59–61). N/m n h Gaussian the and 0.01 ralow a or A 8 | ∆E f eel htan that reveals σ e 2 = o 46 no. nm . g C h mem- The C π/ < −2 g C 0 ρ n the and g Com- . | 0 g nthe in 2 i.e., , from 28617 (1 [8] − 6

CELL BIOLOGY BIOPHYSICS AND COMPUTATIONAL BIOLOGY Eq. 6 under the constraint that α is fixed. We have fixed physiologically realistic values, while = 1.3, σe = 0.5 and = 0.6, = 2.0 and σe = 0.1, which leads to an energy barrier that is σe = 0.9 result in an energy barrier that cannot be overcome similar to our analytical estimate based on the characteristic by thermal fluctuations and thus prevent ILV formation. For ESCRT-recruitment timescales, ∆EB ≈ 0.6 kBT. Fig. 3 shows = 2.0, σe = 0.1 and a vesicle diameter of 46 nm, the open- the quasi-static membrane shape predicted by our mathematical ing angle of the energy barrier (α = 0.04π) corresponds to an model and the ESCRT density together with the experimental ESCRT-free area of just 26 nm2. Hence, only a few ESCRT membrane shapes. From the comparison between the mathemat- proteins have to desorb to overcome the energy barrier and ical model and the experimental data we determine the opening initiate the budding of an ILV. To better understand which angle α = 0.35π, 0.5π, and 0.7π, corresponding to pit shape, U part of the energy dominates during the shape transition we shape, and Ω shape of the membrane. show the individual contributions to the energy together with We turn next to see whether the model can also help us the total membrane energy for = 2.0, σe = 0.1 in Fig. 4B. The to understand the in vitro experimental observations based on changes in both binding energy (Eq. 4) and surface tension GUVs, which have also demonstrated ESCRT-mediated mem- (Eq. 3) are small compared to the overall change in energy. brane budding in the absence of ESCRT-III. The size of the The energy generated by the steric repulsion of proteins (Eq. vesicles formed in these experiments, with a typical vesicle diam- 5) and the bending of the membrane (Eq. 2) both increase eter in the micrometer range, differs largely from that of ILVs continuously, thus opposing ILV formation. The only sizeable (9). According to Eq. 7 the size difference stems from a varia- negative contribution comes from the Gaussian bending term tion of the specific binding energy µ or the Gaussian bending (Eq. 1), which dominates the total membrane energy as the rigidity, quantified by the proportionality factor γg. We expect membrane shape approaches scission (α/π → 1). The diver- the capability of ESCRT proteins to induce a Gaussian bend- gence of the Gaussian energy as the angle approaches α/π → 1 ing rigidity to be similar in both the in vivo and in vitro systems, is a characteristic of the large local variation of the Gaussian giving nearly the same γg. The variation in vesicle size is there- bending rigidity. Theoretical studies have shown that a partially fore attributed to different values of the specific binding energy formed vesicle and a finite vesicle neck are not stable, i.e., the µ, which we suspect is lower in GUV experiments as they lack energy diverges, if the variation in Gaussian bending rigidity is cargo proteins. A significant increase of the ESCRT density was large compared to the bending rigidity associated with mean found in the vesicle neck region on GUVs (9), which is also curvature (65). predicted by our theoretical model (Fig. 3). We observe that Next, we scan the phase space of ∈ [0.5 − 2.5] and σe ∈ [0 − 1] when the membrane has an Ω shape, the ESCRT density profile in numerical simulations, where we minimize the energy (Eq. 6) has a pronounced maximum in the neck region that is six times to determine the magnitude of the energy barrier ∆EB as a func- greater than ρ0. tion of α. In Fig. 4C, we see that ∆EB decreases with increasing and decreasing σe. In Fig. 4C we illustrate the region where an Upstream ESCRT-Mediated Membrane Deformation Does Not Require ILV is formed passively, by drawing a solid line for ∆EB = 2 kBT, Energy Input. In Fig. 4A we show how the membrane energy which for a bending rigidity of κ = 10 kBT (23) is equivalent to depends on α as we go from a flat membrane (α = 0) to an ∆EB/(πκ) ≈ 0.06 in dimensionless units. In addition, we show a Ω-shaped membrane (α/π → 1) for three different combina- dashed line for ∆EB = 0.6 kBT as determined by our analyses as tions of and σe, where = 2.0 and σe = 0.1 correspond to an estimate for the energy barrier of ILV formation. The phase space in Fig. 4C shows that passive ILV formation is feasible for a wide range of values for σe and . Our mathematical model is thus robust and implies that upstream ESCRT-mediated membrane deformation does not require energy.

Neck Closure. In vitro experiments on GUVs have shown that ESCRT-III and VSP4 alone are sufﬁcient to induce vesicle formation with an inverse topology, similar to in vivo ILV budding (66). However, in ILV formation both upstream ESCRTs (ESCRT-0, -I, -II) and ESCRT-III as well as VSP4 and cargo proteins are present, which play different and crucial roles in cargo sorting and ILV formation. The mathematical description of the ILV formation puts us in the position to relate the ESCRT recruitment dynamics, which we have tracked experimentally using ESCRT-0 ﬂuorescence, with the transient membrane shapes. We determine the amount of excess ESCRT proteins (∆n) as the integral over the protein density on the membrane, where the ESCRT density exceeds the baseline value ρ0. In rescaled units ∆n reads as

Z ∞ ρ Z ∞ dφ ∆n = r 1 − ds = sin φds = 1 − cos α, [9] s∗ ρ0 s∗ ds

Fig. 3. We compare the experimentally measured endosome shapes with with s∗ the arc length where the ESCRT-coated region begins. the minimal energy shapes (Eq. 6) for different angles α, when = 2 and σe = Inserting Eq. 8 into the left-hand side of Eq. 9, we find that ∆n is 0.1. (Top row) The experimental membrane shapes (15) are grouped into expressed as an integral over φ, where we know the angle at the three categories: pit shape (α = 0.35π), U shape (α = 0.5π), and Ω shape inner and the outer boundary, with φ(s∗) = α and φ(s → ∞) = 0. (α = 0.7π). For each subgroup the average shape is shown as a solid green line, while the SD is indicated by the shaded area. The ESCRT-free vesicle bud While the membrane undergoes a shape transition from a flat sur- is shown by the solid black line, while the coated region is shown by the solid face to an Ω shape, ∆n appears to increase continuously. The red line. (Bottom row) The ESCRT density is shown along the membrane for experimental measurement of the fluorescent signal of ESCRT- the three characteristic shapes, which exhibits an elevated ESCRT density in 0 is proportional to ∆n. Hence, Eq. 9 enables us to determine the neck region. the shape evolution in terms of the time dependency of α from

28618 | www.pnas.org/cgi/doi/10.1073/pnas.2014228117 Liese et al. Downloaded by guest on September 28, 2021 Downloaded by guest on September 28, 2021 pooigIVfrain nry ti la htbt steric both that clear is negative and It energy. formation) ILV formation) ILV (inhibiting positive (promoting a produces Eq. part energy which total trate the with together 1–5) in to shown set transition shape membrane α/π the to (Inset corresponds 3. Fig. line) red (solid I is tal. et Liese plot we membrane, the of 4. Fig. 2.0, = .4 ocnetteeeg nouiso k of units into energy the convert To 0.04. = κ σ e oudrtn o h nrycagswt h hp changes shape the with changes energy the how understand To (A) 0k 10 = .;II) 0.1; = h aiu fteeeg for energy the of maximum The ) B epo h ifrn otiuin of contributions different the plot We (B) (23). T 1.3, = σ e .;adIII) and 0.5; = ∆E B safnto fteoeigangle opening the of function a as 0.6, = 6 o I) for σ e 2.0, = ..Teeeg landscape energy The 0.9. = B 2.0, = ,tebnigrgdt is rigidity bending the T, σ e . sfudaround found is 0.1 = σ e . ohl illus- help to 0.1 = ∆E α o I) for (Eqs. ftersae asincraueas curvature magnitude Gaussian the rescaled define the we neck, of the ILV quantify the To in (27). preferred curvature membrane the a membrane of exhibits regions ESCRT-III curved and to formation binding ILV of step final accordance 5A (Fig. in observations formation experimental vesicle the jump-like with a The than rather Methods). tinuous and (Materials of evolution time signal corresponding ESCRT-0 fluorescent the natbd eonzn h xrclua ato h epidermal the of part extracellular prebound the we recognizing endosomes, antibody formed an newly mark of ESCRT-III– To absence on cells. the microscopy depleted in electron deformed performed be limiting we formed can ESCRT-III, the newly endosome whether in the depth, of investigate membrane in To process ILV endosomes. com- the EGF-induced study endosomal S7B), to Fig. reach us Appendix, good allowing and (SI cells a endocytosed ESCRT-III–depleted in be of partments still indicative factor could growth as (EGF) epidermal findings the Importantly, these ( efficiency. when knockdown interpret perturbed expected We as is endosomes, (68–70). machinery and to ESCRT endosomes HRS the late of enlarged hyperrecruitment to (SI a led blotting depletion showed Western ESCRT-III experiments pronounced by that Immunofluorescence a verified S7A). to as Fig. led proteins Appendix, Small these CHMP4 endosomes. suf- of and depletion of -II, CHMP2A membrane ESCRT- ,and components the III of limiting -I knockdown (siRNA)-mediated the ESCRT-0, RNA interfering deform i.e., hypoth- our to ESCRTs testing fice upstream (67), ATPase of that VPS4 absence esis the experimen- of precludes the also predictions recruitment which in the components, theoretical ESCRT-III membrane depleted the we endosomal tally, verify mem- the inward To observe in ESCRT-III. to predicts buds constriction abrupt thus brane neck model observed Our to the accumu- scission. leading triggers and the ESCRT-III/VPS4, ESCRTs by of upstream steep obtained recruitment the the curvature that of Gaussian implies lation the of model shape in our the and Moreover, decrease size in ILV. the deformation define forming and membrane the input ESCRT drive energy upstream of can absence in endosome the increase the the on that density implies model Predictions. ematical Theoretical the of Verification Experimental scission membrane to the ILV. leads an that subsequently forming and speculate ESCRT-III we of curvature bly region, Gaussian the the neck and of decrease ESCRT-III the rigid- steep of in bending signal curvature and fluorescent Gaussian the tension comparing membrane By ity. in fluctuations marginally toward only parameters depending shape the membrane on the curve of single feature the a key though on a to even depends that on strongly illustrates barrier collapse It energy points behavior. data same all the exhibiting that see we where 5C and Fig. region. -coated and shows ESCRT-free the between boundary are fIVformation, ILV of by barrier illustrated is logarithmic when formation a line ILV with solid passive map the for color barrier a energy as threshold shown The is scale. barrier barrier energy energy the the on of influence magnitude their determine to simulations cal and curvature) magni- (α/π in place grows takes formation and ILV negative as becomes tude energy costly, bending energetically are Gaussian membrane the the while of bending and proteins of repulsion rtoo h edn iiiisascae ihtema n h Gaussian the and mean the with associated rigidities bending the of (ratio SR-I–eitdsiso ftemmrn eki the is neck membrane the of scission ESCRT-III–mediated σ e ∈ K [0.0−1 g o i obntosof combinations six for σ e PNAS rtoo ufc eso n idn nry nnumeri- in energy) binding and tension surface of (ratio .. ntergm fpsieIVformation, ILV passive of regime the in i.e., 0], ∆E B σ e | . k 2.0 = ∆E and oebr1,2020 17, November B . k 0.6 = B .. h ebaesaei robust is shape membrane the i.e., ; n h nltcletmt o h energy the for estimate analytical the and T B ssonb ahdline. dashed a by shown is T eel con- a reveals 5B, Fig. in shown α, ) ( C →1). h asincraueis curvature Gaussian the , σ e and | K esa h hs pc in space phase the scan We ) B) S7 Fig. Appendix , SI o.117 vol. g K := with , g n e.15). ref. and rgesteassem- the triggers C C 1 | C g 2 o 46 no. 2 ∈ neck u math- Our [2.0−25] ∆E | tthe at B 28619 The .

CELL BIOLOGY BIOPHYSICS AND COMPUTATIONAL BIOLOGY A To analyze the stages of ILV formation, we sorted the budding profiles according to their morphology and observed a reduction in the number of constricted buds (Ω shapes) in the absence of CHMP4 (Fig. 6), which is considered to be the dominat- ing subunit in ESCRT-III filaments (71). Importantly, in the absence of CHMP4 filaments, the limiting membrane of endosomes showed an accumulation of nonconstricted invaginations, likely mediated by the upstream ESCRT machinery (ESCRT-0, -I, and -II). Depletion of CHMP2A, which is the most downstream ESCRT-III filament and which serves to recruit the ATPase VPS4 (67), resulted in a higher proportion of Ω-shaped buds when compared to control or CHMP4 knockdown, arguing for a stalled process of ILV formation at a late stage directly before scission. To summarize, our experimental data support the findings from our mathematical model, where the B upstream ESCRT machinery starts the membrane deformation process through protein crowding, likely involving transmembrane cargo proteins, and this does not require energy in the form of ATP. Discussion By combining mathematical modeling and cell biological data we are able to point to the biophysical determinants that facilitate C ILV budding by upstream ESCRTs, which is a consequence of the interplay between ESCRT-induced Gaussian bending rigidity and their crowding on the membrane. Our mathematical model highlights that while ESCRT dissociation at the budding site is biologically desirable, since it enables the cell to reuse the ESCRT proteins, it is also a physical prerequisite to form an ILV as the systems benefits from the negative energy from Gaussian bending if the Gaussian bending rigidity changes in a step-like manner, i.e., at the transition from an ESCRT-free to an ESCRT-crowded region. The ILV size is set by a balance Fig. 5. (A) The experimental signal intensity of fluorescently labeled HRS between the loss of binding energy in the ESCRT-free region (ESCRT-0) and CHMP4B (ESCRT-III) is shown as a function of time. The aver- and the increase of the Gaussian bending energy in the neck age over 23 isolated signal curves is shown together with the SD (shaded region. Our model predicts 1) a high density of ESCRTs in the area). The fluorescent signal of ESCRT-0 is proportional to the amount of membrane neck, 2) the shape of the endosome membrane dur- adsorbed proteins, which allows us correlate the magnitude of the fluores- ing budding (pit, U, and Ω shape), and 3) vesicle formation to be cent signal with the opening angle α (Materials and Methods). Modified a continuous process in time, all in accordance with experimental from ref. 15, which is licensed under CC BY 4.0.(B) By combining the experimentally measured time evolution of the fluorescent marker with the observations. theoretically predicted membrane shapes, we determine the time evolution By treating the experimental timescales for recruitment of of α according to the fit in A. (C) The Gaussian curvature in the neck, i.e., ESCRTs as a diffusive process we analytically predict the energy at the boundary between the ESCRT-free and -coated region, is determined barrier the system must overcome to form a vesicle ∆EB ≈ 0.6 for six different combinations of , σe, with ∈ [2.0−2.5] and σe ∈ [0.0−1.0]. kBT. Furthermore, our mathematical model that accounts for The Gaussian curvature strongly increases in magnitude as the membrane membrane bending, binding, and crowding of proteins, and the transitions from a flat to an Ω shape. spatial distribution of the upstream ESCRTs, shows that membrane budding is a self-organized passive process that does not need ATP consumption, which explains why it is sufficient for the growth factor receptor (EGFR) and added a secondary antibody ATPase VPS4 to bind to the ESCRT complex in the final stage conjugated to 10-nm gold particles, keeping the cells during the of membrane constriction and scission (19). These findings are whole labeling procedure at 4 ◦C to stall endocytosis. Then we perfectly in line with results from GUV studies showing that ILV stimulated endocytic uptake of EGFR by incubating cells with budding is possible through addition of various ESCRT subunits, EGF ligand for 12 min at 37 ◦C before high-pressure freezing, but does not require VPS4 and ATP (9, 66, 72). By scanning the freeze substitution, and electron microscopy. At 12 min of EGF phase space in the ratio of bending rigidities () and the ratio of stimulation, ESCRTs were shown previously to be very active in the surface tension and the binding energy (σe) in numerical simu- ILV formation (15). From these electron microscopy sections, lations, we show that ILVs may form passively over a wide range we counted the number of formed ILVs relative to the endo- of parameters. Thus, to inhibit membrane budding the system some area. ESCRT-III depletion resulted in a reduced number must be perturbed such that the energy barrier exceeds beyond of abscised ILVs when compared to nontargeting control (SI the range that can be affected by thermal fluctuations. We pre- Appendix, Fig. S7 C–E), indicating that the overall ILV formation dict that a change in membrane tension by a hypertonic shock was impaired. Importantly, we were still able to observe forming (changing σe) would suppress ILV formation. Indeed, it has been ILV buds in ESCRT-III–depleted cells (siControl, 21; siCHMP4, elegantly shown in a GUV model system that low membrane ten- 19; and siCHMP2A, 19 from 47, 47, and 31 endosomes, respec- sion favors ILV formation, while high membrane tension inhibits tively), supporting the notion that buds do form in the absence ILV formation (72). The energy landscape is also sensitive to of ESCRT-III. Given the lower number in abscised ILVs, but the changes of the bending rigidities (changing ), which leads us unchanged or even slightly increased number in buds (Fig. 6), we to speculate about the role of the clathrin layer that is bound to speculated that ESCRT-III depletion might stall or slow down the ESCRT microdomain. The physical properties of the clathrin the progression of ILV formation. layer are not yet understood. We have shown earlier that ILV

28620 | www.pnas.org/cgi/doi/10.1073/pnas.2014228117 Liese et al. Downloaded by guest on September 28, 2021 Downloaded by guest on September 28, 2021 is tal. et Liese TaeVS 1,6,7) hsnto sspotdb our by supported is notion This 71). 67, the (12, com- recruit the to VPS4 recruits functions filament ATPase and CHMP2A ESCRT-III machinery, neck the ESCRT-III membrane plete of the in component oligomerizes dif- Whereas main that process. have the shaping to is membrane expected CHMP4B the are in complex functions to ESCRT-III ferential cases the both of different in the subunits prerequisite assembly, Upon a (27). assembly a likely ESCRT-III of promote effec- is formation neck an the membrane particular, cause curved In proteins morphologically. ESCRT-induced resembles budding Gag 74), ILV (73, the curvature where spontaneous longer tive budding, over a recruitment HIV over spike-like a (73), membrane time. shows budding the ESCRT-III while on HIV span, assemble time in proteins found Gag facilitate been where have to dynamics ESCRT-III assembly similar recruitment which Qualitatively ESCRT-III (27). region, membranes neg- for curved to neck atively preferentially trigger binds the ESCRT-III a since in scission, membrane is curvature believe Gaussian we the in decline upstream process. of budding role the the in underscoring ESCRTs ESCRT-III, of form a absence indeed the profiles forming our in budding Moreover, that by formation. show data ILV penalty microscopy to steric electron energetic pathway the in an in bud evades increase increase. flat membrane energy system significant a an the and a On proteins Instead, to ESCRT minutes. the lead between several repulsion would over this membrane membrane endosome enriched as get the Fluorescence ESCRT-I ESCRTs and at budding: ESCRT-0 upstream that membrane show the GUVs data initial to microscopy on the point work for data earlier membrane determinants our the with Importantly, in accordance also (9). not in but role this remodeling crucial proteins, Together a cargo shape play process. sequestering ESCRTs budding upstream in that the and only argue endosomes in to different us points cat- of leads shape time images these random TEM of the either at for in samples found in egories overweight no an is to there shape, shape, pit U a from a transition to jump-like than rather continuous a a gives to angle which opening the ESCRT-0, of of prediction intensity experimen- the fluorescent in measured increase the tally to it correlate to model theoretical energy binding barrier. the of the decrease cor- in effective binding rationalize clathrin an now of to absence can responds where we model, our which of (15), recruit- framework clathrin endosomes of The absence to total. the ment in in observed impaired profiles severely of is number formation the and indicates identified n and were percentage, endosomes nm.) in the 40 categories bar, different of (Scale the membrane shapes. in membrane limiting profiles representative the of show amount in micrographs the profiles represents budding graph detectable The and “pit”). all U, and pit, sections in into EM described organized 200-nm-thick as microscopy individual electron of by tomograms analyzed and EGFR, gold-labeled nalize 6. Fig. srdcd n tlsIVfraintruhahge energy higher a through formation ILV stalls and reduced) is h ebaesaetasto sacmaidb steep a by accompanied is transition shape membrane The quasi-static the from density ESCRT predicted the used We eaclswr rnfce ihcnrlRAo iN agtn HP sfrso HPAfr4 ,siuae ihEFfr1 i ointer- to min 12 for EGF with stimulated h, 48 for CHMP2A or isoforms CHMP4 targeting siRNA or RNA control with transfected were cells HeLa Ω hp,i ocrac ihtefc that fact the with concordance in shape, Ω hpsb esrn h ai ewe h ekwdhadtebddph(width/depth depth bud the and width neck the between ratio the measuring by shapes α vrtm.Oraayi points analysis Our time. over µ (i.e., GRgl oiieedsmswr olce from collected were endosomes positive EGFR-gold Methods . and Materials a ensonta vngenfloecn rti GP can (GFP) protein fluorescent green It even 78). that 57, shown (56, been deformation has an membrane as now for by help such principle is also established crowding processes may Protein remodeling formation. it ILV membrane as CD63-dependent ESCRTs, other upstream understand by us ILVs understand- vary- that of beyond spatially suggests budding implications have ing membranes a can cell model and on developed (15). crowding the rigidity ILV bending protein forming Gaussian of the ing nature above directly generic weakened The locally ESCRTs/clathrin is the representing coat micrographs density electron electron from the prerequi- examples formation that physical since show ILV ILV a an be of abscise and might rounds to dissociation site membrane new this for endosomal addition, cytosol In the (67). the from into recycled dissociate aided are clathrin, can and in VPS4, ESCRTs disperse while by to membrane, molecules ILV 77). cargo forming (76, the transmembrane weakens machinery the ESCRT gradually allows and which This cargo cargo, between the interaction of the deubiquitination lead- to recruited, In ing are neck. enzymes forming deubiquitinating and the VPS4 in parallel, Gaussian accumulate negative subsequently for will preference endosomal curvature, a particu- with the in subunits ESCRTs, of ESCRT-III deformation. microdomain lar initial a its in causing crowding membrane, protein together to transmem- clathrin leads and ubiquitinated ESCRTs, upstream that of molecules, suggest cargo and consisting We brane molecules. supercomplex process sorting initial remodeling cargo as an clathrin membrane role the with their to in together adding endosome. proteins ESCRTs the cargo upstream at and shape the membrane is implicates in ESCRTs change This upstream the of of signal hallmark fluorescent a the in in revert increase but sured might ESCRTs, profiles some upstream factor, the shape. stabilizing pit of a a the to crowding of In absence the form. the constricted to continuously a due will to buds form stabilizing membrane transition filaments, a CHMP4 the have of in absence could aids filaments which in CHMP4 role and thus addition, and constriction that machinery In ESCRT-III membrane suggests VPS4. complete the in finding membrane of functions recruitment this in the only addition, CHMP4 not of In CHMP4 role (75). This expected constriction shapes. VPS4. the pit neck with mainly of accumula- profiles, consistent absence an budding is observed the unconstricted we of in however, tion cells, scission depleted fail- membrane CHMP4 a In mediate with of consistent to accumulation expected, ure as an buds caused profiles membrane depletion cells budding constricted depleted CHMP2A the CHMP2A versus different: of CHMP4 are in shapes observed be the can that where data, experimental oehrorosrain hwta h xeietlymea- experimentally the that show observations our Together PNAS | oebr1,2020 17, November .,“Ω <0.7, | o.117 vol. ;07t .,“U”; 1.7, to 0.7 ”; | o 46 no. | 28621 >1.7,

CELL BIOLOGY BIOPHYSICS AND COMPUTATIONAL BIOLOGY bend membranes when sufficiently crowded (57), arguing for a ice to decrease the fluorescent signal from the cytosolic pool of proteins rather universal principle. Our data point to a crucial role for before fixation in 3% formaldehyde for 15 min (81). Cells were washed upstream ESCRTs and clathrin (this study and ref. 15) in mam- twice in PBS and once in PBS containing 0.05% saponin before stain- malian cells. In contrast, yeast cells do not require endosomal ing with the indicated primary antibodies for 1 h. After washing three clathrin for ILV formation (79) and might therefore rather rely times in 0.05% saponin in PBS, cells were stained with secondary antibodies for 1 h and washed three times in PBS. The cells were mounted on VPS4 and cargo molecules to achieve sufficient crowding (20). in Mowiol containing 2 mg·mL−1 Hoechst 33342 (Sigma-Aldrich). Confocal Our results highlight that there are many more questions about fluorescence microscopy was done with a Zeiss LSM780 microscope (Carl these configurations that remain to be understood, which require Zeiss MicroImaging GmbH) using standard filter sets and laser lines and a further model refinement and validation with targeted manipu- Plan Apo 63 × 1.4 N.A. oil lens. All images within one dataset were taken at lations of membrane properties and ESCRT/cargo affinities in fixed intensity settings below saturation. either in vitro or in vivo model systems. Electron Microscopy and Measurements. HeLa cells were grown on poly-L- Materials and Methods lysine–coated sapphire discs. To label newly internalized EGFR following Cell Culture and siRNA Transfections. HeLa (Kyoto) cells (obtained from EGF stimulation, cells were first washed with ice-cold PBS and incubated D. Gerlich, Insitute of Molecular Biotechnology, Wien, Austria) were grown on ice with an antibody recognizing the extracellular part of EGFR (mouse according to American Type Culture Collection guidelines in Dulbecco’s anti-EGFR; Pharmingen). After washing four times with ice-cold PBS, cells modified Eagle’s medium (DMEM) high glucose (Sigma-Aldrich) supple- were incubated with Protein A–10-nm gold conjugate (University Medical mented with 10% fetal calf serum (FCS), 100 units·mL−1 penicillin, 100 Center Utrecht Department of Cell Biology) which recognizes the Fc por- −1 ◦ tion of the mouse IgG2b primary antibody. Cells were again washed four µg·mL streptomycin and maintained at 37 C under 5% CO2. The cell −1 line is authenticated by genotyping and regularly tested for mycoplasma times with ice-cold PBS and then stimulated with 50 ng·mL EGF in warm contamination. Cells were transfected using Lipofectamine RNAiMax DMEM for 12 min before high-pressure freezing was done. Sapphire discs transfection reagent (Life Technologies) following the manufacturer’s were high-pressure frozen using a Leica HPM100, and freeze substitution instructions. All siRNAs were purchased from Ambion (Thermo Fisher was performed according to the following schemes: 1) Sapphire discs were Scientific) and contained the Silencer Select modification. The following transferred to sample carriers containing freeze substitution medium (0.1% −1 siRNA sequences were used: CHMP4B1 (50-CAUCGAGUUCCAGCGGGAGtt-30), (wt·vol ) uranyl acetate in acetone, 1% H2O) and placed in a Leica AFS2 for ◦ CHMP4B2 (50-AGAAAGAAGAGGAGGACGtt-30), CHMP4A (50-CCCUGGAG- automated freeze substitution according to the following program: −90 C ◦ ◦ UUUCAGCGUGAtt-30), CHMP4C (50-AAUCGAAUCCAGAGAGAAAtt-30), for 48 h, temperature increase to −45 C for 9 h, −45 C for 5 h, 3× ◦ ◦ and CHMP2A (50-AAGAUGAAGAGGAGAGUGAtt-30). Experiments were wash with acetone, temperature increase from −45 C to −35 C, Lowicryl performed 48 h after transfection. Nontargeting control Silencer Select infiltration with stepwise increase in Lowicryl HM20 concentration (10, 25, ◦ siRNA (predesigned, catalog no. 4390844) was used as a control. The first and 75%, 4 h each) and concomitant temperature increase to −25 C, EM experiment was done with 50 nM siCHMP4B2, and in the other two EM 3 × 10 h with 100% Lowicryl HM20, UV polymerization for 48 h, tem- ◦ experiments a cotransfection of CHMP4B2, siCHMP4A, and siCHMP4C (25 nM perature increase to +20 C, UV polymerization for 24 h. 2) Sapphire each) was done to avoid potential compensation of CHMP4A and CHMP4C. discs were transferred to cryovials containing freeze-substitution medium We did not notice a more penetrant phenotype in the triple knockdown (0.5% uranyl acetate, 1% OsO4, and 0.25% glutaraldehyde in acetone) compared to depletion of CHMP4B only. A total of 50 nM siCHMP2A was and placed in a Leica AFS2 for automated freeze substitution accord- ◦ used for both experiments. ing to the following program: −90 C for 12 h, temperature increase to −60 ◦C for 6 h, −60 ◦C for 2 h, temperature increase to −20 ◦C for 30 min, temperature increase from −20 ◦C to 4 ◦C for 15 min. The Antibodies and Reagents. Antibodies used were as follows: rabbit-anti- ◦ CHMP2A (10477-1-AP from Proteintech, immunofluorescence 1:500), mouse samples were left at 4 C for 30 min before transfer to room tempera- anti–β-actin (A5316 from Sigma-Aldrich, Western blot 1:10,000), rabbit ture, 3× wash in acetone, and epon embedding with stepwise increase anti-CHMP4A (sc-67229 from Santa Cruz, Western blot 1:500), rabbit anti- in epon concentration (25, 50, 75, and 100%). The 200-nm sections were CHMP4B generated as described previously (80) (Western blot 1:1,000), cut on an Ultracut UCT ultramicrotome (Leica) and collected on formvar- rabbit anti-HRS (immunofluorescence 1:100) described previously (17), coated slot grids. Samples were imaged using a Thermo Scientific Talos F200 C microscope equipped with a Ceta 16 M camera. Single-axes tomog- mouse anti-LAMP1 (H4A3 from Developmental Studies Hybridoma Bank, ◦ ◦ ◦ immunofluorescence 1:1,000), and mouse anti-EGFR (555996 from Pharmin- raphy tilt series were recorded between −60 and 60 tilt angles with 2 gen, extracellular labeling of EGFR). All secondary antibodies used for increment. Tomograms were computed in IMOD using weighted back pro- immunofluorescence studies were obtained from Jackson ImmunoResearch jection (82). Measurements of endosome areas and ILV numbers were done Laboratories or from Molecular Probes (Life Technologies). Secondary anti- in FIJI (83). bodies used for Western blotting were obtained from LI-COR Biosciences GmbH. Energy Minimization. We rescale all lengths with Cg giving the dimensionless variables s = Cg · S, r = Cg · R, ψ(S) → φ(s). By introducing this scaling of the Immunoblotting. Cells were washed with ice-cold phosphate-buffered saline variables we rewrite Eq. 6 in dimensionless form as (PBS) and lysed with NP40 lysis buffer (0.1% NP40, 50 mM Tris·HCl, " ∆E Z ∞ dφ sin φ 2 ρ 2 pH 7.5, 100 mM NaCl, 0.2 mM ethylenediaminetetraacetic acid [EDTA], = r + + r − 1 [10] 10% glycerol) supplemented with “Complete EDTA-free protease inhibitor πκ 0 ds r ρ0 cocktail” (05056489001 from Roche). Cell lysates were spun for 5 min ρ dφ at 16,000 × g at 4 ◦C to remove nuclei. The supernatant was mixed +2 sin φ + σe r(1 − cos φ) ds. ρ0 ds with 4× sodium dodecyl sulfate (SDS) sample buffer (200 mM Tris·HCl, pH 6.8, 8% SDS polyacrylamide gel electrophoresis, 0.4% Bromophenol For any angle α the membrane shape in the inner region is described by blue, 40% glycerol) supplemented with dithiothreitol (0.1 M end con- a spherical cap with rescales mean curvature 1. The shape and the energy centration) and then subjected to SDS/PAGE on 12% or 4 to 20% gra- contribution in rescaled units are hence obtained analytically as r = sin φ, 2 dient gels (mini-PROTEAN TGX; Bio-Rad). Proteins were transferred to z = 1 − cos φ, and Einner = πκ(1 − cos α) [4 + (σe + 1)]− σ/e 2 sin α. To min- PVDF membranes (TransBlot Turbo LF PVDF; Bio-Rad) followed by anti- imize the total energy in the outer region numerically, we define a total arc body incubation in 2% BSA in Tris-buffered saline with 0.1% Tween20. length send at which the angle φ reaches zero. We set send = 15 to approxi- Membranes incubated with fluorescent secondary antibodies (IRDye680 mate the limit of an infinitely large surface. Similar to the method described or IRDye800; LI-COR) were developed with an Odyssey infrared scanner by Rozycki et al. (43) we describe the angle φ in the outer region as a (LI-COR). truncated Fourier series:

25 Immunostaining and Confocal Fluorescence Microscopy. Cells grown on cov- s X s φ(s) = α 1 − + φi sin iπ . [11] erslips were incubated with 50 ng·mL−1 EGF-Al647 (E35351; Thermo Fisher s s end i=1 end Scientiﬁc) for 10 min at 37 ◦C. Cells were then placed on ice and per- meabilized with ice-cold PEM buffer (80 mM K-Pipes, pH 6.8, 5 mM The radius r is then obtained from the relation dr/ds = cos φ. The

ethylene glycol tetraacetic acid (EGTA), and 1 mM MgCl2) supplemented prefactors φi are obtained by minimizing the energy Eq. 10 using the python with 0.05% saponin (S7900-25 g from Merck Life Science) for 5 min on basinhopping routine (84) (SI Appendix).

28622 | www.pnas.org/cgi/doi/10.1073/pnas.2014228117 Liese et al. Downloaded by guest on September 28, 2021 Downloaded by guest on September 28, 2021 8 .Scs,S re .Orco,G .Sru,J lmemn iaee ltrncoats clathrin Bilayered Klumperman, J. to Strous, J. clathrin G. recruits Oorschot, HRS V. Urbe, Stenmark, S. H. Sachse, Stang, M. E. Mehlum, 18. A. sort- Bache, protein G. K. Distinct Raiborg, Marks, C. S. M. 17. Berson, F. J. Murphy, M. D. Tenza, D. Raposo, G. 16. Wenzel M. E. 15. Raiborg C. 14. and functions Cellular Stenmark, H. Campsteijn, C. Wenzel, M. E. Raiborg, C. Christ, L. everywhere. are ESCRTs 12. Hurley, H. J. 11. 3 .Dmv,Rcn eeomnsi h edo edn iiiymaueet on measurements rigidity bending of field the in developments Recent Dimova, R. 23. Adell Y. A. M. 19. ESCRTs. of functions many The Stenmark, H. Radulovic, M. Vietri, M. 13. pathway. ESCRT The Emr, D. S. Buchkovich, J. N. Henne, M. W. 10. 8 .Ei,G arkn,M .Kzo,J ipnotShat,Cmuainlmdlof model Computational Lippincott-Schwartz, J. Kozlov, M. M. Fabrikant, G. Elia, curvature N. membrane Negative 28. Hurley, H. J. Groves, T. J. Carlson, A. L. Kai, H. Lee, H. I. 27. Quinney B. K. 24. experiments. possible and Theory - bilayers lipid of properties Elastic Helfrich, W. 22. Schoneberg J. 21. Adell Y. A. M. 20. 6 .E Mierzwa E. B. 26. Chiaruttini N. 25. ecie h hrceitctm vrwihtefloecn inlincreases signal fluorescent the which over time characteristic parameters the describes fit the with to 5A) (Fig. signal how fluorescent predict to us allows h ursetsga ocrepn to correspond to ILV, signal formed fluorescent fully a the to closest are which shapes, are membrane the that recall we signal cent Fit. Signal Fluorescent is tal. et Liese .T olr,J .Hre,Mlclrmcaimo utvsclrbd ignssby biogenesis body multivesicular of mechanism Molecular Hurley, H. J. mediated Wollert, Exosome T. Amiji, 9. M. M. Suresh, M. Mattheolabakis, G. Singh, A. Milane, L. 8. Belov L. 7. required complexes interactions intercellular and sorting secretion, Biogenesis, endosomal Thery, C. Raposo, the G. Colombo, of M. 6. role The Teis, D. Mattissek, ubiquitylated C. of sorting endosomal 5. in machinery ESCRT The Stenmark, molecular endosomes. H. Raiborg, multivesicular Intertwining C. into sorting 4. Protein signaling: Stenmark, H. and Rusten, E. T. Endocytosis Raiborg, C. Zastrow, von 3. M. Sorkin, cell. the into A. entry of 2. portals Regulated Schmid, L. S. Conner, D. S. 1. feryendosomes. early of nedsmlvcoe r novdi rti otn oadlysosomes. toward sorting protein in Cell involved are vacuoles endosomal on endosomes. early pigmented in lysosomes and cells. melanosomes, melanocytic premelanosomes, to localization and ing formation. vesicle intraluminal of (2018). morphology and timing Biol. Cell machinery. membrane-scission Sci. ESCRT the of mechanisms molecular (2011). complexes. ESCRT (2015). microenvironment. tumor the within communication samples. blood from (2016). signatures 25355 diagnostic provide may cells vesicles. extracellular (2014). other and exosomes tumorigenesis. of in (ESCRT) (2014). transport for proteins. membrane Biol. Cell Opin. Curr. networks. (2003). membranes. C. Naturforsch. VPS4. and ESCRT-III formation. vesicle MVB during constriction yoiei bcsindie yECTIIplmrzto n remodeling. (2012). and 2309–2320 polymerization 102, ESCRT-III by driven abscission cytokinetic (ESCRT)-III transport for required complex assembly. sorting endosomal of nucleation catalyzes cytokinesis. (2017). during remodeling membrane mediate to Vps4 deformation. machinery. U.S.A. ESCRT late-acting Sci. the Acad. of recruitment prolonging by biogenesis budding. membrane reverse for implications Ω hpdwt nangle an with shaped 25 (2017). 42–56 42, 3312 (2002). 1313–1328 13, xesv ufc rti rfie fetaellrvsce rmcancer from vesicles extracellular of profiles protein surface Extensive al., et 54 (2020). 25–42 21, f f rc al cd c.U.S.A. Sci. Acad. Natl. Proc. a.Rv o.Cl Biol. Cell Mol. Rev. Nat. (t R ot bqiiae rtisit ltrncae microdomains clathrin-coated into proteins ubiquitinated sorts HRS al., et sasmdt epootoa to proportional be to assumed is ) d.ClodItraeSci. Interface Colloid Adv. Cell = oriae idn fVS oECTIIdie ebaeneck membrane drives ESCRT-III to VPS4 of binding Coordinated al., et yai uui unvri SR-I sebisi euae by regulated is assemblies ESCRT-III in turnover subunit Dynamic al. , et erimn yaiso SR-I n P4t nooe and endosomes to VPS4 and ESCRT-III of dynamics Recruitment al., et ocre SR n ltrnrcutetwvsdfiethe define waves recruitment clathrin and ESCRT Concerted al., et eaaino oddECTIIsia pig rvsmembrane drives springs spiral ESCRT-III loaded of Relaxation al., et rwhfco tmlto rmtsmlieiua endosome multivesicular promotes stimulation factor Growth al., et T-eedn oc eeainadmmrn csinby scission membrane and generation force ATP-dependent al., et 9–0 (1973). 693–703 28, 8866 (2019). 6858–6867 116, 1 .Cl Biol. Cell J. MOJ. EMBO 6–7 (2015). 866–879 163, Nature − Science a.Cl Biol. Cell Nat. Nature (2003). 446–455 15, ∆n* cos h nest fteeprmnal esrdfluores- measured experimentally the of intensity The α(t 6–6 (2010). 864–869 464, 0852 (2001). 5008–5021 20, α α 4312 (2018). 1423–1428 362, 4–5 (2009). 445–452 458, 0 0–2 (2001). 809–823 152, α ) aisi iei h xeietb tigthe fitting by experiment the in time in varies 0.32, = with , 0.7π = 0–2 (2009). 609–622 10, 9–9 (2002). 394–398 4, MOJ. EMBO 59–59 (2015). 15892–15897 112, eteeoeasm h aiu of maximum the assume therefore We . 2–3 (2014). 225–234 208, α(t t 0 = ) .Cl Biol. Cell J. ∆n* = Elife . min, −3.9 3820 (2015). 2398–2407 34, nu e.Cl e.Biol. Dev. Cell Rev. Annu. π o.Mmr Biol. Membr. Mol. ∆n(α = " orlt ohquantities, both relate To ∆n. 362(2017). e31652 6, α 0 .CnrlRelease. Control J. + 34 (2014). 33–49 205, τ = a.Cl Biol. Cell Nat. t . i,where min, 6.6 = a.Commun. Nat. − 0.7π τ .Etael Vesicles Extracell. J. t 0 e.Cell Dev. Nature ) rnsBiochem. Trends 2 = # a.Rv Mol. Rev. Nat. , 9 which 1.59, 111–119 31, 278–294 219, 255–289 30, 787–798 19, 37–44 422, ipy.J. Biophys. rc Natl. Proc. 77–91 21, o.Biol. Mol. 2932 9, [12] 5, Z. τ 6 .Drac .Cpc ebaebnigb rti rwigi fetdb protein by Stachowiak, affected C. is J. crowding protein 57. by bending pro- Membrane Copic, membrane A. curved Derganc, J. coupling by 56. cells Shaping curvature: by endosomes Guided on Gov, coats S. N. clathrin Flat 55. Stenmark, H. Malerod, viral- L. from Wesche, resulting J. budding Raiborg, Arenavirus C. Neuman, W. 54. B. Whittaker, J. R. Schley, D. 53. Carmo, do P. M. 52. Hierro A. 51. Ren F. X. 50. Kostelansky S. M. 49. Mao X. Y. 48. Boucrot E. 47. paths. first-passage Braun and R. transition A. of shape 46. mean The Netz, R. R. Kim, K. in W. force driving 45. possible a as scaffolding Bud-neck coatless Marciniak-Czochra, A. of Mercker, model M. Membrane-elasticity Hummer, 44. G. Hurley, H. J. Boura, E. Rozycki, B. 43. and mechanochemistry surface Cell Allard, J. Lowengrub, J. Yan, M. H. Manakova, K. turgor. with 42. cells in endocytosis of mechanics Membrane Nedelec, endocytic F. Dmitrieff, during S. crosstalk Mechanochemical phase Drubin, 41. lipid G. D. by Oster, F. scission G. vesicle Sun, fluctuations D. Endocytic Y. Liu, Oster, Membrane J. G. Spakowitz, Drubin, 40. G. J. D. A. Kaksonen, M. Mehraeen, Liu, J. S. 39. Lampo, J. T. endocytosis. Cordella, of mechanochemistry The N. Oster, F. 38. G. Drubin, G. D. Sun, D. Y. Liu, J. of 37. fission Adhesion-induced cones: Boura and E. Domes 36. Lipowsky, R. Agudo-Canalejo, stresses J. membrane Protein-driven 35. Chernomordik, V. L. McMahon, T. filaments. H. curved Kozlov, by M. induced M. buckling Membrane 34. Joanny, F. J. Crow, G. J. D. Lenz, M. 33. Arii J. 32. Franceschi De N. 31. Dynamics Lippincott-Schwartz, Bleck J. M. Hurley, 30. H. J. Spurlin, A. T. Sougrat, R. Elia, N. 29. 8 .Syasa .Bdc-amln .Maznk,Edsml“ot fsignaling of “sort” Endosomal Miaczynska, M. Budick-Harmelin, N. Szymanska, E. 58. ellk oprmns h oeFclte o dacdLgtadElec- and providing Light for sup- microscopes. acknowledged relevant Advanced the are to Hospital access for for University Oslo Facilities grateful at Microscopy Programmable Core are tron Environment: The H.S. Convergence Compartments. and Sciences Cell-like A.C. UiO:Life the 263056. from Grant and Council port Research Project processing the Norway from S.L., sample funding experiments. of acknowledge knockdown microscopy gratefully the A.C. electron with and assistance R.R.M., with technical for help Munthe Else expert for Brinch ACKNOWLEDGMENTS. Appendix Availability. Data signal. fluorescent the of and uigHV1budding. HIV-1 during during machinery (ESCRT) abscission. transport in role for (2011). its required and complex cytokinesis sorting endosomal of aea confinement. lateral (2018). forces. cytoskeletal and teins microdomains. dynamic in HRS Sci. scaffolding Cell by sorting protein degradative mediate curvature. membrane cell protein-associated 1976). NJ, Cliffs, Englewood (2004). 221–225 Structure heterotetramer. ESCRT-I yeast transduction. signal and (2000). trafficking membrane in involved protein domains. bar crescent by restricted is and membranes. in curvature Gaussian Phys. budding. membrane ESCRT-induced complexes. ESCRT by induced budding vesicle velocity. travel and healing, formation, (2016). bleb 1636–1647 of determinants the Biol. Comput. formation. vesicle forces. boundary order. lattice protein clathrin destabilize Biol. scission. and budding membrane for proteins. ESCRT by membranes fission. and fusion in Lett. Rev. Phys. integrity. its regulates necks. membrane reconstituted oto:Terl fECTmcieyi euaino eetrmdae signaling receptor-mediated of regulation in machinery ESCRT pathways. of role The control: Biol. α 0 1024(2009). e1000204 7, 4–4 (2012). 944–949 14, oehrwith together 218(2015). 224108 143, SR-I eitsbdigars h ne ula ebaeand membrane nuclear inner the across budding mediates ESCRT-III al., et . 4422 (2006). 2414–2424 119, ouinsrcueo h SR- n I uecmlx Implications supercomplex: -II and ESCRT-I the of structure Solution al., et yrdsrcua oe ftecmlt ua SR- complex. ESCRT-0 human complete the of model structural Hybrid al., et tutr fteECTI nooa rfcigcomplex. trafficking endosomal ESCRT-II the of Structure al., et 0–1 (2009). 406–416 17, eprladsailognzto fECTpoenrecruitment protein ESCRT of organization spatial and Temporal al., et rsa tutr fteVSadFV admdmiso R,a HRS, of domains tandem FYVE and VHS the of structure Crystal al., et ei.Cl e.Biol. Dev. Cell Semin. ebaefiso spooe yisrino mhpti helices amphipathic of insertion by promoted is fission Membrane al., et lh-yuli nue ohpstv encraueadnegative and curvature mean positive both induces Alpha-synuclein al., et 1058(2015). e1004538 11, 311(2009). 038101 103, PNAS ifrnilGoer fCre n Surfaces and Curves of Geometry Differential rc al cd c.U.S.A. Sci. Acad. Natl. Proc. ebaebnigb rti-rti crowding. protein-protein by bending Membrane al., et h SR rti HPBat sadfuinbriron barrier diffusion a as acts CHMP2B protein ESCRT The al., et oeua rhtcueadfntoa oe ftecomplete the of model functional and architecture Molecular al., et ur pn elBiol. Cell Opin. Curr. l td aaaeicue nti ril and article this in included are data study All ici.Bohs caBiomembr. Acta Biophys. Biochim. rnsBohm Sci. Biochem. Trends t a.Commun. Nat. rc al cd c.U.S.A. Sci. Acad. Natl. Proc. 0 | etakMran msa n lik Dahl Ulrikke and Smestad Marianne thank We consfra fstdet h akrudnoise background the to due offset an for accounts oebr1,2020 17, November Cell hls rn.R o.Ln.BBo.Sci. Biol. B Lond. Soc. R. Trans. Philos. LSCmu.Biol. Comput. PLoS .Cl Sci. Cell J. 12 (2018). 11–20 74, 8–9 (2007). 485–498 129, .A.Ce.Soc. Chem. Am. J. ipy.J. Biophys. Structure 39(2018). 3379 9, rc al cd c.U.S.A. Sci. Acad. Natl. Proc. 64 (2010). 36–43 22, ipy.J. Biophys. Cell 9–0 (2010). 699–706 35, 198(2019). 217968 132, 07–08 (2006). 10277–10282 103, .R o.Interface Soc. R. J. LSCmu.Biol. Comput. PLoS 2–3 (2012). 124–136 149, 7–8 (2012). 874–886 20, 3–4 (2015). 833–843 108, | 21–21 (2014). 12211–12216 111, 1042(2018). e1006422 14, 4618 (2014). 1476–1488 106, o.117 vol. 6322 (2012). 2613–2620 134, 1215 (2016). 1152–1159 1858, | 0343(2013). 20130403 10, 1076(2012). e1002736 8, Petc-al Inc., (Prentice-Hall, o 46 no. Cell ipy.J. Biophys. 4846–4851 108, 20170115 373, 447–456 100, Nature | a.Cell Nat. .Chem. J. 28623 PLoS PLoS 431, 110, SI J.

CELL BIOLOGY BIOPHYSICS AND COMPUTATIONAL BIOLOGY 59. A. Roux et al., Role of curvature and phase transition in lipid sorting and fission of 72. A. Booth, C. J. Marklew, B. Ciani, P. A. Beales, In vitro membrane remodeling by ESCRT membrane tubules. EMBO J. 24, 1537–1545 (2005). is regulated by negative feedback from membrane tension. iScience 15, 173–184 60. T. Baumgart, S. T. Hess, W. W. Webb, Imaging coexisting fluid domains in biomem- (2019). brane models coupling curvature and line tension. Nature 425, 821–824 (2003). 73. D. S. Johnson, M. Bleck, S. M. Simon, Timing of ESCRT-III protein recruitment and 61. J. M. Allain, C. Storm, A. Roux, MB Amar, J. F. Joanny, Fission of a multiphase membrane scission during HIV-1 assembly. Elife 7, e36221 (2018). membrane tube. Phys. Rev. Lett. 93, 158104 (2004). 74. A. J. Pak et al., Immature HIV-1 lattice assembly dynamics are regulated by scaffolding 62. H. Lodish et al., “Chemical foundations” in Molecular Cell Biology (W. H. Freeman from nucleic acid and the plasma membrane. Proc. Natl. Acad. Sci. U.S.A. 114, E10056– and Co., New York, ed. 5, 2000), pp. 30–37. E10065 (2017). 63. M. S. Kostelansky et al., Structural and functional organization of the ESCRT-I 75. J. Guizetti, D. W. Gerlich, ESCRT-III polymers in membrane neck constriction. Trends trafficking complex. Cell 125, 113–126 (2006). Cell Biol. 22, 133–140 (2012). 64. R. Lipowsky, Coupling of bending and stretching deformations in vesicle membranes. 76. R. L. Williams, S. Urbe, The emerging shape of the ESCRT machinery. Nat. Rev. Mol. Adv. Colloid Interface Sci. 208, 14–24 (2014). Cell Biol. 8, 355–368 (2007). 65. F. Julicher, R. Lipowsky, Shape transformations of vesicles with intramembrane 77. S. Lata et al., Helical structures of ESCRT-III are disassembled by VPS4. Science 321, domains. Phys. Rev. E 53, 2670–2683 (1996). 1354–1357 (2008). 66. T. Wollert, C. Wunder, J. Lippincott-Schwartz, J. H. Hurley, Membrane scission by the 78. G. Guigas, M. Weiss, Effects of protein crowding on membrane systems. Biochim. ESCRT-III complex. Nature 458, 172–177 (2009). Biophys. Acta Biomembr. 1858, 2441–2450 (2016). 67. M. Babst, D. Katzmann, E. Estepa-Sabal, T. Meerloo, S. Emr, ESCRT-III: An endosome- 79. P. Bilodeau, S. Winistorfer, W. Kearney, A. Robertson, R. Piper, Vps27-Hse1 and ESCRT- associated heterooligomeric protein complex required for MVB sorting. Dev. Cell 3, I complexes cooperate to increase efficiency of sorting ubiquitinated proteins at the 271–282 (2002). endosome. J. Cell Biol. 163, 237–243 (2003). 68. E. B. Frankel et al., Ist1 regulates ESCRT-III assembly and function during multivesicu- 80. A. P. Sagona et al., PtdIns(3)P controls cytokinesis through KIF13A-mediated lar endosome biogenesis in Caenorhabditis elegans embryos. Nat. Commun. 8, 1439 recruitment of FYVE-CENT to the midbody. Nat. Cell Biol. 12, 362–371 (2010). (2017). 81. A. Simonsen, B. Bremnes, E. Ronning, R. Aasland, H. Stenmark, Syntaxin-16, a putative 69. K. Bache, A. Brech, A. Mehlum, H. Stenmark, HRS regulates multivesicular body Golgi t-SNARE. Eur. J. Cell Biol. 75, 223–231 (1998). formation via ESCRT recruitment to endosomes. J. Cell Biol. 162, 435–442 (2003). 82. J. Kremer, D. Mastronarde, J. McIntosh, Computer visualization of three-dimensional 70. J. R. Edgar, E. R. Eden, C. E. Futter, HRS- and CD63-dependent competing mechanisms image data using IMOD. J. Struct. Biol. 116, 71–76 (1996). make different sized endosomal intraluminal vesicles. Traffic 15, 197–211 (2014). 83. J. Schindelin et al., FIJI: An open-source platform for biological-image analysis. Nat. 71. D. Teis, S. Saksena, S. D. Emr, Ordered assembly of the ESCRT-III complex on endo- Methods 9, 676–682 (2012). somes is required to sequester cargo during MVB formation. Dev. Cell 15, 578–589 84. P. Virtanen et al., SciPy 1.0: Fundamental Algorithms for Scientific Computing in (2008). Python. Nat. Methods 17, 261–272 (2020).

28624 | www.pnas.org/cgi/doi/10.1073/pnas.2014228117 Liese et al. Downloaded by guest on September 28, 2021