Stochastic model of endosomal escape of Influenza virus
Thibault Lagache, Christian Sieben, Tim Meyer, Andreas Herrmann, David Holcman
NNoname manuscript No. (will be inserted by the editor)
Stochastic model of endosomal escape of Influenzavirus
Thibault Lagache · Christian Sieben · Tim Meyer · Andreas Herrmann · David Holcman
Received: XX/ Accepted: XX
Abstract Keywords
Mathematical Modeling · Markov Jump Process · WKB approximation · Endosomal Acidification · Influenza
PACS · · · Mathematics Subject Classification (2000) · · · · T. LagacheApplied Mathematics and Computational Biology, IBENS, Ecole Normale Sup´erieure, 46rue d’Ulm 75005 PARIS, FranceE-mail: [email protected]
Present address:
BioImage Analysis Unit, CNRS UMR 3671, Institut Pasteur, France.C. SiebenDepartment of Biology, Molecular Biophysics, IRI Life Sciences, Humboldt-Universit¨at zuBerlin, GermanyT. MeyerInstitute of Chemistry and Biochemistry, Free University Berlin, GermanyA. HerrmannDepartment of Biology, Molecular Biophysics, IRI Life Sciences, Humboldt-Universit¨at zuBerlin, GermanyD. HolcmanApplied Mathematics and Computational Biology, IBENS, Ecole Normale Sup´erieure, 46rue d’Ulm 75005 PARIS, France and Newton Institute,DAMTP Cambridge Cb3 0DS andMathematical Institute, University of Oxford, Andrew Wiles Building, Woodstock Rd, Ox-ford OX2 6GG, United KingdomE-mail: [email protected] a r X i v : . [ q - b i o . S C ] O c t Thibault Lagache et al.
Influenza viruses enter a cell via endocytosis after binding to the surface. Dur-ing the endosomal journey, acidification triggers a conformational change ofthe virus spike protein hemagglutinin (HA) that results in escape of the viralgenome from the endosome to the cytoplasm. A quantitative understandingof the processes involved in HA mediated fusion with the endosome is stillmissing. We develop here a stochastic model to estimate the change of confor-mation of HAs inside the endosome nanodomain. Using a Markov-jump processto model the arrival of protons to HA binding sites, we compute the kineticsof their accumulation and the mean first time for HAs to be activated. Thisanalysis reveals that HA proton binding sites possess a high chemical barrier,ensuring a stability of the spike protein at sub-acidic pH. Finally, we predictthat activating more than 3 adjacent HAs is necessary to prevent a prematurefusion.
The first step of infection by influenza starts when viral particles enter thecell by a process called endocytosis at the host cell surface, where they arecaptured in spherical endosome. The second step is the transport of the virus,trapped inside the endosome. During the third and critical step, the viralgenome, encoding ribonucleoproteins (vRNPs) has to escape from the endo-somal compartment, so that later on, it can translocate into the nucleus [23](Figure 1-A). Fusion between the endosome and influenza virus is mediatedby a low-pH conformational change of the viral envelope glycoprotein hemag-glutinin (HA) (figure 1-A). The goal of this article is to present a new modelbased on endosome acidification and conformational change of the HA to pre-dict the exact timing for initiating fusion between the virus and the endosomemembrane, and thus to release the viral genome.The model accounts for important detailed properties of the glycoprotein HAcomposed by two linked subunits HA1 and HA2, the latter anchoring HA tothe viral envelope. Indeed, at neutral pH, HA is not active (in a non-fusogenicstate), but as the pH decreases due to proton entry into the endosome, a par-tial dissociation of the HA1 subunit results in a spring-loaded conformationalchange of HA2 into an active (fusogenic) state [9]. Consequently, the residencetime of influenza virus genome within the endosome before fusion dependson the kinetics of endosome acidification. Yet, the absence of direct in vivomeasurements of these parameters makes the endosomal step of virus infectiondifficult to analyze. To estimate the pH-driven fusion of influenza viruses in anendosome, we develop a model that accounts for the main kinetics parametersof the fusion process. tochastic model of endosomal escape of Influenza virus 3
We start by developing a kinetics model for endosome acidification, that wecalibrate using experimental data. The model depends on the following pa-rameters: buffering capacity of the endosomal lumen, membrane leakage andproton pumping rate, all together controlling the number of free protons in-side the endosome. Because the proton binding event to HA is discrete, wemodel it here using a classical Markov jump process [27]. Using an asymptoticexpansion of the solution for the mean arrival time equation for the numberof protons to a certain threshold, we obtain an analytical expression for thekinetics of HA conformational change at fix pH values. The model is thencalibrated using the kinetics of the HA conformational change [16]. Finally,by combining the two models for kinetics of endosomal acidification and HAconformational change, we can estimate the number of activated HAs inside anendosome. We predict that at least three adjacent activated HAs are necessaryto trigger membrane fusion [4,11], a cooperativity process that should preventpremature fusion. We confirm some of the predictions using co-labeling virusesand endosomal markers experimental data, showing that intracellular fusionof viruses mainly occur in maturing endosomes.
The model of endosomal acidification follows the free number of protons P e ( t )at time t in the endosomal compartment. The protons enter with an entryrate λ ( t ) S through the V-ATPase proton pumps ( S is the endosomal surfaceand the rate λ ( t ) is associated with to the proton pumps activity) and canescape with a leaking rate L ext ( t ), but can also bind to endosomal buffers.The proton pump rate λ ( t ) is mainly determined by the membrane potential Ψ ( t ) (Figure 11 [7]), which in turn depends on the endosomal concentrationsof several cations ( H + , K + , N a + . . . ) and ( Cl − . . . ). The ionic concentrationinside endosome is tightly regulated by channels, exchangers and leak and inparticular, by raising the interior-positive membrane potential, Na-K ATPaseexchangers have been proposed to limit the acidification of early compared tolate endosomes [5].3.1 Mass action law for free protonsTo derive the time-dependent equations for the free protons, we use the balanceof fluxes. The fast equilibrium between fluxes determines the effective numberof protons P e ( t ) dt entering the endosome which follows the first order kinetics dP e dt = ( λ ( t ) S − L ext ( t )) . (1) Thibault Lagache et al.
Protons are rapidly bound to endosomal buffers. We model the buffer activityusing an ensemble of chemical reactions ( P e B i /B i ) ≤ i ≤ n [6]: P e + B k (cid:10) k ( − P e B , P e + B k (cid:10) k ( − P e B . . . , P e + B n k n (cid:10) k ( − n P e B n , (2)where (cid:16) k i , k ( − i (cid:17) ≤ i ≤ n are the rate constants ( P e B i /B i ) ≤ i ≤ n . Thus the ki-netics equations for the number of free protons P e ( t ) inside an endosome is dP e ( t ) dt = ∆P e ( t ) + n (cid:88) i =1 (cid:18) k ( − i P e B i ( t ) − k i N A V e P e ( t ) B i ( t ) (cid:19) = ( λ ( t ) S − L ext ( t )) + n (cid:88) i =1 (cid:18) k ( − i P e B i ( t ) − k i N A V e P e ( t ) B i ( t ) (cid:19) , where P e B i ( t ) and B i ( t ) are the number of weak acids and bases inside theendosome at time t . We assume that the membrane potential Ψ ( t ) reachesrapidly its steady state value Ψ ( ∞ ) compared to the acidification kinetics [7],we thus approximate the pumping rate λ ( t ) S as λ ( t ) S ≈ λS. (3)where the parameter λ is related to the membrane potential Ψ ( ∞ ). In addition,the protons leak L ext ( t ) is proportional to the endosomal concentration andthe endosomal surface [6] L ext ( t ) = LS P e ( t ) N A V e , (4)where L is a permeability constant, N A is the Avogadro constant and V e is thevolume of the endosome. Consequently, using approximations 3-4 in equation3, we obtain the general dynamics of free protons: dP e ( t ) dt = (cid:18) λ − L P e ( t ) N A V e (cid:19) S + n (cid:88) i =1 (cid:18) k ( − i P e B i ( t ) − k i N A V e P e ( t ) B i ( t ) , (cid:19) . ∆P e ( t ) protons haveentered the endosome, they instantaneously bind to bases, leading to a jump − ∆B i ( t ) on each base ∆P e ( t ) ≈ − n (cid:88) i =1 ∆B i ( t ) . (5) tochastic model of endosomal escape of Influenza virus 5 To estimate the associated pH change d pH with the entry ∆P e ( t ) of protonsand the infinitesimal changes − dB i ( t ) of the number of bases, we use equation2) at equilibrium and time t : k ( − i P e B i ( t ) = k i P e ( t ) B i ( t ) N A V e , for all 1 ≤ i ≤ n. (6)Thus, P e ( t ) N A V e = K i C i − B i ( t ) B i ( t ) , for all 1 ≤ i ≤ n, (7)where K i = k ( − i k i and C i = P e B i (0) + B i (0) are constant. Consequently,pH( t ) = pK i + 1log(10) log (cid:18) B i ( t )( C i − B i ( t )) (cid:19) (8)where pK i = − log( K i ) / log(10). By differentiating equation 8 with respect to B i , the infinitesimal variation dB i of base i is related to d pH of the endosomalpH by d pH = (cid:18) C i B i ( t ) ( C i − B i ( t )) (cid:19) dB i . (9)Using equation 7, we get for equation 9 d pH = (cid:32) N A V e K i + P e ( t )) P e ( t ) C i LN A V e (cid:33) dB i , (10)leading to dB i = N A V e β i ( P e ( t )) d pH , (11)where β i ( P e ( t )) = log(10) C i P e ( t ) K i ( P e ( t ) + K i N A V e ) (12)is the buffering capacity of the weak acid-base ( P e B i , B i ). Finally, using equa-tions 5 and 11 we find that the variation ∆P e ( t ) of protons is related to aninfinitesimal change ∆ pH of the endosomal pH through ∆P e ( t ) = − n (cid:88) i =1 ∆B i = − N A V e (cid:32) n (cid:88) i =1 β i ( P e ( t )) (cid:33) ∆ pH = − N A V e β e ( P e ( t )) ∆ pH , (13)where β e ( P e ( t )) = (cid:80) ni =1 β i ( P e ( t )) is the total buffering capacity of the endo-some, which is approximately constant β e ( P e ( t )) ≈ β e = 40 mM/pH [32].Finally, using the proton extrusion and pumping rates (equation 3 and 4), weobtain the kinetics equation (cid:18) λ − L P e ( t ) N A V e (cid:19) Sdt = − N A V e β e d pH . (14) Thibault Lagache et al.
With d pH( t ) dt = − P e ( t ) dP e ( t ) dt , (15)we obtain that the kinetics equation for free protons accumulation in an en-dosome is dP e ( t ) dt = (cid:18) λ − L P e ( t ) N A V e (cid:19) S log(10) P e ( t ) N A V e β e . (16)When the proton leakage is counterbalanced by the pumps, after a time longenough, the pH reaches an asymptotic value pH ∞ , where the endosome cannotbe further acidified given by P e ( ∞ ) = N A V e − pH ∞ , (17)Consequently, the rate λ is linked to pH ∞ by λ = L − pH ∞ , (18)and equation 16 can be rewritten as dP e ( t ) dt = (cid:18) − pH ∞ − P e ( t ) N A V e (cid:19) LS log(10) P e ( t ) N A V e β e . (19)To conclude, we obtain here a kinetics equation for the endosome acidificationas a function of endosome parameters and permeability. However, equation 19is not sufficient to account for the different stages of the endosomal maturation.Indeed, the final pH ∞ [23] and the permeability L were reported to decreasewith the endosomal maturation [5] and are thus time dependent.3.3 Modeling pH change and acidification of an endosomeWe now relate the acidification dynamics to the change of two proteins that canbe followed experimentally. Indeed, the transition from a first stage endosomecalled early endosome (EE) to a second stage endosome called late endosome(LE) is quantified by a gradual exchange of a protein called Rab5 by anotherone associated to the late phase called protein Rab7 [25]. Kinetics of Rabexchange have been experimentally measured and we approximate here thekinetics of the ratio Rab5/Rab7 obtained from data (Figure 4-C [25]) by asigmoidal function Rab7(t)Rab5(t) + Rab7(t) = 11 + e − ( t − t / ) /τ c , (20)where t / is the half-maturation time and τ c is the time scale of Rab con-version. We then approximate the transition rate from early to late endosome tochastic model of endosomal escape of Influenza virus 7 with the Rab conversion rate and we consider that the steady-state pH ∞ ( t ) rel-ative to the amount of Rab7, that gradually replaces Rab5 during endosomalmaturation is given bypH ∞ ( t ) = pH early ∞ + (cid:16) pH late ∞ − pH early ∞ (cid:17) Rab7(t)Rab5(t) + Rab7(t) . (21)Similarly, the permeability constant follows the equationL( t ) = L early + (cid:0) L late − L early (cid:1) Rab7(t)Rab5(t) + Rab7(t) . (22)3.4 Calibrating the acidification model by live cell imagingWe shall now calibrate the acidification equation to experimental parameters.First, by fitting equation 20 to the experimental data (Figure4-C of [25]) wherethe lag time between initiation and termination of the Rab5/Rab7 replacementis estimated to ≈
10 min., leading to a time constant for τ c ≈
100 s.We use data from endosomal acidification in MDCK cells where the pH insideendosomes decreases very quickly within the first 10-15 min (Figure 1-B) toreach a steady state pH around 5.5 after 20 min, in agreement with previousstudies [35]. The steady state pH is pH early ∞ = 6 . late ∞ = 5 . L and Rab conversion kinetics by solving numerically equation 19 andfitting the experimental acidification curve (Figure 1-C). We found that thepermeabilities of early and late endosomes are L early ≈ . − N A cm s − and L late ≈ . L early = 3 . − N A cm s − , respectively and the half-maturationtime is t / = 10 min.3.5 Accounting for proton influx inside viral core and bufferingThe last step of the kinetics model of protons include the binding and un-binding with various viral components providing buffer capacity. Indeed, thebuffering capacity of the viral proteins and the influx of protons through M2-channels inside the viral core (Figure1-A), the presence of viruses inside en-dosomes changes the overall buffering capacity of the endosome itself and canperturb the overall acidification kinetics. To compute the influx in each virusthrough M2 channels, we use first order transport kinetics [20], summarizedby the chemical reaction P e + M k e (cid:10) k − e M − P k − v (cid:10) k v P v + M . (23) Thibault Lagache et al.
When a proton P e binds a free M k e andunbinding k − e rates, it is transported inside the virus core with an inwardrate k − v , while the forward rate is k v . Thus the inward flux in a single viruscan be computed from equation 23 [20] j M ( P e , P v ) = n M k − e − k e P e α ( P e ,P v ) N A V e α ( P e , P v ) , (24)where n M is the number of M2 channels per viral particle, P v is the numberof free protons inside the viral core and α ( P e , P v ) = k − e + k − v k e (cid:16) P e N A V e + k − v P v k − e N A V v (cid:17) , (25)We extracted the buffer capacity of a virus and accounted for the viral genome,the internal viral proteins and unspecific buffers that can be reached throughthe M2 channels [20]. Most abundant internal proteins are the M1 (3 , n NPP ( pH ) of NP proteins (see Material and Methods) andwe found that n NPP ( pH ) increases almost linearly with pH: n NPP ( pH ) ≈ n NPP ( pH = 7) + 9 (7 − pH ) , (26)indicating that the buffering capacity of NPs is approximatively constant be-tween pH 7 and 5 (equation 11) β NPv ≈ N A V v = 3000 N A V v (27)where V v = πr v is the volume of the viral internal lumen, for a sphericalviral particle with radius r v = 60 nm [19]. The structure of the matrix M1protein is unknown and consequently, we use the cumulative contributions ofAsp, Glu and His residues to estimate the number of M1 proton binding sites. tochastic model of endosomal escape of Influenza virus 9 We thus estimate the fraction P i ( pH ) of occupied residues for a fixed pH usingthe equilibrium constant pKa i for any residue i (Asp, Glu or His) to be P i ( pH ) = (cid:0) pH − pKa i + 1 (cid:1) − . (28)The mean number n M P ( pH ) of protonated site is then given by n M P ( pH ) = n M Asp (cid:0) pH − . + 1 (cid:1) − + n M Glu (cid:0) pH − . + 1 (cid:1) − + n M His (cid:0) pH − . + 1 (cid:1) − . (29)where the number of residue for each group is n M Asp = 12, n M Glu = 12 and n M His = 5. Using equation 29, we plotted n M P ( pH )as function of the pH andobserved that n M P ( pH ) is almost a linear function n M P ( pH ) ≈ n NPP ( pH = 7) + 3 . − pH ) , (30)and obtain that β M v ≈ . N A V v = 10500 N A V v . (31)Additionally to internal M1s and NPs proteins, protons entering the viralcore through M2 channels can also bind to viral nucleic acids and in partic-ular to basic groups in the guanine, adenine and cytosine nucleotides [31].In particular, the buffering capacity β RNA of oligonucleotides in solution, fora concentration c monomers of monomers, has been estimated to be β RNA ≈ . c monomers in the pH range 5-7 (Figure 3-D in [31]). Consequently thebuffering capacity β RNAv of the ≈ β RNAv ≈ . N A V v = 1200 N A V v . (32)Finally, the viral core lumen should also contain other unspecific buffers such ascytoplasmic buffers enclosed during the viral assembly, leading to an unspecificbuffering capacity β v ( pH ) inside the viral lumen that has to be added to thebuffering capacities β NPv and β M v of internal proteins. Due to possible ionicexchange between viral and endosomal lumens, we approximate β v ( pH ) withthe endosomal buffering capacity β e , which is independent of the pH and hasbeen estimated to be [32] β e ≈ mM/ pH . (33)In summary, the proton buffering capacity inside viral cores is equal to β i = β v + β M v + β NPv + β RNAv , (34)and similarly to the flux equation 19, the number of free protons P v ( t ) con-tained in viral core at time t , which determines the influx of protons throughM2 channels (equation 24), follows the kinetics equation dP v ( t ) dt = log(10) N A V v ( β v + β M v + β NPv + β RNAv ) P v ( t ) j M ( P e ( t ) , P v ( t )) . (35) By solving numerically equation 35, we estimate that ≈ ,
000 protons enterthe viral core during endosomal maturation. Using endosomal acidificationkinetics equation 19, we estimate that more than 20 , ,
000 protons bind toendosomal buffers during endosome acidification. Thus, the buffering capacityof a single virus should not influence the endosomal acidification. However thenumber of protons that bind to endosomal buffers drastically decreases withthe endosomal size (e.g. ≈ ,
000 for r e = 100 nm instead of r e = 500 nm).In addition viral particles may accumulate during the endosomal journey [22].Thus, for multiplicity of infection (MOI) and viral accumulation in endosomes,the viral buffering capacity may significantly affect the acidification kineticsof small and intermediate size endosomes. Although the number of protons entering in the endosome is quite huge, asdiscussed in the previous section, the actual number of free protons definingendosomal pH is low ( ∼
300 at pH 6 in an endosome with a radius of 500 nm).In addition, there are few proton binding sites on a single HA that trigger itsconformational change [8], which is the event we shall monitor. This change ofscale between many entering protons and the few free protons and HA bindingsites requires a different description than the previous continuous model. Tocompute the mean time for HA conformation to change as the pH drop, weshall first extract the forward and backward proton binding rates. For thatpurpose, we convert the HA conformational change kinetics, obtained fromexperimental data at various pH [16] into rate constants.At temperature T = 300 K , when the pH decreases from 7 to 4, the numberof bound protons bound to HA1 increases approximatively from 123 to 132(Figure 3 in [8]), suggesting that the number of available number of binding siteis n s = 9 at acidic pH. The influenza virus carries n HA ≈
400 HA trimers [11](Figure 2-A) and thus there are exactly n HA n s sites that can competitivelybind protons. The goal of this section is to compute the mean time that athreshold n T of bound protons to HA1 is reached, which is a model of fusogenicstate, where the protein can engage into the generation of a fusion pore withthe endosomal membrane.4.1 Modeling HA conformational changeTo analyse the conformational changes of a single HA trimer, we follow theoccupied sites X ( t, c ) at time t , for a fix proton concentration c . During time t and t + ∆t , the number of specific bound sites can either increase with aprobability r ( X, c ) ∆t , when a proton arrives to a free site or decreases with tochastic model of endosomal escape of Influenza virus 11 probability l ( X, c ) ∆t when a proton unbinds or remains unchanged with prob-ability 1 − l ( X, c ) ∆t − r ( X, c ) ∆t (Figure 2-A).We estimate hereafter the rates l ( X, c ) and r ( X, c ) and the critical threshold n T . The forward rate depends on the proton concentration c and the numberof free sites n s − X of the HA trimer, thus r ( X, c ) = Kc ( n s − X ) , (36)where K is the forward binding rate of a proton to a binding site.To determine the proton unbinding rate l ( X, c ), we use the values available forthe HA1 protonation [8]. We approximate the number of bound protons ˜ X ( c )with respect to the proton concentration c by a linear function (Figure3 [8])˜ X ( c ) = ˜ X (cid:0) − mol.L − (cid:1) + X ( c ) = ˜ X (cid:0) − mol.L − (cid:1) + (cid:18)
73 + log( c )3 log 10 (cid:19) n s , (37)where ˜ X (cid:0) − mol.L − (cid:1) is the mean number of bound protons at pH=7 and X ( c ) = (cid:18)
73 + log( c )3 log 10 (cid:19) n s (38)is the mean number of HA1 sites that are additionally protonated for a pro-ton concentration c > − mol. L − . Because the unbinding rate does notdepend on the proton concentration l ( X, c ) = l ( X ) and we obtain for theequilibrium ratio l ( X,c ) r ( X,c ) = l ( X ) Kc ( n s − X ) . Using at equilibrium the concentra-tion c ( X ) = 10 Xns − for which X ( c ( X )) = X , the mass-action law leadsto l ( X ( c ) ,c ) r ( X ( c ) ,c ) = 1 or equivalently l ( X ) Kc ( X )( n s − X ) = 1. Finally, we get l ( X ) = K ( n s − X )10 Xns − . (39)In summary, the binding and unbinding rates r and l are given by r ( X, c ) = Kc ( n s − X ) , and l ( X, c ) = l ( X ) = K ( n s − X )10 Xns − . (40)4.2 Rate of HA conformational changeTo compute the mean time that exactly n T protons are bound to a single HAwe use a Markov jump process description. The Master equation is derivedby evaluating during time t and t + ∆t , the variation in the number of boundsites X ( t, c ) among the n s = 9 HA1 proton binding sites. The scaled variableis x ( t, c ) = (cid:15)X ( t, c ) , (41) where (cid:15) = 1 /n s and using the difference ∆x = x ( t + ∆t, c ) − x ( t, c ), thetransition probabilities are P rob { ∆x = (cid:15) | x ( t, c ) = x } = r ( x, c ) ∆t,P rob { ∆x = − (cid:15) | x ( t, c ) = x } = l ( x, c ) ∆t,P rob { ∆x = 0 | x ( t, c ) = x } = (1 − r ( x, c ) − l ( x, c )) ∆t. At a fixed proton concentration, the probability p ( y, t | x, c ) that the numberof protonated sites is equal to y at time t , that is x ( t, c ) = y ), when thereare initially x bound sites ( x ( t = 0 , c ) = x ), is solution of the backward-Kolmogorov equation [21,13,14,27,17] p ( y, t | x, c ) = p ( y, t − ∆t | x + (cid:15), c ) r ( x, c ) ∆t + p ( y, t − ∆t | x − (cid:15), c ) l ( x, c ) ∆t + p ( y, t − ∆t | x, c )(1 − r ( x, c ) ∆t − l ( x, c ) ∆t ) , (42)which has the classical Kramers-Moyal expansion [26] ∂p∂t ( y, t | x, c ) = L x p = r ( x, c ) ∞ (cid:88) n =1 (cid:15) n n ! ( ∂ x ) n p ( y, t | x, c )+ l ( x, c ) ∞ (cid:88) n =1 ( − (cid:15) ) n n ! ( ∂ x ) n p ( y, t | x, c ) . (43)The mean first time τ ( x, c ) that the process X ( t, c ) reaches the threshold x T = n T /n s models the HA1 subunit filled with n s protons. It is precisely themean of the first passage time for the bound protons x ( t, c ) to reach the level x T . The mean first passage time τ ( x, c ) = E [ τ | x ( t = 0 , c ) = x ], satisfies [27]: L x τ ( x, c ) = − x in [0 , x T ] , (44)with the boundary conditions τ ( x, c ) = 0 for x = x T and ∂τ ( x, c ) ∂x = 0 for x = 0 . (45)For (cid:15) (cid:28)
1, a Wentzel-Kramers-Brillouin (WKB) [34,15,3] asymptotic expan-sion of the solution τ ( x, c ) ≈ τ ( c ) is known [14,21,13,27,17] and can be writtenas τ ( c ) ≈ r ( x ( c ) , c ) (cid:115) π(cid:15) ddx (cid:0) lr (cid:1) ( x ( c ) , c ) φ ( x T , c ) , (46)where x ( c ) is the mean number of HA1 sites that are additionally protonatedfor a concentration c > − mol.L − , and the dependency of x with respectto the concentration c in the range 0 < x ( c ) = (cid:15)X ( c ) < x T has been obtainedby a fitting procedure (see equation 38) x ( c ) = 73 + log( c )3 log(10) . (47) tochastic model of endosomal escape of Influenza virus 13 Finally, by definition, φ ( x, c ) = exp (cid:32) − (cid:15) (cid:90) xx ( c ) log (cid:18) l ( s, c ) r ( s, c ) (cid:19) ds (cid:33)(cid:115) l ( x, c ) r ( x, c ) (cid:18) l ( x, c ) r ( x, c ) − (cid:19) . (48)Now replacing the transition rates r ( x, c ) and l ( x ) by their expressions 40 inequation 48 allows us to compute the mean first passage time to the threshold.Indeed, (cid:90) x T x ( c ) log (cid:18) l ( s, c ) r ( s, c ) (cid:19) ds = (cid:90) x T x ( c ) (cid:0) log (cid:0) s − (cid:1) − log( c ) (cid:1) ds = (cid:90) x T x ( c ) (3 log(10) s − (7 log(10) + log( c ))) ds, that is (cid:90) x T x ( c ) log (cid:18) l ( s, c ) r ( s, c ) (cid:19) ds = (cid:90) x T x ( c ) (cid:0) s − log(10 c ) (cid:1) ds = F ( x T ) − F ( x ( c )) , (49)where F ( x ) = 32 log(10) x − log(10 c ) x. (50)leading to φ ( x T , c ) = exp (cid:18) − (cid:15) ( F ( x T ) − F ( x ( c ))) (cid:19) x T − c − (cid:114) x T − c , (51)that is, φ ( x T , c ) = exp (cid:18) − (cid:15) ( F ( x T ) − F ( x ( c ))) (cid:19)(cid:18) x T / − / √ c − √ c / − x T / (cid:19) . (52)Using the expressions for the binding and unbinding rates 40, we get ddx (cid:18) lr (cid:19) ( x ( c ) , c ) = ddx (cid:18) x − c (cid:19) ( x ( c ) , c ) = 3 log(10) c x ( c ) − , (53)which reduces to with formula 47 to ddx (cid:18) lr (cid:19) ( x ( c ) , c ) = 3 log(10) . (54) Finally, with eqs 47-52 and 54, we obtain for the mean conformational changetime τ ( c ) = (cid:15) − Kc (cid:16) + log( c )3 log(10) (cid:17) (cid:114) π(cid:15) (cid:18) (cid:15) (cid:18) F ( x T ) − F (cid:18) / c )3 log(10) (cid:19)(cid:19)(cid:19) x T / − / √ c − √ c / − x T / . (55)Using that (cid:15) = 1 /n s and x T = n T /n s , we finally get the new expression forthe conformational change time: τ ( c ) = √ π exp (cid:18) n s (cid:18) F ( n T /n s ) − F (cid:18) / c )3 log(10) (cid:19)(cid:19)(cid:19) K (cid:112) cn s log(10) (cid:16) log( c )log(10) (cid:17) (cid:16) c / − n T / (2 n s ) − n T / (2 n s ) − / (cid:17) , (56)where F is defined in 50.Equation 56 links the affinities between the ligand (concentration c ) and thebinding sites of a trimer to the conformational change mean time τ ( c ) ofthe trimer. Interestingly, the reciprocal 1 τ ( c ) has been measured for variouspH values [16]: ( τ ( pH = 4 . − = 5 . s − , ( τ ( pH = 5 . − = 0 . s − ,. . . ,( τ ( pH = 5 . − = 0 . s − . Using formula 56 and a least square optimizationprocedure, we have approximated the data (Figure 2-B) and obtain that thecritical threshold is n T ≈ K ≈ . L.mol − s − , (58)These two estimations can also been seen as predictions of the present model.We plotted in Figure 2-B the theoretical rate change for HA-conformational1 τ ( c ) with respect to the proton concentration c and compared it with theexperimental values of [16]. We found a very good agreement (Figure 2-B),validating our jump-Markov model where the cumulative binding of few pro-tons to an activating threshold n T leads to HA conformational change.4.3 A high potential barrier of HA binding sites ensures HA stability atneutral pHWe have seen in section that during endosomal acidification, a huge number ofprotons enter the endosome (more than 20 *10 and bind mostly to endosomebuffers, leaving very few free protons (around 300 at pH 6)). To test whetherHAs buffer entering protons or interact with the remaining few free protons, tochastic model of endosomal escape of Influenza virus 15 we estimate the potential barrier generated at each HA binding site. For thispurpose, we compare the reciprocal of the forward rate constant K (equation58), which is the mean time for a proton to bind a HA protein, with the freeBrownian diffusion time scale. For a fixed proton concentration at a value c ,the proton binding time is τ bind = Kc , while the mean time for a proton todiffuse to the same binding site is [33,30,29,28] τ diff ≈ V πD p η n ( c ) (59). The number of endosomal protons at concentration c is n ( c ) = N A cV , while η is the interacting radius between a proton and a binding site and D p the dif-fusion constant of a free proton ( D p ≈ µm s − measured in the cytoplasm[1]). For η = 1 nm , we find a small ratio τ diff τ bind = K πD p ηN A ≈ − . (60)confirming that the mean time for a proton to bind HA is dominated by a veryhigh activation energy barrier at the HA binding sites, preventing rapid protonbinding. Consequently, the buffering capacity of HAs can be neglected com-pared to the high capacity of other endosomal buffers. In addition, our modelpredicts that high HA1 potential barrier guarantees that the conformationalchange is only triggered after a cumulative binding of n T = 6 protons, ensur-ing a high stability of the protein at pH above 6, as previously characterizedin Table 2 of [16] and confirmed in Figure S1.In summary, we found that the threshold for HA1 conformational change oc-curs when there are n T = 6 bond proton in a total of n s = 9 binding sites. Thebinding is characterized by a very high potential barrier. Thus, when protonsenter an endosome, they will first be captured by endosomal buffers. The re-maining pool of free protons can bind to HA1 sites when they succeed passingthe high potential barrier to ultimately trigger HA conformational change. Combining the kinetics model of endosome acidification with the Markov jumpmodel of HA conformational change, we now propose a kinetics model ofHAs conformational change inside an endosome. We account for the n T = 6protons activating a HA1 trigger leading to HA conformational change. Weshall estimate the numbers HA ( t ) , HA ( t ) . . . HA ( t ) of viral HAs that have0 , . . . t , and compute the number of fusogenic (ac-tive) HA ( t ), responsible for membrane fusion. From relation 36, the forwardrate of a proton to a free HA1 binding site is˜ r ( X ) = r ( X, P e ( t )) /P e ( t ) = Kn s ( n s − X ) N A V e . (61) and the backward rate l ( X ) is given by relation 40, thus the chemical equationsfor protons P e and HA proteins are summarized by HA + P e ˜ r (0 /n s ) (cid:10) l (1 /n s ) HA ,HA + P e ˜ r (1 /n s ) (cid:10) l (2 /n s ) HA , . . . HA + P e ˜ r (5 /n s ) −−−−−→ HA . (62)where the rate constant depends on each stage as given by relation 61. Thestage HA is irreversible and the kinetic rate equations are dHA ( t ) dt = − ˜ r (cid:18) n s (cid:19) P e ( t ) HA ( t ) + l (cid:18) n s (cid:19) HA ( t ) ,dHA ( t ) dt = (cid:18) ˜ r (cid:18) n s (cid:19) HA ( t ) − ˜ r (cid:18) n s (cid:19) HA ( t ) (cid:19) P e ( t )+ l (cid:18) n s (cid:19) HA ( t ) − l (cid:18) n s (cid:19) HA ( t ) , (63). . . dHA ( t ) dt = ˜ r (cid:18) n s (cid:19) HA ( t ) P e ( t ) . (64)Given the proton entry rate (equation 19), these equations can be solved nu-merically.5.1 Modeling the onset of fusion between virus and endosome membranesThe onset of membrane fusion is triggered by the conformational change ofmultiple adjacent trimer in the contact zone between virus and endosome mem-branes [4,11]. However, the number of fusogenic HAs involved in formation andfusion pore enlargement is still an open question.We model the contact zone between the virus and endosome membranes by120 HAs among the 400 covering the virus [11](Figure3-A). Then, using anumerical solution of equation 64, we chose randomly each new fusogenic HAand defines the onset of virus endosome fusion by the stochastic activationof N a adjacent HAs in the contact zone (Figure 3-A). Using 1 ,
000 Monte-Carlo simulations, we estimated the mean and confidence interval at 95% ofthe fusion onset time for different
N a .We found that for
N a = 1 or 2, most viruses fuse in EE, whereas for
N a = 3or 4 viruses fuse in ME. Finally, for
N a = 5 or 6, viruses mostly fuse in LE(Figure 3-B,C). The common prediction is that
N a = 3 − tochastic model of endosomal escape of Influenza virus 17 N a = 3 or 4 adjacent fusogenic HA are needed tomediate fusion .
Influenza viruses are internalized into endosomes via receptor-mediated en-docytosis. During their transport along microtubules, the endosomes accu-mulate protons, which eventually enable virus-endosome fusion mediated bythe influenza HA, resulting in release of the viral genome in the cell cyto-plasm. Hence, the duration of endosomal transport as well as the localizationof fusion critically depend on endosomal acidification and HA conformationalchange at low pH. Here we presented a a new model to investigate the role ofkey parameters that shape the endosomal residence time of influenza viruses.By associating a kinetics model of endosomal acidification with a Markov-jump process model of HA conformational change, we estimated the number offusogenic HAs as function of time inside endosomes, and we modeled the onsetof fusion with the stochastic activation of
N a adjacent HAs. Using the model,we predict the high HA stability at neutral pH due to the high activationbarrier of protons binding sites. In association with
N a ≥
3, this ensuresthat fusion occurs in ME, preventing a premature fusion in EE. As endosomalmaturation is associated with retrograde transport of endosomes along MTs,this should increase the nuclear targeting of viral genome and pathogenicityof the virus.
Acknowledgements
This research was supported by an Marie Curie grant (D.H.), by theDeutsche Forschungsgemeinschaft (HE 3763/15-1) (A.H.) and the Bundesministerium furBildung und Forschung (eBio: ViroSign) (CS and AH). T.L. is funded by a Bourse Rouxfrom Institut Pasteur.
References
1. al Baldawi, N.F., Abercrombie, R.F.: Cytoplasmic hydrogen ion diffusion coefficient.Biophys J (6), 1470–9 (1992). DOI 10.1016/S0006-3495(92)81953-72. Bayer, N., Schober, D., Prchla, E., Murphy, R.F., Blaas, D., Fuchs, R.: Effect ofbafilomycin a1 and nocodazole on endocytic transport in hela cells: implications forviral uncoating and infection. J Virol (12), 9645–55 (1998)3. Brillouin, L.: La m´ecanique ondulatoire de schrodinger: une m´ethode g´en´erale der´esolution par approximations successives. Comptes Rendus de l’Academie des Sciences(183)4. Danieli, T., Pelletier, S.L., Henis, Y.I., White, J.M.: Membrane fusion mediated by theinfluenza virus hemagglutinin requires the concerted action of at least three hemagglu-tinin trimers. J Cell Biol (3), 559–69 (1996)5. Fuchs, R., Schmid, S., Mellman, I.: A possible role for na+,k+-atpase in regulatingatp-dependent endosome acidification. Proc Natl Acad Sci U S A (2), 539–43 (1989)6. Grabe, M., Oster, G.: Regulation of organelle acidity. J Gen Physiol (4), 329–44(2001)7. Grabe, M., Wang, H., Oster, G.: The mechanochemistry of v-atpase proton pumps.Biophys J (6), 2798–813 (2000). DOI 10.1016/S0006-3495(00)76823-88. Huang, Q., Opitz, R., Knapp, E.W., Herrmann, A.: Protonation and stability of theglobular domain of influenza virus hemagglutinin. Biophys J (2), 1050–8 (2002).DOI 10.1016/S0006-3495(02)75464-79. Huang, Q., Sivaramakrishna, R.P., Ludwig, K., Korte, T., B¨ottcher, C., Herrmann,A.: Early steps of the conformational change of influenza virus hemagglutinin to afusion active state: stability and energetics of the hemagglutinin. Biochim BiophysActa (1), 3–13 (2003)10. Hutchinson, E.C., von Kirchbach, J.C., Gog, J.R., Digard, P.: Genome packaging ininfluenza A virus. The Journal of general virology (Pt 2), 313–328 (2010)11. Ivanovic, T., Choi, J.L., Whelan, S.P., van Oijen, A.M., Harrison, S.C.: Influenza-virusmembrane fusion by cooperative fold-back of stochastically induced hemagglutinin in-termediates. Elife , e00,333 (2013). DOI 10.7554/eLife.0033312. Kieseritzky, G., Knapp, E.W.: Optimizing pka computation in proteins with ph adaptedconformations. Proteins: Structure, Function, and Bioinformatics (3), 1335–1348(2008). DOI 10.1002/prot.21820. URL http://dx.doi.org/10.1002/prot.21820
13. Knessl, C., Mangel, M., Matkowsky, B., Schuss, Z., Tier, C.: Solution of kramers-moyalequations for problems in chemical physics. Journal of Chemical Physics (3), 1285–1293 (1984)14. Knessl, C., Matkowsky, B., Schuss, Z., Tier, C.: An asymptotic theory of large deviationsfor markov jump-processes. Siam Journal On Applied Mathematics (6), 1006–1028(1985)15. Kramers, H.A.: Wellenmechanik und halbzahlige quantisierung. Zeitschrift fur Physik (1926)16. Krumbiegel, M., Herrmann, A., Blumenthal, R.: Kinetics of the low ph-induced confor-mational changes and fusogenic activity of influenza hemagglutinin. Biophys J (6),2355–60 (1994). DOI 10.1016/S0006-3495(94)80721-017. Lagache, T., Danos, O., Holcman, D.: Modeling the step of endosomal escape duringcell infection by a nonenveloped virus. Biophys J (5), 980–9 (2012). DOI 10.1016/j.bpj.2011.12.03718. Lamb, R., Krug, R.: Orthomyxoviridae: The viruses and replication. In: D. Knipe,P. Howley, D. Griffin (eds.) Fields Virology, 4th edn. Lippincott Wiliams and Wilkins(1996)tochastic model of endosomal escape of Influenza virus 1919. Lamb, R.A., Choppin, P.W.: The gene structure and replication of influenza virus. AnnuRev Biochem , 467–506 (1983). DOI 10.1146/annurev.bi.52.070183.00234320. Leiding, T., Wang, J., Martinsson, J., DeGrado, W.F., Arsk¨old, S.P.: Proton and cationtransport activity of the m2 proton channel from influenza a virus. Proc Natl Acad SciU S A (35), 15,409–14 (2010). DOI 10.1073/pnas.100999710721. Matkowsky, B., Schuss, Z., Knessl, C., Tier, C., Mangel, M.: Asymptotic solution of thekramers-moyal equation and first-passage times for markov jump processes. Phys. Rev.A , 3359–3369 (1984)22. Matlin, K.S., Reggio, H., Helenius, A., Simons, K.: Pathway of vesicular stomatitis virusentry leading to infection. J Mol Biol (3), 609–31 (1982)23. Mercer, J., Schelhaas, M., Helenius, A.: Virus entry by endocytosis. Annu Rev Biochem , 803–33 (2010). DOI 10.1146/annurev-biochem-060208-10462624. Ng, A.K.L., Zhang, H., Tan, K., Li, Z., Liu, J.h., Chan, P.K.S., Li, S.M., Chan, W.Y., Au,S.W.N., Joachimiak, A., Walz, T., Wang, J.H., Shaw, P.C.: Structure of the influenzavirus a h5n1 nucleoprotein: implications for rna binding, oligomerization, and vaccinedesign. FASEB J (10), 3638–47 (2008). DOI 10.1096/fj.08-11211025. Rink, J., Ghigo, E., Kalaidzidis, Y., Zerial, M.: Rab conversion as a mechanism ofprogression from early to late endosomes. Cell (5), 735–49 (2005). DOI 10.1016/j.cell.2005.06.04326. Risken, H.: The Fokker-Planck Equation: Methods Of Solution And Applications.Springer (1996)27. Schuss, Z.: Theory and Applications of Stochastic Processes. Springer (2010)28. Schuss, Z., Holcman, D.: Stochastic Narrow Escape in Molecular and Cellular Biology:Analysis and Applications. Springer (2015)29. Schuss, Z., Singer, A., Holcman, D.: The narrow escape problem for diffusion in cellularmicrodomains. Proc Natl Acad Sci U S A (41), 16,098–103 (2007). DOI 10.1073/pnas.070659910430. Singer, A., Schuss, Z., Holcman, D.: Narrow escape, part iii: Non-smooth domains andriemann surfaces. J Stat Phys (3), 491–509 (2006)31. Stoyanov, A., Righetti, P.: Buffer properties of biopolymer solutions, as related to theirbehaviour in electrokinetic methodologies. Journal of Chromatography A , 11–18(1999)32. Van Dyke, R.W., Belcher, J.D.: Acidification of three types of liver endocytic vesicles:similarities and differences. Am J Physiol (1 Pt 1), C81–94 (1994)33. Ward, M., Keller, J.: Strong localized perturbations of eigenvalue problems. Siam Jour-nal On Applied Mathematics (3), 770–798 (1993)34. Wentzel, G.: Eine verallgemeinerung der quantenbedingungen fur die zwecke der wellen-mechanik. Zeitschrift fur Physik (1926)35. Zaraket, H., Bridges, O.A., Duan, S., Baranovich, T., Yoon, S.W., Reed, M.L., Sa-lomon, R., Webby, R.J., Webster, R.G., Russell, C.J.: Increased acid stability of thehemagglutinin protein enhances h5n1 influenza virus growth in the upper respiratorytract but is insufficient for transmission in ferrets. J Virol (17), 9911–22 (2013).DOI 10.1128/JVI.01175-130 Thibault Lagache et al. Figures and tables
Fig. 1 Structure and endosomal trafficking of the Influenza virusA - Structureof the Influenza virus.
Influenza is an enveloped virus. Main spike proteins anchored inthe envelope are the neuraminidase (NA) and the HA (HA). Protons can access the core ofthe virus through M2 channels. Main matrix protein is M1 protein. Viral genome of the virusis composed by eight viral ribonucleoproteins (vRNPs).
B Endosomal trafficking of thevirus.
Influenza virus enters the cell via receptor-mediated endocytosis and progress rapidlytowards an Early Endosome (EE). Then, maturation of EE into a Maturing Endosome (ME)and Late Endosome (LE) is associated with an acidification of the endosome lumen and aretrograde transport of the endosome along the microtubules (MTs) of the cell towards thenucleus, the destination of vRNPS for virus replication. Fusion between the virus and theendosome membrane is critically controlled by the low pH conformational change of HAs,but the kinetics of in vivo escape remains largely unknown.
C- Kinetics of endosomalpH decrease obtained from intracellular fluorescence microscopy (red line. Mean ± SEM)and coarse-grained modeling (equation 19, black line. Model parameters are summarized intable 1).
Fig. 2 Free protons in the endosome triggers HA conformational change A -Schematic representation of the influenza virus inside an endosome.
The right-hand side shows a scheme of an isolated HA trimer. Free protons in the endosome canbind to HA trimers. The protons binding rates r ( X, c ) and l ( X ) depend on the number ofoccupied sites X and on the concentration c of free protons in the endosome. When thenumber of bound protons reaches a given threshold, the HA trimer changes conformationinto a fusogenic state. B Rate of the HA conformational change as a function of thepH.
The theoretical curve (solid line) for the rate of HA conformational change ( τ ( c )) − approximate well the experimental data (red circled crosses) [16]. The region inside thedashed box is magnified in the upper inset.tochastic model of endosomal escape of Influenza virus 21 Fig. 3 Model and fluorescence experiments of the intracellular onset of virus-endosome fusion A - Model of fusion onset.
The fusion between virus and endosomemembranes is triggered by the conformational change of Na adjacent HAs in the contact zonebetween virus and endosome ( ≈
120 among the 400 HAs covering the virus envelope [11]).
B- Modeling the stochastic activation of HAs on virus envelope during endosomaltrafficking.
Solving equation 64 we estimated the time window (95% confidence interval)of intracellular fusion for 1 ≤ Na ≤ C - Localization of fusion events as function ofNa.
Using time windows of fusion onset and endosome maturation kinetics (equation 20),we estimated the localization (EE, ME or LE) of fusion onset as function of the number Na of adjacent fusogenic HAs needed for the fusion onset. D - In vivo monitoring of fusionbetween virus and endosomes.
MDCK cells expressing Rab5-CFP and Rab7-GFP wereincubated with R18-labeled influenza A viruses. Fusion was observed as a strong increase ofR18 signal due to de-quenching after dilution. Scale bar = 1 µm E - In vivo localizationof fusion events.
Fusion events were counted and categorized regarding their localizationin EE (Rab5), ME (Rab5 + Rab7) or LE (Rab7).2 Thibault Lagache et al.
Table 1 Parameters of the endosome acidification modelParameters Description Value r e Radius of the endosome r e = 500nm [25] V e Volume of the endosome V e = πr e = 5 .
22 10 − L r v Radius of the influenza virus r v = 60nm [19] V v Volume of the viral internal lumen V v = πr v = 9 10 − L N A Avogadro constant N A = 6 .
02 10 mol − β e Buffering capacity of the endosomal lumen β e = 40mM/pH [32] β v Buffering capacity of the viral lumen β v = β e = 40mM/pH (this study) β M v Buffering capacity of viral M1s β M v ≈ N A V v mM/pH (this study) β NPv
Buffering capacity of viral NPs β NPv ≈ N A V v mM/pH (this study) β RNAv