Journal of Biological Physics 26: 219–234, 2000.

© 2001 Kluwer Academic Publishers. Printed in the Netherlands.


A Statistical Mechanics Model for Receptor
Dept. of Physics, University of California San Diego, La Jolla, CA 92093-0319, U.S.A.
∗ Corresponding author

Abstract. We introduce and study a simple lattice statistical mechanics model for the clustering of
tumor necrosis factor receptor I (TNFR1). Our model explains clustering under over-expression of
the cytoplasmic signal transducer as well as the clustering induced via extracellular ligand binding;
also we explain why the loss of transducer leads to a rapid break-up of the clusters. The basic
mechanism at work is a first-order (cooperative) phase transition caused by the multimeric binding
capability of the receptor-transducer complex. Using cooperativity of this type, the cells are found
to have an enhanced sensitivity and robustness. In general, our method can be applied to other
receptor-clustering related signaling system.
Key words: Signal transduction, receptor clustering, statistical mechanics
Abbreviations: TNFR1 – Tumor Necrosis Factor Receptor I; TNFα – Tumor Necrosis Factor;
SODD – Silencer of Death Domain; TRADD – TNF Receptor Associated Death Domain; UV –
Ultra-Violate; TRAF2 – TNFR Associated Factor 2; FADD – Fas Associated Death Domain; RIP –
Receptor Interacting Protein; JNK – c-Jun N terminal Kinase; NFκB – Nuclear Factor-κB

1. Introduction
Cell growth, differentiation, migration, and apoptosis are regulated in part by extracellular polypeptide cytokines [1]. Since most peptides are unable to pass through
the hydrophobic cell membrane, executing the biological functions requires specific surface receptors to recognize the peptides and thereby transmit the signals.
In many cases, it has been observed that clustering of receptors is the essential part
of the mechanism whereby proper intracellular signals can be generated. This leads
to two fundamental questions: (1) why is the receptor clustering required for signal
transduction, and (2) how is the clustering regulated under physiological conditions (i.e., the dose response curve)? Here, via simple modeling, we investigate the
second question for one particular example, that of tumor necrosis factor receptor
I (TNFR1) signaling.
TNFR1 regulates cell death and survival through its association with a cytoplasmic transducer, ‘TNF receptor associated death domain’ (TRADD) and via
the clustering of TRADD-TNFR1 complexes [2]. Specifically, the clustering is
accomplished by the multiple self-binding motifs on TRADD (different from its

tumor necrosis factor α (TNFα ) which is capable of oligomerizing three receptors regardless of their cytoplasmic condition [2. if applying ligand can properly modulate the effective binding between nearest-neighbor TNFR1-TRADD complexes. ‘silencer of death domain’ (SODD) which was found to bind to the TNFR1-TRADD binding site [5].220 C. we have made several assumptions. One molecular hypothesis is that trimerizing receptors brings them into close proximity and thus ‘squeezes’ out the bulky inhibitors (molecular weight of SODD = 62KD. Here. the experimental data suggested that TNFR1-TNFα has a binding lifetime longer than hours [6]. however. Then. a firstorder phase transition can occur with the surface molecules spontaneously segregated into dilute and dense phases [7]. From the biochemical prospective. 5]. 5]. The recruited TRADD is then phosphorylated by activated downstream enzymes and released from TNFR1. is inconsistent with the observed time course of SODD-TNFR1 association during ligand treatment and the low TNF-TNFR1 dissociation rate. TNFR1TRADD association is normally blocked by a cytoplasmic inhibitor. based on a simple lattice model that incorporates the most relevant parameters. Thus. however. if trimerizing receptors could block the inhibitor binding. this occurs about 10 min after the treatment [5]. to examine the validity of this statistical mechanics approach. we are interested in the short-term behavior (the onset of clustering). TRADD is recruited to (and SODD is released from) receptors after 5 min of ligand treatment. Figure 1(a). 2. This behavior clearly requires enhanced TRADD recruitment to liganded receptors. there would be no inhibitor-receptor re-association in 10 min since most receptors would still be di/trimeric. BASIC ASSUMPTIONS To simplify the model. it is not totally understood how trimerizing TNFR1 can restore the clustering. Our results show that (a) simple modeling does predict TNFR1 clustering in a relevant parameter range. receptor internalization. and (b) using cooperativity of this type. To prevent this. TRADD = 32KD) [5]. GUO AND H. [3]. The Model 2. This. Normally. we study the statistical mechanics of TNFR1 clustering. the cells can enhance both their sensitivity and robustness. a biophysical property of this receptor system that has not been addressed. Clustering is restored in the presence of a trimeric extracellular ligand. and the long-term consequences of clustering (such as downstream signaling. The system has an intrinsic cooperativity arising from TRADD self-multimerization capacity. we estimate the time scale of system nucleation rate. To see if this idea works. thus over-expression of TRADD can lead to spontaneous clustering and constitutive signaling [4. Our goal is to construct a system Hamiltonian that can be easily simulated and analyzed. . There is.1. Now. followed by an SODD-TNFR1 re-association. LEVINE receptor binding site). our finding agrees well with the experimental data. Finally.

. we treat separately the cases of clustering caused by ‘over-expression’ of TRADD. we ignore the obvious fact that dimers and trimers are bigger than monomers. Also. n as the content number of TNFR1 (R). Second. Figure 1(b). or caused by ligand binding. (b) If we assume that on average each recruited TRADD can associate with 4 nearest neighbors. This could of course happen without TNFα . when the receptors are liganded to di/trimer. this should be a reasonable approach. Then. SODD (S) & TRADD (T )). in the time scale of interest. TNFα (L). m. Thus our first assumption is that the system can reach a ‘quasi-equilibrium’ state via rapid receptor diffusion. but certainly can be subsumed as a change in the effective binding energy. adaptation/desensitization) will not be included. Fourth and finally. each lattice site i can be occupied by either zero or one effective molecule. we define a state label ti for the occupying molecule (i. For instance. the leading effect of ligand binding is to change the number of effective nearest neighbors and hence the system cooperativity. Here. we assume that the binding strength between TRADD-TNFR1 complex is much weaker than that between TNFR1 and TNFα . denoted as ni = 0 or 1. . (a) The cytoplasmic domain of receptor (TNFR1) can associate with either the inhibitor (SODD) or transducer (TRADD). given the long TNFR1TNFα binding lifetime. if we assume that every TRADD can selfassociate with four nearest-neighbor TRADD. ti ≡ Rk Ll Sm Tn with k. This requires the predicted nucleation rate (for the onset of clustering) to be within a reasonable time scale. Then. the ‘effective molecules’ that allow for clustering will be either solitary TNFR1 monomers or liganded di/trimers. the number of effective nearest neighbors that a solitary TNFR1 can bind to will be different from that of a liganded TNFR1. Third. this approximation overvalues the entropy of the multimeric state. l. however. this is the closest that neighboring surface molecule complex (solitary or liganded TNFR1) can get to each other. TNF treatment can enhance the cooperativity from 4 to 6. each TNFR1-TRADD complex unit will be able to bind to 6 similar ones. we treat the cell surface as a lattice with a spacing a0 ≈ 1nm. whereas the binding of SODD prevents clustering. given the TRADD multiple self-binding capacity. Then. Also.A STATISTICAL MECHANICS MODEL FOR RECEPTOR CLUSTERING 221 Figure 1. the entropy cost of co-localizing two complexes would be too high without the offsetting ligand binding energy.e. this allows the clustering of TNFR1-TRADD complex. TRADD has multiple motifs for self association.

TRADD-TRADD binding energies. Here a(ti . f (3) = 3 (see Figure 1) corresponds to the number of TRADD self-bindings allowed inside a TNFR1 oligomer. This implies that the molecules responsible for clustering are liganded TNFR1 oligomers with at least two TRADD associated. we need (m + n) ≤ k. Next. we have µ(Rk Ll Sm Tn ) = kµR + l(µL + kgL ) + m(µS + gS ) +n(µT + gT ) + f (n)gE (2) with µL as the ligand chemical potential. This takes the general form 1X Hint ({n. [SODD]) range where spontaneous clustering does not occur. GUO AND H. µ(R1 L0 S0 T1 ) = µR + µT + gT . → a single free TNFR1 → a single TNFR1-SODD complex → a single TNFR1-TRADD complex with µR . Explicitly. t}) = − µ(ti )ni ≡ − µ ((Rk Ll Sm Tn )i ) ni (1) i i with µ(ti ) as the chemical potential for the state of the occupied site. −gT as the binding energies. we have θ(ti ) = 1 if the site has a TNFR1-TRADD complex. and k > 1 if l = 1 (the liganded case). tj ) is a ‘state’dependent function indicating the nearest-neighbor interaction. LEVINE THE HAMILTONIAN Next. or trimeric TNFR1. on the other hand. First. the simplest form of a(ti . Here f (1 ≤ n ≤ 2) = n − 1. the possible states includes k ∈ {1. dimeric. t}) = − Jij a(ti . Since the cooperativity arises from the capacity of TRADD self-multimerization. we assume that each TNFR1-TRADD . tj ) is the product a(ti . Simulations To see if our model can generate clustering. [TRADD]. j are nearest-neighbors and 0 otherwise. however. µT as the molecule chemical potential and −gs . corresponding to a solitary. 2. 1}. In the TNF present case. and −gL . Overall. we have µ(R1 L0 S0 T0 ) = µR . µS . In the TNFα -present case. l ∈ {0. 3. µ(R1 L0 S1 T0 ) = µR + µS + gs . −gE as the TNFR1-TNFα . 3}. we are interested in the molecular concentration (i.e. tj ) = gE θ(ti )θ(tj ). we perform a Metropolis Monte Carlo simulation.2. we construct the system Hamiltonian. Obviously. tj )ni nj (3) 2 ij with Jij = 1 if i. Thus θ(ti = Rk Ll Sm Tn ) = 1 if n ≥ 2 and 0 otherwise.. we add an interaction term into the Hamiltonian. in the absence of TNF.222 2. Our first step is to define a linear term X X H0 ({n. in the TNFα -free case. C. In the TNF free case.

e   β(µT +gT )      P e β µR + β1 ln 1+eβ(µS +gS ) βgE j Jij nj θ(tj ) → e {1}. 5]. The jumping probability for a surface molecule to move to another lattice site and the probability for surface molecules to react and change their type/states are determined by the Hamiltonian and obey detailed balance [7].e.A STATISTICAL MECHANICS MODEL FOR RECEPTOR CLUSTERING 223 complex can bind to four similar ones (i. as a simple estimate. and the brackets [.]eq indicates the equilibrium concentration of the respective molecule. . i. the relevant parameters that switch TNFR1 between non/effective states are gE and the ratio B = eβ(µT +gT ) /[1 + eβ(µS +gS ) ]. Since −gE is the binding energy between each TRADD self-binding motif. . In detail. we immediately see that for a given receptor density. roughly the energy of a single hydrogen bond at room temperature. tuning the B factor moves the system from a non-clustering to a clustering phase.. Next. aside from the ‘shifted’ µR (which can be fixed by the given receptor density anyway). Obviously. the ensemble for a TNFR1 occupying a single lattice i will be  h i P β µT +gT +gE j Jij nj θ(tj ) βµR β(µS +gS ) e {1. Using the procedure introduced in our previous work [7]. . . Thus our model can predict spontaneous clustering as seen in experiment [4. clusters are not allowed to move collectively. In our simulation. we need two additional parameters. which might not be the case in reality.. we might take gE ≈ 4kB T . e }. The crucial B factor is related to the molecular concentrations and dissociation constants via     TNFR1(m) eq [TRADD]eq = KdTRADD TNFR1(m) • TRADD eq     TNFR1(m) eq [SODD]eq = KdSODD TNFR1(m) • SODD eq Here the notation ‘TNFR1(m)’ means TNFR1 molecules distributed on the membrane. and perform simulations on an 100 × 100 square lattice with periodic boundary condition. we have −1  eβ(µS +gS ) = [SODD] KdSODD . e 1 + eβ(µS +gS ) where we have separated the subensemble not effective for clustering into {. From Figure 2.e. from Equation (2). In the ligand-absence case. To perform the simulations.}. B= [TRADD]/KdTRADD 1 + [SODD]/KdSODD −1  eβ(µT +gT ) = [TRADD] KdTRADD (4) Obviously a larger B factor implies a higher transducer to inhibitor concentration ratio. the ligand chemical potential µL and binding energy −gL . we consider the ligand-presence case. the system cooperativity D = 4 as illustrated in Figure 1(a)). we can connect these parameters with real experimental variables through the .

The open diamond represent free TNFR1 or TNFR1 associated with SODD. Here hni is the given receptor density. GUO AND H. (6) h2 with mTNF as the mass of TNFα [7]. and [TRADD]/KdTRADD are fixed quantities during the Metropolis Monte Carlo simulation 1+[SODD]/KdSODD (4 × 108 steps). The presence of TNFR1-TRADD clustering at (A) high inhibitor or (B) transducer concentrations in the absence of TNFα . LEVINE Figure 2. B = and a filled square indicates a TNFR1-TRADD complex. formula  TNFR1(m)    [TNFα ] = KdTNF TNFα • TNFR1(m) eq −1  = [TNF] KdTNF eq → eβ(µL +gL ) (5) Treating the ligand as an ideal gas [7] allows us to find an estimate of the ligand binding energy # " 3 2π mTNF kB T 2 gL = kB T ln /KdTNF ≈ 60kB T . This estimate is far larger than the order of gE and thus supports our assumption that binding between TNFR1-TRADD .224 C.

whereas the liganded TNFR1 oligomers are not allowed to dissociate. the filled/open circles indicate liganded TNFR1 with/out TRADD associated. this is a reasonable constraint. Thus most TNFR1 will be driven into oligomeric structures after long term simulation. The open diamond represent unliganded free TNFR1 or TNFR1-SODD complex. we can take 15 < gL /kB T < 60. a filled square indicates a TNFR1-TRADD complex. In the simulation.A STATISTICAL MECHANICS MODEL FOR RECEPTOR CLUSTERING 225 Figure 3. and the filled/open triangles indicate TNFR1 trimer with/out TRADD associated. the TNFα concentration is a fixed quantity (consistent with the experimental procedure [5]) and the monomeric TNFR1s are allowed to update their un/liganded states according to the TNFα binding energy and concentration. In general. ligand concentration [TNFα ] and the B factor are fixed quantities. The presence of TNFR1-TRADD clustering at (C) moderate and (D) low B conditions with the presence of TNFα . given the long binding lifetime. we perform the simulation on a 100 × 100 triangular lattice. Here the receptor density hni. Then. i. but clustering can occur with the presence of ligand. this entire range yields similar numerical results. assuming the cooperativity D = 6 caused by ligand binding (Figure 1(b)).e. complex is much weaker than that between TNFR1 & TNFα . . 3 to 10 times the value of gE . Note that (C) has the same cytoplasmic parameters as (A).

This leads to the self-consistent mean-field equation ∗ eβ(µφ +gE Dφ ) φ = ∗ 1 + eβ(µφ +gE Dφ ) ∗ (11) . MEAN . we introduce an auxiliary Gaussian field [7.e. This is consistent with the experimental result [5] and confirms that our model can capture the basic physics of TNFR1 clustering.t })] (7) {ni . LEVINE In the presence of ligand. We now approximate the partition function by assuming that the dominant contribution arises near a spatiallyuniform saddle-point. followed by a SODD re-association and break-up of the existent of clusters. we next utilize a mean-field approach to analyze the conditions under which clustering can take place. the TNFR1-associated TRADD are phosphorylated and dissociated.226 C. GUO AND H. TNFα treatment is not sufficient as the SODD binding dominates (Figure 3D). if the downstream components of the signaling cascade are activated by clustering.t })+Hint ({n. At the very low [TRADD] (small B factor). This yields Z P P β Z = C Dφe− 2 gE ij Jij φi φj + i 9i ({φ}) (8) Q Here. Dφ = i dφi with φ ranged from −∞ to +∞. Analytical Approach 4. 4. we have   9i ({φ}) = ln 1 + eβµR (1 + eβ(µS +gS ) + eβ(µT +gT ) eβgE Jij φj ) (9) Defining the effective chemical potential for the order parameter φ via eβµφ = eβ(µR +µT +gT )  1 + eβµR 1 + eβ(µS +gS ) (10) P we can simplify 9i ({φ}) → ln[1+eβ(µφ +gE j Jij φj ) ]. while most TNFR1s remain trimeric. For the case of no ligand present. To proceed.. we found that the same parameters that previously caused no clustering (Figure 2A) now can induce clustering (Figure 3C). Thus. To have a more quantitative view.FIELD APPROACH The partition function for this model equals X Z= e−β[H0 ({n.ti } P where {ni .ti } means ensemble summation over the different configurations as detailed earlier. however. and C is a normalization constant which does not affect the thermodynamic properties of the partition function. i. 8] to decouple the quadratic term in Hint into a linear term such that we can sum over the single site ensembles independently.1. one can imagine a ‘dynamical’ process progressing from Figure 3C to Figure 3D.

And. and D = 4. Figure 4 shows how we use graphical intersection to determine φ ∗ and µφ . the mean-field equations remain the same. and only the parameter µφ gets redefined. the coordination neighbor D is now set to 6. ∗ x(1 + BeβgE Dφ ) hni = 1 + x(1 + BeβgE Dφ∗ ) (12) where x = eβµR [1 + eβ(µS +gS ) ] and B is defined in Equation (4). For the case at hand. for a fixed receptor density.. From the saddle-point equations. βgE = 4. This implies that the van der Waals loop exists only if gE ≥ gEmin = 4kB T /D.A STATISTICAL MECHANICS MODEL FOR RECEPTOR CLUSTERING 227 P Here. Similarly. Again. In detail. we introduce a field φ to decouple the quadratic term.1. one can find an expression for the mean receptor density in terms of the derivative of the effective free energy with respect to the chemical potential µR .2. Working this out explicitly. we introduce a φ chemical potential. we have chosen hni = 0. the two unknowns φ ∗ and x can be determined by solving Equation (11). Up to a constant. γ3 = B 2 eβ(µL +3gL +gE ) [3 + Be2βgE ]. we notice that the solution curve of the second equation intersects the first curve past the lower binodal curve. and x and B are defined previously. Then. PHASE DIAGRAMS Using this mean-field approach. we find that H ({φ}) is reduced to the same expression as that given above for the ligand-free case. In Figure 4. This means that the system will attempt . There also exist two binodal points. (12). because of the nature of the liganded receptor. we can find the phase diagram by solving a pair of coupled equations for two unknowns. the variables simply related to the receptor density and chemical potential. we find φs± = (1/2)± (1/4) − (1/βgE D) and φb± = (1/2) ± ξ with ξ = 2 tanh[gE Dξ /2]. Just free case. the partition function becomes # " 3 3 X X P 9i ({φ}) = ln 1 + (13) αk x k + γk x k eβgE ij Jij φj k=1 k=2 where the coefficients are given by αk = (kB + 1)(δk0 + eβ[µL +(k−1)gL] ). this means that there exist two spinodal points φs± at which the solution branch determining the chemical potential µφ become unstable. Next. we proceed to the model with TNF present. here P as the TNFαP via eβµφ = 3k=2 γk x k /[1 + 3k=1 αk x k ]. 4. So.e. D = j Jij is the number of nearest neighbors which we take as 4 in this TNFα -free case. i. γ2 = B 2 eβ(µL +2gL +gE ) . the order parameter φ ∗ and x. Notice that we have a van der Waals loop in the solution diagram of Equation (11). there is a minimal TRADD self-binding strength that allows for clustering. labeled φb± which correspond to places at which the different solutions for the chemical potential (the two possible phases of the system) √ have equal free energy.

Consequently. from Equation (11). This requires a minimal receptor density hnimin at a given set of transducer. GUO AND H.1. with nc = e−βgE Dξ 1+e−βgE Dξ (14) where ξ and the B factor are defined as before. . Thus the system is metastable and can segregate into a dilute phase and a dense phase (the long-dashed curve) corresponding to hni = 0. The last curve shown verifies that the solution curve for hni = 0. as one can see from Equation (11)). (12) we have  hnimin = nc + e βgE D 2  +e βgE D 1 2 −ξ  B −1 . (via nucleation) to spontaneously segregate into a dilute and a dense phase so as to generate the lowest overall free energy. Likewise. the minimal receptor density in the presence of ligand can be computed.999979. LEVINE Figure 4.228 C. For the ligand-free case. The mean-field solution for Equation (11) (the solid curve) and Equation (12) (the dashed curve). showing that these two phases can coexist. the system becomes metastable if the receptor density is large enough such that the saddle point solution φ ∗ is located between the binodal φb− and spinodal φs− points. Here the density is given by hni = 0.999979 intersects the upper branch of the van der Waals loop at exactly the same value of µφ . The consequences of these results are shown in Figure 5. Note that the solid curve ends at the point φ (u) which (u) satisfies BeβgE Dφ [1/φ (u) − 1] = 1 (and hence x = ∞. The uniform solution given by the intersection of the two curves lies above the binodal point φb− . inhibitor concentrations.

(a): The dependence of the (dimensionless) minimal receptor density hnimin on transducer.A STATISTICAL MECHANICS MODEL FOR RECEPTOR CLUSTERING 229 Figure 5. In general. the dashed curve will depend on [TNFα ]. The solid curve is for the ligand-independent and the dashed curve is for the ligand-induced case. . here fixed at [TNFα ]= KdTNF . The upper curve corresponds to φb+ and the lower one is related to φb− . (b): The receptor density corresponding to the two binodal points at given strength of the B factor (related to [TRADD]. [SODD]). inhibitor concentrations (determined by the B factor). as a function of the TNFα concentration. Here a0 ≈ 1nm.

Jnuc ≈ νe−1F with 1F as the free energy cost (as compared to the metastable initial phase) and ν as the growth rate of this droplet. the droplet of nucleation. all is known is that the order of magnitude of the clustering rate is estimated to lie in the range of 0. between the binodal and spinodal points). Likewise. driving the system from a relatively dilute uniform phase to coexisting dilute & dense phases. e. or indirect measurement from the receptor chemical modification (phosphorylation & adaptor association. 11. none of them can provide explicit data.g. 5. 12. this can be done via finding an inhomogeneous solution of the equation which results from ifunctional. unphosphorylated TRADD can be increased by applying osmotic pressure [10]. a nucleation process [14] will take place. LEVINE What can we learn from these phase diagrams? First. Then. electron microscopy [11]. TRADD by mRNA (antisense RNA). 5.230 C. Obviously. there are many other ways to modulate the molecular concentrations including over-expressing (suppressing) SODD. This equation takes the form h varying the free energy φ ln 1−φ µφ a02 ∇ 2 φ = −φ− (15) βgE D gE D . Typically. The Nucleation Rate Next. fluorescence imaging [12]. Second. it is known that the receptor density can be down-regulated by antibody treatment and the active. For instance. To find out 1F . 1[TRADD] ∼ [1P ]γ with γ > 0. measuring a precise clustering rate on living cells for a cooperative signaling system of this type ( TNF and other growth factor receptor families) is rather a big challenge. Unfortunately. The rate of this process Jnuc is determined by the fact that the system must pass through a spatially inhomogeneous transition state. these diagrams reveal qualitatively how the signaling responds to a change in any of the molecular concentrations. this gives the cells a sensitivity and a robustness.) provided that these reactions are fast after clustering [5. and our results can be directly tested by experiment. compared to Figure 5(a) (the constitutive signaling). GUO AND H. Experimentally. we test the validity of statistical mechanics approach of this type by estimating the time scale of nucleation. The available methods to date are direct observation by immuno-staining [10].e. assuming a simple scaling between the osmotic pressure overload 1P and the increment of active TRADD. when the receptor density n is down-regulated to nc . 10. 13]. the deep-‘well’-like phase boundaries in Figure 5(b) indicates that the variation of ligand at small (moderate) concentration ranges can cause a strong (weak) response in ligand-induced clustering. the existence of a firstorder phase transition has an immediate implication that the cells use an ‘all-ornone’ reaction to execute the programmed death or proliferation. & UV irradiation [4. as the receptor system is pushed into the metastable part of the phase diagram (i. both are required for functional biology [9]. we find from Equation (14) a power law divergence 1P ∗ ∼ [n − nc ]−γ for the pressure necessary to induce clustering. 13]. Here. Also. 10].2 ∼ 1 min−1 [5. we need to determine the transition structure (droplet).

Figure 6(a)C. φ0 and δφ. and when φmet a → φb− . 13]. dφ/dr in between. Discussion We have presented a lattice statistical mechanics model for TNFR1 clustering. the free energy barrier 1F ≡ d 2 r[f (φ) − f MF (φmet a )] is shown in Figure 6(a)D. In general. Requiring this to be less than zero leads to φmet a < φ0 < φ(dense) . The interaction between receptors can lead to a first-order . e. Here µφh is the chemical potential fixed by the ‘homogeneous’ metastable state βµφ = i − βgE Dφmet a . Equation (15) can be solved numerically. we have two unknowns. Clearly.) endows the system with a desired robustness. 6. Interestingly. 5×10−3 for the B factor 5×10−3 . manifesting the onset of a long wavelength R instability for this type of calculation. in which the growth rate ν ≈ φmet a /t¯c where φmet a is the density of metastable state molecule surrounding the droplet and t¯c is the mean ‘capture’ time for wandering molecules to adhere to the droplet [15]. Our basic idea is simple. with κ = βg1E D ln 1−φ − 0 ln φmeta 1−φmeta µ φ0 − gEφD . The droplet profile is plotted in Figure 6(a)A. Also.2 ∼ 1 min−1 ) in TNFR1 [11. 10−2 . r = ∞ ends and matching φ. near the origin φ(r) ≈ φ0 + 14 κr 2 .231 A STATISTICAL MECHANICS MODEL FOR RECEPTOR CLUSTERING 2 d 1 d where ∇ 2 is the Laplacian which in polar coordinates becomes dr 2 + r dr . we find λ2 = [βgE Dφmet a (1 − φmet a )]−1 − 1. 1F → ∞. To obtain the growth rate ν.g. the droplet becomes more and more flattened as φmet a approaches the spinodal point φs− . 1F → 0. to have a qualitative range of the nucleation rate. at the maximal sensitivity level. from Equation (15). 10−1 . Then. we get that the corresponding TNFR1 densities hni are roughly 5×10−1 . For a large system size L. so as to give the system a maximal sensitivity. this implies that the metastable φmet a < φs− . we obtain Jnuc = φmet a ln[L/Rc ]e−1F 2Dm (16) where Dm ≈ 10−8 cm2 /sec is the membrane diffusionR constant and Rc is the effective cluster size defined via π Rc2 [φ(dense) − φmet a ] ≡ d 2 r[φ(r) − φmet a ]. i h φ0 On the other hand. 5] and other receptor system [12. Thus. which can be fixed by integrating Equation (15) from both the r = 0. respectively. from Figure 5B. when φmet a → φs− . the receptor density is just below the binodal point for the ligand-absent case. and thus confirms the validity of our model. we use the Becker-Doering theory. Since λ2 must be greater than zero. we find that rm → ∞ as φmet a → φs− . This quantity is very close to the experimentally estimate (0. As r → ∞ we have φr→∞ ≈ φmet a + δφe−λr . Also. If we denote the location of the peak in the free energy density as rm . from Figure 6(b). 5×10−2 . the lack of dependence of the nucleation rate on the details of parameters (the system size and cytoplasmic conditions such as the B factor. we assume that in physiological condition. we found that the nucleation rate for these three values are almost in the same range (Jnuc ≈ 10−2 ∼ 1min−1 ). as expected. Finally.

LEVINE Figure 6. the transducer. Here the B factor (i.232 C. indicated by the solid.. Clearly. with different system sizes (≡ (La0 )2 × nm2 . The droplet size Rc corresponding to an effective cluster and the free energy barrier are also estimated in (C) and (D). and long-dashed curves) and B factors. Here [TNFα ]= KdTNF = 10 ng/ml is set as the usual application dosage [10]. the system size does not greatly affect the nucleation rate. GUO AND H. inhibitor concentration ratio) is defined in Equation (4). (a): (A) The calculated droplet density and (B) corresponding free energy density profiles in different metastable states. dashed. (b): The calculated nucleation rate at molecular unit.e. respectively. .

was found to self-associate into a trimer [16]. Heldin. 80 (1955). it is known that clustering TRADD-TNFR1 complex can recruit adaptors such as TRAF2. at the level of maximal sensitivity. such as programmed cell death or cell proliferation. Turning this around. . Also. our calculations on the nucleation rate show an independence on the details of cellular parameters. and caspase [2]. one of the adaptors. 2. Curr. our results indicate that compared to the constitutive one. leading to gene expressions for the cell death or proliferation. i. FADD. 2-state-like behavior is also popular in small-scale biochemistry such as functional protein folding and the corresponding chemical reactions. we expect that our modeling can be easily extended to other systems. and Jones. Cell. V. But. Then. TRAF2. during evolution. Recently. 5 (1995). 735–743. which in turn can activate downstream elements including JNK. D. this implies that the receptor system will spontaneously phase separate for some range of parameters. A. Biol. A. that clustering TNFR1-TRADD complex provides a molecular scaffold to facilitate signaling. E.M. the cells have developed receptor clustering as a ‘large-scale’ digital machinery to execute an ‘all-or-none’ response. and Dixit. 9858–9862. NFκB.R. and Baichwal. V. Dimerization of cell surface receptors in signal transduction. References 1. Ashkenazi.: 1995. the ligand-regulated signaling has a rather larger sensitivity (robustness) in small (moderate) ligand concentration. Acknowledgement HL and CG acknowledge the support of the US NSF under grant DMR98-5735. Science 281 (1998). Perhaps after long-term evolution.: Systematic mutational analysis of the death domain of the tumor necrosis factor receptor 1-associated protein TRADD. transducer. and RIP. 3. in an analog world. Thus our model successfully explains how ligand-induced TNFR1 clustering can take place. This again provides the cells a robustness. Opin. two very questions that should be addressed are (a) what is the physiological role of ligand treatment in functional biology? and (b) what design principles might have been used. and/or the ligand concentration.Y. 213–223.A STATISTICAL MECHANICS MODEL FOR RECEPTOR CLUSTERING 233 phase transition with a discontinuous jump in the receptor density as a function of the receptor. Such a ‘digital’ (all-or-none) response is different from the ‘continuous’ one seen in some type of sensory such as amoeba chemotaxis [17] in which an ability to continuously monitor the relative change of stimuli is necessary.H.. Given that a cooperative clustering mechanism of this type has been widely adopted by other families of growth factor signaling [1]. From the phase diagrams. for the cells to acquire the advantage of clustering? In the current system.: Death receptors: signaling and modulation. inhibitor densities.: Recognition at the cell surface: recent structural insights. Stuart. Biol. Park. 271 (1996). This might shed a light on the possible purpose for clustering.e. 1305–1308. C. J. Struct. Chem.I.

Acad. Zimmermann. W. Lett. 264 (1989). LEVINE Boldin. D. Y. New York.: Nucleation.. Gadella. 137 (1997). 270 (1995). J. J.. USA. J. Theo. Varfolomeev. Zhang.. and Pound. V.W. Sci.J. S.: Fluorescence resonance energy transfer reveals Interleukin (IL)-1 dependent aggregation of IL-1 type I receptors that correlates with receptor activation.M.. S. Tong. 16227–16238.K. 72 (1997).G. Phys. Biol.. 1543–1558. H. 365–374. H. and Jovin. J. J. 3096–3100. E.A. Shemer-Avni. Mosselmans. Karbach. Chem. Rev. 21 (1968).: Self-association of the ‘death domains’ of the p55 tumor necrosis factor (TNF) receptor and Fas/APO1 prompts signaling for TNF and Fas/APO1 effects. Y. Nature 387 (1997). N. Science 274 (1996). 20264–20272.. 2358–2365.. 9. 14. Phys. 244 (1998). GUO AND H. J. J. 27562–27568. 270 (1995). 129 (1995). Koland. Biol. and Fiers. A. Marcel Dekker. Nucleation and Growth Kinetics. C.. J.M. J. M. 141 (1988). T. P. Villa. Biol. Y. Ann. Wofsy.: Condensation and Evaporation. 77 (1999).. D. Abraham.. J.: Robustness in simple biochemical networks. 139 (1997).. Cell Res.M.C.D. Simm... and Goldstein. U.C. R. 270 (1995).: Dynamic distribution of chemoattractant receptors in living cells during chemotaxis and persistent stimulation. 543–546. T. M. Murphy. Chem. N.: The type 1 receptor (CD120a) is the high-affinity receptor for soluble tumor necrosis factor. Nature 398 (1999). 10. Immunol. J. A.: Physics of chemoreception.: World Scientific. Leicht. 533–538. J. E. G. Amit. 379–393. H. Garraway III. Sci. Barkai.. 108–157.S.: Ann. 913–917. 6. Woronicz.. 1194–1197. M. Guo. 15. Camonis.L. and Hoppe.: Induced expression of trimerized intracellular domains of the human tumor necrosis factor (TNF) p55 receptor elicits TNF effects. 16.C. Biophys. Vandevoorde. and Wallach. Z. H. Natl. New Jersey. Jiang..F.: Prevention of constitutive TNF receptor 1 signaling by silencer of death domains. I.. 193–219.. Academic Press. Langer. Matr. I. Zettlemoyer. Chem. B. Fenn. Cell Biol.: A thermodynamic model for receptor clustering. 387–391.A.. B. and Wu. Biophys.. Chem.L.M. Liu.: Late signals from PDGF receptors leading to the activation of p70S6 kinase are necessary for the transition from G1 to S phase in AKR-2B cells. 973–976. Cell Biol. Pro. 95 (1998). 54 (1969). Dumont. G.N. Biophys. D.. and Galand. and Levine.E.E. Wajant. W. and Leibler. D. Sup. Rosette. New York.. II.. I.P. 12. and Cerione. and Purcell. MacMillan. 41 (1967). 1974. 1963. Haugh. . J. A. Hirth. and Baird. M. C. R.: Endocytic pathway of recombinant murine tumor necrosis factor in L-929 cells. Proc..: Physical modulation of intracellular signaling processes by locational regulation. Berg.. River edge. P. Metzger. and Scheurich. F. Park. 11. Fiers. J. A. J. J. V.J. C. Mett.234 4.M.: Structural basis for self-association and receptor recognition of human TRAF2.B. Mao. 7. Exp. Dower. Grell. 2014–2031. 5. 8699–8707. Ad. Chem. G. W. J.H. L. 20 (1977). V.: Visualization of epidermal growth factor (EGF) receptor aggregation in plasma membrane by fluorescence resonance energy transfer. C. New York. S. J. and Goeddel. A. and Karin. K. Kent.R. C. Burkitt. Holowka. P..: Ultraviolet light and osmotic stress: activation of the JNK cascade through multiple growth factor and cytokine receptors. Hoppe. Guo. Cell Biol. J.P. Phys.V. 8. 13. D.. 1993. and Devreotes. D. H. Chumakov. Science 283 (1999). Biol. Xiao.: Oligomerization of epidermal growth factor receptors on A431 cells studied by time-resolved fluorescence imaging microscopy. 1969.: Kinetics of tyrosine phosphorylation when IgE dimers bind to FC receptors on rat basophilic leukemia cells. 17. Haegeman. 570–575. D. Hepburn.: Homogeneous Nucleation Theory. and Lauffenburger. 258–275..