Alternating host cell tropism shapes the persistence, evolution and coexistence of Epstein-Barr virus infections in human

Epstein-Barr virus (EBV) infects and can persist in a majority of people worldwide. Within an infected host, EBV targets two major cell types, B cells and epithelial cells, and viruses emerging from one cell type preferentially infect the other. We use mathematical models to understand why EBV infects epithelial cells when B cells serve as a stable refuge for the virus and how switching between infecting each cell type affects virus persistence and shedding. We propose a mathematical model to describe the regulation of EBV infection within a host. This model is used to study the effects of parameter values on optimal viral strategies for trans-mission, persistence, and intrahost competition. Most often, the optimal strategy to maximize transmission is for viruses to infect epithelial cells, but the optimal strategy for maximizing intrahost competition is for viruses to mainly infect B cells. Applying the results of the within-host model, we derive a model of EBV dynamics in a homogeneous population of hosts that includes superinfection. We use this model to study the conditions necessary for invasion and coexistence of various viral strategies at the population level. When the importance of intrahost competition is weak, we show that coexistence of diff erent strategies is possible.

cc cc University of Utah Institutional Repository Author Manuscript Alternating host cell tropism shapes the persistence, evolution and coexistence of Epstein-Barr virus infections in human Ciao T . Huynh Frederick R. Adler September 10, 2010 Abstract Epstein-Barr virus (EBV) infects and can persist in a majority of people worldwide. Within an infected host, EBV targets two major cell types, B cells and epithelial cells, and viruses emerging from one cell type preferentially infect the other. We use mathematical models to understand why EBV infects epithelial cells when B cells serve as a stable refuge for the virus and how switching between infecting each cell type affects virus persistence and shedding. We propose a mathematical model to describe the regulation of EBV infection within a host. This model is used to study the effects of parameter values on optimal viral strategies for trans-mission, persistence, and intrahost competition. Most often, the optimal strategy to maximize transmission is for viruses to infect epithelial cells, but the optimal strategy for maximizing intrahost competition is for viruses to mainly infect B cells. Applying the results of the within­host model, we derive a model of EBV dynamics in a homogeneous population of hosts that includes superinfection. We use this model to study the conditions necessary for invasion and coexistence of various viral strategies at the population level. When the importance of intrahost competition is weak, we show that coexistence of different strategies is possible. 1 Introduction Epstein-Barr virus (EBV) belongs to the herpesvirus family, infects over 90% of humans world­wide and persists for the lifetime of the person [18]. Most individuals infected with EBV are 1 University of Utah Institutional Repository Author Manuscript asymptomatic, but the virus has been associated with many diseases and cancers including infec-tious mononucleosis (1M), Burkitt's lymphoma, Hodgkin's lymphoma, and nasopharyngeal carci-noma (NPC). EBV is transmitted by intimate contact, mainly through saliva and oropharyngeal § secretion [2]. The virus primarily targets two cell types, B cells and epithelial cells. In B cells, H ~ EBV can establish a long-term infection. EBV can drive an infected B cell through different stages :> g of latent infection where the viral genome remains inside the cell. The virus turns off most of its :::J o M gene expression, stays quiescent, and remains invisible to the immune response within memory B ~b cells. These memory B cells can be activated and become plasma-like B cells within which virions C ~ can replicate and burst out (lytic infection). In epithelial cells, an EBV infection often results in M ~. rt virus replication and producion. Latent infection of epithelial cells is rare and has been observed only in the case of NPC. In vitro, viruses that emerge from one cell type preferentially infect the other [3]. The viral glycoprotein gp42 on EBV's envelope serves as the molecular switch that EBV uses to alternate infections between the two cell types. This viral protein is required for infection of B cells, but inhibits infection of epithelial cells. To enter a B cell, gp42 needs to bind to HLA class II molecules, antigen presenting proteins on the surface of antigen presenting cells, on the cell to trigger fusion. In B cells, gp42 interacts with HLA class II molecules, which are being synthesized before being presented on the cell surface, and becomes a target for degradation. Hence, viruses produced by B cells express low levels of gp42. Because epithelial cells do not express HLA-II, no such intracellular interactions occur and viruses produced by epithelial cells express higher levels of gp42 compared § to viruses produced by B cells [3]. As a result, viruses emerging from B cells are approximately ~ 5 times more efficient at infecting epithelial cells than the viruses emerging from epithelial cells, ~ which are 100 times more efficient in infecting B cells than viruses derived from B cells [4, 11]. rt :::J :; The viral infection of B cells and the maintenance of a persistent infection within the host have ~ been intensively studied [19]. The dynamics of EBV infection of B cells and the T cell responses ~ C ~ have been investigated using simulations of agent-based models, C-ImmSim and PathSim [6, 20]. M ~. The infection of epithelial cells, however, has not been included in these models. Simulations of these models reproduce qualitative features of real infections where the maximum number of 2 University of Utah Institutional Repository Author Manuscript infected cells and viruses occur sometimes between one to a few weeks after initial infection, and a persistent infection is established within months. PathSim shows that EBV cannot result in a persistent infection unless it can maintain a latent infection within the memory B cell population. § The simulation results of this agent-based model also highlight the sensitivity of the dynamics of H ~ infection to variation in the reactivation rate of lytic infection from infected memory B cells. A :> g small increase in this rate causes the number of infected cells to expand quickly and remain at a :::J g high level [20]. However, the sensitivity may be due to the lack of T cell expansion and proliferation ~b in this agent-based model. C ~ A mathematical model describing the within-host dynamics of EBV infection has been con- H ~. rt structed to study the T cell responses to persistent virus [7]. This model uses ordinary differential equations to track the number of latently and lytically infected cells, viruses, and T cells. However, it does not identify a specific class of target cells, such as B cells or epithelial cells. The model allows newly infected cells to produce viruses without going through latent stages of infection and allows lytically infected cells to become latently infected. The first of these two assumptions only applies for infection of epithelial cells. The second assumption does not reflect the biology of EBV infection of either cell type. Differential equation models have also been developed to study the dynamics of other herpesvirus infections like cytomegalovirus (CMV or HHV -5) [27] and HHV-6 [26]. Both CMV and HHV-6 can infect a wide range of cells in human and share the ability to establish persistent infection within the host cells with other herpesviruses [16, 28]. In this study, we develop and analyze mathematical models describing the within-host and § between-host dynamics of infection to explore the effects of switching between host cell types on ~ viral shedding, persistence, and evolution. Within a host, EBV must infect B cells to establish a ~ long-term infection. However, infection of epithelial cells plays an important role in transmitting the rt :::J g infection. In fact, EBV is transmitted mainly through saliva and most viruses found in the saliva of ~ infected hosts show characteristics of deriving from epithelial cells [13]. We propose a mathematical ~ C ~ model to capture the within-host dynamics of virus and host-cell interaction, including the dynamics H ~. of epithelial cell infection. Assuming EBV can modify its ability to infect B cells and epithelial cells, we use this model to compute viral strategies that maximize transmission and the total virus 3 University of Utah Institutional Repository Author Manuscript population being produced (intrahost competition). We then determine how changes in parameters affect the optimal viral strategy for transmission and for intrahost competition. While infection of epithelial cells plays a key role in transmission, intrahost competition emphasizes the role of § infection of B cells. Finally, we apply the results of our within-host model to derive a model H ~ of EBV infection at the population level to study the conditions for invasion and coexistence of :> g different viral strategies. :::J o H ~ b 2 Model of the within-host dynamics of an EBV infection c (fJ n H ~. 2.1 Model rt cc Naive Cell B1 oo~ Latently Infected B2 Infection ~ ------I.~ ~_ Differentiation ~ Plasma cell B4 o iffe rentiation Memory Cell B3 Figure 1: Model of EBV infection of B cells and epithelial cells. Our model adds the infection of epithelial cells to the model of EBV infection of B cells proposed by Thorley-Lawson [23]. Our mathematical model (Figure 1, and Equation 1) describing the dynamics of EBV infection within an infected host follows two types of target cells (B cells and epithelial cells), two types of viruses (B-cell-derived, VB , and epithelial-ceIl-derived, VE) , and two types of cytotoxic T cells or CTLs (attacking latently infected B cells, T2 , and lytically infected cells, T4, respectively). B cells break into four state variables: naive B cells (B1 ) , latently infected B cells (B2 ), latently infected 4 University of Utah Institutional Repository Author Manuscript memory B cells (B3), and lytically infected B cells or plasma cells (B4). B2 and B3 represent different stages of latency. B2 are newly infected naive B cells and expressing EBV latent genes (up to 9 of them). Because of this latent gene expression, T cells can recognize and kill these B2 § cells. The latent gene expression helps driving a B2 cell into a memory infected stage (B3), in H ~ which no viral gene being expressed. There is no T-cell response to B3 cells. Epithelial cells do :> g not ordinarily harbor latent virus and require only two state variables: uninfected epithelial cells :::J o H (Ed, and lytically infected epithelial cells (E4). The model consists of a system of 10 ordinary cc differential equations: (1) T2 (/J2T1 W(B2) + ()2T2W(B2) - 5T2 T4 ¢4Tl[W(B4 + E4)] + ()4T4[W(B4 + E4)]- 5T4. The dynamics of B cells obey these assumptions: • Naive B cells begin with an initial population of Bo and turn over at rate d1. They can encounter and be infected by the virus populations VB and VE with rates J.LBb VB and J.LEb VE, respecti vely. • An infection of a naive cell, B 1 , gives rise latently infected cells, B2 , by proliferation of these newly infected cells, where p is the proliferation factor. These B2 cells die at a rate d2, can be recognized and killed by effector T cells at a rate k2T2, and can enter the latently infected 5 cc cc University of Utah Institutional Repository Author Manuscript memory state at rate c, which is driven by EBV gene expression switching off. • Infected memory cells, B3, obey homeostatic regulation similar to normal memory B cells. They are invisible to the T cell responses, undergo division with a rate r. On average, one cell becomes lytically infected and one remains in the memory state. The rate sr represents the death of B3 due to homeostatic regulation of memory cells, where s is the regulation factor. For normal homeostasis, s = 2 balances the proliferation rate of 2r [15]. The frequent turnover of the memory B cell population helps maintain the supply of plasma B cells that produce antibodies against various types of antigens. EBV takes advantage of this mechanism to activate its lytic cycle. • Lytically infected B cells, B4 , arise from lytic reactivation of infected memory B cells, B3 , at a rate r, die and release viruses at a rate d4 , and can be killed by effector T cells at a rate The dynamics of epithelial cells assume the following: • Uninfected epithelial cells start with initial population Eo, and turn over at a rate de. They can encounter and be infected by the virus populations VB and VE with rates /-LBe VB and /-LEe VE , respectively. • Lytically infected epithelial cells, E4 , die naturally at a rate de, die from virus bursting out at a rate ",(, and can be killed by effector T cells at a rate k4T4. Free viruses, VB and VE, are produced from B cells and epithelial cells at rates nd4 and n"'( respec-tively, where n is the average burst size, and are cleared at a rate dv . To model the CTL response, we assume that the naive population, Tl, is regulated at a constant level and, upon stimulation by viral antigens, become effector cells against latent or lytic infection with a rate (h or ¢4, respectively. The activated effector cells, T2 and T4 , can proliferate further with rates ()2 and ()4, respectively, upon stimulation by viral antigens from infected cells. Each type of effector cell dies at a rate 8. Activation and proliferation of CTLs are saturating function of the 6 cc cc University of Utah Institutional Repository Author Manuscript available infected cells, w(Bj ) B· J K+Bj ' (2) where K is the number of infected cells at which activation or proliferation is half maximal. Table 1 presents the parameter values used for simulation and analysis of the model. This system of Equation 1 has two equilibria. The infection-free equilibrium is given by B~ = Bo , E~ = Eo, and every other state variable equals zero. If virus replication is sufficiently efficient, there is also a persistent equilibrium, where all state variables take on positive values. The stability of the infection-free equilibrium is determined by the basic reproductive ratio, Ro n (Pf.-LBbBOC f.-LEeEo'Y) 2d~ (8 - 1) (d2 + c) + de + 'Y n (Pf.-LBbBOC f.-LEeEo'Y) 2 4Pf.-LEbBOCf.-LBeEO'Y + 2d~ (8 - 1)(d2 + c) - (de + 'Y) + (8 - 1)(d2 + c)(de + 'Y) , (3) of EBV in a naive host [9] that can be found by the next generation matrix [25]. Infections of both B cells and epithelial cells contribute to the basic reproductive ratio of EBV. If Ro < 1, the infection-free equilibrium is stable and a long term infection cannot establish within a host. If Ro > 1, the infection-free equilibrium is unstable and EBV can establish a persistent infection. The dynamics of infected cells, viruses, and T cell responses for the case when Ro > 1 are shown in Figure 2. The populations of infected cells and viruses peak during the second week of infection and then resolve down to low levels at equilibrium. Infection of epithelial cells plays a larger role during the primary infection than during the long-term infection. Simulation of EBV infection using the agent-based model, PathSim, produces clearance of virus and infected-cells in the case of a small viral burst size or the inability of EBV to enter latency within memory B cells [20]. Furthermore, 7 cc cc University of Utah Institutional Repository Author Manuscript EBV infection also cannot be established within people with X-linked agammaglobulinemia because they do not have mature B cells. Our model produced similar results; small virus burst size (small value of n), no latency within memory B cells (c = 0) , or no mature susceptible B cells (Bo = 0) leads to Ro < 1 and virus clearance. 4 ~ 6~x_1_0 __ ~ ______ ~ ____ ~ ____ ~ Coo ~ 0)= 2 _1 0 U0) 4 .•. - - - 83 -"'0 .. o 0) .. ~....... II ~ ~ 2 ~ I E 'E ~I II :::J .- \ - Z Oo~ -'~-1-0-0 -'-'--2'-0-~0-- -------3--0-0- -------4-~0 0 days ~ 1500~----~----~----~-----. ~ ~1000 1----~: 1 -o "0'0) .£) - I E.~ I Q5 ~ 500 I~ ~ °O~\ ~----~~~~~-~-~~~-~-~-~-~ 100 200 300 400 days 00 .s2; 4 0) ~ 2 LL oo ~ --100 200 days 300 400 15000~----~----~----~--~ PJ ~ .....J I- ~ 10000 o ~ 0) ~ 5000 :::J Z 100 200 days 300 400 Figure 2: Dynamics of infected cells, viruses, and T cell responses using parameter values in Table 1 2.2 The effects of change in parameter values on transmission, persistent infec­tion and the dynamics of EBV infection Changes in the parameter values can alter the transmission efficiency, equilibrium viral loads, and short-term dynamics after initial infection. Transmission, as we have seen, depends on the amount of viruses produced by epithelial cells. Most EBV-infected individuals are healthy carriers and can transmit the virus throughout their lifetime. Although the amount of transmissible epithelial cell viruses varies during the course of the infection, over the long term, maximizing transmission is equivalent to maximizing the equilibrium value of epithelial cell viruses (VE) produced in the long 8 cc cc University of Utah Institutional Repository Author Manuscript run. A host can be infected by multiple strains of EBV, at least during primary infection [21]. In the case of a perfectly cross-reacting T cell response (i.e., one epitope of T cells responds to all strains of EBV) and a single cell type to infect , the principle of competition exclusion implies that multiple strains cannot coexist within a host [5]. In our model, there are two types of target cells and this principle does not necessarily apply. Numerical solutions of our model, however, do not display equilibrium coexistence in a host. Furthermore, the strain with higher total number of viruses being produced at equilibrium (V; = V~ + V~) wins out to establish a persistent infection within a host. Assuming EBV can modify its ability to infect B cells and epithelial cells, we investigate the optimal strategy the virus would used to maximize transmission or to maximize viral load in persistent infection. ....... t.:I -Q:I t:: ..o.... . .0 « ( a) \\ ........................ Ability to infect epithelial cells aB 0 ,-, ,', , I , I I :r-':Y, , I ~ B E ,', ",-~ v *lC..J* B E (b) ,-, I , I I .:. I *1'. " I B E [\---"1 *1': ,,,' B E aE 1 Figure 3: Trade-off in the ability to infect B cells and epithelial cells. (a) For a virus, increase in infection of B cells results in decrease in infection of epithelial cells and vice versa. (b) The strategies of VB and VE are defined as aB and a E , respectively; as these two parameters increase, both types of viruses prefer to infect epithelial cells. We first develop a model to describe constraints in viral strategies within a host. We have assume that J-LEb > J-LBb and J-LBe > J-LEe (Table 1) since VE are more efficient at infecting B cells and VB are more efficient at infecting epithelial cells [3]. This same study also found that infection efficiency varies among EBV strains. Assuming EBV can modify its ability to infect B cells and epithelial cells, it can only do so if there is a trade-off between the ability to infect B cells and 9 University of Utah Institutional Repository Author Manuscript epithelial cells (Figure 3a). For example, VB can improve its ability to infect B cells by increasing its expression of gp42. By doing so, however, it decreases its ability to infect epithelial cells. To model this tradeoff, we constrain the infection rates to be weighted averages of the pure VB § and VE strategies, which are represented by jiB and JiE respectively. Let ryB and ryE be the modified cc infection rates of VB and VE, respectively, given by ( TJBb ) TJBe ( aB J.LBb ) + (1 - aB) ( J.LEb ) J.LBe J.LEe (4) ( TJEb ) TJEe ( = aE J.LBb ) + (1 - aE) ( J.LEb ) . J.LBe J.LEe (5) The weighted parameters aB and aE define strategies of VB and VE, respectively. If aB = 1 and aE = 0 (Figure 3b, upper left), the TJ'S take on the same values as the J.L 's. As aB decreases, VB switch from infecting epithelial cells to infecting B cells. As aE increases, VE switch from infecting B cells to infecting epithelial cells. When aB and aE are small, both VB and VE prefer to infect B cells (Figure 3b, bottom left). In contrast, when aB and aE are large, both VB and VE prefer to infect epithelial cells (Figure 3b, upper right). At the bottom right corner of Figure 3b, where aB is small and aE is large, VB prefer to infect B cells and VE prefer to infect epithelial cells. Replacing J.L by TJ in the system of Equation 1, the equilibrium values VE(ai) and V;(ai) can be found numerically for any given strategy (aB' aE). Our next step is to analyze the effects of parameters on viral strategies that maximize VE and V;. Natural variation between individuals in the expression of EBV receptors (HLA class II on B cells and CR2 or integrins on epithelial cells) can affect the susceptibility of these cells to EBV infection [8, 12]. The susceptibility of B cells and epithelial cells to virus classes VB and VE are described by the parameters J.LBb , J.LBe, J.LEb, and J.LEe. We thus focus our sensitivity analysis on these four parameters and the other three parameters that are related to the production of viruses within a host: the rate of virus replication and production from B cells (d4 ) and from epithelial cells (r), and the viral burst size (n). We consider each of these parameters at three different values: the 10 University of Utah Institutional Repository Author Manuscript value listed in Table 1 (middle), this value divided by two (small) and multiplied by two (large). For example, the three values for n are 500, 1000, and 2000. We thus have 37 (or 2187) parameter combinations to study. We ran the same sensitivity analysis on all other parameters in Table 1 but § found minimal effects on the optimal strategy for transmission and establishment of a persistent H ~ infection. :> C rt o::: J 2.2.1 Identifying strategies that maximize transmission H ~b We assume that EBV maximizes transmission to new hosts by maximizing equilibrium amount of C ~ virus produced by epithelial cells (VE). We divide the (\:B(\:E-plane in Figure 3b into a 21 x 21 H ~. rt grid where each of the 441 points on the grid represents a viral strategy. For every parameter set, cc we find VE numerically for all strategies and locate the strategy where VE is maximized. Figure 4 shows the influence of parameter values on the viral strategy that maximizes VE. We focus our presentation of the results at the four corners where VB and V E prefer to switch infections between the two cell types (Region I) , to primarily infect epithelial cells (Region II) , to mainly target B cells (Region III), and to infect their own producing cells (Region IV). • A minority of parameters (180 out of 2187 sets) favor a strategy similar to what has been observed in vitro where VB preferentially infect epithelial cells and V E preferentially infect B cells (Region I) [3]. The parameter sets that maximize VE in this region tend to have f..J,Be, n , and '"Y small and f..J, Ee and d4 large. • A large number of parameter sets (1548 out of 2187) favor strategies where both VB and VE preferentially infect epithelial cells (Region II). • The remaining parameter sets (459 out of 2187) favor strategies that lie in a range where (\:B is large and (\:E is at an intermediate value. • There are no parameter sets favoring strategies in Region III and IV where (\:B is small, implying that variation in parameter values has minimal effect on the strategy of B cell viruses in infecting epithelial cells to maximize transmission. 11 cc cc 1 (]) :::J (ij > o University of Utah Institutional Repository Author Manuscript Region I: etB 2 0.75 and etE ~ 0.25 180 sets (]) ~- 0 o o o 0 ....J E :::J '5 - 0 (]) o 0 000 0 :2: 'iii E - 0 (j) 00 0 00 !-lEb !-lBb !-lBe !-lEe n d4 Y Parameters rfJ ~ <l) rfJ OJ I.i.> ~ (]) :::J (ij > Region II: etB 2 0.75 and etE 2 0.75 1548 sets (]) e> ro ....J E :::J '5 (]) :2: 'iii 0000 E 0 (j) !-lEb !-lBb !-lBe !-lEe n d4 Y Parameters Region III: etB ~ 0.25 and etE ~ 0.25 Region IV: etB ~ 0.25 and etE 2 0.75 No set of parameters No set of parameters 1 Figure 4: Effects of parameter values on optimal transmission (maximizing VB)' Each parameter takes on three different values: small, medium, and large. The four panels summarize parameter sets that give optimal VB at the four corners in Figure (3b). Region I: aB 2: 0.75 and aE :S 0.25; Region II: aB 2: 0.75 and aE 2: 0.75; Region III: aB :S 0.25 and a E :S 0.25; Region IV: aB :S 0.25 and a E 2: 0.75. The dot size represents the number of times the parameters take on particular values. Figure 5 shows the time-course dynamics of free viruses and the T cell responses for sample sets of parameters from Region I and Region II in Figure 4. The viral load initially elevates, peaks during the second week of infection, and then resolves down to a low equilibrium level. Strategies in Region II of the etBetE-plane give a higher VB at the expense of a lower VB' 2.2.2 Identifying strategies that maximize total viral load We now make the assumption that a host gets infected by two or more EBV strains that differ only in their strategies on the etBetE-plane. In the case of a perfectly cross-reacting T cell response, our numerical results show that a virus with a higher V; would win out to establish a persistent infection with no stable coexistence of multiple strains within a host. We thus assume that the 12 cc 1 o University of Utah Institutional Repository Author Manuscript Region I: CtB 2 0.75 and CtE ~ 0.25 ~ 106 .... "> Q) , - VB ____ .., ---VE ~ 104 \ '-" " ,,-- ... _----------- o ~ ; 0> ' .... ' .Q 102~----~----~----~--~ o 1 00 200 300 400 :J days 105 I- () '.0... Q) ..0 -.:r- - T2 , ---.--------- - - -T4 E :::J E-Region II: CtB 2 0.75 and CtE 2 0.75 o T'" 0> .Q --------- o 100 200 300 400 days :J 105~--~----~----~====~ I­() '.0... Q) ..0 E :::J c .;:..::..,.- ---- --- - ------- 6r100 ~~100~----~--~~--~~--~ .Q 0 1 00 200 300 400 0> 0 1 00 200 300 400 ------------------~~~~-----------------------~----------------~~~~--------------- Region III: CtB ~ 0.25 and CtE ~ 0.25 Region IV: CtB ~ 0.25 and CtE 2 0.75 No set of parameters No set of parameters 1 Figure 5: Effects of parameter values on dynamics of the model for sample sets of parameters from the two upper corners in Figure 4. § intrahost competition will select the virus that maximizes the total number of virions produced at H ?:J equilibrium. For each of the 2187 sets of parameters, we found V; numerically and located the :> g strategy where V; is maximized. ~ o H ~ Figure 6 shows the influence of parameter values on viral strategy that maximizes V;. The optimal ~ ~ strategies lie predominantly in Region III (1449 out of 2187 sets of parameters), 294 sets in Region H ~. II, with only a few conditions favoring a strategy in Region I (12 sets) or Region IV (33 sets). The remaining parameter sets (389) favor strategies in the middle and between the four regions. 13 cc cc 1 o Q) :::J ro > University of Utah Institutional Repository Author Manuscript Region I: CtB 2 0.75 and CtE ~ 0.25 Region II: CtB 2 0.75 and CtE 2 0.75 OJ ~ C1l ....J E :J ii OJ ~ ro E (j) !--lEb !--lBb 12 sets !--lBe !--lEe Parameters n y 1 1 1 1 1 1 I Q) :::J ro > OJ ~ C1l ....J E :::J ii OJ ~ ro E (j) 0 0 0 0 o o I I !--lEb !--lBb 294 sets 0 0 0 0 0 0 0 0 0 o o o o I I I I !--lBe !--lEe n y Parameters _____________________________________ 1 ___________________________________________ _ Region III: CtB ~ 0.25 and CtE ~ 0.25 1449 sets jOOOQO 0 OOOOCXXJ ]0 !--lEb !--lBb !--lBe !--lEe n d4 Y Parameters I I 1 I CtE Q) :::J ro > Region IV: CtB ~ 0.25 and CtE 2 0.75 33 sets OJ ~ 0 C1l ....J E :::J ii OJ ~ ro E (j) !--lEb !--lBb !--lBe !--lEe n d4 Y Parameters 1 Figure 6: Effects of parameter values on a persistent infection (maximizing V;) with similar notation as in Figure 4. • In Region II, both VB and VE prefer to infect epithelial cells. The parameters values for which V; is maximized in this region are characterized by a high rate of virus production in epithelial cells (rr) and a low rate of virus production in B cells (d4 ) . • The majority of parameter values favor strategies in Region III where both VB and VE prefer to infect B cells. This implies that the infection of B cells plays a key role in the intrahost competition and the establishment of a long term infection. Time-course dynamics of free viruses and the T cell responses for the four sample sets of parameters from the four Regions I-IV are shown in Figure 7. Higher levels of V~ often come at the expense of a lower VB' 14 cc cc 1 o University of Utah Institutional Repository Author Manuscript Region I: CtB ~ 0.75 and CtE ~ 0.25 -.".-------- o 100 200 300 400 days ~ 105~----~--~~--~~==~ a ~ _ - T _ " _____ --------- 2 ~ t --- T4 ..0 E ::J C --- Region II: CtB ~ 0.75 and CtE ~ 0.75 o ~ 0> .Q ::J I- a- 0 .... Q) ..0 E ::J E-o 105 ~B ,I " .. .. ... -' .. - - - - - - - - VE 100 200 300 days 400 , ______ ---------- - T2 ~ T I~ ---------------~ --- 4 o 0 0 o ~10 g110 0 100 200 300 400 ~ 0'--------1-0~0--2~0~0--3~0-0--4-----'00 -~---------------_Qqy~---------------------------------------_Q~y~--------------- Region III: CtB ~ 0.25 and CtE ~ 0.25 o 0> .Q ~ - VB , ---VE \ \ .. .. --;----------------- o 100 200 300 400 days ~ 105~--~----~----~==~ ~ ~_ - T o " --- ___ -------- 2 ~ --- T4 ..0 E ::J C --- ~100~--------'----~----~----~ ~ 0 100 200 300 400 da s Region IV: CtB ~ 0.25 and CtE ~ 0.75 o ~ 0> .Q o ::J 105 I- a '.0... Q) ..0 E ::J C --- ,.".--._---------- 100 200 300 400 days T ~~- ~- ~- ~- -------------------------, -- 2 - - - T4 o d)100~--------'----~----~----~ .Q 0 100 200 300 400 da s 1 Figure 7: Effects of parameter values on dynamics of the model for sample sets of parameters from the four Regions I-IV in Figure 6. 15 cc cc University of Utah Institutional Repository Author Manuscript 3 A model of between-host dynamics For most sets of parameters, maximizing VE for transmission and maximizing V; for intrahost competition favor different strategies. To predict the outcome of this conflict, we develop an implicit model of EBV dynamics in a homogeneous population of hosts that includes both transmission and superinfection. This model tracks susceptible hosts and hosts infected by one of the m strains of viruses. We make the assumption that virus strains do not coexist within a single host. A more competitive virus, with a higher value of V; , can take over a host from the one with a lower value of V; although the takeover requires time [1 , 17]. We first analyze the model for the case where m = 2 and find the conditions for coexistence of different viral strategies within a population. We study the case where m > 2 primarily through simulation. A host population infected with two strains of virus described by ct1 and ct2, where cti = (0:1,,0:1;), is captured by the following system of equations with three state variables: susceptible hosts (S), hosts infected by strain 1 (h), and hosts infected by strain 2 (h). aN - p,(ct1) ~S - p,(ct2) ~ S - as , p,(ctI) ~ S - p,(ct2) V(ct2' ct1) ~ h + p,(ctI)V (ct1 , ct2) ~ h - ah , (6) ( ..... )h S ( ..... ) ( .......... ) h ( ..... ) ( ..... ..... )h P, 0:2 N + P, 0:2 V 0:2,0:1 h N - P, 0:1 V 0:1 , 0:2 N h - ah. This model assumes that • There is no infection-induced death because mortality due to EBV infection is very rare even in the case of severe infectious mononucleosis. The total population, N = S + h + 12 , is conserved by setting the per capita birth rate equal to the death rate, a. • Susceptible individuals are infected by strain i at a rate p,(cti)1dN, where p,(cti) = aVE(cti). This implies that the infection rate of strain i is proportional to the number of epithelial cell viruses of that strain being produced within an infected host. • Hosts infected by strain i can be superinfected, or taken over, by virus strain j at a rate 16 cc cc University of Utah Institutional Repository Author Manuscript J-L(aj)v(aj , ai)Ij/N. Because intrahost competition is determined by V; , we assume that where 9 is an increasing function with g(O) = 0 and g( -x) = -g(x). We set 2 g(x) = -arctan(x/A) , 7r (7) which is scaled so that g' (0) = 1/ A and lim g(x) = 1. x---+oo • For all i, the transmission rate is larger than the birth and death rates (J-Li > a). Equation 6 can be rewritten as (8) where summarizes the competitive ability of strain 2 relative to strain 1. If Ve(aI) > Ve(a2) and V;(aI) > V;(a2), strain 1 is better at both transmission and intrahost competition and will thus always win. If Ve(a1 ) > Ve(a2) and V;(aI) < V;(a2), strain 1 is better at transmission while strain 2 is better at intrahost competition. Then J-Ll = J-L( aI) > J-L2 = J-L( ( 2) and 121 > o. 17 cc cc University of Utah Institutional Repository Author Manuscript Equation 8 has four equilibria: Po (0,0), ( N(l - 1~1)' 0) , (O,N(l- :2))' and ( (/11 - (J)N - (/11 + hd1:;' ,1:;' = /11N _ (J N) . /11 /11 - /12 + 121 121 Linear stability analysis implies that: • The disease-free equilibrium, Po, is unstable if /1i > (J. • The strain 1 equilibrium, PI, is positive and stable if 121 < O"~~~ =~)) . • The coexistence equilibrium, P12, is positive and stable if O"~~~ =~)) < 121 < O"~~~=~)) . • The strain 2 equilibrium, P2, is positive and stable if 121 > O"~~~=~)) . (9) If 121 is small, the force of superinfection is weak and strain 1 wins since it has a higher transmission rate. As 121 increases, strain 2 can coexist, and will eventually outcompete strain 1 for larger values of 121. The model with m > 2 strains of viruses can be written as (10) Let m 1. F· - "f··-.2 2 - L 2J N' j=l where Iij = 0 when i = j. Using the assumption of constant population size, Equation 10 can be 18 cc reduced to University of Utah Institutional Repository Author Manuscript (11 ) ~ We can think of Fi as the total force of strain i taking over other strains. If V;(ai) is small, Iij :c> can be negative and, hence, Fi can have a negative value. If V;(ai) is large, Iij is more likely to rt :::J g be positive and so is Fi . At equilibrium, we have cc It = 0 N (,"i (1 -t. Ij / N ) + F[ - if ) or It = f..-li Strain i thus can exist at equilibrium if Ft + f..-li (1 -f Ij / N) > (j. j=l-i (12) The condition for invasion can be expressed in terms of Equation 12. Let G be a set of strains (G C {I, 2, ... , k}) that can exist in the equilibrium state. Strain i, which is not in G, can invade the equilibrium if Equation 12 is satisfied. This implies that if strain i is not intrahost competitive (low V;(ai) and hence Ft is negative), then it can only invade the equilibrium if it has a high transmission rate (high VE(ai) and hence f..-li). The force of infection, Ft, depends on the scaling factor A (Equation 7). When A is small, IIij I is large and the force of superinfection is strong making it easier for a strain with a higher V; to win via intrahost competition (Figure 8: row 1, col. 1). When A is large, the force of superinfection is weaker. It is more difficult for a strain with a higher V; to take over a host from a strain with a lower V;. Instead, strains with a lower V; but a higher VE (higher transmission) can invade and exist at equilibrium (Figure 8). The pressure of intrahost competition thus favors strains with small D:B and small D:E to maxi-mize V;. The pressure of transmission, however, supports strains with large D:B and large D:E such 19 cc cc University of Utah Institutional Repository Author Manuscript I I 1 I . (jaJ O.Sf I 1 I · 1 I 1 at A = 1 O.S A = 3.9*107 . I . . . . I I .• I j I . . I . . . . I I 1,--------~--------, I · 1 I · I I · 1 (jaJ O.Sf • . j 1 1 1 . * ~51 1 ! * 0.4p I · 1 01 • ~ o O.S A = 5.38*108 1 ,...-:~'1-t-O:"-------' II .0...1 1 0· .11~ ·*··. * . 0..1 1 I1 1 * * 0.111 I . . . 0.1 .1. O· 11 * 1 1 • * 1 (jaJ O.Sf • . . . . . . . . 0.11 I . . 1 1 ! ! ! OL _.J o O.S 1 A = 5.71*108 11---~it-+~--------, I . 0.12 * 0.12 • • • 1 !. ..•. 0.12 ~ *- ~.12 . I I · ... 0.12 * 0*12 I . 0.12 · . I (jaJ o.Sf • • •. j ! . ! I I II OL . . J o O.S 1 A = 9.99*108 m l ' 1 >l (j o.sr . 1 1 ! . . oL~~----~~~~~J o O.S 1 uE A = 9*106 r > ~ O.S~ j I · 0.49 · • . J I· .. •• ** 0:51 ...• OL--------~-------- o O.S 1 A = 9.9*107 1i---:-----~--------l 1 1 ! . . . · 1 1 1 1 * 1 1 O.Sf .. . .... . .. . .. j I . . . . 1 1 1 I · 1 oL _ ______ ·~·~·_·_·_·_ __ •__ J o O.S 1 A = 5.43*108 11--~~'-------l I 0.1 0~11 ¥.1* 0.11 .... ! . . . 0.11 * * ~ 11 I O.S~ . . .. . • 0.11 O~1 I . · •••• 0 1 . J o O.S A = 5.81*108 1r--~~ )~I(f2~~1-------'1 1 . • •• * 9.11 .... I 0.11 * 0.11 I ·. 0.11 · * . *- o.sf . .• • • . .. 0.11 1 • • • • • • • • • I . II . . . .. .... . ! 1 • • • • • • OL--------~---------.J o O.S 1 A = 4.37*109 1,---------~-------, I. . . . . . .O.S* .* ! • . . . . · ··· · ··· 0:51 1 1 I . . ... 1 O.Sf • • . . ... • j I · 1 I · : 1 1 1 I . . . I 01 ----.J o O.S 1 uE A = 1.9*107 1 i--------;---------l I I 1 1 1 1 1 1 O.Sf j I • I II · • • *1 II OL- ~ o O.S A = 4.9*108 1i-------~~-------l 1 1 1 . . * 1 . . . 1 1 1 1 1 1 1 O . S~ . j 1 1 1 1 oIL ___._ ____ ~ ______._ _~ 1 o O.S 1 A = 5.48*108 11---~-*~--------, 10.12 . * 0.12 · .. 1 I ·· · 012 *" ·* 0.12 · 1 ! . . . . , 0:12 * * 0.1~ 1 • . • • • 0.12 * 1 O.Sf j I . . 1 1 1 ! ! o L_· ___ · __ ~~ __ · ___ · __ J o O.S 1 A = 5.88*108 1I, --·-·-·~1- -~----. -. -. -. -,I I · ! I1 · • • I1 o.Sf . . .. . j I . I II • • II 1 . 1 OL J o O.S 1 A = 1*1010 ~, • .. · 1 1 . . . . . . I • • • • • • 1 • oj ! • · 1 • I Figure 8: Snapshots of t he winning and coexisting strains with t he corresponding density of infected hosts at equilibrium (red '*') as A increases (Ifij l decreases) for a sample set of parameters from Region III in Figure 6. We consider 441 strains (blue ,*, ) representing different viral strategies on t he CtBCtE plane. 20 cc University of Utah Institutional Repository Author Manuscript that VIE is maximized. When the importance of intrahost competition is weak, strategies of higher transmission become dominant and coexistence of multiple strains is possible at the population level. H ~ 4 Discussion :> c rt 5 We have developed models of the within-host and the between-host dynamics of EBV infection to H ~ study the effects of switching host cell tropism on transmission, persistence, and viral evolution. The ~ ~ c model of the within-host dynamics tracks B cells, epithelial cells, T cells, and viruses. Although our (fJ n H ~. model does not explicitly include spatial dynamics, the numerical solution of the dynamics of B cell rt cc infection (Figure 2) are consistent with simulations of the agent-based model, PathSim, such that small virus burst size or exclusion of a memory stage produces clearance [20]. This result supports the key role of virus latency within memory B cells in the establishment of a persistent EBV infection within a host. The agent-based models, however, have not yet address the importance of epithelial cell infection. Viruses derived from B cells (VB) preferentially infect epithelial cells and viruses derived from epithelial cells (VE ) preferentially infect B cells [3]. Viruses derived from epithelial cells are the main type that is shed in saliva, and play a key role in transmission [13]. Assuming EBV can modify its ability to infect B cells and epithelial cells, we compute viral strategies that maximize transmission and persistent infection. Maximizing transmission is assumed to be equivalent to maximizing the equilibrium amount of virus produced from epithelial cells. Maximizing intrahost competitive ability, in contrast, favors strains that produce the highest total number of viruses at equilibrium. Under most conditions, the optimal strategy for transmission is having both types of viruses (VB and VE ) to preferentially infect epithelial cells. Under conditions when the rate of virus replication and production from B cells is high and production from epithelial cells is low, maximizing transmission favors viral strains that have viruses produced by one cell type infecting the other, in accord with observations of viral behavior in vitro [3]. In contrast, the optimal strategy for establishing and maintaining a persistent infection (intrahost competition) is for both types of 21 University of Utah Institutional Repository Author Manuscript viruses to mainly infect B cells, an observation consistent with reports that, at least in healthy carriers, EBV primarily targets and maintains its persistent infection within the population of B cells [10]. a For the majority of parameter combinations we studied, maximizing V~ for transmission and H ~ maximizing V; for intrahost competition favor different strategies. No stable coexistence of multiple :> g strategies is observed within a host. This may be due to the assumption of perfect cross-reactivity of :::J o H T cells in our model. To further understand the conflict of strategies, we combine these results of the ~b within-host model with a model at the population level and show that coexistence is possible. This C ~ between-host model includes both transmission and superinfection. In this case, when the force of H ~. rt superinfection is strong, viruses that preferentially infect B cells survive within and between hosts. When the importance of intrahost competition is weak, strategies of higher transmission become dominant and coexistence of multiple strains is possible at the population level. Multiple strains of EBV, indeed, have been detected within healthy carriers for periods of more than 300 days [21]. There are at least two possible explanations for this. If multiple strains coexist at the host population level, as in our model, but within-host takeovers are slow, individuals could have transient multiple infections. Alternatively, coexistence of multiple strains within a host could result from either partial or no cross-reactivity of T cell responses. Our model can be modified to allow for either of these complications. The existence of multiple strains, T cell responses, and the switching of infection between cell types may not only play important roles in the evolution of EBV but also in the pathology of a EBV-associated diseases like infectious mononucleosis and nasopharyngeal carcinoma. Infectious ~ mononucleosis (1M) is thought to be caused by primary infection of EBV in teenagers and young ~ adults with symptoms including fever, fatigue, and sore throat. The overwhelming number of rt :::J :; T cells, especially against viral lytic proteins cause these symptoms [10]. With the existence of ~ multiple strains, it is possible that 1M may be caused by a secondary infection with a strain that is ~ C ~ particularly efficient at lytic replication and production of new viruses. For example, an infection H ~. with a strain that is highly efficient at infecting epithelial cells may generate an extreme T cell response against EBV lytic proteins and cause symptoms of 1M. 22 cc cc University of Utah Institutional Repository Author Manuscript Nasopharyngeal carcinoma is a cancer of epithelial cells in the nose and pharynx. EBV infection of epithelial cells normally results in lytic infection. In NPC patients, however, EBV can maintain latency within epithelial cells. Expression of virus latent proteins within these cells contributes to cell proliferation, cell survival, and inefficient T cell responses, all of which can accelerate tumor development. The shift of tropism from B cell to epithelial cell disease may be induced by im-munoglobulin A (IgA) [22], which can enhance the viral entry into epithelial cells while interfering with the infection of B cells. Ongoing research extends the basic model of within-host dynamics (Equation 1) to study the association of EBV infection with these two important pathologies, 1M and NPC. Acknow ledgements We would like to thank Dr. Thorley-Lawson and other members of his laboratory for the opportunity to visit their lab and study the biology of EBV infection of B cells, and Dr. Hutt- Fletcher for insightful discussions on EBV infection of epithelial cells. Funding for this work was provided by the National Science Foundation (NSF) Research Training Group (RTG) (Award Number DMS0354259) and a 21st Century Science Initiative Grant from the James S. McDonnell Foundation. References [1] F .R. Adler and J. Mosquera. Is space necessary? Interference competition and limits to biodiversity. Ecology, 81(11):3226- 3232, 2000. [2] W.A. Andiman. Epidemiology of primary Epstein-Barr virus infection and infectious mononu-cleosis. In A. Tselis and H.B. Jenson, editors, Epstein-Barr Virus, volume 1, pages 39- 57. Taylor & Francis Group, New York, NY, 1st edition, 2006. [3] C.M. Borza and L.M. Hutt-Fletcher. Alternate replication in B cells and epithelial cells switches tropism of Epstein-Barr virus. Nat. Med., 8(6):594-599, June 2002. 23 cc University of Utah Institutional Repository Author Manuscript [4] C.M. Borza, A.J. Morgan, S.M. Turk, and L.M. Hutt-Fletcher. Use of gHgL for attachment of Epstein-Barr virus to epithelial cells compromises infection. J. Viral., 7(10):5007- 5014, May 2004. [5] H.J. Bremermann and H.R. Thieme. A competitive exclusion principle for pathogen virulence. J. Math. Biol., 27(2):179- 190, 1989. [6] F. Castiglione, K. Duca, A. Jarrah, R. Laubenbacher, D. Hochberg, and D.A. Thorley-Lawson. Simulating Epstein-Barr virus infection with C-ImmSim. Bioinformatics, 23(11):1371- 1377, Mar. 2007. [7] M. Davenport, C. Fazou, A.J. McMichael, and M.F.C. Callan. Clonal selection, clonal senes-cence, and clonal succession: the evolution of the T cell response to infection with a persistent virus. J. Immunol. , 168(7):3309- 3317, 2002. [8] K.M. Haan, W.W. Kwok, R. Longnecker, and P. Speck. Epstein-Barr virus entry utilizing HLA-DP or HLA-DQ as a coreceptor. J. Virol. , 74(5):2451-2454, Mar. 2000. [9] J.M. Heffernan, R.J. Smith, and L.M. Wahl. Perspectives on the basic reproductive ratio. J.R. Soc. Interface, 2(4) :281-293, 2005. [10] A.D. Hislop, G.S. Taylor, D. Sauce, and A.B. Rickinson. Cellular responses to viral infection in humans: lessons from Epstein-Barr virus. Annu. Rev. Immunol. , 25(1):587-617, 2007. C [11] L.M. Hutt-Fletcher. EBV entry and epithelial infection. In Robertson ES, editor, Epstein-Barr C ~ Virus, volume 1, pages 359-378. Caister Academic Press, Norfolk, England, 1st edition, 2005. :&> [12] L.M. Hutt-Fletcher. Epstein-Barr virus entry. J. Virol., 81(15):7825- 7832, Aug. 2007. o H ~ [13] R. Jiang, R.S. Scott, and L.M. Hutt-Fletcher. Epstein-Barr virus shed in saliva is high in ~ ~ C B-cell-tropic gp42. J. Viral., 80(14):7281- 7283, 2006. (fJ n H ~. rt [14] L. Jones and A. Perelson. Opportunistic infection as a cause of transient viremia in chronically infected HIV patients under treatment with HAART. B. Math. Biol., 67(6):1227- 1251, 2005. 24 cc cc University of Utah Institutional Repository Author Manuscript [15] D.C. Macallan, D.L. Wallace, Y. Zhang, H. Ghattas, B. Asquith, C. Lara, A. Worth, G. Panayiotakopoulos, G.E. Griffin, D.F. Tough, and P.C. Beverley. B-cell kinetics in hu­mans: rapid turnover of peripheral blood memory cells. Blood, 105(9):3633- 3640, 2005. [16] E.S. Mocarski, T. Shenk, and R.F. Pass. Cytomegaloviruses. In D.M. Knipe and Howley P.M, editors, Field's Virol, volume 2, pages 2701- 2772. Lippincott Williams and Wilkins, Philadelphia, PA, 5th edition, 2007. [17] J. Mosquera and F.R. Adler. Evolution of virulence: a unified framework for coinfection and superinfection. J. Theor.Biol, 195(3):293-313, 1998. [18] A. Rickinson and E. Kieff. Epstein-Barr Virus. In D.M. Knipe and Howley P.M, editors, Field's Virol, volume 2, pages 2575-2627. Lippincott Williams and Wilkins, Philadelphia, PA, 4th edition, 2001. [19] E.S. Robertson, editor. Epstein-Barr virus, volume 1. Caister Academic Press, Norfolk, England, 1st edition, 2005. [20] M. Shapiro, K.A. Duca, K. Lee, E. Delgado-Eckert, J. Hawlins, A.S. Jarrah, R. Laubenbacher, R. Laubenbacher, N.F. Polys, V. Hadinoto, and D.A. Thorley-Lawson. A virtual look at Epstein-Barr virus infection: Simulation mechanism. J. Theor. Biol., 252(4):633- 648, Feb. 2008. [21] D. Sitki-Green, M. Covington, and N. Raab-Traubb. Compartmentalization and transmission of multiple Epstein-Barr virus strains in asymptomatic carriers. J. Viral. , 77(3):1840-1847, Feb. 2003. [22] J.W. Sixbey and Q.Y. Yao. Immunoglobulin A-induced shift of Epstein-Barr virus tissue tropism. Science, 255(5051):1578-1580, 1992. [23] D.A. Thorley-Lawson. EBV persistence and latent infection in vivo. In Robertson ES, editor, Epstein-Barr Virus, volume 1, pages 309-349. Caister Academic Press, Norfolk, England, 1st edition, 2005. 25 cc cc University of Utah Institutional Repository Author Manuscript [24] S.M. Turk, R. Jiang, L.S. Chesnokova, and L.M. Hutt-Fletcher. Antibodies to gp350/220 enhance the ability of Epstein-Barr virus to infect epithelial cells. J. Viral., 80(19):9628- 9633, Oct. 2006. [25] P. van den Driessche and J. Watmough. Reproduction numbers and sub-threshold endemic equilibria for compartmental models of disease transmission. Math. Biasci., 180(1-2):29- 48, 2002. [26] G. Wang, G.R.F. Krueger, and L.M. Buje. Mathematical model to simulate the cellular dynamics of infection with human herpesvirus-6 in EBV-negative infectious mononucleosis. J. Med. Viral., 71(4):569- 577, 2003. [27] D. Wodarz, S. Sierro, and P. Klenerman. Dynamics of killer T cell inflation in viral infections. J.R.S. Interface, 4(14):533- 543, 2007. [28] K. Yamanishi, Y. Mori, and P.E. Pellett. Human herpesviruses 6 and 7. In D.M. Knipe and Howley P.M, editors, Field's Viral, volume 2, pages 2701-2772. Lippincott Williams and Wilkins, Philadelphia, PA, 5th edition, 2007. 26 cc cc University of Utah Institutional Repository Author Manuscript Table 1: Parameters used in simulations of the within-host model (Equation 1). The rates are given in a unit of per minute. We use many parameters from PathSim where the rates are estimated and given in a unit of per 6 minutes [20]. We convert them into the unit of per minute. Parameter Description Value Reference d1 Turnover rate of naive B cells 1/6000 [20] /-lEb B cell infection rate per epithelial cell virus 3.3 x 10- 10 [20] a /-lBb B cell infection rate per B cell virus /-lEb/ 1OO [11] P Proliferation factor 2 [20] d2 Death rate of latently infected B cell 1/11520 [20] c Rate of latently infected cell 0.001 [20] b going into memory stage k2 Rate of latently infected B cells 3.8 x 10- 8 [20] c killed by activated T cell r Rate of reactivation of 8.3 x 10- 5 [20] lytic infection from latent infection s Regulation factor of memory B cells 2 [15] d4 Death rate of lytically infected cell 1/4320 [20] due to viruses bursting out k4 Rate of lytically infected B cell 7.6 x 10- 8 [20]C killed by activated T cell de Turn-over rate of epithelial cell 1/6000 d /-lBe Epithelial cell infection rate per B cell virus 3 x 10- 11 e /-lEe Epithelial cell infection rate per epithelial cell virus /-lBe/5 [11] 'Y Death rate of infected epithelial cell 1/6000 due to viruses bursting out f n Viral burst size 1000 [20] dv Death rate of virus 1/2160 [20] ¢2 Rate of CTL activation against latent infection 1.95 x 10- 5 [20] 9 ¢4 Rate of CTL activation against lytic infection 4.48 X 10- 5 [20]g ()2 Rate of effector CTL proliferation 3.25 x 10- 5 [20] h against latent infection ()4 Rate of effector CTL proliferation 3.25 x 10- 5 [20]h against lytic infection K Number of infected cells when 105 [14] T cell activation is half maximal b Death rate of T cells 1/156000 [20] aProbability of virus and cell encounter per minute multiplied by the probability of infection and divided by the number of viruses (~ 107 ). bWe take this to be the same rate as the estimation of .1% of lymphocytes leaving the Waldeyer's ring per minute. cProbability of lymphocytes encounter per minute multiplied by the probability that Ti kills its target and divided by the number of Ti(~ 104 ). dEstimated, taken to be the same as d1 . eEstimated, taken to be less than J1 E b [24]. iEstimated, taken to be less than d4 [3]. 9Probability of lymphocyte encounter per minute multiplied by the probability of Ti activation by Bi, where i = 2 or 4. hProbability of lymphocyte encounter given per minute multiplied by the frequency of cell division (every 8-12 hours). 27