Mathematical Modeling of Interleukin-27 Induction of Anti-Tumor T Cells Response

Interleukin-12 is a pro-inflammatory cytokine which promotes Th1 and cytotoxic T lymphocyte activities, such as Interferon- secretion. For this reason Interleukin-12 could be a powerful therapeutic agent for cancer treatment. However, Interleukin-12 is also excessively toxic. Interleukin-27 is an immunoregulatory cytokine from the Interleukin-12 family, but it is not as toxic as Interleukin-12. In recent years, Interleukin-27 has been considered as a potential anti-tumor agent. Recent experiments in vitro and in vivo have shown that cancer cells transfected with IL-27 activate CD8+ T cells to promote the secretion of anti-tumor cytokines Interleukin-10, although, at the same time, IL-27 inhibits the secretion of Interferon- by CD8+ T cells. In the present paper we develop a mathematical model based on these experimental results. The model involves a dynamic network which includes tumor cells, CD8+ T cells and cytokines Interleukin-27, Interleukin-10 and Interferon-. Simulations of the model show how Interleukin-27 promotes CD8+ T cells to secrete Interleukin-10 to inhibit tumor growth. On the other hand Interleukin-27 inhibits the secretion of Interferon- by CD8+ T cells which somewhat diminishes the inhibition of tumor growth. Our numerical results are in qualitative agreement with experimental data. We use the model to design protocols of IL-27 injections for the treatment of cancer and find that, for some special types of cancer, with a fixed total amount of drug, within a certain range, continuous injection has better efficacy than intermittent injections in reducing the tumor load while the treatment is ongoing, although the decrease in tumor load is only temporary.


Introduction
Interleukin-12 (IL-12) is a pro-inflammatory cytokine that plays a central role in the connection of the innate resistance and adaptive immunity, by promoting Th1 and cytotoxic T lymphocyte (CTL) activities, such as IFN-c secretion.IL-12 could be a powerful therapeutic agent to eradicate tumor or to prevent the development of metastasis [1][2][3][4].However, IL-12 has also been shown to be excessively toxic [5,6], although there is at least one ongoing clinical trial with IL-12 using a new delivery method (IL-12 DNA plasmid) that is intended to overcome toxicity problems.In recent years there has been increasing interest to investigate the role of another member of the IL-12 family, namely, Interleukin-27 (IL-27), which is less toxic than IL-12, as a potential anti-tumor agent [7].IL-27 is a cytokine capable of regulating Th1, Th2, Th17, and T reg responses [8].In autoimmune diseases, Murugaiyan et al. [9] have shown that IL-27 promotes production of IL-10 and IFN-c by naive human CD4 + T cells, and Stumhofer et al. [10] demonstrated that IL-27 promotes production of IL-10 by CD4 + and CD8 + T cells.Reviewing the role of IL-27 in anticancer immunotherapy, Swarbrick et al. [11] asserted that IL-27 may have both pro-inflammatory and anti-inflammatory functions, and it promotes Th1 immune response and CD8 + cell activation.Since Hisada et al. [7] first reported the anti-tumor efficacy of IL-27 in 2004, the potent anti-tumor activity of IL-27 has been verified in various tumor models [11][12][13].Many studies suggest a role of IL-27 in enhancing anti-tumor CD8 + T cell responses [7,[14][15][16][17].The enhancing role of IL-27 in generating anti-tumor CTL response was also demonstrated using IL-27R deficient mice [18,19].
IL-10 has inhibitory and stimulatory effects on human CD8 + T cells [20], and in viral infection it is known to inhibit effector and memory CD4 + T cell responses but not memory CD8 + T cells [21].IL-10 may have positive or negative effect on tumor suppression (Asadullah et al. [22]).Numerous studies (e.g.[23,24]) show that increase in IL-10 produced by macrophages is associated with tumor progression, while other studies [25][26][27][28] suggest that IL-10 plays a positive role in tumor rejection.IL-27 can induce production of IL-10 in CD8 + T cells [10,29].In a recent study, Liu et al. [30] used P1CTL TCR transgenic mouse model and mouse plasmacytoma tumor system to investigate how IL-27 enhances the anti-tumor responses.They found that IL-27 significantly enhances the survival of activated tumor antigen specific CD8 + T cells in vitro and in vivo, and induces IL-10 upregulation in these T cells.It was also suggested in [30], and demonstrated in [25][26][27][28], that CTL IL-10 production contributes to tumor rejection.These results have important implications for designing IL-27-based immunotherapy against human cancer.
In the present paper, we develop a mathematical model that describes the anti-tumor activity of CD8 + T cells in terms of IFN-c and IL-10 productions, when these T cells are activated by IL-27 from the tumor microenvironment.The model is based on the experiments by Liu et al. [30] (with mice infected with plasmacytoma) whereby cancer cells are transfected with an IL-27 vector so that IL-27 is released in the tumor microenvironment.We show that the model simulations agree qualitatively with the experimental results of [30].We next extend the model to include therapeutic treatment of cancer in wild type mice by injection of IL-27.We note however, that in this case, the tumor microenvironment includes both CD4 + and CD8 + T cells (whereas in the experiments with transgenic mice of [30] the CD8 + T cells were taken to be dominant).As mentioned above, IL-27 promotes the secretion of IL-10 and IFN-c by CD4 + T cells [9][10][11], and we assume that these cytokines have the same effect on tumor rejection as those secreted by CD8 + T cells.We then use CD8 + T cells to represent both cells, CD4 + and CD8 + .In the modified model the only source of IL-27 comes from the drug, since cancer cells do not generally secrete IL-27.We compare the efficacy of different strategies of IL-27 injections.For example, we found that continuous injection of IL-27 for 24 weeks at a fixed amount F , within a certain range, is more effective than intermittent injection of the amount 2F, full three weeks at a time with three weeks spacing between injections, for 24 weeks.These predictions however must be viewed just as suggestions since they may only apply to special types of cancer, such as plasmacytoma in bone or soft tissue, and since, furthermore, the model does not include other important factors in tumor progression such as angiogenesis and the immune response.

Mathematical model
In this model, we assume that the tumor is spherical and that it initially lies in a spherical tissue of radius L. The variables that will be used in the model are listed below and we assume that all the variables are radially symmetric: I 27 (r,t) : concentration of Interleukin{27 in pg=cm 3 , I 10 (r,t) : concentration of Interleukin{10 in pg=cm 3 , T(r,t) : (tumor antigen specific) activated CD8 z T cell density, cell=cm 3 , where r is the distance from a point x~(x 1 ,x 2 ,x 3 ) to the origin: q and 0ƒrƒL.These variables satisfy a system of partial differential equations based on the network exhibited in Figure 1.
The first term represents the diffusion of I 27 with coefficient D I27 .Although diffusivities of cytokines and cells may depend on the concentrations of the tumor cells and normal healthy cells, for simplicity, here and in the sequel all diffusivities are assumed to be spatially uniform.In the experiment in [30] 1) are given in Table 1.
IL-10.The Interleukin-10 (IL-10) in Figure 1 is pro-inflammatory, in accordance with the experiments of [30] The first term is the diffusion of I 10 .The second term accounts for the production of I 10 by CD8 + T cells for the absence of I 27 [30].The experiments in [30] indicate that I 27 significantly increases the production of I 10 by CD8 + T cells, and this is accounted by the third term.The last term is the degradation of I 10 .The parameter values of Equation ( 2) are listed in Table 2.
CD8 + T cells.The equation for the density of (activated) CD8 + T cells, T(x,t), is given by The first term is a dispersion of CD8 + T cells with coefficient D T .The second term accounts for activation of CD8 + T cells by P1A antigen from the tumor cells.I 27 promotes survival of CD8 + T cells, and so does also I 10 , but to a smaller degree [30].We present these two facts by correspondingly decreasing the death rate m T of T cells in the last term of Equation (3).The parameter values in Equation (3) are given in Table 3.Although I 27 contributes more than I 10 to promote the half-life of CD8 + T cells, we take b T ~9 since the concentration of I 10 is much smaller than the concentration of I 27 .
IFN-c.Interferon-c (IFN-c) is a cytokine with diffusion coefficient D c and degradation rate m c .It is produced by T cells and, as shown in [30], the production is inhibited by I 27 .Thus, I c (x,t) satisfies the equation: The second and third terms represent the proliferation and death of cells, respectively.Generally, I 10 is regarded as an antiinflammatory cytokine.However, in different experimental models, I 10 could suppress or promote the functions of immune system [30,39].Liu et al. [30] found that the I 10 produced by CTL contributes to tumor rejection.Hence, the fourth term accounts for the indirect inhibition of tumor cells by I 10 .Cytokine, I c promotes the anti-tumor response, such as increase production of IL-12, and induces natural killer cells to kill cancer cells [40,41].For simplicity, we take the fifth term in (5) to represent the  5.
The dimensional and dimensionless values of all the parameters of Tables 1-5 are listed in Table 6.
Initial conditions.We assume that tumor cells are initially concentrated near r~0, taking c(r,0)~c 0 (e {r=E {e {L=E ), where L~1 cm or L~1:5 cm, and E is a positive number less than or equal to 1.In the simulations, we shall take E~1 but the results do not change qualitatively with smaller values of E. Since I 27 is produced by J558-IL-27 tumor cells, the initial concentration of I 27 should be similar to the density of tumor cells; we take Initially, there are no activated CD8 + T cells, hence T(r,0)~0: Since I 10 and I c are produced by CD8 + T cells, we take I 10 (r,0)~0 and I c (r,0)~0: Boundary conditions.Since all variables are radially symmetric, the first r-derivative at r~0 is equal to zero.We assume no-flux condition for all variables at r~L.This is justified by the fact that L is large enough so that the exterior of the ball of radius L lies completely within the healthy tissue, initially.
Parameters nondimensionalization.We nondimensionalizate the Equations ( 1) -( 5): For nondimensional variables and parameters, we consider the tumor growth in a ball fr rƒ2g or fr rƒ3g.The nondimensional PDE model is given by the following system of equations:
(iv) J558-IL-27 tumor cells with large production rate of IL-27.It has been reported in [30] that I 27 can enhance the population of CD8 + T cells.Moreover, I 27 also enhances I 10 produced by CD8 + T cells to inhibit the tumor growth, but at the same time it suppresses the pro-inflammatory cytokine I c secreted by CD8 + T cells.In spite of its inhibition of I c , I 27 still promotes CD8 + T cells to suppress the tumor growth.
In view of these experimental results we expect the total mass of I 10 to increase from cases (i) to (iv), the total CD8 + T cell population to increase from cases (i) to (iv), and the total population of cancer cells to decrease from cases (i) to (iv), as time progresses.
Correspondingly, we associate with the four cases (i) -(iv) increasing values of the parameter a 27 :  In Figures 2-5, we took L~1cm such that tumor cells are not visible near the boundary r~1cm for all time t[½0,90days.Figure 2 shows the time-dependent profiles of the total mass of I 27 ,I 10 ,I c , and total populations of CD8 + T cells and cancer cells.We see that growth/decrease of these variables, as â a 27 varies, corresponds qualitatively to the experiments in [30].Figures 3-5 show significant spatial variations of these variables at days 3,9, and 15, with or without production of I 27 .We also see the effect of I 27 on cancer cells and CD8 + T cells densities at different distances from the point of origin of the cancer.For example, at the origin, at day 3 the cancer cells density changed from 4:95|10 8 cell=cm 3 with no treatment by I 27 to 4:65|10 8 cell=cm 3 with largest production of I 27 , while at day 15 it changed from 4:2|10 8 cell=cm 3 with no treatment to 1:7|10 8 cell=cm 3 with largest production of I 27 .Similarly, at the origin, the CD8 + T cell density increased at day 3 from 2:3|10 6 cell=cm 3 without treatment to 3:3|10 6 cell=cm 3 with largest production of I 27 , whereas at day 15 the density increased even more significantly from 3:5|10 6 cell=cm 3 with no I 27 treatment to 9|10 6 cell=cm 3 with maximal production of I 27 .Note that Figure 5E and 5F show that the tumor cell density is almost zero near the boundary r~1cm and the tumor cells concentrate in the region r~0:5cm.
Tumor initiating in internal organs can also be treated by I 27 , but the mechanism for introducing I 27 will depend on the location of the tumor.For example, in colitis induced colon cancer, one could use yeast which were programmed to express I 27 [42].Oncolytic virus which are engineered to produce I 27 within tumor cells could turn the tumor into immunogenetic, thus enabling the immune system to reject the tumor.
We want to use our model in order to design treatments for a wild type mouse by I 27 injection.We recall, as noted in the Introduction, that for wild type mouse, both CD4 + and CD8 + T cells produce IL-10 and IFN-c [9][10][11] and we assume that IL-10 secreted by CD4 + T cells has the same tumor rejection quality as the IL-10 secreted by CD8 + T cells.We then use CD8 + T cells to represent both cells, CD4 + and CD8 + .We also note that in vivo tumor cells do not generally secrete I 27 , so we take â a 27 ~0 in Equation (1).But we also need to include an injection term in Equation ( 1 We make the pharmacokinetic assumption that f (r,t) decreases in r from the outer boundary of the tumor (r~1:5cm) towards the inner core (r~0), and take where a is some positive constant; F is viewed as the ''amount'' of injection.
We consider here, for illustration, two strategies of treatment: (i) continuous injection of I 27 at a fixed amount F for 24 weeks, and (ii) intermittent injections, at double amount 2F , full three weeks at a time with three weeks spacing between injections.Accordingly, for the continuous strategy for 0ƒrƒL and 0ƒtƒ24 weeks (in real time) , 0, for 0ƒrƒL and 24vtƒ30 weeks (in real time) , and for the intermittent strategy for 0ƒrƒL and t 2i ƒtvt 2iz1 , i ~0,1,2,3 0, for 0ƒrƒL and t 2iz1 ƒtvt 2(iz1) , i~0,1,2,3 and 24vtƒ30 weeks in case (ii), where the length of each interval ½t j ,t jz1 is three weeks (the drug is injected only during the intermittent intervals ½t 2i ,t 2iz1 ) and t 0 ~0.In the following simulations, we take a~2:25cm 2 ; however the same results remain qualitatively the same for other values of a (not shown here).
In Figures 6-9, we take L~1:5cm so that the tumor cell density remains negligible near the boundary r~L, during the entire simulation time which is 30 weeks and hence the boundary conditions are not affecting the results during the entire simulation (For longer simulation time, e.g. 60 weeks, we need to take L~3cm (not shown here.)).We also take the simulation mesh size dr~0:01cm and dt~0:01day.In Figure 6, we compare the results of the two strategies in case F ~3|10 4 pg=cm 3 =day.We see that continuous injection yields better results in reducing the tumor level and slightly delaying relapse after the drug is withdrawn.and c, respectively, for model (6) with L~1:5cm which the first equation for I 27 is replaced by (7) and all parameter values are taken from Table 6.In (E), the upper curve is for J558-Ctrl tumor cells, the dotted-dashed curve ({ : { : {) is for intermittent injection, and the dashed curve is for continuous injection with F ~3|10 4 pg=cm 3 =day and a~2:25cm 2 , for the first 24 weeks.doi:10.1371/journal.pone.0091844.g0066) with L~1:5cm and a~2:25cm 2 which the first equation for I 27 is replaced by (7) and all parameter values are taken from Table 6.The upper curve is for J558-Ctrl tumor cells, the dotted-dashed curve ({ : { : {) is for intermittent injection, and the dashed curve is for continuous injection.doi:10.1371/journal.pone.0091844.g008Figure 9. Tumor growth and migration for shorter injection schedule.(A) is the concentration profile of c at 4 weeks and (B) is the profile of total number of c for t[½0,4weeks, under F ~3|10 4 pg=cm 3 =day, for model (6) with L~1:5cm and a~2:25cm 2 which the first equation for I 27 is replaced by (7) and all parameter values are taken from Table 6.The upper curve is for J558-Ctrl tumor cells, the dotted-dashed curve ({ : { : {) is for intermittent injection, and the dashed curve is for continuous injection.(A) shows that the concentration of tumor cells are not visible near the boundary r~1:5cm, for all t[½0,4weeks.doi:10.1371/journal.pone.0091844.g009 Figure 7, for the same experiment as in Figure 6, shows the concentration profiles of tumor cells at times 4 weeks, 24 weeks, and 30 weeks for J558-Ctrl, intermittent injection, and continuous injection cases.Notice from Figures 7A, 7B, and 7C that the tumor has progressed during the periods of 4 weeks, 24 weeks, and 30 weeks to r~0:6cm, r~1:05cm, and r~1:2cm, respectively.Figures 6 and 7 show that I 27 injection slows down tumor growth during drug injection, but it does not change the migration speed of tumor cells.Figure 8 compares the results of the above two strategies for smaller values of F , namely, F~1|10 4 pg=cm 3 =day and F ~0:5|10 4 pg=cm 3 =day.We see that continuous injection is still more effective, but, for smaller amount of injection, the relative advantage of continuous injection is decreased.Simulations of these two strategies for other values of F in the range of 10 3 {10 6 pg=cm 3 =day (not shown here) give the same results, namely, that continuous injection is preferable to intermittent injections.In order to make a definite recommendation on continuous versus intermittent injection one would need to consider also possible side-effects that may arise from these two strategies.
Although the expected lifespan of the mouse in the experiments of Liu et al. [30] was one month, for the purpose of therapy we performed simulations for the longer period of 30 weeks.But it is also interesting to consider the case of treatment for one month only.This is done in Figure 9 where we have taken in Equation ( 9) 0ƒtƒ4 weeks for the continuous treatment, and, in Equation ( 10), intermittent time t j ~j weeks.Figure 9 shows that continuous treatment is again preferable to intermittent treatment.Figure 9A shows the concentration profile of tumor cells at 4 weeks; note that tumor cells do not reach the boundary within 4 weeks.Figure 9B displays the profile of total number of tumor cells.We see that the continuous treatment still has better efficacy than intermittent treatment.

Sensitivity analysis
In order to provide support to the robustness of the simulation results, we ran sensitivity analysis on parameters which appear in Equations ( 1) - (5).The parameters chosen for the sensitivity analysis are either those whose baseline was crudely estimated, or those that seem to play more important role in the model predictions.
We list these parameters with their ranges, baselines, and units, in Table 7.In this analysis, â a 27 varies from 32:4 to 12960.Following the sensitivity analysis method described in [43], we performed Latin hypercube sampling and generated 5000 samples to calculate the partial rank correlation coefficients (PRCC) and pvalue, with respect to the ratio Ð 2 0 ĉ c(r r, t t)r r 2 dr r= Ð 2 0 ĉ c { (r r, t t)r r 2 dr r, where ĉ c(r r, t t) (resp.ĉ c { (r r, t t)) accounts for the J558-IL27 (resp.J558-Ctrl) tumor cell density, at t t~30 and r r[½0,2.The PRCC and their pvalues are listed in Table 8.A negative PRCC (i.e.negative correlation) means increase in the parameter value will decrease the ratio Ð 2 0 ĉ c(r r, t t)r r 2 dr r= Ð 2 0 ĉ c { (r r, t t)r r 2 dr r; that is, it will increase the rejection of tumor treated by IL-27 versus untreated tumor.Conversely, positive PRCC means that increased rejection of the tumor (treated by IL-27 versus untreated) will occur if this parameter is decreased.
The sensitivity analysis data are shown in Figures S1-S4 in Supplementary Material and summarized in Table 8.The most significant negatively correlated parameters in promoting rejection of tumor treated by IL-27 versus untreated tumor are a 27 ,s T ,s T ; less significant parameters are g c ,g c ,a 10 ,b T .The effect of a 27 has already been displayed in Figures 2-5.The negative correlations of s T , g c , and g c are not surprising, since s T is the rate by which tumor activates T cells (while T cells are increased with IL-27 treatment; see Figures 2-5) and g c and g c are, respectively, the killing rates of tumor cells by I 10 and I c (while I 10 and I c increase with IL-27 treatment; see Figures 2-5).The negative correlations of a 10 ,s T , and b T are also not surprising, since I 27 promotes the production of I 10 to inhibit tumor cells, and larger s T and b T promote survival of CD8 + T cells.
The most significant parameters in promoting tumor are s c ,c T ,s c and, to a smaller degree, s 10 .This also is not surprising, since increasing s c results in decreased inhibition of tumor cells by I 10 , increasing c T results in decreased number of CD8 z T cells, increasing s c results in decreased inhibition of tumor cells by I c , and increasing s 10 results in decreased I 10 .We note that the parameters s 10 and s c , in Table 8, have small PRCC with p-values that are larger than 0:01; this means that they are not sensitive to the ratio Ð 2 0 ĉ c(r r, t t)r r 2 dr r= Ð 2 0 ĉ c { (r r, t t)r r 2 dr r.

Discussion
IL-12 plays a central role in linking the innate resistance and adaptive immunity, and could be a powerful anti-tumor agent.However, since IL-12 is excessively toxic, the cytokine IL-27, which is a less toxic member of the IL-12 family, has been considered as a possible replacement of IL-12 as anti-tumor agent [7,8,[14][15][16][17].It was demonstrated by Liu et al. [30] that IL-27 enhances the survival of tumor antigen specific CD8 + T cells and induces their upregulation of IL-10, which acts as an anti-tumor cytokine.This suggests that IL-27 could play an important role in immunotherapy against human cancer.
The aim of the present paper was to develop a mathematical model that can be used to explore and predict the efficacy of different protocols of IL-27 treatment.To do that we first set up a dynamical system of partial differential equations whereby IL-27 is produced by transfected J558-IL-27 tumor cells, as demonstrated in the experiments of Liu et al. [30].The model included IL-27induced CD8 + T cells and cytokines IL-10 and IFN-c.By carefully estimating the parameters of the equations we showed that the model simulations agree with the experimental data of Liu et al. [30].
The model can be used to examine the effect of injecting IL-27 into the microenvironment of cancer in a mouse, and design strategies for such injections.We illustrated this by comparing the efficacy of two protocols: (i) continuous injection (e.g., daily) of IL-27 for 24 weeks at a fixed amount F , and (ii) intermittent injections during the first 24 weeks with three weeks injection at a fixed amount 2F followed by three weeks spacing, and withdrawing the drug after the 24 weeks for both protocols (i) and (ii).We found that the continuous injection has better efficacy in reducing the tumor load, and also in delaying relapse after the drug is withdrawn, while the treatment is ongoing.However, in establishing these results we made the assumption that IL-10 produced by IL-27 activated CD4 + T cells has the same pro-inflammatory property as the IL-10 produced by CD8 + T cells.In addition, we made the pharmacokinetic assumption that the drug density decreases toward the inner core of the tumor, and we also took the drug ''amount'' F in the range of 10 3 {10 6 pg=cm 3 =day.
We note that our model was based on the experiments by Liu et al. [30] with plasmacytoma, but not with other tumor cells.Furthermore, the model did not include the effects of lymphoid and vascular compartments, as these were not reported in [30].Hence the present paper should be considered only as an initial building block for a more comprehensive model which should include angiogenesis as well as the immune response of macrophages, dendritic cells, and T cells (Th1, Th2, Th17, and T reg s).We note in particular that pro-inflammatory macrophages (M 1 ) secrete a family of IL-12 cytokines including IL-27 [44], and the IL-12 family attracts CTLs which kill tumor cells, so that M 1 macrophages suppress tumor growth.On the other hand, antiinflammatory macrophages (M 2 ) secrete IL-10 which promotes tumor growth [23,24].Regulatory T cells promote tumor growth and are inhibited by IL-27 [45,46].Thus the present paper's prediction of the efficacy of different protocols of treatment of plasmacytoma in bone or soft tissue with IL-27 will need to be reexamined when more data become available that will enable us to include the important compartments of the immune and vascular systems.
We note also that the proposed intervention with IL-27 in our paper shows benefits only while the treatment is ongoing.The treatment has neither significant short-term benefits nor any longterm benefits after the drug has discontinued.It is becoming increasingly common to treat tumors with several drugs.In addition to tumor specific drugs, a generic mitotic inhibitory drug, which disrupts microtubules that pull the cell apart, is often usedsince cancer cells are more sensitive to inhibition of mitosis than normal healthy cells.In our model, the effect of such a drug is to increase the death rate parameter in the equation for cancer cells.Further work should also include the combined effect of treatment of IL-27 with a mitotic inhibitory drugs.

Estimates of the densities of tumor cells and T cells
Many of the parameters are based on experiments reported in [30].In [30], the volume of the tumor was measured in days 12,14,16, and 18, but the number of CD8 + T cells and concentrations of I 10 and I c were measured only in the first 5 days.
From Figure 5D in [30], the volume of the tumor at days 12,14,16, and 18 were approximate 30,60,100, and 180 in mm 3 .Hence, where A is the number of tumor cells in per mm 3 .If we consider a simplified equation for ( 5) then, for any two times t 1 and t 2 , c(t 2 )~c(t 1 )e (l 1 {m c )(t 2 {t 1 ) : If we apply this formula to the 6 pairs of the numbers from (11) to compute l 1 {m c and take the mean value, we get l 1 {m c ~0:295=day: Since the half-life of melanoma tumor cells is approximate 4 days, [47], m c ~ln2=4=day&0:173=day, and then l 1 ~0:468=day.
In the experiments in [30], there were two kinds of tumor cells: J558-IL-27 which generates IL-27, and J558-Ctrl which does not generate IL-27.The antigen P1A on J558 tumor cells is recognized by receptors TCRs on cytotoxic T cells, P1CTL.Liu et al. [30] used P1CTL with glycoprotein CD8 (which is called CD8 + T cells) to investigate the immune response for IL-27.Their P1CTL cells were of four different types: (i) P1CTL which can recognize J558-Ctrl tumor cells and generate IL-10 to inhibit tumor growth; (ii) IL-10 -/-P1CTL which can recognize J558-Ctrl tumor cells but cannot generate IL-10; (iii) P1CTL/IL-27 which can recognize J558-IL-27 tumor cells and generate IL-10; and (iv) IL-10 -/-P1CTL/IL-27 which can recognize J558-IL-27 tumor cells but cannot generate IL-10.
The number of tumor cells at t~0 day (in [30]) was 5|10 6 cells, and we assume (see Figure 5D in [30]) that they occupy volume 10mm 3 .Hence There is no data in [30] on the density of the tumor in day 5.We assume that this density is larger than c(0) but substantively smaller than the maximal capacity 10 9 cell=cm 3 .We take for J558-Ctrl with P1CTL or J558-IL-27 with P1CTL/IL-27, but for J558-Ctrl with IL-10 -/-P1CTL or J558-IL-27 with IL-10 -/-P1CTL/IL-27, since the last two types of T cells do not generate I 10 .
Next, we choose s c ~150pg=cm 3 and proceed to estimate g c by considering T cells P1CTL and P1CTL/IL-27 which generate I 10 .For (25), if I 10 and I c are constants, then the solution is c(t)~e The concentration of I c with I 10 is smaller than the concentration of I c where I 10 is blocked [30].We take Ic Iczsc & 1 3 .In [30], the concentration of I 10 vary from 1pg=cm 3 to 2500pg=cm 3 .We take s c ~150pg=cm 3 , so that I10zsc I10 &3:5; hence g c ~0:345=day.

Supporting Information
, Liu et al. used gene transfected tumor cells, J558-IL-27, to produce I 27 in the tumor microenvironment.Accordingly, we use the second term to describe the production of I 27 by the transfected J558-IL-27 tumor cells.The last term stands for the degradation of I 27 .The parameter values of Equation ( 27 zb T s T I 10 T |fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl} proÀsurvival by I 27 and I 10 :

Figure 7 .
Figure 7. Concentration profiles of tumor cells at different times.(A), (B), and (C) are the concentration profiles of c at times 4 weeks (short time), 24 weeks (time at which injections are withdrew), and 30 weeks (the final time for simulation), respectively, under drug amount F ~3|10 4 pg=cm 3 =day and a~2:25cm 2 .The upper curve is for J558-Ctrl tumor cells, the dotted-dashed curve ({ : { : {) is for intermittent injection, and the dashed curve is for continuous injection.The concentration of tumor cells are not visible, when r is close to 1:5cm, for all t[½0,30weeks; L~1:5cm.doi:10.1371/journal.pone.0091844.g007

Figure 8 .
Figure 8.Comparison of continuous versus intermittent treatment for different drug amount.(A) and (B) are the profiles of total number of c with F ~1|10 4 pg=cm 3 =day and F ~0:5|10 4 pg=cm 3 =day, respectively, for model (6) with L~1:5cm and a~2:25cm 2 which the first equation for I 27 is replaced by(7) and all parameter values are taken from Table6.The upper curve is for J558-Ctrl tumor cells, the dotted-dashed curve ({ : { : {) is for intermittent injection, and the dashed curve is for continuous injection.doi:10.1371/journal.pone.0091844.g008
5): dc(t) dt ~l1 c(t){m c c(t){g c I 10 I 10 zs c c(t){g c I c I c zs c c(t): ): dc(t) dt ~l1 c(t){m c c(t){g c I c I c zs c c(t): If I c is constant, then the solution is c(t)~e (l1{m c {g c Ic Ic zsc )t c(0) which leads to g c ~½(l 1 {m c )

(l1{m c {g c I 10 I
10 zsc {g c Ic Ic zsc )t c(0), and hence g c ~½l 1 {m c {g c I c I c zs c

Figure
Figure S1 Sensitivity analysis.Sensitivity analysis on a 27 ,s 10 ,a 10 , and s T .(PDF) Figure S2 Sensitivity analysis.Sensitivity analysis on s 10 ,g c ,g c , and s c .(PDF) Figure S3 Sensitivity analysis.Sensitivity analysis on s c ,s c ,s T , and b T .(PDF) Figure S4 Sensitivity analysis.Sensitivity analysis on c T .(PDF) The following equation describes the evolution of I 27 : degradation :

Table 2 .
Parameters for the IL-10 equation.

Table 5 .
Parameters for tumor cell equation.

Table 7 .
Parameters chosen for sensitivity analysis.
cm 2 =day, [48][48], we have m 10 ~16=day.To estimate s 10 , we consider the case of J558-Ctrl tumor cells, for which the term with I 27 is removed from (19):