dournal o[ J. Math. Biology (1981) 12:363-373 Mathematical Biology by Springer-Verlag1981 A Model of the Role of Natural Killer Cells in Immune Surveillance - I* Stephen J. Merrill Division of Applied Mathematics, Lefschetz Center for Dynamical Systems, Brown University, Providence, RI 02912, and Department of Mathematics, Statistics and Computer Science, Marquette University, Milwaukee, WI 53233, USA Summary. The theory of immune surveillance of Thomas and Burnet stated in part that antigenic differences between neoplastic and normal cells provide the stimulus for their destruction by cells of the immune system. Burnet pointed to the T lymphocyte as the cell which mediated this surveillance. The existence of some form of surveillance in cases of no Tlymphocyte functioning presents the possibility that surveillance, if present at all, is mediated by non T cells. Cells identified as naturally cytotoxic killer (NK) cells appear to have properties required of a surveillance effector population. This paper utilizes properties of N K cells and the effects of interferon on this population to construct a mathematical model of the characteristics that an NK cell surveillance would have. A two level theory of immune surveillance is proposed. Key words: N K cells - Immune surveillance O. Introduction The theory of immune surveillance of cancer suggested by Thomas 1"35]and Burnet 12] and developed by Burnet 1,3], 14] proposed that changes which take place on the surface of a neoplastic cell are utilized by cells of the immune response to eliminate those neoplastic colonies in much the same way as transplants are rejected. Although Tlymphocytes (T-cells), which are primarily responsible for the rejection of most transplants, have been shown to be cytotoxic to tumors 1'131 resistance to tumors whose surface antigens (TAA) are weakly antigenic (e.g., spontaneous tumors 1191) would not readily be recognized and eliminated by T-cells. In nude mice, where T-cells are naturally absent due to lack of a thymus, spontaneous tumors are exceedingly rare 1"30] and occur no more frequently than spontaneous tumors in mice with functioning T-cells. Taken together, this suggests that at least for spontaneous tumors and others whose TAA are weak antigens, other mechanisms are responsible for antitumor defense. * This research has been supported in part by the National Science Foundation under grant # NSFEng. 7904852 0303 -- 6812/81/0012/0363/$02.20 364 S.J. Merrill The discovery of the existence of naturally cytotoxic cells [14], [26] pointed to the possibility of a new mechanism for resistance against tumors. Natural cytotoxicity is defined as cytotoxicity by a population of cells from peripheral blood and lymphoid organs against various tumor ceils which is independent of any antigenic priming. The recognition of strong natural cytotoxicity in nude mice and the susceptibility of most spontaneous tumors to N K lysis gives a reasonable explanation for in vivo data concerning tumor development and growth in these immunologicaUy compromised individuals. In this paper, a model of surveillance as it would be mediated by N K ceils is proposed and analyzed. As a result, a modification of the theory of immune surveillance is suggested which recognizes the (nearly) independent contributions to tumor resistance by both T and N K populations. Other mathematical models which describe interactions of tumor cells with cells of the immune system include Lefever and Garay [21], [22], Garay and Lefever [10], Grossman and Berke [11], Albert, Freedman and Perelson [1], DeLisi and Rescigno [7], Reseigno and DeLisi [28] and Merrill [24]. Several of these efforts were directed at the description of surveillance and the failure of surveillance ("sneaking through"). Mathematics used ~n cancer research has been surveyed by Swan [33], Eisen [9] and Thames [34]. 1. Properties of N K CellsNatural killer (N/Q cells are described as a collection of cells primarily found in the spleen and peripheral blood with the ability to lyse some tumor cells without antigen priming. The kinetic difference between NKcells and T~cells in their action against in vitro tumor cells illustrates the differences. If N K cells are incubated with susceptible tumor cells, lysis of tumor cells will begin almost immediately and continue at a fairly constant rate. T-cells incubated with tumor will however not immediately lyse tumor. Lysis will begin only following a delay (the priming) after which the rate of lysis will rise quickly, reach a peak and then decline. The parallels between this and the mechanism involved in the rejection of transplants gives one of the arguments for the existence of some kind of T-cell mediated surveillance in vivo. Because N K cells are functionally defined, their description morphologically has been difficult. They are now known to be mononuclear cells originating in the bone marrow [29]~ Surface markers which would indicate the hemopoietic origin as lymphoid are lacking or deficient, although one theory of their origin is as prethymic T-cells [15]. Other evidence suggests that N K cells may be promonocytes [23]~ Recent evidence from the mouse N K cell [20] suggests that N K cells are a heterogeneous class of cells which need not have a single origin. The identification of "large granular lymphocytes" (LGL) as the primary human N K cell by Timonen and Saksela and the description of their response to contact with tumor cells [36] gives the biological basis for the construction of this model. Timonen et al. [36] found that LGL from human peripheral blood produced interferon when in contact with tumor cells. One effect of this interferon is to augment the cytotoxic activity of inactive "pre-NK" cells. The amplification of the The Role of Natural Killer Cells in Immune Surveillance 365 cytotoxic response and the reduced tumor growth rate due to this endogenous interferon [8] is assumed in this model to be the main mechanism of N K mediated tumor surveillance. Interferon may also reduce the NK killing of tumor cells, especially evident at concentrations which may be present in clinical trials. It has been suggested [37] that this may be due to difficulty in NK-tumor binding caused by the interferon. Although in vitro anti-tumor action by N K cells is not questioned, the in vivo role of these ceils in tumor defense is not established~ The question has been primarily examined by correlating resistence to tumor challenge to in vitro NK activity. High in vitro N K activity correlates strongly with tumor resistance in vivo if the tumor is small (103 - 104 cells in the mouse) [ 18]. Tumor types which escape in vivo in general have not been susceptible to N K lysis in vitro [15]. These findings suggest that in vivo, NK cells are at least participating in the activity against tumors recogn~ed by these cells. N K ~ctivity in vitro can be suppressed by suppessor T-cells and macrophagelike cells [6]. As the dynamics of the suppressor cell role in the sneaking through phenomenon are still cloudy [25], the model constructed here assumes very general terms for the dynamics of tumor growth and elimination as well as the dynamics of interferon production due to tumor-NK cell contact so that most types of interaction will be included. N K cells seem to include K cells, those cells which mediate antibody dependent cytotoxicity (ADCC) [15]. This model will assume there is no anti-tumor antibody present in the system (or equivalently, if present, that its concentration is constant in time). 2. Construction of the Model The assumptions used in the construction of this model are: 1) The source of pre-NK ceils is the bone marrow. In the absence of any stimulation or suppression of the marrow, pre-NKcells appear at constant rate $1. 2) The rate of maturation of pre-NK to NK cells is an increasing function of interferon concentration. In the absence of interferon, this rate is proportional to pre-NK cell concentration. 3) N K cells produce interferon in contact with tumor. This interferon acts on pre-NK cells and the tumor. Interferon concentration is nearly homogeneous spatially (well-mixed body). 4) Tumor growth is affected by interferon present and the rate of cytotoxicity by N K cells. Other mechanisms which affect tumor growth are assumed to be approximately unchanging in relative magnitude during the lifetime of the model. By this is meant that all other effects are approximately in the form ~Tfor cceither a positive or negative constant. Assuming the law of mass action, the equation governing the pre-NK population, NKv, is dNKp. = $I - k I N K p l F - kzNK v. dt (1) In (1), IF is the concentration of interferon, St, kl and k2 are positive constants. 366 S~ J. Merrill k2NK v is the rate of pre-NK death plus baseline pre-NK maturation (not interferon dependent). N K f l F is proportional to the rate of pre-NK contact with interferon, which is assumed proportional to the increase in pre-NK maturation. The "mature" N K cell population, NKm, similarly has dynamics governed by dNKm dt = k t N K p l F - k3NK~ + k'2NKp (2) where k~ and k3 are positive constants, k3NKm gives the rate of exit from the mature NK class (through death or further differentiation) and k'2NKp is the entrance rate into the mature class by natural maturation (not interferon dependent). The equation governing interferon concentration is dlF - - ~ --- Sz - k j F + ks(NK., T) - k'INKplF - kv(IF, T) (3) where $2 and k] are non-negative constants and k4 is a positive constant. Sz is the rate of (external) supply of interferon, k~ the removal or decay rate, k'~NKgF gives the rate of interferon removal due to interaction with pre-NK cells, ks(NK~, T) is a non-negative function which is proportional to the rate at which interferon is produced when NKm mature NKcells and T tumor cells are present, k7 is a general function which gives the rate at which interferon is removed by tumor-interferon contact. A basic assumption of this model is that the arena of the interactions is a wellmixed body (by blood and lymph circulation) and local variations in interferon concentration is assumed to be of short duration and not essential in the process. The functions ks and k~ in the analysis will be assumed to satisfy 0ks aks ks(O, T) = O, ks(NK,~, O) = O, ~ > O, > O, ONI~ aT kT(o, 73 = o, kT(ZF, O) = O. (4) The equation governing the population of tumor cells, T, is assumed to be dT - ~ m r(IF, T) -- k6(NK., IF, T) (5) where 7 is a function of interferon concentration and tumor population expressing the growth rate and ks is a non-negative function of NK,., IF and T expressing the removal rate by mature N K cells. The functions y and k6 are assumed to satisfy y( IF, O) = k6(O, IF, 73 = ks(NK., IF, O) = O, Ok6 ONK,,, > 0 and Ok6 0---T> 0. 3. Analysis The model developed in the previous section is dNKp - ' ~ = $1 - kxNKvlF - k2NKp, dlF <~O, (6) The Role of Natural Killer Cells in ImmuneSurveillance 367 dNKm dt = k l N K p l F - k3NK,, + k'2NKp, dlF = $2 - k, I F + ks(NKm, ~ - k'INKplF- kT(IF, 7~, ,it dT - - = ~(IF, T) - k6(NK., IF, T) dt (7) with constants St, $2, kt, k't, k2, k~, k3, k,/> 0. Functions k~, k6, k7 and ~,satisfying (4) and (6) and initial conditions NKp(O), NK.(O), IF(O), T(O) >10. The analysis of (7) will follow the natural functioning o f the system. As the natural state has no tumor present, the analysis begins by examining the three equations derived from (7) when T - - 0 with the same assumptions on the parameters" dNKp d--~- = $1 - k I N K p I F - kzNKp, dNK~ dt = k l N K p l F - k3NKm + k'2NKp, dlF - ~ = $2 - k , I F - k]NKpIF. (8) Define a box, B, in NKp, NK,,, IF space by B= {(NK,,NK.,,IF)IO<~NK, <<~,O<< NK. < ktStS______22 k'2St k2kak,+k-~'O'lF<'~,}" Theorem 1. Solutions to (8) whose initial conditions (NK~(O), NK,(O), IF(O)) ~ B exist, are unique and for all t > O, (NKp(t), NK.,(t), IF(t))eB. Proof Follows from the standard existence and uniqueness theorems [5], ['12] and from the vector field defined by (8) always pointing inward on surface of the boundary of B. As the solutions beginning in the positively invariant region B are constrained to that compact set for all t > 0, as t--* oo, each solution must approach some connected, compact set, its co-limit. The simplest possible is an equilibrium point, and there is a unique equilibrium point to (8) lying in the box, call it (Xo,Yo, Zo). Linearizing (8) about this point we find that ( NKp- xo,' ( - ( k l z o + k2) NK., - Yo] ,~ klzo + k'2 i F - Zo / - k' Zo 0 - k3 0 -ktxo. ,/NK~- xo, k,xo l [ N K . - Yo] - (k, + V, x o ) l \ X F / when (NKp, NK., IF) is close to (xo, Yo, Zo). The eigenvalues of the Jacobian matrix are the roots of (2 + k3)(~,2 + (k~zo + k2 + k'xxo + k,)2 + klk,zo + k'lk2x o + kzk,) = O. 368 S.J. Merrill By the assumption on the parameters, all roots have negative real parts. The critical point (Xo, Yo, Zo) is thus locally asymptotically stable. In fact, the following holds: Theorem 2. I f (NKp(t), NKm(t), IF(t)) is a solution o f (8) with (NKp(O), NKm(O), 1F(O)) >>,O, then (i) ( N K p ( t l ) , N K ~ ( t l ) , I F ( t l ) ) ~ B for some tl >i 0 and for all t > tl, and (ii) lira,_ ~(NKp(t), NKm(t), IF(t)) = (xo, Yo, Zo). Theorem 2 implies that (xo,Yo, Zo) attracts everything in the positive octant. Proof Let 2 = N K v - Xo, = NKm - Yo, ~,=IF-zo ~. and define v(2,37, z') = where kl gz k• b= 2 z + b2~ + k'~ ' 2k~xo + 2klzo k l g o + k 2 + k,, -b k'lX o " In coordinates 2,)~, L (8) becomes = - k12~. - klzo2 - klxo~. - k22, = k12~. + k : o 2 + k~xo3. - k3P + k'22, (9) = - k , 2 - kt2~. - k'lzo2 - k'~xo2 and (0, 0, 0) is the unique critical point of the system in the region of interest (3 >/ - Xo, 37 t> - Yo, z >t - zo). As the derivative of the positive semi-definite function V along a solution of (9) [5] is dV dt dV(2(t),fi(t),~.(t) dt (2)2 2ki(z o + ~ bki(zo + ~ + (z')2[ - 2 k l ( x o + 2 ) - b k l ( x 2 k2 o +2)-2~k,~] ~<0 in the region of interest, the lim,_~ V(t) exists and is 0. Let (2(t),y(t), 3.(0) be a solution of (9) in this region, then lim,_~(2, 37,z') = {(0, 37,0)137>t - Yo} = E. As the a~-limit of (2(t),y(t),3.(t)) must itself be invariant and (m-limit) = E, the w-limit must be an invariant set inside E. By examining the second equation of (9) when 2 = 5 = 0, we find that 37 = 0 is the only invariant set contained in Eo We have shown that limt_~(2(t),5(t))= (0,0). Thus 2 and 5 must stay bounded. This implies in (9) that )7 also stays bounded, and )7(t) ~ 0 as t --, oo. In coordinates The Role of Natural Killer Cells in Immune Surveillance 369 (NKp, NK~,, IF), as (Xo, Yo, Zo) corresponds to (0, 0, 0) and lies inside B, the theorem is proven. We now examine (7). Critical points of this system must satisfy = 7(IF, T) - k6(NK,,, IF, T) --- 0. Expanding this difference in a Taylor series about T = 0 (using (6)) v(IF, T) - k6(NK--,n,IF, T) = v(IF, O) - k6(NKm, IF, O) ~k6 T ( dV (IF, O) --~(NKm, IF, O)) + kOT / a2y Tz a2k6 + I (IF, r - -ff-fT (NXm, IF, r for some r e (0, T), { a~'(/F, O) = r {.dr Ok~ T az~, c32k6 ---~(NK,,,,IF, O)+-.~.(-~--s162162 Thus T = 0 is always a solution of T = 0 and as a result, (Xo, Yo, Zo, 0) is always one critical point of (7). Linearizing about this point the Jacobian matrix is r-(k'z~ -k30 -k~xo ktxo Ok.~ -ktzo +k'2 3k~ ~ /Ok5 ] Ok~ "~ From (4) and (6), ak5 Ok6 Ok7 ..-3V (zo, 0) = ak6 O---~(y o, O) = O---~=(yo, Zo, O) = -~(Zo, O) = ale -b-~(yo, Zo, O) = O. As a result, the eigenvalues of A consist of the eigenvalues from the linearization of (8) about (Xo, Yo, zo) with the addition of aV . Ok6 ,to = b--f tzo, 0) - - ~ (yo, zo, 0). We have Theorem 3. The critical point (xo, Yo, Zo, 0) of(7) is locally asymptotically stable if 2o < 0 and unstable if;to > 0. Moreover the stable manifold of (xo, Yo, zo, O) always contains the manifoM T = O. Biologically Theorem 3 says that if (xo, Yo, Zo) lies in a region where 2o < 0 (20 being determined by the growth characteristics of the tumor and its susceptibility to NK lysis), small tumors will be eliminated with the result that the system returns to (Xo, Yo, Zo, 0). Description of the curve 2o = 0. Set 20 = h(IF, NK,,,) = 0 Ok6 (IF, O) - --~ (NKm, IF, 0). 3~ S.J. Memll IF ~o < 0 ).o > 0 | ~r | NKm Fig. 1. A typical curve ;to = 0 separating the positive octant into regidns where (NK,,,,IF) -- (Yo,Zo) would determine the stability of (Xo,Yo, Zo, 0) according to Theorem 3 h is determined by the response of a particular tumor to interferon and its susceptibility to NK lysis. These parameter functions should be able to be approximated in vitro and this model would then be able to predict the probability that a tumor of that type would escape the N K surveillance once an individual's status (Xo,Yo, Zo) is determined. Fix IF > 0o As ~h/ONK,, < 0, either there exists a unique number ~ ( I F ) > 0 such that h(IF, ~ ( I F ) ) = 0 or no such S~ exists. If h([F, 0) > 0 and no ~ ( I F ) exists for any IF > 0, (Xo,Yo, zo) is always unstable and the tumor will evade extinction~ If h(IF, 0) < 0, for all IF and no ~ ( I F ) exists, (xo, Yo, zo, 0) is the only critical point and Tis always decreasing. On the other hand, if for some IF = IFo, ,9'(IFo) exists, h(IF, ~ ) = 0 can be solved for ~ as a function of IF near (IFo, ~(IFo)) since c~h/O~ < O. Figure 1 displays the relationship between NK,, and IF along such a curve ;to = 0 if h(IF, 0) > 0 for some IFo > 0 and c~h/alF (IFo, SP(IFo)) < 0~ 4. Discussion According to the model presented here, tumors satisfying NK susceptibility parameters of Theorem 3, ensuring ;to < 0, would be eliminated by this natural cytotoxic mechanism. A second part of that theorem (that T = 0 comprises the stable manifold in the case ;to > 0) predicts that a tumor which has ;to > 0 will never be totally eliminated by this mechanism naturally. It is expected that the parameters Yo and Zo can be determined by standard methods outlined in the cited literature while Yo + xo should be the total number of NK active cells after treatment withsuitable levels of interferon~ The computation of ;to for a particular tumor in a particular individual may be accomplished by estimating the relative growth rate for small tumors in the presence of interferon at concentration Zo but no N K cells as (gy/aT)(zo, 0) y(Zo, s)/a for small tumor population s. The assumption of exponential growth for small tumor size is most likely appropriate. (ak6/~T)(yo, Zo, 0) involves estimating The Role of Natural Killer Cells in Immune Surveillance 371 the mean survival time of a tumor in the presence of the appropriate concentration of N K cells and interferon as (Ok6/OT)(yo, Zo, 0) ~ k6(Yo, Zo, ~)/e. This model may be tested experimentally in vivo by artificially altering parameters xo,Yo,.Zo to attempt to modulate tumor viability in a nude mouse. N K Role in Immune Surveillance In the model developed here, N K cells would effectively eliminate tumors with growth and N K susceptibility parameters satisfying the conditions given in Theorem 3. Any tumor which escapes this mechanism of defense would still have to deal with the immune system side, mediated primarily by T cells. Tumors that lack sufficient antigenicity, however, would most likely encounter only the N K cells and conversely, tumors resistant to N K cells would primarily encounter the T cell mediated mechanism. In this light, "immune surveillance" can best be seen as the result of two essentially independent anti-tumor mechanisms, each recognizing and responding to different structures on the surface of a neoplastic cell. Given an individual whose N K status and immune status are known, the types of tumors which could escape both N K and T surveillance can be commented on. Each tumor has a susceptibility to N K lysis and a susceptibility to lysis by T-ceU mediated mechanisms (roughly the antigenicity if the tumor has not been previously encountered and blocking effects ['24] are not considered). The tumor also has growth kinetics determined by the characteristics of the tumor in the environment in which it is placed (including interferon present, natural antibody present and other parameters). In order to survive, the tumor must have the N K status of the individual in the ,lo > 0 region of (NK~, IF) plane and must be able to grow to a critical mass before the priming of T-cells would enable the elimination by that mechanism. How the antigenicity affects the time that the tumor has to reach critical mass is a fundamental question in this area. Also, the stochastic nature of this event most likely also plays a critical role. An analogy of Prehn [27] given in 1970 can be easily extended to explain the interaction between the two systems. The analogy can be made to a very large warehouse with multitudinous roomsfilled with combustible materials. Spontaneous combustion is frequent and, in addition, arsonists (represented by viruses, radiation and chemical oncogens) lurk in the halls and passageways. In this analogy, N K defense can be represented as smoke detectors whose triggering immediately starts an automatic sprinkling system. The T-cell defense is represented by heat detectors, which are connected to alarms in the fire department some blocks away. In each case, there are certain fires that are detected sufficiently early by one but not the other. The independent nature of the systems enhances the reliability and for most challenges the systems will cooperate. References 1. Albert, A., Freedman, M., Perelson, A.: Tumors and the immune system: The effects of a tumor growth modulator. Math. Biosci. SO, 2 5 - 5 8 (1980) 372 2. 3. 4. 5. 6. S.J. Merrill Burnet, F. M.: C a n c e r - a biological approach. Br. Med. J. I) 779-786 and 841-847 (1957) Burnet, F. M.: The concept of immune surveillance. Prog. Exp. Tumor Res. 13, 1 - 2 3 (1970) Burnet, F. M.: Immunological surveillance in neoplasia. Trans. Rev. 7, 3 - 2 5 (1971) Cronin, Jane: Differential equations, introduction and qualitative theory. Marcel Dekker, 1980 Cudkowicz, G., Hochman, P. S.: Do natural killer cells engage in regulated reaction against self to ensure homeostasis? Immunolo Rev. 44, 13-41 (1979) 7. DeLisi, C., Rescigno, A.: Immune surveillance and neoplasia, I. A minimal mathematical model. Bull. Math. Biol. 39, 201-221 (1977) 8~ DeMaeyer, E., DeMaeyer-Guignard, J.: Interferons. In: Comprehensive virology, Vol. 15 (H. Fraenkel-Conrat, R. R, Wagner, eds.), pp. 205-284. Plenum 1979 9. Eisen, Me: Mathematical models in cell biology and cancer chemotherapy. Springer-Verlag 1979 10o Garay, R. P., Lefever, R.: A kinetic approach to the immunology of cancer: Stationary states properties of effector-target cell reactions. J. Theor~ Biol. 73, 417 -438 (1978) 11o Grossman, Z., Berke, G.: Tumor escape from immune elimination. Jo Theor. Biol. 83, 2 6 7 - 296 (1980) 12o Hale, J. K.: Ordinary differential equations. Wiley-Interscience 1969 13- Hellstr6m, K. E., Hellstr6m, L: Lymphocyte mediated cytoxicity and blocking serum activity to tumor antigens. Adv. Immunol. 18, 209-277 (1974) 14. Herberman, Re B., Nunn, M. E., Lavrin, D. H., Asofsky, R.: Effect of antibody to 0 antigen on cellmediated immunity induced in syngeneic mice by routine sarcoma virus. Jo NatL Cancer Inst. 51, 1509-1512 (1973) 15~ Herberman, R. B., Holden, H. T.: Natural cell-mediated immunity. Adv. Cancer Res. 27, 305 - 377 (1978) 16. Herberman, R. B., Djeu, Jo Yo) Kay, H. D~ Ortaldo, J. R~ Riccardi, C., Bonnard, G. D., Holden, H. T., Faguani, R.., Santoni, A., Puccetti, P.: Natural killer cells: Characteristics and regulation of activity. Immunoi. Rev~ 44, 4 3 - 7 0 (1979) 17. Kiessling, R., Hochrrian, P. S., Hailer, O., Wigzell, H~ Cudkowicz, G.: Evidence for a similar or common mechanism for natural loller cell activity and the resistance to hemopoietic grafts. Euro J. Immunolo 7, 655-663 (1977) 18. Klessling, R., Wigzell, H." An analysis of the murine NKcell as to structure, function and biological relevance. ImmunoL Revo 44, 165-208 (1979) 19~ Klein, G., Klein, E.: Immune survdllance against virus-induced tumors and nonrejectability of spontaneous tumors: Contrasting consequences of host versus tumor evolution. Prec. Nat. Acad. Sci. USA 74, 2121-2125 (1977) 20. Koo, G~ C., Jacobsen, Jo B., Hammerling, G. J., Hammerling, U.: Antigenic profile of routine natural killer cells. J. lmmunoL 125, 1003-1006 (1980) 21. Lefever, R., Garay, R. P.: A mathematical model of the immune surveillance against cancer. In: Theoretical immunology (Go L Bell, Ao S~Perelson, G~He Pimbley, Jr., eds.), pp. 481 - 518. Marcel Dekker 1978 22. Lefever, Re, Garay, R. P.: Local description of immune tumor rejection. In: Biomathematics and cell kinetics (Ao J. Valleron, P. D~ M~ Macdonald, eds.), pp. 333-344. Elsevier/North-Holland 1978 23. Lohmann-Matthes, M. L., Roder, J.: Promonocytes have the functional characteristics of natural killer ceils. J. Immunol. 123, 1883 - 1886 (1979) 24. Merrill, S.J.:Amathematicalmodeloftumorgrowthandcytotoxicbiockingactivityo Math.Biosci. 47, 7 9 - 8 9 (1979) 25. Naor, D.: Suppressor ceils: Permitters and promoters of malignancy. Adv. Immunot. 29, 45 - 125 :(1979) 26. Oldham, R. K., Siwarski, D., McCoy, J. L., Plata, E..I., Herberman, R~ B.: Evaluation of a cellmediated eytotoxicity assay utilizing lZSIododeoxyuridine-labeled tissue culture target cells~Natl. Cancer Inst. Monograph 37, 4 9 - 5 8 (1973) 27. Prehn, R. T.: Discussion. In: Immune surveillance (R. T. Smith, M. Laody, eds.), pp. 451-462. Academic Press 1970 28. Rescigno, A., Del,isi, C.: Immune surveillance and neoplasia, II. A two-stage mathematical model. Bull Math. Biol. 39, 487-497 (1977) 29. Roder, J. C., Lohmann-Matthes, M. L., Domzig, W., Wigzell, H.: The beige mutation in the mouse, II. Selectivity of the natural (NK) cell defect. J. Immunoi. 123, 2174-2181 (1979) The Role of Natural Killer Cells in Immune Surveillance 373 30. Rygaard) J., Poulsen, Co O.: The nude mouse versus the hypothesis of immunological surveillance. Trans~ R~v. 28, 43-61 0976) 31. Saksela, E., Timonen, T., Ranki, A., Hayre, P.: Morphological and functional characterization of isolated effector cells responsible for human natural killer activity to fetal fibroblasts and to cultural cell line targets. Immunol. Rev. 44, 71 - 123 (1979) 32~ Santoli, D., Koprowski, H." Mechanisms of activation of human natural killer cells against tumor and virus-infected cells. ImmunoL Rev. 44, 125-163 (1979) 33. Swan, G. W.: Some current mathematical topics in cancer research. University Microfilms 1977 34. Thames, H. D. : Mathematical models of dose and cell cycle effects in multifraction radiotherapy. In: Modeling and differential equations in biology (T. A. Burton, ed.), pp. 51 - 105. Marcel Dekker 1980 35. Thomas, L.: Reactions to homologous tissue antigens and relation to hypersensitivity. In: Cellular and humoral aspects of the hypersensitive states (Ho S. Lawrence, ed.), pp. 529-532. Hoeber 1959 36. Timonen, T., Saksela, E., Virtanen, I., Canteil, K.: Natural killer cells are responsible for the interferon production induced in human lymphocytes by tumor cell contact. Ear. J. Immunol. 10, 422-427 (1980) 37. Welsh, R. M., Karre, K., Hansson, M., Kunkd, L. A., Kiessling, g. W. : Interferon-mediated protection of uormat and tumor target ceils against lysis by mouse natural killer ceils. J. Immunol. 126, 219-225 (1981) Received November 18, 1980/Revised February 5, 1981