Proc. Nat. Acad. Sci. USAVol. 69, No. 6, pp. 1596-1600, June 1972

Specific Fractionation of Immune Cell Populations(cell fractionation/antigens/immunoglobulin receptors/clonal selection theory)

U. RUTISHAUSER, C. F. MILLETTE, AND G. M. EDELMAN

The Rockefeller University, New York, N.Y. 10021

Contributed by Gerald M. Edelman, March 31, 1972

ABSTRACT Antigen-binding cells from spleens ofimmune and nonimmune mice were isolated by themethod of fiber fractionation. Binding of the lymphoidcells to derivatives of nylon fibers made with various anti-gens was prevented by the presence of the respective freeantigen, as well as by antibodies to mouse immunoglob-ulins. Antigen-binding cells specific for dinitrophenylgroups were separated from direct and indirect plaque-forming cells of the same specificity. Spleen cells from im-mune and nonimmune animals were fractionated ac-cording to their relative affinities for antigen, and the per-centage of antigen-binding cells in the spleens of non-immune animals was estimated. A comparison of thenumbers and relative affinities of immunoglobulin recep-tors of immune and nonimmune populations indicatedthat after immunization only those antigen-binding cellsof higher affinities were increased in number. This findingsuggests that the specificity of clonal selection depends notonly upon the binding of antigen to a lymphoid cell butalso upon the capacity of that cell to be triggered to ma-ture and replicate.

Lymphoid cells in various stages of maturation differ in theclass and specificity of their immunoglobulin receptors, aswell as in their biological functions. Fractionation of thisheterogeneous population of cells is essential to an under-standing of their role in the selective immune response.

Useful methods have been described for the fractionation ofimmune cell populations (1-3) by the use of antigen-coatedbeads to bind cells through their surface receptors. Bead col-umns do not permit direct visualization of the bound cells forquantitation, however, and isolation of a pure population ofthese cells for further studies is difficult. The detection ofantigen-binding by autoradiography and rosette formation (seerefs 4 and 5) permits direct quantitation of specific cells but isnot convenient for fractionation and subsequent study of thecells.We have devised a procedure, called fiber fractionation (6),

that permits both separation and quantitative study ofimmune cell populations. Cells specific for a given antigen canbe removed from the general population, visually quantitated,and characterized with respect to the specificity, number,affinity, and distribution of their surface receptors. In thepresent paper, we report the use of this method to studyantigen-specific populations from the spleens of immune andnonimmune mice.

MATERIALS AND METHODS

Cell Suspensions and Immunization. Spleen cells of 2- to5-month-old Balb/c mice (Jackson Lab., Bar Harbor, Me.),were prepared as described (6). Sheep erythrocytes were ob-tained from Microbiological Associates, Bethesda, Md.Mice were injected intraperitoneally with 200 Ag of antigen

prepared with complete Freund's adjuvant or adsorbed ontobentonite (7). Primary immunization consisted of a singleinjection. Secondary immunizations consisted of one or twoadditional injections with bentonite-adsorbed antigen atmonthly intervals. In each case, spleens were removed 4-7days after the final injection.

Preparation of Derivatized Fibers and Beads. Derivatizationof nylon monofilaments and meshes was carried out asdescribed (6). The nylon was incubated for 30 min at roomtemperature (25°) in a solution containing 0.25 mg/mlDnp8-BSA (8), Tosyl2o-BSA, or BSA and 1.25 mg/ml of watersoluble carbodiimide [1-cyclohexyl-3-(2-morpholinoethyl)-carbodiimide metho-p-toluenesulfonate; Aldrich Chemical Co.,Milwaukee, Wis.]. The extent of derivatization was deter-mined by use of lu2I-labeled antigen (8). For coupling ofsonicated sheep erythrocyte stroma (9), 2.5 mg/ml of antigenand 12.5 mg/ml of carbodiimide were used.The Dnp8-BSA derivative of agarose (Sepharose 4B;

Pharmacia, Uppsala, Sweden) was prepared by the method ofcyanogen bromide activation (3).

Cell Binding to Derivatized Fibers. Derivatized fibers strungin polyethylene collars were incubated with 108 spleen cellsin 4 ml of Hank's balanced salt solution, without NaHCO3Grand Island Biological Co.) for 1 hr at room tempera-ture on a reciprocal shaker having a 3.3-cm horizontalstroke at 78 oscillations per minute. The fibers were alignedperpendicular to the direction of shaking. Inhibition ofbinding was accomplished by adding the indicated concen-trations of the inhibitor to the cell suspension prior to in-cubation with the fibers. Unbound cells were removed byimmersion of the entire assembly in phosphate-bufferedsaline (pH 7.4).Two methods were used to deplete spleen cell populations

of specific fiber-binding cells (FBC). 1 g of nylon fiber was cutinto 2- to 4-mm segments, derivatized, and placed in a60 X 15-mm petri dish containing 2.5 X 108 cells in 8 ml ofHank's solution. After incubation for 1 hr at 4° on a reciprocalshaker, supernatant cells were removed for study. Alterna-tively, cells were incubated in a 35 X 8 mm air-free chamberseparated into two compartments by a derivatized nylon

1596

Abbreviations: BGG, bovine gamma globulin; BSA, bovine serumalbumin; FBC, fiber-binding cells; PFC, plaque-forming cells;RFC, rosette-forming cells; Tosyl, the p-toluenesulfonyl group.

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 2,

202

0

Proc. Nat. Acad. Sci. USA 69 (1972)

mesh (308 gauge, square weave, Nitex; Tobler, Ernst, andTraber, New York, N.Y.). 8 ml of a cell suspension (1.25-2.5 X 107 cells/ml) was added and the chamber was placedon a horizontal rotary shaker at 200 rpm for 1 hr at 4°. Thechamber was inverted every 15 min so that the cells filteredthrough the mesh under unit gravity. The mesh could beremoved from the liquid without loss of bound cells and wasthen washed by immersion in medium. Bound cells werecounted in a hemocytometer after their removal by pipetting4 ml of medium repeatedly through the mesh. With bothmethods, the extent of depletion was tested by assay of thesupernatant cells for fiber-binding cells (FBC) with strungfibers.

Assay of Fractionated Cell Populations. The relative numberof immunoglobulin receptors on fiber-fractionated cells wasestimated by incubation of cells bound to Dnp-derivatizedfibers with increasing amounts of 125I-labeled rabbit anti-mouse immunoglobulin (8) for 30 min at room temperature.After the unbound antibody was washed away, the cells wereremoved from the fibers by plucking (6) into medium con-taining 10% fetal-calf serum, after which 107 carrier cellsfrom fresh spleens were added to facilitate further washing bycentrifugation. The radioactivity of the cells was determinedin a model 4320 Nuclear Chicago gamma spectrometer.

Cells specific for Dnp groups were first isolated by fiberfractionation on strung fibers in order to identify those cells(5) that could form specific rosettes (RFC). After removal ofunbound cells, 0.3 ml of a 33% suspension of Tnp-coupledsheep erythrocytes (7) was added to the dish, which was in-cubated for 1 hr at room temperature with gentle recipro-cal shaking; free erythrocytes were then washed away. Specificinhibition of rosette formation was performed by addition ofthe appropriate inhibitor to the dish before incubation withthe erythrocytes.

TABLE 1. Binding of mouse spleen cells toantigen-derivatized fibers

Number of cells boundt

Fiber antigen* Immunized: Unimmunized

Dnp-BSAExp. 1 802 285

2 1004 3013 654 283

Tosyl-BSAExp. 1 353 143

2 297 130BSAExp. 1 173 65

2 112 58StromaExp. 1 160 75

2 145 70

* See Methods for details of derivatization. The Dnp-BSA andTosyl-BSA fibers were coated with 1011 and 2 X 1011 antigenmolecules per cm, respectively.

t Expressed as number of cells bound to both edges of a 2.5-cmfiber segment.

$ Secondary responses to Dnp3rBGG [bovine gamma globulin(fraction II), Armour Pharmaceutical], Tosyl3o-BSA, BSA, andsheep erythrocytes, respectively. Cells from three mice werepooled.

a

b

FIG. 1. Mouse spleen cells bound to Dnp-BSA nylon mono-filament (a, X200 magnification) and mesh (b, X 125 magnifica-tion).

The number of antibody-secreting or plaque-forming cells(PFC) in fractionated cell populations was determined withTnp-coupled sheep erythrocytes in a modified Jerne assay(7). IgG-secreting cells were assayed in the presence of rabbitantiserum to mouse immunoglobulin, at a dilution of 1:200.This antiserum suppressed over 90% of the IgM plaques.

RESULTSQuantitation and specificity of cell bindingThe specific binding of mouse spleen cells to an antigen-derivatized fiber and mesh is illustrated in Fig. 1. Variousantigens including haptens, proteins and cell membraneswere used to derivatize the fibers (Table 1). The cells werefirmly attached to the fibers and their number was determinedby counting those bound to the edges of a 2.5-cm segment offiber. This number represented 1% of the total cells bound todishes containing 25 cm of fiber. High binding capacitiescould be achieved by the use of fiber meshes (up to 5 X 101cells/mesh). The number of FBC was not appreciably affectedby cell viability as long as over 70% of the cells were livingand all cell aggregates were removed.The binding of cells from immunized and unimmunized

mice to antigen-derivatized fibers was specifically inhibited bythe presence of the soluble antigen. Unrelated antigenmolecules did not inhibit binding (Table 2). Prior incubationof the spleen cells with the IgG fraction or Fab-fragments (8)

Willw1k. wa-.

Fiber Fractionation of Cells 1597

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 2,

202

0

1598 Immunology: Rutishauser et al.

0

b

FIG. 2. Rosettes formed byr spleen cells bound to a Dnp-BSAfiber after incubation with Tnp-derivatized sheep erythrocytes.(a) X330, (b) X850 magnification.

derived from a rabbit antiserum directed against mouse

immunoglobulin prevented the specific binding of cells toantigen-substituted fibers. No inhibition was obtained whennormal rabbit IgG was used. The presence of 5 mM sodiumazide also did not affect fiber binding. These data stronglysuggest that the attachment is specific and is mediated bymmunoglobulin receptors on the cell surface.

Rosette formation and distribution of receptors

In order to examine the distribution and specificity of antigenreceptors on FBC, spleen cells bound to fibers derivatizedwith Dnp-BSA were tested for their ability to form rosettesin situ (fiber-RFC) with Tnp-coated sheep erythrocytes.The results are shown in Table 3 and typical fiber-rosettes are illustrated in Fig. 2. The uniform appear-ance of the fiber-RFC suggests that antigen receptors are

distributed over the entire surface of FBC. About 54% of theFBC from immunized mice and 17% from unimmunized micewere capable of forming rosettes. In contrast, an unfrac-tConated spleen cell population from immune and nonimmunemice contained 3-5% and 0.3-0.4% RFC respectively as

determined by the centrifugation-resuspension rosette assay(5). Since this assay was performed at 40, antibody secretingcells were not detected as RFC (10). Formation of the fiberrosettes could be inhibited by both I)np-BSA and anti-immunoglobulin, but not by tosyl-BSA (Table 3). Un-derivatized sheep erythrocytes did not form rosettes.

Relative number and affinity of receptors on FBC

Fiber fractionation allowed a comparison of the number ofspecific antigen-binding receptors on the surface membranesof cells from immune and nonimmune animals. At saturation,FBC from immunized mice bound 8 X 104 anti-immuno-globulin molecules, whereas those from unimmunized micebound 2.5 X 105 molecules. These numbers are consistent withthe number of molecules bound (about 4 X 104) in experi-merits with unfractionated spleen cell populations (11).The effect of various free antigen concentrations on the

number of spleen cells bound to Dnp-BSA-derivatizedfibers is shown in Fig. 3. The inhibition of binding dependsul)on the fact that the receptors of the cells are blocked byfree antigen so that binding to the fiber cannot occur. Oncecells are bound to the fibers, however, they cannot be removedby presence of the soluble antigen (6).Although the total number of fiber-binding cells detected in

single dishes with cell suspensions from immunized mice wasonly 2-4 times that obtained with nonimmune animals(Fig. 1), the experiments illustrated in Fig. 3 revealed thatthe two cell populations differ greatly in their susceptibility toinhibition by Dnp8-BSA. A clear-cut change in the number ofcells of higher affinity was found after immunization. In theimmune animal, over 60% of the FBC are inhibited by antigenconcentrations of less than 4 gg/ml, whereas at the sameconcentrations, the number of FBC from an unimmunizedanimal decreases by less than 3%. At higher antigen con-centrations the curves are identical. Similar results were ob-tained with the monovalent antigen, e-Dnp-lysine, as aninhibitor.

Separation of FBC from plaque-forming cells (PFC)and estimation of total FBC

A cell population depleted of specific FBC was obtained byincubation of spleen cells with a large amount of antigen-derivatized fiber and collection of the unbound cells. Thisprocess removed less than 5% of the total number of cells.When the number of Dnp-specific antibody secreting cells inthe unabsorbed and the depleted cell populations was com-pared by the plaque assay, no difference was observed forboth IgG- and IgM-secreting cells from immunized andunimmunized animals. In contrast, incubation of spleen cellswith antigen-coated agarose beads (3) under the same con-ditions resulted in a partial depletion of PFC (Table 4).

TABLE 2. Specific inhibition of spleen cell binding toderivatized fibers

Inhibitor*

Stro- Anti-Fiber antigen Immunogen Dnp Tosyl ma Ig

Dnp-BSA Dnp-BGG 91t 1 - 93Dnp-BSA None 73 2 - 72Tosyl-BSA Tosyl-BSA 3 87 - 90Tosyl-BSA None 6 59 - 63Stroma Stroma <5 <5 70 80Stroma None <5 <5 50 45

* Dnp8-BSA and Tosyl20-BSA present at 200 Ag/ml; sonicatedstroma and rabbit anti-mouse immunoglobulin at a concentrationof 1 mg/ml.

t All values are expressed as % inhibition.

Proc. Nat. Acad. Sci. USA 69 (1972)

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 2,

202

0

Proc. Nat. Acad. Sci. USA 69 (1972)

The percentage of specific Dnp-BSA, tosyl-BSA, and BSAbinding cells in spleens from unimmunized mice could bedetermined directly by counting the number of bound cellsby the mesh depletion method. Cell binding to the antigen-derivatized meshes was 50-70% inhibitable by antigen or

anti-immunoglobulin. Six separate experiments for eachantigen, including a control with inhibitor to correct forbackground, yielded reproducible results. Hapten-BSA and

hapten-BGG conjugates inhibited binding to hapten-BSAfibers equally well, suggesting that the specific binding tothe fiber was largely via the haptenic groups. The percentagesof cells in unimmunized mouse spleens that bound specificallyto Dnp-BSA-, tosyl-BSA-, and BSA-derivatized fibers were

1.1-1.5%, 0.6-0.8%, and 0.4-0.6%, respectively. UsingDnp-BSA derivatized meshes, the depletion procedureremoved 90% of the inhibitable FBC and 65-75% of therosettes as detected with Tnp-sheep erythrocytes.

DISCUSSION

Fiber fractionation provides a rapid and direct means for theisolation and quantitative study of cells having specificantigen receptors. The method can be used with a widevariety of antigens and extensive comparisons may be madeamong cells from a single animal. Binding of cells to antigenfibers is inhibited by prior treatment of the cells with antibodyto mouse immunoglobulin, and it therefore seems likely thatthe cell-fiber linkage is mediated by immunoglobulin receptorson the cell surface. The inhibition with soluble antigens indi-cated that these receptors have a high degree of specificity.The number of cells that bind to fibers can easily be deter-

mined by visual inspection (Fig. 1). Several factors influencethe number of FBC, including the valence of the antigen, thesurface concentration of antigen molecules on the fiber, theincubation conditions, and the immunological history of thecells.The precise biological role of FBC has not been defined.

In vivo experiments are in progress to determine whether FBCinclude precursors of antibody secreting cells and thymus-derived cells. These experiments are facilitated by theability to remove cells from the fibers (6) and to separateFBC from PFC.A substantial proportion (30-50%) of FBC from non-

immune mice are not inhibitable by either antigen or anti-immunoglobulin (Table 2). These cells probably representnonspecific adherence to the fibers, but the possibility that

TABLE 3. Formation of Dnp-specific rosettes by spleen cellsbound to Dnp-erivatized fibers

% Fiber-RFCtInhibitor of rosette formationt

Anti-FBC* None Dnp Tosyl Ig

Immunized§ 603 54 6 49 2Unimmunized 296 17 5 18 1

* 2.5-cm edge count.t Expressed as (fiber-RFC/FBC) X 100.I Dnp8-BSA and Tosyl2o-BSA at 200 ,g/ml; rabbit anti-mouse

immunoglobulin at 0.5 mg/ml added during rosette formation.§ Secondary response to Dnp38-BGG.

800j

60

O 400

O 200

0 0.01 0.1 10 100Inhibitor concentration (p)g/ml)

FIG. 3. Inhibition by free Dnp8-BSA of spleen-cell binding toDnp8-BSA derivatized fibers. Cell numbers represent fiber edgecounts for a 2.5-cm fiber segment. Spleens from immunized micewere removed at the height of a secondary response to Dnp-BGGand cells from several mice were pooled. * *, immunized;A /A, unimmunized.

binding occurs through low affinity or unusual receptorscannot be ruled out. In all experiments, however, the numberof these cells was reproducible and the corresponding back-ground corrections can be made.Although absorption experiments using derivatized fibers

depleted the cell population of FBC and of most RFC, it isnotable that no PFC were removed. Several investigators(1-3) have reported that antigen-coated beads can be usedto remove some portion of specific antibody-secreting cellsfrom immune cell populations. Using Dnp-BSA derivatizedagarose in a batch absorption experiment (3), we were able toremove 40% of the PFC. The reason for the differencebetween the fiber and the bead methods has not been deter-mined. It is pertinent, however, that many antibody-secretingcells do not have immunoglobulin receptors on their surfaces(5, 10). The relative sensitivity of the antibody assays usedmust also be considered.A striking aspect of our results is the high percentage of

F13C found in nonimmune animals. During the course ofimmunization the number of FBC in a mouse spleen increasedby a factor of only 2-4 (Fig. 1). However, the same animalsshowed 15- and 600-fold increases when assayed by therosette and indirect plaque assays, respectively. The lowamplification of the total specific FBC poIpulation afterimmunization is due to the detection, in both immunized and

TABLE 4. Depletion of Dnpfiber-binding cells and Dnpplaque-forming cells

Assay*

Direct IndirectPFC/106 PFC/106

Absorbed with: FBC cells cells

Unabsorbed 481 t 25 668Dnp-fiber 64 24 676Tosyl-fiber 462 24 640Dnp-Sepharose 120 16 396

* Each assay was performed on cells remaining after the var-ious absorption procedures. Cells were obtained from a pool ofmice after secondary immunization.

t Expressed as the number of FBC on the edges of a 2.5-cmfiber segment.

Fiber Fractionation of Cells 1599

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 2,

202

0

1600 Immunology: Rutishauser et al.

unimmunized animals, of a relatively large number of FBCthat are neither RFC nor PFC. Additional experiments arerequired to prove that the surface immunoglobulin mediatingthe specific binding of a cell to a fiber is actually synthesizedby that cell and is not passively adsorbed. The latter possi-bility seems unlikely for FBC from unimmunized animals,however.A comparison of the percentages of RFC and fiber-RFC

allows an estimate of the enrichment of specific antigen-binding cells by fiber fractionation. From 0.3 to 0.4% of thespleen cells from an unimmunized mouse are RFC and 70%of these cells are capable of binding to the fibers. In immunizedanimals, the number of RFC reached 3-5%. In contrast,fiber-fractionated cells from unimmunized and immunizedmice contain 17 and 54% fiber-RFC, respectively. Theenrichment, therefore, ranges from 14- to 50-fold.The percentage of FBC in an unfractionated cell population

can be estimated by multiplying the percentage of RFC inthis population by the ratio of FBC to fiber-RFC afterfractionation. Assuming that rosette formation is equallyefficient in both methods, and given the fact that 70% of theunfractionated RFC are also FBC, the spleen cells from anunimmunized animal are estimated to contain 1.2-1.7% FBCcapable of binding to Dnp-derivatized fibers. In corroborationof this estimate, direct depletion studies with Dnp-nylonmeshes indicated that the fraction of FBC in an unimmunizedmouse was 1.1-1.5%. Immunized animals had 4-8% FBC ascalculated by the rosette methods.The relatively high percentage of antigen-binding cells in

an unimmunized animal is consistent with clonal selectiontheories of immunity. Assuming that FBC include the pre-cursor cells for PFC, it would appear, however, that onlythose cells that have sufficiently high binding constants for aparticular antigen would be stimulated to undergo divisionand differentiation. This is in accord with the conclusions ofother workers (12, 13).

Indirect evidence for this suggestion was obtained inexperiments to determine the effect of antigen concentrationon the inhibition of FBC from immune and nonimmuneanimals (Fig. 3). The dependence of binding upon freeantigen concentration may result from variation in eitherthe binding constants of the receptors or the number ofreceptors per cell. Two observations appear to exclude thelatter possibility: (a) the number of anti-immunoglobulinmolecules bound to fiber-fractionated cells from unimmunizedmice (2.5 X 105/cell) was somewhat higher than that foundwith cells from immunized mice (8 X 104/cell). Althoughthese results do not account for possible differences in theclasses of immunoglobulin receptors (e.g., IgG versus IgM),it is evident that immunization does not result in a largeincrease in the number of receptors per cell. (b) The inhibitioncurves obtained with the monovalent antigen e-Dnp-lysine

were similar to those obtained with the multivalent Dnp8-BSA inhibitor. If immunization had resulted only in anincrease in the number of receptors, then the binding ofcells from nonimmune animals should have been inhibited bylower concentrations of monovalent inhibitor.The fact that 1% of the spleen cells from an unimmunized

animal can bind to a single antigenic determinant would posea serious paradox if it is found that FBC include a significantnumber of precursor cells. Can the immune system commitso many precursor cells to a single antigen and also maintaina large enough library of specificities for proper function?Our results suggest that although many cells can bind theantigen to some extent, only a few cells with receptorshaving sufficiently high binding constants actually respond.The remaining cells may also be capable of binding to otherantigens, some of which may bind strongly enough to causetriggering.

This work was supported by Grants Al 09273 and AM 04256from the National Institutes of Health.

1. Wigzell, H. & Andersson, B. (1969) "Cell separation onantigen-coated columns. Elimination of high rate antibody-forming cells and immunological memory cells," J. Exp.Med. 129, 23-36.

2. Truffa-Bachi, P. & Wofsy, L. (1970) "Specific separation ofcells on affinity columns," Proc. Nat. Acad. Sci. USA 66,685-692.

3. Davie, J. M. & Paul, W. E. (1970) "Receptors on immuno-competent cells. I. Receptor specificity of cells participatingin a cellular immune response," Cell. Immunol. 1, 404-418.

4. Nossal, G. J. V. & Ada, G. L. (1971) Antigens, LymphoidCells and the Immune Response (Academic Press, New York).

5. McConnell, I. (1971) "Antigen receptors on the surface ofantibody-secreting cells," Nature New Biol. 233, 177-179.

6. Edelman, G. M., Rutishauser, U. & Millette, C. F. (1971)"Cell fractionation and arrangement on fibers, beads, andsurfaces," Proc. Nat. Acad. Sci. USA 68, 2153-2157.

7. Rittenberg, M. B. & Pratt, K. L. (1969) "Antitrinitro-phenyl (TNP) plaque assay. Primary response of Balb/cmice to soluble and particulate immunogen," Proc. Soc.Exp. Biol. Med. 132, 575-581.

8. Williams, C. A. & Chase, M. W. (eds.) (1967) Methods inImmunology and Immunochemistry (Academic Press, NewYork), Vol. 1.

9. Dodge, J. T., Mitchell, C. & Hanahan, D. J. (1963) "Thepreparation and chemical characteristics of hemoglobin-free ghosts of human erythrocytes," Arch. Biochem. Bio-phys. 100, 119-130.

10. Wilson, J. 0. (1971) "The relationship of antibody-formingcells to rosette-forming cells," Immunology 21, 233-245.

11. Yahara, I. & Edelman, G. M. (1972) "Restriction of themobility of lymphocyte immunoglobulin receptors byconcanavalin A," Proc. Nat. Acad. Sci. USA 69, 608-612.

12. Siskind, G. P. & Benacerraf, B. (1969) "Cell selection byantigen in the immune response," Advan. Immunol. 10,1-50.

13. Davie, J. M. & Paul, W. E. (1972) "Receptors on immuno-competent cells. V. Cellular correlates of the maturation ofthe immune response," J. Exp. Med. 135, 660-674.

Proc. Nat. Acad. Sci. USA 69 (1972)

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 2,

202

0