1: \documentclass[12pt]{article}

2: \usepackage{graphicx}

3: \usepackage{epsf}

4: \addtolength{\textheight}{100pt}

5: \addtolength{\textwidth}{70pt}

6: \addtolength{\topmargin}{-30pt}

7: \addtolength{\oddsidemargin}{-35pt}

8: %\addtolength{\evensidemargin}{-35pt}

9: %

10: %Gleichungen

11: \newcommand{\be}[1]{\begin{equation}\label{#1}}

12: \newcommand{\ee}{\end{equation}}

13: \newcommand{\bea}[1]{\begin{eqnarray}\label{#1}}

14: \newcommand{\eea}{\end{eqnarray}}

15: %

16: \newcommand{\nn}{\nonumber\\}

17: %

18: %

19: %Referenz auf Gleichungen und Abbildungen

20: \newcommand{\gl}[1]{Eq.\,(\ref{#1})}

21: \newcommand{\abb}[1]{Fig.\,\ref{#1}}

22: \newcommand{\tab}[1]{Tab.\,\ref{#1}}

23: \newcommand{\kap}[1]{Sec.\,\ref{#1}}

24: %

25: \newcommand{\eps}{\varepsilon}

26: \newcommand{\question}{{\bf ???}}

27: %

28: %

29: \begin{document}

30: \vspace*{2cm}

31: \begin{center}

32: {\LARGE \bf 

33: Recycling probability and dynamical\\[2mm]

34: properties of germinal center reactions}\\

35: \vspace{1cm}

36: %

37: %

38: Michael Meyer-Hermann$^1$, 

39: Andreas Deutsch$^2$

40: and

41: Michal Or-Guil$^2$\\

42: \vspace{1cm}

43: %

44: $^1$Institut f\"ur Theoretische Physik, TU Dresden,

45: D-01062 Dresden, Germany\\

46: E-Mail: meyer-hermann@physik.tu-dresden.de\\

47: \vspace{2em}

48: $^2$Max-Planck-Institut f\"ur Physik komplexer Systeme,\\

49: N\"othnitzer Str.38, D-01187 Dresden, Germany

50: \end{center}

51: 

52: 

53: \vspace{2cm}

54: \noindent{\bf Abstract:}

55: We introduce a new model for the dynamics of

56: centroblasts and centrocytes in a germinal center. 

57: The model reduces the germinal center reaction to the elements

58: considered as essential and

59: embeds proliferation of centroblasts, point mutations of the

60: corresponding antibody types 

61: represented in a shape space, differentiation to centrocytes,

62: selection with respect to initial antigens, differentiation of positively

63: selected centrocytes to plasma or memory cells and recycling of

64: centrocytes to centroblasts. We use exclusively

65: parameters with a direct biological interpretation such that,

66: once determined by experimental data, 

67: the model gains predictive power. 

68: Based on the experiment

69: of Han et al.\,(1995b) we predict that a high rate of recycling 

70: of centrocytes to centroblasts is necessary for the 

71: germinal center reaction to work reliably. 

72: Furthermore, we find a delayed start of the

73: production of plasma and memory cells with respect to the start

74: of point mutations, which turns out to be necessary for 

75: the optimization process during the germinal center reaction.

76: The dependence of the germinal center 

77: reaction on the recycling probability is analyzed.

78: \vspace*{\fill}

79: \eject

80: \newpage

81: 

82: \section{Introduction}

83: 

84: Germinal centers (GCs) are essential

85: elements of the humoral immune response

86: (for a review see MacLennan, 1994). In a primary

87: immunization some B-cells respond to the presented antigen

88: and are activated. These active cells migrate to follicular

89: dendritic cells (FDCs) present in lymphoid tissues. 

90: Unprocessed fragments of the antigen, so called epitopes, 

91: are expressed on the FDCs.

92: In the milieu of the FDCs the B-cells are in a phase of

93: intense proliferation and show the essential features of

94: B-blasts such that they are denoted as centroblasts. These

95: proliferating centroblasts together with the FDCs develop

96: to a GC. 

97: 

98: This morphological development is

99: initiated through an unidentified differentiation signal

100: after about three days. The resulting GC is

101: characterized by two zones (Nossal, 1991): 

102: % \question Eventuell noch Mac90 zitieren \question

103: the dark zone filled with 

104: continuously proliferating and mutating centroblasts 

105: and the light zone

106: containing the FDCs and centrocytes. The latter are

107: generated from the centroblasts (Liu et al., 1991), 

108: which are available in great diversity (Weigert et al., 1970)

109: because of  a very high 

110: mutation rate (Nossal, 1991; Berek \& Milstein, 1987).

111: % \question Eventuell noch Koc89 zitieren \question

112: In contrast to

113: centroblasts the centrocytes express antibodies 

114: of cell specific types on their surface.

115: In this way

116: they are able to undergo a specific selection process 

117: through the interaction with the epitopes presented on the

118: FDCs in the light zone. The diversity of the proliferating

119: and mutating centroblasts together with the specificity of 

120: the selection process in the light zone allow an 

121: optimization of the affinity between the antibodies

122: and the epitopes. This affinity maturation in the GC leads to a

123: dominant population of cells with high affinity of the

124: corresponding antibodies to the epitopes within 7-8 days

125: after immunization (Jacob et al., 1993). 

126: 

127: Centrocytes undergo

128: apoptosis if they are not positively selected in an appropriate 

129: range of time. They are

130: rescued from apoptosis (Liu et al., 1989; Lindhout et al., 1993;

131: Lindhout et al., 1995; Choe \& Choi, 1998)

132: if they are able to bind to the

133: antigen on the FDCs {\it and} to interact successfully

134: with T-helper cells present in the external

135: region of the light zone. Even antigen independent signals

136: may be required for the centrocytes to survive (Fischer et al., 1998).

137: Successfully selected centrocytes

138: receive a differentiation signal which determines their

139: fate. They may differentiate into memory cells that are

140: important in the case of a second immunization.

141: Alternatively, they differentiate to antibody producing

142: plasma cells. These are intensifying the immune response

143: in two ways: the number of antibodies increases and their

144: specificity with respect to the direct unspecific

145: immune response without GC reaction is optimized.

146: In both cases the corresponding cell is leaving the

147: GC environment. 

148: 

149: A third possibility exists for the

150: positively selected centrocytes: they may be

151: recycled to centroblasts and reenter the highly proliferating

152: and mutating stage of development in the dark zone 

153: (Kepler \& Perelson, 1993). 

154: This recycling hypothesis has been neither directly checked

155: experimentally (for an indirect check see Han et al., 1995b)

156: nor -- if true -- the probability 

157: of recycling is known. It would be of great interest to

158: know about the necessity of a centrocyte recycling

159: process for the affinity maturation. The

160: most evident argument to support the recycling hypothesis is

161: based on the occurrence of multi-step somatic hypermutations

162: in the GC reaction. Under the assumption of random

163: hypermutations, recycling of already mutated cells to

164: fast proliferating centroblasts appears to be a necessary

165: process.

166: 

167: In this paper we estimate how large the

168: recycling probability should be, i.e.\,which proportion of

169: the positively selected centrocytes

170: should be recycled to get a reliable GC reaction.

171: For this purpose we introduce a simplified model of the

172: GC dynamics based on a functional view on it.

173: The model differs from

174: others (Oprea \& Perelson, 1996; Oprea \& Perelson, 1997;

175: Rundell et al., 1998) by omitting some

176: details of the GC reaction 

177: which we consider as unnecessary to understand

178: the obligatory occurrence of the recycling procedure. 

179: On the other hand compared to the model introduced in

180: (Kesmir \& de Boer, 1999) we added a formalism to treat

181: somatic hypermutation in a shape space, which seems

182: to be essential to understand the process of affinity

183: maturation.

184: A very important difference to related models is that

185: we embed the experiment (Han et al., 1995b) for the first time.

186: This experiment 

187: provides an indirect evidence for the existence of the recycling

188: process and we will interpret it in a quantitative way in order to

189: extract information on the recycling probability from it.

190: We use exclusively parameters with a direct biological

191: interpretation to ensure the predictive

192: power of the model. 

193: 

194: In the next section we will define the shape space, in which the

195: cell population dynamics takes place, and the corresponding

196: dynamical quantities and parameters 

197: including their dynamical interdependence.

198: All parameters introduced in the model

199: are determined by experimental observations.

200: The third section is dedicated to the presentation and

201: interpretation of the model

202: in comparison with experimental GC reactions. 

203: We will analyze the stability of our results to verify the

204: notion of a prediction. Finally, we investigate the dependence

205: of some characteristics of the germinal center reaction on the

206: variation of the recycling probability and give some concluding

207: remarks.

208: 

209: 

210: \section{Definition of the model}

211: 

212: \subsection{Shape space and affinity space}

213: 

214: \paragraph{Postulates}

215: Before defining a model for the GC dynamics it is

216: necessary to specify the space in which the dynamics takes

217: place. As we want to represent cells corresponding to

218: different types of antibodies we use the well known

219: shape space concept (Perelson \& Oster, 1979).

220: This ansatz is based on the following assumptions:

221: \begin{itemize}

222: \item Antibodies can be represented in a phenotypical shape space.

223: \item Mutation can be represented in the shape space by next-neighbor jumps.

224: \item Complementarity of antibodies and antigen.

225: \item Homogeneity of the affinity weight function on the shape space.

226: \end{itemize}

227: 

228: 

229: \paragraph{Representation of antibody phenotype}

230: The shape space is taken as a $D$-dimensional finite size 

231: lattice of discrete equidistant points,

232: each of them representing one specific antibody type.

233: Thus the shape space becomes a phenotype space, i.e.\,it is

234: not primarily a representation of genetic codes but of the

235: resulting features due to specific genetic codes.

236: 

237: \paragraph{Representation of mutation}

238: Nevertheless, we want to represent point mutations defined

239: in an unknown genotype space in the shape space. To this

240: end we are compelled to formulate the action of a point

241: mutation in the shape space:\\

242: {\it A point mutation of a given antibody is represented

243: in the phenotypical shape space by a jump to one of the

244: nearest neighbor points.}\\

245: This assumption is not very spectacular as it only requires

246: that a point mutation does not lead to a random 

247: jump of the antibody phenotype, i.e.\,that conformation, electrical properties,

248: etc. are not dramatically changed. Surely, the mutation of 

249: some key base pairs may exist which imply fundamental changes

250: of the encoded antibody features. However, 

251: these exceptions will not alter the dynamics of the GC

252: until the number of such key base pairs remains small with respect

253: to the number of {\it smooth} mutation points.

254: 

255: \paragraph{Antigen representation and complementarity}

256: A central quantity is the affinity of a given antibody and an antigen epitope.

257: Having a representation of antibodies in the shape space, a counterpart 

258: for antigens is necessary. 

259: We emphasize that the number of possible antibodies is finite, whereas the

260: diversity of antigens is principally unbounded. Therefore, 

261: we represent an antigen on the B-cell shape space at the position of

262: the B-cell with highest affinity to the antigen, i.e.\,at the position of

263: an antibody with optimal complementarity to the antigen. 

264: 

265: \paragraph{Property of smooth affinity}

266: Starting from these assumptions (phenotype shape space, 

267: representation of mutation by next-neighbor jumps, and

268: complementarity) we have reached a representation

269: of antibodies and antigens in a shape space,

270: which has the property of {\it affinity neighborhood}, i.e.\,that

271: neighbor points in the shape space have comparable affinity to

272: a given antigen. This property is a direct consequence of the

273: definition of mutations in a (phenotype) shape space. 

274: Starting from a 

275: non-directed mutation as base for affinity maturation in GCs,

276: we estimate the affinity neighborhood as a necessary

277: property of the underlying shape space. An initial B-cell with a

278: particular affinity to a certain antigen must have the possibility of

279: successively optimizing the affinity, i.e.\,of stepwise climbing

280: an {\it affinity hill}. If no affinity neighborhood would exist, 

281: i.e.\,if there was a random affinity distribution on the shape space,

282: each mutation would lead to a random jump in the affinity.

283: An optimization of the affinity to the antigen may occur

284: accidentally in this scenario, but should be a very rare event.

285: 

286: \paragraph{Homogeneous affinity weight function}

287: Affinity neighborhood allows for the definition of an

288: {\it affinity weight function} which determines the affinity

289: of antigen and antibody in dependence of their distance 

290: in the shape space. We assume this weight

291: function $a(\phi,\phi^*)$

292: to apply equally well in all regions

293: of the shape space which corresponds to a homogeneous

294: affinity distribution over the shape space, and to be of

295: exponential form:

296: \be{affinity}

297: a(\phi,\phi^*) \;=\; \exp\left(-\frac{||\phi-\phi^*||^\eta}{\Gamma^\eta}\right)

298: \ee

299: where $\Gamma$ is the width of the affinity weight function

300: and $||\cdot ||$ denotes the Euclidean metric.

301: The question remains with which exponent $\eta$ the distance enters

302: the exponential weight function and it will be argued that $\eta=2$ is

303: a reasonable value (see \abb{Gamma}).

304: 

305: 

306: \subsection{Formulation of the dynamics}

307: \label{dynamics}

308: 

309: Having a shape space at hand it is possible to define distributions

310: $B(\phi)$, $C(\phi)$, $A(\phi)$, and $O(\phi)$

311: of centroblasts, of centrocytes,

312: of the presented antigen epitopes, and of the

313: plasma and memory cells on the shape space

314: (see Tab.~\ref{variables}).

315: We focus on the centroblast distribution $B$

316: and analyze its dynamical behavior in the different phases

317: of the GC reaction.

318: %

319: \begin{table}[ht]

320: \begin{center}

321: \begin{tabular}{|l|c|l|} \hline

322: Cell type & Variable & Processes \\ \hline

323: antigen & $A(\phi)$ & interaction with centrocytes \\

324: centroblasts & $B(\phi)$ & proliferation\\

325: && mutation\\

326: && differentiation to centrocytes\\

327: centrocytes & $C(\phi)$ & selection and apoptosis\\

328: && recycling to centroblasts\\

329: && differentiation to plasma and memory cells\\

330: plasma and memory cells & $O(\phi)$ & removed from GC\\

331: \hline

332: \end{tabular}

333: \caption[]{\sf The variables in the mathematical model of 

334: the GC reaction and the processes that are considered.}

335: \label{variables}

336: \end{center}

337: \end{table}

338: %

339: \begin{figure}[ht]

340: \includegraphics[height=15cm,angle=-90]{phases.eps}

341: %\center{\hspace*{0mm} \epsfxsize=16cm \epsfysize9cm

342: %            \epsfbox{phases.eps}}

343: \caption[]{\sf 

344: The three model phases of the germinal center reaction: 

345: (I) proliferation phase (starting at $t_0=-3\,d$), 

346: (II) optimization phase (starting at $t_1=0\,d$, the duration

347: $\Delta t_2$ is determined in Sec.~\ref{outputdyn}), 

348: (III) output phase (ending at $t_3=18\,d$). 

349: The parameters in brackets

350: refer to the model equations \gl{pmodell}, 

351: \gl{modell}, \gl{centrocytes}, and \gl{output}

352: and can be determined from experimental data 

353: (see Sec.~\ref{dynamics}-\ref{outputdyn}).}

354: \label{phases}

355: \end{figure}

356: %

357: 

358: There were several attempts to divide the germinal center

359: reaction in different working phases. From the

360: point of view of our model we are led to a new functional

361: phase distinction of the GC reaction (see Fig.\,\ref{phases}).

362: Our three-phase description of the GC reaction is

363: in accordance with most of the models established so far

364: (see e.g.\,MacLennan, 1994; Liu et al., 1991; Kelsoe, 1996) but

365: is in contradiction to the time scales found in 

366: (Camacho et al., 1998), which differ strongly.

367: It is not intended to find mechanisms of transition between the phases 

368: of the germinal center reaction. These dynamical phases are

369: assumed according to experimental observations.

370: 

371: The first phase is the already mentioned phase of fast

372: proliferation of a low number of seeder cells in the

373: environment of FDCs. In this phase it is likely that 

374: somatic hypermutation is not taking place to a relevant

375: amount (Han et al., 1995b; Jacob et al., 1993; 

376: McHeyzer-Williams et al., 1993; Pascual et al., 1994b)

377: such that it may be understood as an enlargement phase 

378: of the cell pool

379: available for the following process of affinity maturation.

380: The corresponding dynamical behavior of the centroblast

381: distribution is described as

382: %

383: \be{pmodell}

384: \frac{dB}{dt}(\phi) \;=\; p B(\phi) 

385: \quad {\rm for} \quad t-t_0<\Delta t_1

386: \quad,

387: \ee

388: %

389: where $p$ denotes the proliferation rate and $t_0$ is the

390: time of first immunization.

391: This phase lasts $\Delta t_1=3$ days and is well 

392: established by experiment (Liu et al., 1991).

393: 

394: After three days the optimization phase starts. 

395: The GC gets its morphological form, i.e.\,dark and light zones emerge.

396: In the dark zone centroblasts continue to proliferate but,

397: additionally, somatic hypermutations are broadening

398: the initial centroblast distribution on the shape space.

399: At the same time the selection process operates

400: in the light zone and the selected centrocytes are either recycled

401: to centroblasts or differentiate to plasma and memory cells.

402: The development of the centroblast distribution

403: on the shape space is now described by

404: %

405: \bea{modell}

406: \frac{dB}{dt}(\phi) &=&

407: (p - 2pm -g) B(\phi) 

408: + \frac{pm}{D} \sum_{||\Delta\phi||=1} B(\phi+\Delta\phi)

409: \nn

410: && + (1-q) \sum_{\phi^*} C(\phi) A(\phi^*) \,a_0 \,a(\phi,\phi^*)

411: \quad,

412: \eea

413: %

414: for $t-t_0\ge\Delta t_1$.

415: Here, $g$ is the rate of centroblast differentiation

416: into centrocytes. $m$ is the mutation probability, 

417: $a_0$ is the probability of an optimal

418: centrocyte to be positively selected,

419: $q$ is the probability

420: of differentiation into plasma and memory cells 

421: for positively selected centrocytes, 

422: and $D$ is the dimension of the underlying shape space.

423: The different contributions to \gl{modell} are

424: discussed in detail below.

425: Note that this equation does not include eventual finite size

426: effects due to small populations.

427: 

428: It is not clear a priori, if differentiation into antibody producing

429: plasma and memory cells is triggered already in this

430: second phase. To allow a start of the output cell production

431: delayed by the time interval $\Delta t_2$

432: we divide the optimization phase into two sub-phases

433: which differ in the output probability $q$:

434: \bea{phase2}

435: q&=&0 \quad {\rm for} \quad 0 \le t-\Delta t_1 < \Delta t_2 \nn

436: q&>&0 \quad {\rm for} \quad 0 \le t-\Delta t_2 < \Delta t_3

437: \quad.

438: \eea

439: The time delay $\Delta t_2$ will be fixed by experimental

440: data (see \kap{outputdyn}).

441: The output phase include optimization and production of

442: plasma and memory cells and lasts for the remaining

443: GC life-time, which is about

444: $\Delta t_3=21{\rm days}-3{\rm days}-\Delta t_2$

445: (Choe \& Choi, 1998; Kelsoe, 1996; Jacob et al., 1991).

446: Then, only a few proliferating B-blasts remain in the

447: environment of the FDCs (Liu et al., 1991).

448: 

449: 

450: 

451: \subsubsection*{Proliferation}

452: 

453: During the whole GC reaction a fast proliferation

454: of B-cells takes place. After the activation of B-cells by an

455: interaction with an antigen, these move to FDCs and undergo

456: a fast proliferation phase in their environment 

457: (Liu et al., 1991; Hanna, 1964; Zhang et al., 1988; Grouard et al., 1995). 

458: It is likely that the fast proliferation is triggered by the

459: dendritic cells (Dubois et al., 1999).

460: After about three days (Liu et al., 1991)

461: when the GC starts to develop its characteristic light and

462: dark zone (Camacho et al., 1998),

463: a fast proliferation of centroblasts is still observed in the dark zone.

464: The extremely high rate of 

465: proliferation could be determined 

466: to be (Liu et al., 1991; Zhang et al., 1988)

467: \be{p}

468: \frac{p}{\ln(2)}\;\approx\; \frac{1}{6\,h}

469: \quad,

470: \ee 

471: a result known already for a long time (Hanna, 1964). 

472: The population of centroblasts at $\phi$ grows exponentially, 

473: which is represented

474: by the term $pB(\phi)$ in \gl{pmodell} and \gl{modell}.

475: 

476: 

477: \subsubsection*{Mutation}

478: 

479: It is likely that somatic hypermutation does not occur in the

480: proliferation phase of the GC reaction 

481: (Han et al., 1995b; Jacob et al., 1993; 

482: McHeyzer-Williams et al., 1993; Pascual et al., 1994b).

483: On the other hand the growth of the centroblast population is reduced 

484: by possible mutations from state $\phi$ to its neighbors

485: $\phi+\Delta\phi$, where $||\Delta\phi||=1$. 

486: In a continuous space this corresponds to a diffusion process

487: as used in (Perelson \& Wiegel, 1999) to represent mutation.

488: Each pair

489: of cells produced by proliferation will populate a neighbor

490: state with a probability of $m$. This mutation probability turns out

491: to take extremely high values of $m\approx 1/2$ 

492: (Nossal, 1991; Berek \& Milstein, 1987),

493: % \question Eventuell noch Koc89 zitieren \question

494: which corresponds

495: to a factor $10^7$ larger probability compared to mutations outside

496: the GCs (Janeway \& Travers, 1997). We want to point out that here only

497: point mutations of phenotypical relevance are taken into account.

498: As a consequence the population

499: growth by proliferation is reduced by the important

500: amount $2pm$, which results in an effective proliferation rate 

501: at $\phi$ of $p(1-2m)$.

502: One observes that for $m=1/2$ the two first terms in \gl{modell}

503: cancel. This corresponds to the situation that in each centroblast

504: replication one new cell remains in the old state and the second new cell

505: mutates to a neighbor state such that the total number of cells in

506: state $\phi$ remains unaltered.

507: 

508: In the same way as mutation depopulates the state $\phi$ in the 

509: shape space and populates its neighbors, $\phi$ is populated by

510: its neighbors. This gives rise to the non-diagonal term in \gl{modell}.

511: Each neighbor of $\phi$ mutates with the same rate $2pm$ from 

512: $\phi+\Delta\phi$ to one of its $2D$ neighbors. 

513: Only the mutation from $\phi+\Delta\phi$ to the particular neighbor

514: $\phi$ enhances the population of centroblasts at $\phi$

515: such that the rate is weighted by the inverse number of neighbors

516: $1/(2D)$ and is summed over all possible neighbors of $\phi$.

517: 

518: \subsubsection*{Selection, recycling and cell production}

519: 

520: After the establishment of the light and dark zones in

521: the GC the differentiation of centroblasts to

522: antibody presenting centrocytes is triggered to allow

523: a selection process in the light zone. The centroblasts

524: in the shape space state $\phi$ are diminished 

525: with the rate of differentiation $g$, 

526: leading to the term $-gB(\phi)$ in \gl{modell}.

527: The centrocyte population $C(\phi)$ of type $\phi$ is enhanced

528: simultaneously by the same amount of cells:

529: \be{centrocytes}

530: C(\phi) \;=\; +gB(\phi)

531: \quad.

532: \ee

533: The centrocytes move to the light zone where they

534: undergo a selection process. 

535: Their further development is splitted

536: three-fold. The non-selected centrocytes die through apoptosis

537: and are eliminated from the GC dynamics. 

538: It is known for a long time that the centrocytes

539: undergoing apoptosis were in cell cycle

540: a few hours ago (Fliedner, 1967) such that it is likely that they

541: differentiated from the centroblasts. On the other hand

542: apoptosis takes place in the light zone

543: (Hardie et al., 1993) where centroblasts are not present

544: in high density.

545: The selected centrocytes are emitted from the GC

546: with probability $q$ and differentiate either into

547: antibody producing plasma cells or to memory cells.

548: The model does not distinguish between plasma and

549: memory cells. Only their sum is taken into account

550: and is denoted by {\it output cells} $O(\phi)$.

551: Nevertheless, the dynamics of both types of

552: output cells may be different (Choe \& Choi, 1998).

553: Also, the degree of affinity maturation differs (Smith et al., 1997).

554: In addition, we do not consider any further proliferation

555: of output cells in or outside of the GC, 

556: which may be important for a quantitative

557: comparison with experimental measurements.

558: 

559: Alternatively the selected

560: centrocytes are recycled to become centroblasts and

561: to reenter the proliferation process in the dark zone.

562: This happens with probability

563: $1-q$ and contributes to the centroblast distribution, 

564: corresponding to the last term in \gl{modell}.

565: 

566: We want to emphasize that we do not resolve the

567: time course of the selection process, which is regarded

568: to be fast with respect to the centroblast proliferation

569: time scale. This procedure is justified

570: by a minimal duration of a typical selection process

571: of $1-2$ hours (van Eijk \& de Groot, 1999), 

572: which is about one fourth of the

573: proliferation time scale. Nevertheless, one should be

574: aware of possible implications due to the fact that

575: the selection process does not occur instantaneously.

576: We effectively enclose the selection time in the parameter

577: $g$, which in this way

578: becomes a measure of a complete selection

579: cycle including the differentiation into centrocytes, a first

580: selection at the FDCs, a second selection at T-helper

581: cells and finally the recycling process.

582: This has two consequences for the model: 

583: The number of selectable centrocytes for each shape

584: space state $\phi$ is given at every time by \gl{centrocytes},

585: i.e.\,the number of centrocytes just being in the selection

586: process. Furthermore, the details of the multi-step

587: selection process (Lindhout et al., 1997) are omitted. 

588: The selection is modeled by a sum over the shape space of

589: the antigen distribution presented on the FDCs

590: weighted by the affinity function \gl{affinity}. In other

591: words the centrocytes $gB(\phi)$ are selected if an 

592: antigen is close enough in the shape space.

593: The meaning of {\it close enough} is determined by the

594: width of the affinity function and by the probability $a_0$

595: of a positive selection for an optimal centrocyte with

596: respect to one presented antigen. 

597: 

598: One should be aware

599: that the probability $a_0$ is not necessarily equal to one, 

600: as centrocytes are in the state of activated apoptosis (Cohen et al., 1992)

601: such that their life time is finite and they have to be selected within

602: this life time to be rescued from apoptosis. This maximum

603: selection probability is determined by the relation of

604: the centrocyte life time of about $10\,h$ 

605: (Liu et al., 1989, 1994)

606: and the duration of the selection process of $1-2\,h$

607: (van Eijk \& de Groot, 1999).

608: A rough estimate using two Gaussian distributions with

609: reasonable widths for the life time and the selection time

610: peaked at $10\,h$ and $1.5\,h$ respectively leads to

611: $a_0\approx 0.95$. Thus, if a GC contains only

612: centrocytes of optimal complementarity to a presented

613: antigen, still about $5\%$ of them will undergo apoptosis.

614: 

615: In conclusion, selection can be described by

616: the convolution term

617: \be{selection}

618: \sum_{\phi^*} C(\phi) A(\phi^*) \,a_0 \,a(\phi,\phi^*)

619: \quad,

620: \ee

621: where the number of selectable centrocytes is given by

622: \gl{centrocytes} and $A(\phi^*)$ represents the distribution of

623: antigens on the shape space. This selection term

624: contributes with probability $q$ to the production of 

625: plasma and memory cells

626: \be{output}

627: \frac{dO}{dt}(\phi) \;=\;

628: q \,\sum_{\phi^*} C(\phi) A(\phi^*) \,a_0 \,a(\phi,\phi^*)

629: \quad.

630: \ee

631: All selected centrocytes not emitted from the GC

632: are recycled and in this way enhance the centroblast

633: population giving rise to the last term in \gl{modell}.

634: 

635: \subsection{Initial conditions}

636: \subsubsection*{Antigen distribution}

637: 

638: Throughout the paper the antigen distribution is assumed

639: unequal to zero at exactly one point in the

640: shape space:

641: \be{antigenini}

642: A(\phi^*) \;=\; A(\phi_0) \;=\; \delta(\phi^*-\phi_0)

643: \quad,

644: \ee

645: where $\delta(\cdot)$ is one for its argument equal 

646: to zero and zero otherwise. Consequently, 

647: the sum in the convolution term

648: in \gl{modell} is reduced to one contribution.

649: This implies that we consider only one type of antigen epitope

650: to be present in the GC.

651: What seems to be a restriction at first sight

652: turns out to be an assumption without any loss of generality.

653: In view of about $10^{11}$ possible states 

654: in the naive repertoire of humans (Berek \& Milstein, 1988) 

655: two randomly chosen antigen epitopes

656: will not be in direct neighborhood of each

657: others in the shape space.

658: Both antigen epitopes will only have an influence

659: on each other during the GC reaction when centroblasts

660: corresponding to one of the antigen epitopes have a non-negligible

661: affinity to the other epitope.

662: In the shape space language this situation corresponds to an 

663: overlap of two spheres centered at both antigen epitopes.

664: The radius of these spheres is determined by the maximum

665: number of mutations which may occur during a GC reaction.

666: This number will rarely exceed $9$ (K\"uppers et al., 1993;

667: Wedemayer et al., 1997) such

668: that in a $4$-dimensional shape space (a more detailed

669: discussion of the shape space dimension follows in 

670: Sec.~\ref{ssdim})

671: one sphere of this size covers about $10^{-8}$th

672: part of the space. An overlap is unlikely as

673: an unreasonable large number of $10^8$ antigen epitopes 

674: is needed to get an important probability for a mutual influence.

675: Therefore we consider one-antigen epitope GC reactions only. 

676: A multiple-antigen or a multiple-antigen epitope

677: GC reaction has to be considered as

678: sum of one-antigen epitope GCs.

679: 

680: There exists experimental evidence that the amount of presented antigens

681: is reduced during a GC reaction

682: (Tew \& Mandel, 1979). Since the amount of antigen

683: is only halved during $30$ days 

684: (Tew \& Mandel, 1979; Oprea et al., 2000)

685: we do not expect an important recoil effect on the

686: selection probability, which would become time dependent

687: otherwise. Furthermore, we believe that this antigen diminution

688: is too weak to be responsible for the vanishing cell

689: population in GCs after about $3$-$4$ weeks. Anyhow, 

690: we aim to show that our three-phase model allows for

691: a vanishing population solely due to the dynamics of the system, 

692: without the inclusion of any antigen distribution dynamics.

693: 

694: 

695: \subsubsection*{B-cell distribution}

696: 

697: The B-cells which have moved to FDCs are in a fast proliferation

698: phase. In this first phase of GC development the

699: cell population grows exponentially until it has reached

700: about $10^4$ cells after three days. One may ask

701: how many seeder cells are necessary and sufficient

702: to give rise to this large cell population within such a short

703: space of time. One finds experimentally that the follicular

704: response is of oligoclonal character 

705: (Liu et al., 1991; Jacob et al., 1991; K\"uppers et al., 1993;

706: Kroese et al., 1987) and the number of

707: seeder cells is estimated between two and six.

708: This result is in accordance with the proliferation

709: rate of $\ln(2)/(6\,h)$ cited above.

710: 

711: The number of seeder cells $\sum_\phi B(\phi,t=0)$ 

712: determines the initial

713: B-cell distribution $B(\phi)$ in our model. Corresponding

714: to the experimental results cited above, we

715: start with the reasonable number of 

716: three seeder cells for the follicular response

717: to an immunization with a specific antigen.

718: These cells are chosen randomly in the shape space 

719: within a sphere of radius $R_0$ around the presented

720: antigen. Due to the assumed affinity neighborhood

721: this reflects a threshold affinity of antigen and

722: activated initial B-cells. This was 

723: already found experimentally (Agarwal et al., 1998):

724: The activation of a B-cell by

725: an antigen is necessarily connected with a minimum

726: affinity of the corresponding epitopes and

727: antibodies. The radius of the sphere can be

728: determined by the maximum number of mutations

729: occurring during the process of affinity maturation

730: in GCs: 

731: \be{r0}

732: R_0=N_{\rm max}

733: \quad,

734: \ee

735: which is found to be $8$ or $9$ 

736: (K\"uppers et al., 1993; Wedemayer et al., 1997).

737: In this picture, B-cells with an affinity above threshold

738: are activated by a specific antigen, giving

739: rise to a fast immune response with antibodies

740: of low but non-vanishing affinity. The

741: germinal center reaction has to be understood

742: as an optimization process for these initially activated

743: B-cells leading to a second slower primary immune 

744: response of high specificity.

745: 

746: \subsection{Width of affinity function}

747: 

748: Due to affinity neighborhood in

749: the shape space the affinity function may be chosen

750: according to \gl{affinity}. 

751: The width $\Gamma$ has to be consistent with the

752: affinity enhancement that is achieved during a GC 

753: optimization process.

754: Let the factor of affinity enhancement be $10^\xi$

755: and the number of corresponding somatic hypermutations

756: be $N$. Then from \gl{affinity} we get

757: \be{width}

758: \Gamma \;=\; \frac{N}{\left(\xi \ln(10)\right)^{\frac{1}{\eta}}}

759: \quad.

760: \ee

761: Typical numbers of somatic hypermutations are $N=6$

762: (K\"uppers et al., 1993) corresponding to an affinity enhancement

763: of $100$ ($\xi=2$) (Janeway \& Travers, 1997), while $N=8$ 

764: is considered to be a large number of

765: mutations (K\"uppers et al., 1993)

766: corresponding to a high factor of affinity enhancement 

767: of $2000$ ($\xi=3.3$) (Janeway \& Travers, 1997). A huge

768: affinity enhancement by a factor of $30000$ ($\xi=4.5$)

769: is reached with $N=9$ mutations (Wedemayer et al., 1997). 

770: As can be

771: seen in \abb{Gamma}, these values for the affinity

772: enhancement with growing number of somatic

773: hypermutations are consistent with a Gaussian affinity

774: function ($\eta=2$) with width of $\Gamma\approx 2.8$.

775: %

776: \begin{figure}[ht]

777: \vspace{-1cm}

778: \center{\hspace*{0mm} \epsfxsize=10cm \epsfysize10cm

779:         \epsfbox{2gamma.eps}}

780: \caption[]{\sf The width of the Gaussian affinity function 

781: as a function of the factor of affinity enhancement 

782: for different numbers of somatic hypermutations.

783: Vertical lines denote the positions of the used

784: experimental values of affinity enhancement. One

785: can observe that consistently for all three values the resulting

786: width becomes approximately $2.8$.}

787: \label{Gamma}

788: \end{figure}

789: %

790: It should be mentioned that a consistent value

791: for the width of the affinity function can not be

792: obtained for other integer values of $\eta$.

793: But one should be aware that the used

794: affinity enhancements are very rough estimates and

795: thus our result should be understood as a first

796: guess -- even if the compatibility with the experimental

797: data is convincing.

798: 

799: 

800: \subsection{Shape space dimension}

801: \label{ssdim}

802: 

803: The shape space dimension $D$ enters the model through

804: the mutation term in \gl{modell} and has to be chosen

805: appropriately to our model. 

806: We already mentioned that somatic hypermutation is likely to

807: occur randomly (Weigert et al., 1970; Radmacher et al., 1998), 

808: i.e.\,that the direction of mutation in the shape space

809: is not governed by the position of the antigen.

810: The term describing mutation in \gl{modell} is in

811: analogy to a diffusion term in a continuous space

812: as used in (Perelson \& Wiegel, 1999). 

813: Let us assume for a moment that the

814: development of the GC reaction is governed 

815: by a diffusion process

816: and let the number of maximum point mutations

817: be $N_{\rm max}$. Then the diffusion process

818: operates in a sphere around a seeder cell

819: of radius $N_{\rm max}$

820: in the shape space. In order to match the position

821: of the antigen with at least one centroblast, all positions

822: in the sphere have to be reached by diffusion.

823: This means that the total number of

824: centroblasts $N_c$ present in the GC $3$ days after immunization, 

825: i.e.\,when selection is triggered, should be of the

826: order of magnitude of the number of points in the sphere.

827: The number of points in the sphere grows exponentially

828: with the dimension of the shape space.

829: Thus, for $N_{\rm max}=8$ and $N_c=12000$ it follows that

830: the shape space dimension should be $4$. From the point

831: of view of a pure diffusion process this is an upper bound for

832: the shape space dimension.

833: 

834: However, the above argument does not include

835: the proliferation rate, the selection rate, and the

836: recycling probability. 

837: The effective number of centroblasts 

838: available to reach the antigen may become larger than the

839: $12000$ cells assumed above through proliferation and

840: recycling. As we want to describe a decreasing population

841: for large times without further assumptions,

842: the effective rate of production of additional centroblasts

843: during the GC reaction cannot

844: exceed some small value, since the centroblast population would

845: explode otherwise for large times. So it seems unlikely that

846: the shape space dimension appropriate for our model is

847: substantially larger than $4$. We will start with

848: $D=4$ and check our results for the

849: dimensions $5$ and $6$.

850: We want to remind that we

851: are dealing with a phenotype space in this model and that

852: a genotype space probably requires a substantially

853: higher dimension.

854: 

855: 

856: 

857: \subsection{Long term behavior}

858: \label{gvalue}

859: 

860: The parameter $g$ was introduced in the model as the

861: rate of differentiation from centroblasts in the dark zone

862: into antibody presenting centrocytes ready to

863: undergo a selection process. 

864: As we do not consider

865: the centrocyte dynamics during the multi-step 

866: selection process (Lindhout et al., 1997),

867: this parameter effectively describes the total duration

868: of the selection process including the differentiation

869: mentioned above, the selection at the FDCs and

870: with T-helper cells, and the recycling to centroblasts.

871: Only those cells that have successfully finished this program

872: are contributing to the centroblast population described

873: in \gl{modell}.

874: 

875: Thus, being an effective measure of the speed of the

876: selection process, we are able to deduce an upper

877: bound for $g$. Solely the inhibition of apoptosis

878: during the selection process at FDCs and

879: with T-helper cells takes at least $1$-$2$ hours 

880: (Lindhout et al., 1995; van Eijk \& de Groot, 1999).

881: This gives us an upper bound for the rate $g$ of

882: \be{gupper}

883: \frac{g}{\ln(2)} \;<\; \frac{1}{2\,h}

884: \quad.

885: \ee

886: Nevertheless, this upper bound is not sufficient to

887: fix $g$. As $g$ governs the whole reaction speed it will

888: play a crucial role for the final state $21$ days after 

889: immunization. As we know that at the end of the GC

890: reaction only a few cells remain in the environment of 

891: the FDCs (Liu et al., 1991; Kelsoe, 1996),

892: the parameter $g$ will be tuned in order to get this result.

893: 

894: Mathematically, this requirement has the form of a

895: long term condition. Let us look for an equilibrium

896: condition at the point $\phi_0$

897: representing the antibody type of optimal complementarity to

898: a presented antigen:

899: \be{equilibrium}

900: \frac{dB}{dt}\left(\phi=\phi_0,t\right)\;=\;0

901: \quad.

902: \ee

903: Then, the selection term in \gl{modell}

904: becomes simply $(1-q)ga_0 B(\phi_0,t)$ and

905: for a non-vanishing centroblast distribution 

906: at $\phi_0$ we get

907: \be{equilcondition}

908: \eps\;\equiv\;

909: p-2pm-g+\frac{pm\beta}{D}+(1-q)ga_0\;=\;0

910: \quad,

911: \ee

912: where we have introduced $\eps$ as a measure for the

913: vicinity to the equilibrium condition and

914: \be{betadef}

915: \beta(\phi_0,t) \;=\; 

916: \frac{\sum_{||\Delta\phi||=1} B(\phi_0+\Delta\phi,t)}

917:        {B(\phi_0,t)}

918: \quad.

919: \ee

920: $\beta(t)$ is a measure for the sharpness of the centroblast

921: distribution in the shape space around the antigen position

922: $\phi_0$. 

923: $\beta$ becomes constant when the

924: selection process starting at day $3$ after immunization

925: has reached a phase in which the

926: centroblasts at the position of the antigen dominate.

927: This is already the case $8$ days after immunization

928: according to experiments (Jacob et al., 1993) and

929: is confirmed in our model (see \abb{beta46}).

930: 

931: Thus, at large times $\beta$ can already be considered as

932: constant in the equilibrium condition \gl{equilcondition} 

933: and the same applies to $\eps$. 

934: In this stage of GC development

935: the centroblast population at $\phi_0$ has a pure 

936: exponential form

937: \be{Bexp}

938: B(\phi_0,t) \;=\; e^{\eps t}

939: \quad.

940: \ee

941: The sign of $\eps$ determines whether the centroblasts

942: population explodes ($\eps>0$) or dies out

943: ($\eps<0$). 

944: Because of the finite life times of GCs, $\eps$

945: should adopt values slightly smaller than zero. 

946: 

947: 

948: \subsection{Recycling probability}

949: 

950: The recycling probability $1-q$ determines the fate of the

951: positively selected centrocytes in the light zone. Apoptosis

952: was inhibited for these cells and they may either become

953: plasma or memory cells or return to the dark zone to reenter

954: the fast proliferation phase. This recycling hypothesis 

955: (Kepler \& Perelson, 1993)

956: has been intensively discussed and the main position is

957: that random somatic hypermutation

958: -- and it is likely that somatic hypermutation 

959: occurs randomly 

960: (Weigert et al., 1970; Radmacher et al., 1998), 

961: i.e.\,that the direction

962: of mutation in the shape space does not depend on the

963: position of the antigen

964: -- does not lead to a sustained optimization of affinity

965: in a one-pass GC reaction (Oprea et al., 2000).

966: %!!! \question weitere Zitate darin \question

967: To reach a specific

968: position in the shape space a multi-step optimization and 

969: verification process is required to avoid an arbitrary aimless

970: walk through this high dimensional space.

971: 

972: To propose a new perspective onto this question we consider the most

973: convincing experimental evidence for the existence of such a

974: recycling process (Han et al., 1995b). A GC reaction was

975: initiated in mice with one antigen. $9$ days after immunization, 

976: i.e.\,in a stage

977: of the reaction in which the {\it good} cells already dominate

978: (Jacob et al., 1993), a second

979: different, but related antigen was injected. 

980: This change of the reaction conditions 

981: in the course of the GC reaction leads to some very interesting

982: observations: The rate of apoptosis of centrocytes is enhanced by a factor

983: of about $\nu\approx 5$ compared to the non-disturbed GC reaction.

984: This can be explained by the fact that the antigen distribution 

985: in the shape space

986: was changed such that most of the cells present

987: after $9$ days may fit to the first antigen but not to the

988: second one. The probability of a positive selection is diminished and

989: the inhibition of apoptosis does not occur to the same extent

990: as before. On the other hand the total cell population in

991: the GC vanishes with a $\omega\approx 8$-fold higher 

992: speed compared to the unperturbed GC reaction. 

993: This is very difficult to explain without

994: recycling: As we know that the centrocytes (only very small

995: numbers of centroblasts) undergo apoptosis

996: if not rescued in time, the change of the apoptosis rate

997: is of relevance only for the centrocyte population. The faster

998: reduction of the whole GC cell population can not be explained

999: by the enhanced apoptosis rate, if there exists no recoil effect

1000: of the centrocyte apoptosis to the centroblast population.

1001: This is an indirect confirmation of the recycling hypothesis.

1002: 

1003: We want to consider this argument in

1004: a quantitative way by translating this experiment

1005: into the language of our model. We know from our

1006: consideration in Section \ref{gvalue} that

1007: at day $9$ after immunization the optimization process

1008: is already completed and that the centroblast population

1009: at the antigen position $\phi=\phi_0$

1010: in the shape space behaves according to \gl{Bexp}.

1011: But due to the modification of the antigen distribution

1012: at day $9$ after immunization the GC reaction returns 

1013: back to its dynamical

1014: selection phase. Thus, until day $9$ the centroblast

1015: population $B_1(\phi,t)$ evolves with respect to

1016: the antigen distribution $A_1(\phi^*)$ 

1017: as in \gl{antigenini}

1018: and after day $9$ two GC developments have to

1019: be compared: One continues its usual reaction

1020: as before. The second $B_2(\phi,t)$ 

1021: starts with the distribution

1022: $B_1(\phi,t=9\,d)$ and evolves with respect to

1023: the new antigen distribution

1024: \be{anti2}

1025: A_2(\phi^*)\;=\;

1026: \rho_1 \delta(\phi^*-\phi_0)

1027: +\rho_2 \delta(\phi^*-(\phi_0+\Delta\phi^*))

1028: \quad,

1029: \ee

1030: where $\Delta\phi^*$ denotes the shift of the additional

1031: second antigen with respect to the first one. $\rho_{1,2}$

1032: take into consideration the (possibly) different concentrations 

1033: of both antigens.

1034: 

1035: Now we are able to calculate the apoptosis rate in both

1036: scenarios for the cells of some fixed but arbitrary type $\phi$. 

1037: The number of centrocytes available for

1038: selection at every moment is $gB(\phi)$. The apoptosis

1039: rate is given by this number reduced by the selected

1040: cells

1041: \be{apoprate}

1042: V_i(\phi) \;=\; 

1043: gB(\phi)

1044: \left(1-a_0\sum_{\phi^*} A_i(\phi^*) a(\phi-\phi^*)\right)

1045: \quad,

1046: \ee

1047: where $i$ denotes the two scenarios. For the original

1048: GC reaction this becomes

1049: \be{apopeins}

1050: V_1(\phi)

1051: \;=\;

1052: gB(\phi) \left(1-a_0 a(\phi-\phi_0)\right)

1053: \quad,

1054: \ee

1055: while in the GC with the new antigen distribution

1056: we find

1057: \be{apopzwei}

1058: V_2(\phi)

1059: \;=\;

1060: gB(\phi)

1061: \left(1-a_0 a(\phi-\phi_0) \rho_1

1062:           \left(1+ \alpha(\phi)\right)

1063: \right)

1064: \quad,

1065: \ee

1066: where we have shifted $\phi^*$ in the second term

1067: of \gl{anti2} and introduced the ratio

1068: \be{alpha}

1069: \alpha(\phi) \;=\; 

1070: \frac{\rho_2 a(\phi-\phi_0-\Delta\phi^*)}

1071: {\rho_1 a(\phi-\phi_0)}

1072: \ee

1073: valid for each point $\phi$ separately.

1074: The factor $\nu$ of the apoptosis enhancement found

1075: in the experiment above is then given by the ratio of

1076: both apoptosis rates

1077: \be{nu}

1078: \nu \;=\;

1079: \frac{V_2}{V_1} \;=\;

1080: \frac{1-a_0 a(\phi-\phi_0) \rho_1

1081:             \left(1+ \alpha(\phi)\right)}

1082: {1-a_0 a(\phi-\phi_0)}

1083: \quad.

1084: \ee

1085: 

1086: On the other hand, this enhancement of apoptosis leads

1087: to a faster population reduction by a factor $\omega$,

1088: which in our model is given by the ratio of the changes

1089: of the cell distributions over the shape space

1090: at the time of the presentation of the new antigen.

1091: According to \gl{modell} and using the new

1092: antigen distribution $A_2$ this ratio is given by

1093: \bea{omega}

1094: \omega 

1095: &=&

1096: \frac{dB_2}{dt}

1097: \left[\frac{dB_1}{dt}\right]^{-1}(\phi,t=9\,d)

1098: \nn

1099: &=&

1100: \frac{p-2pm-g+\frac{pm}{D}\,\beta(\phi)

1101:        +(1-q)ga_0 a(\phi-\phi_0)\,

1102:           \rho_1 \left(1+\alpha(\phi)\right)}

1103: {p-2pm-g+\frac{pm}{D}\,\beta(\phi)

1104:        +(1-q)ga_0 a(\phi-\phi_0)}

1105: \quad,

1106: \eea

1107: where the cell distributions $B(\phi,t=9\,d)$ cancel as they

1108: are equal in both scenarios for $t=9\,d$.

1109: 

1110: Taking the results for $\nu$ (\gl{nu}) and for $\omega$ (\gl{omega})

1111: at the shape space point of

1112: the primary antigen $\phi=\phi_0$

1113: the affinity function becomes equal to one. $\beta$

1114: is already a constant for $t=9\,d$ such that we may

1115: eliminate $\rho_1(1+\alpha(\phi_0))$ from both equations

1116: to get

1117: \be{nutoomega}

1118: \frac{\nu-1}{\omega-1}

1119: \;=\;

1120: \frac{p-2pm-g+\frac{pm}{D}\,\beta(\phi_0)+(1-q)ga_0}

1121:       {(1-q)g(a_0-1)}

1122: \ee

1123: or for the recycling probability

1124: \be{recycle}

1125: 1-q \;=\;

1126: -\frac{p-2pm-g+\frac{pm}{D}\,\beta(\phi_0)}

1127: {g\left[a_0+(1-a_0)\frac{\nu-1}{\omega-1}\right]}

1128: \quad.

1129: \ee

1130: This relation allows us to determine the parameter $q$ and in

1131: this way to give a quantitative prediction for the recycling 

1132: probability for selected centrocytes to reenter the dark zone and to

1133: continue proliferation. 

1134: The recycling probability has to fulfill \gl{recycle}

1135: in order to be consistent with the experiment 

1136: of Han et al., (1995b).

1137: We want to emphasize, that the condition \gl{recycle} is independent

1138: of the concentrations of the two antigen types used in the

1139: experiment.

1140: 

1141: \subsection{Output dynamics}

1142: \label{outputdyn}

1143: 

1144: The production of plasma and memory cells optimized with

1145: respect to the presented antigen in

1146: the course of the GC reaction is governed by

1147: \gl{output}. According to our discussion of the different phases

1148: of the GC reaction in \kap{dynamics}, the starting time

1149: of {\it output cell} production is not necessarily

1150: correlated with the starting time of the mutation

1151: and selection process about $3$ days after immunization. 

1152: There may occur

1153: a time delay of the output production in relation to the

1154: selection process (see Fig.\,\ref{phases}). 

1155: This possibility is encountered by 

1156: comparing the dynamics of 

1157: output cell production with experiment 

1158: (Han et al., 1995a).

1159: Here, the number of output cells of high affinity to the presented

1160: antigen $\phi_0$ is compared at day $6$ and

1161: day $12$ after immunization and their relation is found

1162: to be approximately 

1163: %

1164: \be{out126}

1165: v_O

1166: \equiv

1167: \frac{\int_{0d}^{12d}dt\,O(\phi_0)}

1168:        {\int_{0d}^{6d}dt\,O(\phi_0)}

1169: \approx

1170: 6

1171: \quad.

1172: \ee

1173: As output cells are already present at day $6$ after immunization,

1174: the starting point of output cell production has to fulfill

1175: $0\,h<\Delta t_2<72\,h$ 

1176: ($\Delta t_2$ denotes the time window

1177: between the start of mutation and selection and the start

1178: of output cell production, see Fig.~\ref{phases}).

1179: It is intuitively clear that for large $\Delta t_2$ the speed

1180: of output cell production $v_O$ will be enhanced, for 

1181: the cell population of the GC will already be peaked at the 

1182: position of the antigen in the shape space. Thus, we 

1183: are able to adjust $\Delta t_2$ to the correct 

1184: output cell dynamics.

1185: 

1186: 

1187: 

1188: % =============================================

1189: 

1190: \section{Results and Predictions}

1191: 

1192: \begin{table}[ht]

1193: \begin{center}

1194: \begin{tabular}{|l|r@{$\;=\;$}l|} \hline

1195: Parameter & quantity & value \\ \hline

1196: Proliferation rate & $p/\ln(2)$ & $1/(6\,h)$ \\

1197: Somatic hypermutation probability & $m$ & $0.5$\\

1198: Rate of differentiation of centroblasts to centrocytes & $g/\ln(2)$ & 0.352\\

1199: Dimension of shape space & $D$ & 4\\

1200: Selection probability for optimal centrocytes & $a_0$ & $0.95$\\

1201: Output probability of selected centrocytes & $q$ & 0.2\\

1202: Number of seeder cells at FDCs & 

1203: $\sum_\phi B(\phi,t=0)$ & $3$\\

1204: Radius of B-cell activation around antigen & $R_0$ & $8-10$\\

1205: Time duration of phase I of GC reaction &

1206: $\Delta t_1$ & $72\,h$\\

1207: Time duration of phase II of GC reaction &

1208: $\Delta t_2$ & 48\,h\\

1209: Time duration of the whole GC reaction &

1210: $\sum_{i=1}^3 \Delta t_i$ & $504\,h$\\

1211: \hline

1212: \end{tabular}

1213: \caption[]{\sf Summary of all parameters of the model.

1214: They were determined by experimental data. 

1215: For explanations and references see the last section. 

1216: The parameters $g$, $q$, and $\Delta t_2$ are determined 

1217: in the course of the solution of \gl{modell} with respect to experimental 

1218: data discussed before.}

1219: \label{parameter}

1220: \end{center}

1221: \end{table}

1222: 

1223: \begin{figure}[ht]

1224: \center{\hspace*{0mm} \epsfxsize=10cm \epsfysize10cm

1225:             \epsfbox{bsum108.eps}}

1226: \caption[]{\sf Time course of the centroblast population integrated 

1227: over the whole sphere in the shape space around the antigen

1228: $\phi_0$. In the first phase the initial B-cells proliferate. The

1229: large population reached at $t=0\,h$ is then reduced by the

1230: selection process. At

1231: $t=48\,h$

1232: the production of plasma and

1233: memory cells starts, leading to an exponential decrease in

1234: the overall centroblast population.}

1235: \label{bsum46}

1236: \end{figure}

1237: %

1238: Starting from a randomly chosen antigen epitope and --

1239: following the oligoclonal character of the B-cell population 

1240: of a GC -- three randomly chosen activated B-cells in a

1241: sphere of radius $R_0$ according to \gl{r0}, we let the

1242: GC reaction develop as described by \gl{pmodell} and

1243: \gl{modell}.

1244: The different phases (see \kap{dynamics}) of the GC reaction

1245: are respected. The parameters are determined from experimental

1246: data as summarized in \tab{parameter}. The parameters

1247: $g$, $1-q$ and $\Delta t_2$, describing the differentiation rate

1248: of centroblasts to centrocytes, the recycling probability for

1249: the selected centrocytes and the time delay for the production

1250: of output cells, respectively, are iteratively adjusted 

1251: in order to fit with the experiments

1252: described in the previous section. The time variable 

1253: is running from $t_0=-72\,h$ to accentuate the first phase of 

1254: pure proliferation as a {\it preGC-phase}. The mutation and

1255: selection process is started at $t=0\,h$ and the production

1256: of plasma and memory cells begins at $t=\Delta t_2$. The

1257: differential equation is solved numerically with a modified Euler

1258: method in a subspace, in which we assume Dirichlet boundary

1259: conditions with $B=0$.

1260: 

1261: The best results are obtained by an iteration procedure for

1262: \be{bestpar}

1263: \frac{g}{\ln(2)}=\frac{0.355}{h} 

1264: \quad , \quad

1265: 1-q=0.8

1266: \quad ,\quad

1267: \Delta t_2=48\,h

1268: \quad.

1269: \ee

1270: At $t=0\,h$ we find a large oligoclonal population of 

1271: $12288$ centroblasts as a result of the proliferation phase

1272: started with $3$ initial cells 

1273: (see \abb{bsum46}). These cells are of different type compared to

1274: the cells of optimal affinity to the presented antigen and in a 

1275: typical distance from them (varying between

1276: $3$ and $8$ point mutations to reach the optimal cell type).

1277: 

1278: \paragraph{Optimization phase}

1279: The selection process at the FDCs and with the T-helper cells

1280: starts and reduces the centroblast population with low affinity

1281: to the presented antigen. At the same time

1282: somatic hypermutation

1283: leads to a spreading of the centroblast distribution

1284: over the shape space, resulting in a small but non-vanishing amount

1285: of centroblasts at the shape-space position of the antigen

1286: (see \abb{boe46}).

1287: %

1288: \begin{figure}[ht]

1289: \center{\hspace*{0mm} \epsfxsize=10cm \epsfysize10cm

1290:             \epsfbox{boe108.eps}}

1291: \caption[]{\sf The centroblast population and the 

1292: integrated plasma/memory cell production at the position

1293: of the antigen in the shape space. The centroblast population

1294: shows a growth due to proliferation and

1295: somatic hypermutation starting at $t=0\,h$. 

1296: The population growth is slowed down by

1297: the production of output cells starting at $t=48\,h$. Now

1298: the number of produced plasma and memory cells is increased

1299: steadily until the end of the GC reaction. Then, only

1300: $2$ cells (of optimal type) remain in the environment of the 

1301: FDCs.}

1302: \label{boe46}

1303: \end{figure}

1304: %

1305: The large majority of B-cells does not survive the 

1306: selection process (compare the summed cell population 

1307: decrease in \abb{bsum46}), dies through apoptosis and

1308: is rapidly phagocytosed by macrophages present in the GC.

1309: 

1310: \paragraph{Secondary proliferation}

1311: As all positively selected centrocytes reenter the proliferation

1312: process in the optimization phase of the GC reaction, the number of 

1313: centroblasts at the antigen position grows. Note

1314: that $\eps>0$ follows for the equilibrium condition \gl{equilcondition}

1315: because of $q=0$, according to \gl{phase2}. As a consequence,

1316: the overall centroblast population would restart to grow

1317: if this growth was not inhibited by the production of plasma 

1318: and memory cells after $48\,h$. Especially, 

1319: this {\it secondary proliferation} process is found for the centroblasts

1320: encoding antibodies of high affinity to the antigen

1321: (see \abb{boe46}). The centroblasts at the antigen position

1322: are practically all recycled (beside the reduction due to

1323: $a_0<1$) such that no reduction process exists any more.

1324: This gives rise to a new perspective on the GC reaction:

1325: We need just as many initially proliferated cells, that a

1326: spreading of the distribution through somatic hypermutation

1327: leads to at least one or two cells matching the antigen position in the

1328: shape space. Then a considerable amount of cells of this

1329: optimized type is reached within the time delay $\Delta t_2$ of the production

1330: of output cells, opening the possibility of a full proliferation

1331: process for these cells.

1332: 

1333: \paragraph{Recycling probability}

1334: When the output production is turned on at $t=48\,h$

1335: the dynamics of the germinal center is characterized by

1336: $\eps<0$ in \gl{equilcondition}, such that an exponential decrease

1337: of the whole centroblast population is initiated -- including

1338: the population of the optimized cells 

1339: (see \abb{bsum46} and \abb{boe46}). 

1340: Nevertheless, this decrease is slow enough

1341: to allow the production of a considerable amount of plasma

1342: and memory cells (see \abb{boe46} and \abb{showo46}).

1343: %

1344: \begin{figure}[ht]

1345: \center{\hspace*{0mm} \epsfxsize=10cm \epsfysize10cm

1346:             \epsfbox{ocut108.eps}}

1347: \caption[]{\sf A cut through the sphere in the shape space around the 

1348: presented antigen at position $2$. The seeder cells are at a distance of

1349: at least $5$ mutation steps.

1350: The number of produced plasma and memory

1351: cells grows considerably during the GC reaction. The

1352: rapidity of growth diminishes in course of time because of the decreasing

1353: number of optimized centroblasts at the antigen position.}

1354: \label{showo46}

1355: \end{figure}

1356: %

1357: It is an interesting observation that a rather small number of 

1358: centroblasts of the optimal type with respect to the antigen

1359: is sufficient to produce this large number of plasma and 

1360: memory cells -- even with a small output probability of

1361: $q=0.2$. On the contrary, a larger output probability may

1362: decrease the integrated number of plasma and memory cells,

1363: as less cells reenter the proliferation process. In conclusion,

1364: we claim that a large recycling probability of $80\%$ 

1365: of the positively selected centrocytes is

1366: not only necessary to achieve an optimization by 

1367: multi-step mutations but also to achieve a high

1368: number of resulting plasma and memory cells.

1369: 

1370: \paragraph{The end of the GC reaction}

1371: Finally, at day $21$ after immunization, only a few cells

1372: with maximum affinity to the antigen remain

1373: in the environment of the FDCs, as required from experiment 

1374: (Liu et al., 1991; Kelsoe, 1996). So the dynamics of our

1375: model allow for a vanishing of the germinal center population

1376: without any reduction of the antigen concentration or other

1377: additional requirements. The length of the

1378: germinal center reaction is determined by the interplay of 

1379: the centroblast (into centrocyte) differentiation rate $g$ and the

1380: delay of output production $\Delta t_2$, which is determined

1381: by \gl{out126}, so that the GC

1382: life time is basically controlled by $g$. 

1383: So from the perspective of our model a diminishing cell

1384: population may result solely

1385: from the interplay of proliferation, differentiation of

1386: centroblasts to centrocytes, and output production.

1387: 

1388: \paragraph{Dependence of initially activated seeder cells}

1389: The total number of remaining cells depends on the 

1390: distance of the initial B-cells from the antigen $\phi_0$.

1391: For numbers of mutation steps (necessary to reach the antigen)

1392: between $N=3$ and $N=8$, the number of remaining

1393: B-blasts varies from $43$ to $0$. For a typical number

1394: of mutations of $N=5$ about $10$ B-blasts remain in the

1395: environment of the FDCs at day $21$ after immunization.

1396: In other words, the GC life time depends on the initial

1397: distribution of the seeder cells with respect to the presented

1398: antigen. It would be interesting to check this statement

1399: experimentally by observing GCs with a variable number of

1400: mutations occurring during the optimization process.

1401: One could imagine for real GCs a self-regulating process 

1402: which depends on the already positively selected centrocytes and thus

1403: prolongs the GC reactions with large distances between

1404: seeder cells and antigen, and shortens correspondingly

1405: the ones with small distances.

1406: However, there is experimental evidence that the intensity of the

1407: GC reaction depends on the quality of the initially activated

1408: B-cells (Agarwal et al., 1998). Also in quasi-monoclonal mice

1409: (Cascalho et al., 1996) it was found that the volume of

1410: GCs 4 days after immunization depends on the average affinity

1411: of the reservoir B-cells to the antigen (de Vinuesa et al., 2000,

1412: Fig.~6). A quantitative comparison of these data with our model results

1413: would be interesting.

1414: 

1415: 

1416: \subsection{Result stability}

1417: 

1418: The parameter set in \gl{bestpar} is determined such 

1419: that the experimental conditions \gl{recycle}

1420: and \gl{out126} are fulfilled for typical initial conditions. 

1421: Therefore, the prediction

1422: of the recycling probability to be $80\%$ is a statement

1423: which -- in the framework of our model --

1424: is justified to the same degree as the experiments

1425: are exact. 

1426: Nevertheless,

1427: the sensitivity of the predicted high recycling probability

1428: to such modifications is weak.

1429: 

1430: It should be

1431: mentioned that in addition there is a certain freedom of 

1432: choice concerning the number of remaining cells at day

1433: $21$ after immunization. For example, the parameter set

1434: with a still higher recycling probability

1435: \be{par2}

1436: \frac{g}{\ln(2)}=\frac{0.52}{h} 

1437: \quad , \quad

1438: 1-q=0.9

1439: \quad ,\quad

1440: \Delta t_2=55\,h

1441: \quad.

1442: \ee

1443: leads to a result of comparable quality as under the conditions

1444: of \gl{bestpar}

1445: -- it is clear that a higher recycling probability slows

1446: down the production of output cells such that simultaneously

1447: the time delay of the production start has to be larger --

1448: with the difference that the number of remaining cells

1449: is reduced with respect to 

1450: the previous result such that no cells remain at the end

1451: of typical GC reactions. For this reason and because

1452: $g$ is on its upper bound \gl{gupper},

1453: this parameter set was not considered as reasonable.

1454: 

1455: If one requires a smaller recycling probability the

1456: parameter set

1457: \be{par3}

1458: \frac{g}{\ln(2)}=\frac{0.286}{h} 

1459: \quad , \quad

1460: 1-q=0.7

1461: \quad ,\quad

1462: \Delta t_2=42\,h

1463: \quad.

1464: \ee

1465: leads to correct values for $v_0$ and 

1466: \gl{recycle} is also fulfilled.

1467: The reduced recycling probability accelerates the

1468: output cell production such that it has to be decelerated

1469: by a reduction of the time delay $\Delta t_2$. Even if

1470: the differentiation rate stays in a reasonable range,

1471: the number of remaining cells in the GC at day $21$ 

1472: after immunization is about ten-fold with respect to

1473: \gl{bestpar}. More generally one may consider the number

1474: of remaining cells for typical GC reactions with an average

1475: number of mutations necessary to reach the antigen.

1476: %

1477: \begin{center}

1478: \begin{tabular}{|l|c|c|c|c|c|} \hline

1479: Recycling probability $1-q$ & $0.5$& $0.6$& $0.7$& $0.8$& $0.9$\\

1480: \hline

1481: $\sum_\phi B(\phi,t=21\,d)$ & 1476& 515& 134& 10& 0\\

1482: \hline

1483: \end{tabular}

1484: \end{center}

1485: %

1486: The mentioned experimental data are all reproduced for each

1487: value of $1-q$, solely the number of remaining cells is left open.

1488: To ensure that this number is in accordance 

1489: with the experimental statement that a {\it a few} proliferating

1490: B-blasts remain at the FDCs at this stage of GC development,

1491: we are led to a recycling probability of $80\%$. We want

1492: to emphasize that this result is determined very clearly, as

1493: for the other recycling probabilities the remaining number of

1494: cells in the GC final state differs by at least an order of magnitude.

1495: 

1496: This variation of parameters shows the range of

1497: stability of our prediction:\\

1498: First, the window of

1499: possible time delays $\Delta t_2$, which are in

1500: accordance with \gl{out126} is very narrow.

1501: Already for $\Delta t_2>55\,h$ the condition

1502: \gl{gupper} will be violated. On the other hand

1503: for $\Delta t_2<38\,h$, despite the immense 

1504: number of remaining cells in the final GC state,

1505: the condition \gl{out126} can not be fulfilled

1506: anymore. In consequence, it is not possible

1507: that the production of plasma and memory cells

1508: already starts with the establishment of light

1509: and dark zone, i.e.\,that it starts with the existence of

1510: a fully developed GC. The emphasized

1511: experimental bounds require necessarily a time

1512: delay for the production of output cells of

1513: at least $42\,h$. This means that the production

1514: process should be initiated by a separate signal

1515: and is not present in a working GC automatically.\\

1516: %

1517: Secondly, the window of possible recycling

1518: probabilities is strongly bound by the

1519: experiments (Han et al., 1995a, 1995b) and by the

1520: experimentally observed number of remaining

1521: cells at day $21$ after immunization. This leads

1522: to the conclusion that the recycling probability

1523: should be at least $70\%$ and not larger

1524: than $85\%$. 

1525: Even ignoring the required number of cells in the

1526: final state, \gl{out126} does not allow a 

1527: recycling probability smaller than $60\%$.

1528: We would like to emphasize that this

1529: result differs extremely  from the value for the recycling

1530: probability of $0.15$ assumed in (Kesmir \& de Boer, 1999). 

1531: Nevertheless,

1532: we do not regard this discrepancy as a contradiction

1533: because in the model of Kesmir \& de Boer not all parameters

1534: were fixed by experimental data.

1535: 

1536: \paragraph{Shape space dimension}

1537: The analysis presented above

1538: was performed with a shape space dimension of $4$,

1539: (compare \kap{ssdim}). It is an important remark that our

1540: results remain essentially unaffected for $D=5$ or $6$.

1541: Using for instance $D=5$, only the parameter

1542: $g$, controlling the selection speed,

1543: has to be reduced slightly by $7\%$ in order to stay in

1544: accordance with the experimental bounds. The total

1545: number of remaining cells at day $21$ after immunization

1546: does not change but its distribution is spread out on the

1547: shape space.

1548: For $D=6$, the

1549: parameter $g$ is again reduced by $6\%$, while all

1550: other parameters remain unchanged to stay in accordance

1551: with the experimental data.

1552: Again, the cell distribution at day $21$ after immunization

1553: becomes less centered at the antigen

1554: position. We conclude that, for the values tested,

1555: the dependence on the

1556: shape space dimension is weak enough such that

1557: our results are not affected by it. Even, the slower

1558: selection speed for higher dimensions of less than

1559: $10\%$ per dimension remains in the range of

1560: accuracy of our results. 

1561: 

1562: Note, that the dependence on the shape space dimension

1563: is restricted to an intermediate stage (phase II) of the GC reaction.

1564: In the first phase of pure proliferation $D$ does not

1565: enter the dynamics at all.

1566: In the late phase of selection,  

1567: when the distribution $B(\phi)$ becomes approximately

1568: spherically symmetric around the point representing the

1569: antigen, $D$ cancels exactly with the factor $1/D$ in

1570: \gl{modell} as there are $D$ equal terms contributing to

1571: the sum such that $D$ factorizes.

1572: A dependence of our results on the shape space

1573: dimension may only occur during the early phase of selection,

1574: in which the distribution $B(\phi)$ is clearly not symmetric

1575: around the position of the antigen.

1576: Since the distribution of centroblasts

1577: develops into a spherically symmetric distribution around $\phi_0$

1578: already before day $8$ after immunization (see \abb{beta46}),

1579: we deduce that a dependence on the shape space dimension

1580: may occur between day $3$ and day $8$ of the GC reaction only.

1581: 

1582: \paragraph{Mutations with large changes of affinity}

1583: We have assumed so far that mutations are represented in the

1584: shape space by next-neighbor jumps.

1585: To check the robustness of the result for the recycling probability

1586: against the possibility of mutations that lead to large distance

1587: jumps in the shape space, we consider a worst case argument.

1588: Let us suppose that $10\,\%$ of the mutations lead to a large jump

1589: in the shape space. This is translated into the model by reducing

1590: the number of next-neighbor-mutations by a factor of $0.9$. The

1591: result is a lower efficiency of the mutation process

1592: as the probability for a random mutation to improve affinity

1593: with respect to the antigen is very small. Therefore, the final

1594: number of cells is reduced by a factor of 10, and the

1595: centroblast to centrocyte differentiation rate has to be adjusted

1596: (from $0.244$ to $0.255$) in order to fulfill condition \gl{recycle}.

1597: A similar result as before is obtained by an adjustment

1598: of the recycling probability from

1599: $0.8$ to $0.78$. As this is a worst case argument, we consider

1600: our result as robust against large jump mutations.

1601: 

1602: 

1603: \subsection{Optimization speed}

1604: 

1605: %

1606: \begin{figure}[ht]

1607: \center{\hspace*{0mm} \epsfxsize=10cm \epsfysize10cm

1608:             \epsfbox{beta108.eps}}

1609: \caption[]{\sf The time course of $\beta(\phi_0,t)$ (see \gl{betadef}) 

1610: at the shape space position of the antigen for the parameter set

1611: \gl{bestpar}.

1612: $\beta$ becomes constant shortly after $t=96\,h$, 

1613: i.e.\,at day $7$ after immunization.}

1614: \label{beta46}

1615: \end{figure}

1616: %

1617: From experiment it is expected that the cells

1618: with high affinity to the antigen dominate already

1619: at day $8$ after immunization (Jacob et al., 1993). 

1620: This is verified in the model by checking if the asymptotic regime

1621: of the GC reaction is reached at day $8$ in the sense that $\beta$

1622: (defined in \gl{betadef}) becomes constant.

1623: As can be seen in \abb{beta46}, $\beta$ becomes

1624: constant in the course of day $7$ after immunization

1625: for the parameter set \gl{bestpar}. This behavior of $\beta$

1626: is not altered in a great range of parameter variation. 

1627: Besides the reproduction of the experimental evidence

1628: for the optimization process to be accomplished after

1629: $8$ days, this result is an a-posteriori

1630: check for the derivation of the recycling

1631: probability, which is based on a constant value

1632: of $\beta(\phi_0)$ at day $9$ after immunization

1633: (see \gl{nutoomega}).

1634: 

1635: 

1636: \subsection{Optimization quality}

1637: 

1638: It may be interesting to verify if larger or smaller recycling

1639: probabilities lead to better optimization of the antibody

1640: types. The statement that recycling may be necessary

1641: to achieve a considerable affinity enhancement in a multi-step 

1642: mutation process is widely discussed (Oprea et al., 2000)

1643: and enforced by the present work. We would like to point

1644: out that the relative number of {\it good} outcoming plasma

1645: or memory cells increases with larger recycling probabilities

1646: (see \abb{resqrun}).

1647: %

1648: \begin{figure}[ht]

1649: \center{\hspace*{0mm} \epsfxsize=10cm \epsfysize10cm

1650:             \epsfbox{qrun.eps}}

1651: \caption[]{\sf Comparison of the total plasma and memory cell

1652: distribution at day $21$ after immunization

1653: on a cut along one of the coordinates in the shape space 

1654: with the antigen at coordinate position $2$.

1655: The parameters are uniquely determined by the 

1656: experimental data with the exception of the remaining number of cells at

1657: the end of the GC reaction, which is undetermined to allow different

1658: recycling probabilities $1-q$. The distribution of the output cells

1659: becomes sharper for larger $1-q$.}

1660: \label{resqrun}

1661: \end{figure}

1662: %

1663: In the shape space this corresponds

1664: to a sharper peak of the distribution $\int_{0\,d}^{21\,d} dt\,O(\phi,t)$.

1665: A too small recycling probability leads to a spreading of the

1666: outcoming plasma and memory cells in the shape space, 

1667: i.e.\,to a weaker specificity of the GC output.

1668: 

1669: On the other hand, we have already seen that a larger recycling probability

1670: lowers the absolute number of produced plasma and memory cells

1671: such that we are confronted with two competitive tendencies. Large

1672: recycling probabilities lead to specific but weak GC reactions, while

1673: small probabilities lead to unspecific but intense GC reactions. One may

1674: agree, that the value of $80\%$ calculated here and supported by

1675: experimental data, is a good compromise between these two

1676: extremes.

1677: 

1678: 

1679: 

1680: \subsection{Start of somatic hypermutation}

1681: 

1682: There is experimental evidence that somatic

1683: hypermutation does not occur during the first phase 

1684: of proliferation of centroblasts in the environment

1685: of the FDCs 

1686: (Han et al., 1995b; Jacob et al., 1993; 

1687: McHeyzer-Williams et al., 1993; Pascual et al., 1994b).

1688: However, we checked if our model allow somatic hypermutation

1689: to start during the proliferation phase.

1690: The primary reason for the retardation of the output 

1691: production is to give time to the GC to develop sufficient

1692: good centroblasts before the GC population is weakened 

1693: through an additional output. Therefore, one may

1694: in principle think of a mutation starting

1695: long before the selection process. In this scenario

1696: the output production would start simultaneously with

1697: selection and one is led to a two phase

1698: process only: A first one with proliferation and 

1699: somatic hypermutation and a second one with additional 

1700: selection and output production.

1701: The centroblast-types would already be spread out in the shape

1702: space when selection and output start. 

1703: %

1704: \begin{figure}[ht]

1705: \center{\hspace*{0mm} \epsfxsize=10cm \epsfysize10cm

1706:             \epsfbox{boe120.eps}}

1707: \caption[]{\sf The centroblast population and the 

1708: integrated plasma/memory cell production at the position

1709: of the antigen in the shape space. Except for the

1710: time phases we used the same parameters as in \gl{bestpar}.

1711: The centroblast population

1712: shows a smooth enhancement due to somatic hypermutation

1713: starting at $t=-48\,h$. The population growth is stopped by

1714: the production of output cells starting at $t=0\,h$. The

1715: slope of the summed output is substantially smaller than

1716: expected from experiment. $21$ days after immunization

1717: only $1$ cell remains. }

1718: \label{boe61}

1719: \end{figure}

1720: %

1721: There is no parameter set in accordance

1722: with all experimental data. Especially, it is impossible

1723: to get the correct output dynamics as described in

1724: \gl{out126}. Typically we get $v_0=2.9$ being strongly

1725: different from the required value and a much

1726: smoother result (compare \abb{boe61}

1727: to \abb{boe46}). In addition the distribution of plasma

1728: and memory cells is not as well pronounced at the antigen 

1729: position such that the GC reaction

1730: gives rise to output cells with weaker specificity. Thus,

1731: we deduce from our model that the dynamic of plasma 

1732: and memory cell production in real GC reactions

1733: does not allow somatic

1734: hypermutation to occur considerably before the selection

1735: process is started.

1736: 

1737: 

1738: 

1739: \section{Conclusions}

1740: 

1741: We developed a new model for the GC reaction

1742: using its main functional elements. The GC reaction

1743: is described by the evolution of a centroblast distribution

1744: on an affinity shape space with respect to an initially

1745: fixed antigen distribution.

1746: On this shape space an affinity

1747: function, modeling the complementarity of

1748: antibody and antigen, was defined and its

1749: functional behavior as well as its width were

1750: deviated from affinity enhancements known

1751: for real GC reactions.

1752: 

1753: In the model the reaction is decomposed into three phases

1754: (see \abb{phases}). 

1755: In the proliferation phase a few seeder cells multiply. 

1756: Our model excludes somatic hypermutation 

1757: to occur already at this stage of the GC development.

1758: An optimization phase follows,

1759: in which in addition to proliferation, mutation of the

1760: antibody type and selection take place.

1761: This leads to a competition of spreading and

1762: peaking of the centroblast distribution in the

1763: shape space.

1764: In this phase, all positively selected cells reenter the

1765: proliferation phase in the dark zone, i.e.\,no plasma

1766: or memory cells are produced.

1767: This occurs solely in the output production phase. In this

1768: third phase all elements of the GC reaction are active.

1769: 

1770: The dynamical evolution of the cell population in

1771: a GC is described by a set of coupled

1772: linear differential equation. 

1773: All parameters are determined

1774: with experimental data and especially using

1775: (Han et al., 1995b). The model shows the

1776: typical behavior of GCs without any further

1777: adjustments.

1778: Due to the oligoclonal

1779: character of GCs a large amount of identical 

1780: centroblasts is produced in the proliferation phase.

1781: These diffuse over the shape space in the second

1782: phase, in which only those cells are retained that

1783: exhibit a high affinity to the presented antigen.

1784: As the {\it good} cells reenter the proliferation

1785: process, this leads to a dominance of good cells

1786: within $7$ to $8$ days after immunization in

1787: accordance with experiments. The already

1788: optimized cell distribution then gives rise to

1789: plasma and memory cells leaving the FDC

1790: environment in the output production phase, leading

1791: to large numbers of output cells of optimized

1792: type due to the long duration of this third phase.

1793: The GC reaction is rather weak $21$ days after

1794: immunization and only a small number of

1795: cells remains in the environment of the FDCs.

1796: This shows that the end of a GC reaction is

1797: not necessarily coupled with an antigen consumption

1798: or additional signals, but may occur simply due

1799: to the (unchanged) dynamical development of the

1800: GC.

1801: 

1802: Our main result is that the experimental data and

1803: especially the evaluation of (Han et al., 1995b) lead

1804: to the following view of the GC reaction.

1805: The optimization phase lasts for not less than

1806: $42\,h$ and not longer than $55\,h$. During this

1807: time {\it all} selected centrocytes are recycled

1808: and reenter the proliferation phase.

1809: Without such a {\it non-output

1810: phase} the number of optimal cells with respect

1811: to the presented antigen is not large enough to

1812: allow a considerable output rate during the

1813: remaining life time of the GC. 

1814: We looked for an experimental evidence for this

1815: delayed production of output cells. Indeed, comparing

1816: two experiments 

1817: (Jacob et al., 1993; Pascual et al., 1994a) we found that 

1818: plasma and memory cells were first observed more than 

1819: two days later than mutated cells during a GC reaction. 

1820: This confirms experimentally the existence of the 

1821: non-output phase that we found in the model.

1822: As output cells are not produced from the beginning

1823: of the GC reaction, the output should be triggered by a special

1824: signal, related to the affinity of the B-cells and

1825: the corresponding antigen. But even in the

1826: output production phase, the recycling probability

1827: turns out to lie between $70\%$ and $85\%$

1828: and thus is substantially larger than values

1829: expected until now (Kesmir \& de Boer, 1999). 

1830: We want to emphasize that such large recycling 

1831: probabilities turned out to be crucial for a GC 

1832: outcome of high specificity. Larger recycling probabilities

1833: lead to specific GC reactions of low intensity, while

1834: smaller ones to less specific but more intense 

1835: GC reactions. The value of about $80\%$ that we

1836: found represents a good compromise between these

1837: competitive tendencies. On the other hand it is

1838: not surprising that these recycled cells have not been

1839: found experimentally: The absolute number of recycled

1840: cells is very small compared to the total number

1841: of cells present in the GC, so that a measured signal

1842: for recycled cells is suppressed by more than two

1843: orders of magnitudes. A better signal should be

1844: observed in later stages of the GC reaction when

1845: the cells of high affinity to the antigen are

1846: already dominating.

1847: 

1848: Furthermore, we conclude that the whole selection

1849: process including differentiation of centroblasts

1850: to antibody presenting centrocytes, the multi-step

1851: selection process itself and finally the recycling

1852: take around $3$ or $4\,h$. This result is compatible

1853: with the experimental observation that the inhibition

1854: of apoptosis during the selection may take about

1855: $1$ or $2\,h$.

1856: 

1857: It is interesting, that the produced cells

1858: possess a certain broadness in the shape space, 

1859: i.e.\,not only the centrocytes of optimal affinity to the antigen

1860: differentiate to plasma and memory cells.

1861: This may be of relevance for the resistance of an

1862: immune system against a second immunization

1863: with a mutated antigen. A considerable number

1864: of memory cells with high affinity to the mutated

1865: antigen may still be present for this reason.

1866: 

1867: The results of our model suggest new experiments.

1868: A systematic and quantitative analysis of the GC

1869: reaction in dependence of the initial conditions, 

1870: i.e.\,the number of necessary point mutations to reach

1871: the cell type of optimal affinity to the antigen, would

1872: lead to new insights into the interdependence of the

1873: essential elements of the GC. This may yield

1874: hints about self-regulating processes.

1875: Furthermore, it would be interesting to

1876: analyze quantitatively

1877: the total number of cells in the course

1878: of the GC reaction.

1879: Finally, it may be interesting to compare our model

1880: with its continuous counterpart. 

1881: The translation of \gl{modell} into continuous

1882: space leads to a differential equation of Schr\"odinger

1883: type with a Gaussian potential. 

1884: 

1885: %\vspace*{5mm}

1886: %{\bf Acknowledgements.}

1887: 

1888: 

1889: 

1890: \vfill

1891: \eject

1892: \newcounter{fig}

1893: \section*{References}

1894: \begin{list}{\normalsize

1895: {\rm }}

1896: {\usecounter{fig}

1897: \setcounter{fig}{0}

1898: \labelwidth0mm

1899: \leftmargin8mm

1900: \rightmargin0mm

1901: \labelsep0mm

1902: \topsep0mm

1903: \parsep0mm

1904: \itemsep0mm}

1905: \itemindent-8mm

1906: \normalsize

1907: 

1908: \item

1909: {\sc Agarwal, A., Nayak, B.P. \& Rao, K.V.S.}

1910: {\rm (1998)}.

1911: {\rm B-Cell Responses to a Peptide Epitope -- VII -- Antigen-Dependent 

1912: Modulation of the Germinal Center Reaction},

1913: {\it J. Immunol.\/}

1914: {\bf 161},

1915: {\rm 5832--5841}.

1916: 

1917: \item

1918: {\sc Berek, C. \& Milstein, C.}

1919: {\rm (1987)}.

1920: {\rm Mutation drift and repertoire shift in the maturation 

1921: of the immune response},

1922: {\it Immunol. Rev.\/}

1923: {\bf 96},

1924: {\rm 23--41}.

1925: 

1926: 

1927: 

1928: \item

1929: {\sc Berek, C. \& Milstein, C.}

1930: {\rm (1988)}.

1931: {\rm The Dynamic Nature of the Antibody Repertoire},

1932: {\it Immunol. Rev.\/}

1933: {\bf 105},

1934: {\rm 5--26}.

1935: 

1936: 

1937: 

1938: \item

1939: {\sc Camacho, S.A., Koscovilbois, M.H. \& Berek, C.}

1940: {\rm (1998)}.

1941: {\rm The Dynamic Structure of the Germinal Center},

1942: {\it Immunol. Today\/}

1943: {\bf 19},

1944: {\rm 511--514}.

1945: 

1946: \item

1947: {\sc Cascalho, M., Ma, A., Lee, S., Masat, L. \& Wabl, M. }

1948: {\rm (1996)}.

1949: {\rm A Quasi-Monoclonal Mouse},

1950: {\it Science\/}

1951: {\bf 272},

1952: {\rm 1649--1652}.

1953: 

1954: 

1955: \item

1956: {\sc Choe, J. \& Choi, Y.S.}

1957: {\rm (1998)}.

1958: {\rm IL-10 Interrupts Memory B-Cell Expansion in the Germinal 

1959: Center by Inducing Differentiation into Plasma-Cells},

1960: {\it Eur. J. Immunol.\/}

1961: {\bf 28},

1962: {\rm 508--515}.

1963: 

1964: 

1965: 

1966: \item

1967: {\sc Cohen, J.J., Duke, R.C., Fadok, V.A. \& Sellins, K.S.}

1968: {\rm (1992)}.

1969: {\rm Apoptosis and programmed cell death in immunity},

1970: {\it Annu. Rev. Immunol.\/}

1971: {\bf 10},

1972: {\rm 267--293}.

1973: 

1974: 

1975: 

1976: \item

1977: {\sc Dubois, B., Barth\'el\'emy, C., Durand, I., Liu, Y.-J., Caux, 

1978: C. \& Bri\`ere, F.}

1979: {\rm (1999)}.

1980: {\rm Toward a Role of Dendritic Cells in the Germinal Center 

1981: Reaction -- Triggering of B-Cell Proliferation and Isotype 

1982: Switching},

1983: {\it J. Immunol.\/}

1984: {\bf 162},

1985: {\rm 3428--3436}.

1986: 

1987: 

1988: 

1989: \item

1990: {\sc Eijk, M. van \& Groot, C. de},

1991: {\rm (1999)}.

1992: {\rm Germinal Center B-Cell Apoptosis Requires Both Caspase 

1993: and Cathepsin Activity},

1994: {\it J. Immunol.\/}

1995: {\bf 163},

1996: {\rm 2478--2482}.

1997: 

1998: 

1999: 

2000: \item

2001: {\sc Fischer, M.B., Goerg, S., Shen, L.M., Prodeus, A.P., Goodnow, 

2002: C.C., Kelsoe, G. \& Carroll, M.C.}

2003: {\rm (1998)}.

2004: {\rm Dependence of Germinal Center B-Cells on Expression 

2005: of Cd21/Cd35 for Survival},

2006: {\it Science\/}

2007: {\bf 280},

2008: {\rm 582--585}.

2009: 

2010: 

2011: 

2012: \item

2013: {\sc Fliedner, T.M.}

2014: {\rm (1967)}.

2015: {\rm On the origin of tingible bodies in germinal centers 

2016: in immune responses},

2017: {\rm in: H. Cottier (Hrsg.)},

2018: {\rm Germinal Centers in Immune Responses}.

2019: {\rm Springer},

2020: {\rm Berlin},

2021: {\rm pp.\,218--224}.

2022: 

2023: 

2024: 

2025: \item

2026: {\sc Grouard, G., De Bouteiller, O., Banchereau, J. \& Liu, Y.-J.}

2027: {\rm (1995)}.

2028: {\rm Human follicular dendritic cells enhance cytokine-dependent 

2029: growth and differentiation of CD40-activated B cells},

2030: {\it J. Immunol.\/}

2031: {\bf 155},

2032: {\rm 3345--3352}.

2033: 

2034: 

2035: 

2036: \item

2037: {\sc Han, S.H., Hathcock, K., Zheng, B., Kelper, T.B., Hodes, 

2038: R. \& Kelsoe, G.}

2039: {\rm (1995a)}.

2040: {\rm Cellular Interaction in Germinal Centers: Roles of 

2041: CD40-Ligand and B7-1 and B7-2 in Established Germinal 

2042: Centers},

2043: {\it J. Immunol.\/}

2044: {\bf 155},

2045: {\rm 556--567}.

2046: 

2047: 

2048: 

2049: \item

2050: {\sc Han, S.H., Zheng, B., Dal Porto, J. \& Kelsoe, G.}

2051: {\rm (1995b)}.

2052: {\rm In situ Studies of the Primary Immune Response to 

2053: (4--Hydroxy--3--Nitrophenyl) 

2054: Acetyl IV. Affinity-Dependent, Antigen-Driven B-Cell 

2055: Apoptosis in Germinal Centers as a Mechanism for Maintaining 

2056: Self-Tolerance},

2057: {\it J. Exp. Med.\/}

2058: {\bf 182},

2059: {\rm 1635--1644}.

2060: 

2061: 

2062: 

2063: \item

2064: {\sc Hanna, M.G.}

2065: {\rm (1964)}.

2066: {\rm An autoradiographic study of the germinal center in 

2067: spleen white pulp during early intervals of the immune 

2068: response},

2069: {\it Lab. Invest.\/}

2070: {\bf 13},

2071: {\rm 95--104}.

2072: 

2073: 

2074: 

2075: \item

2076: {\sc Hardie, D.L., Johnson, G.D. \& MacLennan, I.C.M.}

2077: {\rm (1993)}.

2078: {\rm Quantitative analysis of molecules which distinguish 

2079: functional compartments in germinal centers},

2080: {\it Eur. J. Immunol.\/}

2081: {\bf 23},

2082: {\rm 997--1004}.

2083: 

2084: 

2085: 

2086: \item

2087: {\sc Jacob, J., Kassir, R. \& Kelsoe, G.}

2088: {\rm (1991)}.

2089: {\rm In situ studies of the primary immune response 

2090: to (4-hydroxy-3-nitrophenyl)acetyl. 

2091: I. The architecture and dynamics of responding cell 

2092: populations},

2093: {\it J. Exp. Med.\/}

2094: {\bf 173},

2095: {\rm 1165--1175}.

2096: 

2097: 

2098: 

2099: \item

2100: {\sc Jacob, J., Przylepa, J., Miller, C. \& Kelsoe, G.}

2101: {\rm (1993)}.

2102: {\rm In situ studies of the primary response 

2103: to (4-hydroxy-3-nitrophenyl)acetyl. 

2104: III. The kinetics of V region mutation and selection 

2105: in germinal center B cells},

2106: {\it J. Exp. Med.\/}

2107: {\bf 178},

2108: {\rm 1293--1307}.

2109: 

2110: 

2111: 

2112: \item

2113: {\sc Janeway, C.A. \& Travers, P.}

2114: {\rm (1997)}.

2115: {\rm Immunologie}.

2116: {\rm Spektrum Akademischer Verlag},

2117: {\rm Heidelberg, Berlin, Oxford}.

2118: 

2119: 

2120: 

2121: \item

2122: {\sc Kelsoe, G.}

2123: {\rm (1996)}.

2124: {\rm The germinal center: a crucible for lymphocyte selection},

2125: {\it Semin. Immunol.\/}

2126: {\bf 8},

2127: {\rm 179--184}.

2128: 

2129: 

2130: 

2131: \item

2132: {\sc Kepler, T.B. \& Perelson, A.S.}

2133: {\rm (1993)}.

2134: {\rm Cyclic re-entry of germinal center B cells and the 

2135: efficiency of affinity maturation},

2136: {\it Immunol. Today\/}

2137: {\bf 14},

2138: {\rm 412--415}.

2139: 

2140: 

2141: 

2142: \item

2143: {\sc Kesmir, C. \& Boer, R.J.\,de},

2144: {\rm (1999)}.

2145: {\rm A Mathematical Model on Germinal Center Kinetics and 

2146: Termination},

2147: {\it J. Immunol.\/}

2148: {\bf 163},

2149: {\rm 2463--2469}.

2150: 

2151: 

2152: 

2153: \item

2154: {\sc Kroese, F.G., Wubbena, A.S., Seijen, H.G. \& Nieuwenhuis, 

2155: P.}

2156: {\rm (1987)}.

2157: {\rm Germinal centers develop oligoclonally},

2158: {\it Eur. J. Immunol.\/}

2159: {\bf 17},

2160: {\rm 1069--1072}.

2161: 

2162: 

2163: 

2164: \item

2165: {\sc K\"uppers, R., Zhao, M., Hansmann, M.L. \& Rajewsky, K.}

2166: {\rm (1993)}.

2167: {\rm Tracing B Cell Development in Human Germinal Centers 

2168: by Molecular Analysis of Single Cells Picked from Histological 

2169: Sections},

2170: {\it EMBO J.\/}

2171: {\bf 12},

2172: {\rm 4955--4967}.

2173: 

2174: 

2175: 

2176: \item

2177: {\sc Lindhout, E., Lakeman, A. \& Groot, C. de},

2178: {\rm (1995)}.

2179: {\rm Follicular dendritic cells inhibit apoptosis in human 

2180: B lymphocytes by rapid and irreversible blockade of 

2181: preexisting endonuclease},

2182: {\it J. Exp. Med.\/}

2183: {\bf 181},

2184: {\rm (1985--1995}.

2185: 

2186: 

2187: 

2188: \item

2189: {\sc Lindhout, E., Mevissen, M.L., Kwekkeboom, J., Tager, J.M. \& Groot, 

2190: C. de},

2191: {\rm (1993)}.

2192: {\rm Direct evidence that human follicular dendritic cells 

2193: (FDC) rescue germinal centre B cells from death by 

2194: apoptosis},

2195: {\it Clin. Exp. Immunol.\/}

2196: {\bf 91},

2197: {\rm 330--336}.

2198: 

2199: 

2200: 

2201: \item

2202: {\sc Lindhout, E., Koopman, G., Pals, S.T. \& Groot, 

2203: C.\,de},

2204: {\rm (1997)}.

2205: {\rm Triple check for antigen specificity of B cells during 

2206: germinal centre reactions},

2207: {\it Immunol. Today\/}

2208: {\bf 18},

2209: {\rm 573--576}.

2210: 

2211: 

2212: 

2213: \item

2214: {\sc Liu, Y.-J., Barth\'el\'emy, C., De Bouteiller, O. \& Banchereau, 

2215: J.}

2216: {\rm (1994)}.

2217: {\rm The differences in survival and phenotype between centroblasts 

2218: and centrocytes},

2219: {\it Adv. Exp. Med. Biol.\/}

2220: {\bf 355},

2221: {\rm 213--218}.

2222: 

2223: 

2224: 

2225: \item

2226: {\sc Liu, Y.-J., Joshua, D.E., Williams, G.T., Smith, C.A., Gordon, 

2227: J. \& MacLennan, I.C.}

2228: {\rm (1989)}.

2229: {\rm Mechanism of antigen-driven selection in germinal centres},

2230: {\it Nature\/}

2231: {\bf 342},

2232: {\rm 929--931}.

2233: 

2234: 

2235: 

2236: \item

2237: {\sc Liu, Y.-J., Zhang, J., Lane, P.J., Chan, E.Y. \& MacLennan, 

2238: I.C.M.}

2239: {\rm (1991)}.

2240: {\rm Sites of specific B cell activation in primary and 

2241: secondary responses to T cell-dependent and T cell-independent 

2242: antigens},

2243: {\it Eur. J. Immunol.\/}

2244: {\bf 21},

2245: {\rm 2951--2962}.

2246: 

2247: 

2248: 

2249: \item

2250: {\sc MacLennan, I.C.M.}

2251: {\rm (1994)}.

2252: {\rm Germinal Centers},

2253: {\it Annu. Rev. Immunol.\/}

2254: {\bf 12},

2255: {\rm 117--139}.

2256: 

2257: 

2258: 

2259: \item

2260: {\sc McHeyzer-Williams, M.G., McLean, M.J., Labor, P.A. \& Nossal, 

2261: G.V.J.}

2262: {\rm (1993)}.

2263: {\rm Antigen-driven B cell differentiation in vivo},

2264: {\it J. Exp. Med.\/}

2265: {\bf 178},

2266: {\rm 295--307}.

2267: 

2268: 

2269: 

2270: \item

2271: {\sc Nossal, G.}

2272: {\rm (1991)}.

2273: {\rm The molecular and cellular basis of affinity maturation 

2274: in the antibody response},

2275: {\it Cell\/}

2276: {\bf 68},

2277: {\rm 1--2}.

2278: 

2279: 

2280: 

2281: \item

2282: {\sc Oprea, M. \& Perelson, A.S.}

2283: {\rm (1996)}.

2284: {\rm Exploring the Mechanism of Primary Antibody Responses 

2285: to T-Cell-Dependent Antigen},

2286: {\it J. Theor. Biol.\/}

2287: {\bf 181},

2288: {\rm 215--236}.

2289: 

2290: 

2291: 

2292: \item

2293: {\sc Oprea, M. \& Perelson, A.S.}

2294: {\rm (1997)}.

2295: {\rm Somatic mutation leads to efficient affinity maturation 

2296: when centrocytes recycle back to centroblasts},

2297: {\it J. Immunol.\/}

2298: {\bf 158},

2299: {\rm 5155--5162}.

2300: 

2301: 

2302: 

2303: \item

2304: {\sc Oprea, M., Nimwegen, E.\,van \& Perelson, A.S.}

2305: {\rm 2000)}.

2306: {\rm Dynamics of One-pass Germinal Center Models: Implications 

2307: for Affinity Maturation},

2308: {\it Bull. Math. Biol.\/}

2309: {\bf 62},

2310: {\rm 121--153}.

2311: 

2312: 

2313: 

2314: \item

2315: {\sc Pascual, V., Cha, S., Gershwin, M.E., Capra, J.D. \& Leung, 

2316: P.S.C.}

2317: {\rm (1994a)}.

2318: {\rm Nucleotide Sequence Analysis of Natural and Combinatorial 

2319: Anti-PDC-E2 Antibodies in Patients with Primary Biliary 

2320: Cirrhosis},

2321: {\it J. Immunol.\/}

2322: {\bf 152},

2323: {\rm 2577--2585}.

2324: 

2325: 

2326: 

2327: \item

2328: {\sc Pascual, V., Liu, Y.-J., Magalski, A., De Bouteiller, 

2329: O., Banchereau, J. \& Capra, J.D.}

2330: {\rm (1994b)}.

2331: {\rm Analysis of somatic mutation in five B cell subsets 

2332: of human tonsil},

2333: {\it J. Exp. Med.\/}

2334: {\bf 180},

2335: {\rm 329--339}.

2336: 

2337: 

2338: 

2339: \item

2340: {\sc Perelson, A.S. \& Oster, G.F.}

2341: {\rm (1979)}.

2342: {\rm Theoretical Studies of Clonal Selection: Minimal Antibody 

2343: Repertoire Size and Reliability of Self-Non-self Discrimination},

2344: {\it J. Theor. Biol.\/}

2345: {\bf 81},

2346: {\rm 645--670}.

2347: 

2348: 

2349: 

2350: \item

2351: {\sc Perelson, A.S. \& Wiegel, F.W.}

2352: {\rm (1999)}.

2353: {\rm Some Design Principles for Immune System Recognition},

2354: {\it Complexity\/}

2355: {\bf 4},

2356: {\rm 29--37}.

2357: 

2358: 

2359: 

2360: \item

2361: {\sc Radmacher, M.D., Kelsoe, G. \& Kepler, T.B.}

2362: {\rm (1998)}.

2363: {\rm Predicted and Inferred Waiting-Times for Key Mutations 

2364: in the Germinal Center Reaction -- Evidence for Stochasticity 

2365: in Selection},

2366: {\it Immunol. Cell Biol.\/}

2367: {\bf 76},

2368: {\rm 373--381}.

2369: 

2370: 

2371: 

2372: \item

2373: {\sc Rundell, A., Decarlo, R., Hogenesch, H. \& Doerschuk, P.}

2374: {\rm (1998)}.

2375: {\rm The Humoral Immune-Response to Haemophilus-Influenzae 

2376: Type-B -- A Mathematical-Model Based on T-Zone and Germinal 

2377: Center B-Cell Dynamics},

2378: {\it J. Theor. Biol.\/}

2379: {\bf 194},

2380: {\rm 341--381}.

2381: 

2382: 

2383: 

2384: \item

2385: {\sc Smith, K., Light, A., Nossal, G. \& Tarlington, D.}

2386: {\rm (1997)}.

2387: {\rm The extent of affinity maturation differs between the 

2388: memory and antibody-forming cell compartments in the 

2389: primary immune response},

2390: {\it EMBO J.\/}

2391: {\bf 16},

2392: {\rm 2996--3006}.

2393: 

2394: 

2395: 

2396: \item

2397: {\sc Tew, J. \& Mandel, T.}

2398: {\rm (1979)}.

2399: {\rm Prolonged antigen half-life in the lymphoid follicles 

2400: of specifically immunized mice},

2401: {\it Immunology\/}

2402: {\bf 37},

2403: {\rm 69--76}.

2404: 

2405: 

2406: \item

2407: {\sc de Vinuesa, C.G., Cook, M.C., Ball, J., Drew, M., Sunners, Y.,

2408: Cascalho, M., Wabl, M., Klaus, G.G.B. \& MacLennan, C.M.}

2409: {\rm (2000)}.

2410: {\rm Germinal Centers without T Cells},

2411: {\it J. Exp. Med.\/}

2412: {\bf 191},

2413: {\rm 485--493}.

2414: 

2415: 

2416: \item

2417: {\sc Wedemayer, G.J., Patten, P.A., Wang, L.H., Schultz, 

2418: P.G. \& Stevens, R.C.}

2419: {\rm (1997)}.

2420: {\rm Structural insights into the evolution of an antibody 

2421: combining site},

2422: {\it Science\/}

2423: {\bf 276},

2424: {\rm 1665--1669}.

2425: 

2426: 

2427: 

2428: \item

2429: {\sc Weigert, M., Cesari, I., Yonkovitch, S. \& Cohn, M.}

2430: {\rm (1970)}.

2431: {\rm Variability in the light chain sequences of mouse antibody},

2432: {\it Nature\/}

2433: {\bf 228},

2434: {\rm 1045--1047}.

2435: 

2436: 

2437: 

2438: \item

2439: {\sc Zhang, J., MacLennan, I.C.M., Liu, Y.-J. \& Land, P.J.L.}

2440: {\rm (1988)}.

2441: {\rm Is rapid proliferation in B centroblasts linked to 

2442: somatic mutation in memory B cell clones},

2443: {\it Immunol. Lett.\/}

2444: {\bf 18},

2445: {\rm 297--299}.

2446: 

2447: 

2448: 

2449: \end{list}

2450: 

2451: 

2452: 

2453: 

2454: \end{document}

2455: 

2456: 

2457: 

2458: 

2459: 

2460: 

2461: 

2462: 

2463: 

2464: 

2465: 

2466: 

2467: 

2468: 

2469: 

2470: 

2471: 

2472: 

2473: