ARTICLES

Evidence from the generation of immunoglobulin G–

© 2004 Nature Publishing Group http://www.nature.com/natureimmunology

secreting cells that stochastic mechanisms regulate

lymphocyte differentiation
Jhagvaral Hasbold1,2,3, Lynn M Corcoran2, David M Tarlinton2, Stuart G Tangye1 & Philip D Hodgkin1,2,3

Naive B lymphocytes undergo isotype switching and develop into immunoglobulin-secreting cells to generate the appropriate class
and amount of antibody necessary for effective immunity. Although this seems complex, we report here that the generation of
immunoglobulin G–secreting cells from naive precursors is highly predictable. The probabilities of isotype switching and development
into secreting cells change with successive cell divisions and interleave independently. Cytokines alter the probability of each
differentiation event, while leaving intact their independent assortment. As a result, cellular heterogeneity arises automatically as
the cells divide. Stochastic division-linked regulation of heterogeneity challenges the conventional paradigms linking distinct
phenotypes to unique combinations of signals and has the potential to simplify our concept of immune complexity considerably.

The generation of antibody-secreting cells (ASCs) and the secretion of
antibody after antigen challenge in vivo is a complex process that is still
not completely understood. An historical view attributes most antibody
to long-lived, quiescent plasma cells lodged in bone marrow, which were
generated earlier in germinal centers in secondary lymphoid tissues1.
More recently it has become apparent that much of the antibody formed
arises early from short-lived ASCs generated before the development of
the germinal center2,3. ASCs that develop rapidly in extrafollicular sites
after primary immunization disappear after a few days through apopto-
sis4. In contrast, the long-lived ASCs residing in the bone marrow origi-
nate from germinal centers5,6, and their numbers are maintained by
continuous differentiation of memory B cells7,8 and their intrinsically
long half-life9–11. The development of ASCs is associated with the regu-
lated loss of lineage-specific surface markers. In mice, ASCs formed in
vivo can be identified by their high expression of syndecan-1 (Synd1;
CD138) in conjunction with low B220 (refs. 4,12,13).
B cells can be induced to secrete immunoglobulin in vitro with a
range of mitogenic stimuli, including T cells, T cell–derived compo-
nents, T cell–independent stimuli such as lipopolysaccharide, and
cross-linking surface immunoglobulin–specific antibodies14. Thus, in
vitro systems afford an opportunity to dissect B cell activity in con-
trolled conditions to provide insight into the more complex in vivo
systems. Over the last several years, quantitative examination of these
processes has shown a remarkable regularity in cell activity and a
strong reliance on the division history of the cells. Particularly notable
is the relationship between cell division number and the appearance
of isotype switched B cells after activation with T cell–dependent or T
cell–independent stimuli15–19. The proportion of cells expressing a
switched isotype is constant within a cell division and independent of
the time taken before entering the division, which varies considerably.
This quantitative regularity allows division number to serve as a
framework for monitoring the rate of cell differentiation and for mea-
suring the influences of cytokines on each aspect of the process18,20.
Here we show a similar division-linked regularity during the com-
mitment of mouse B cells to the ASC lineage and show that interleukin
5 (IL-5) increases immunoglobulin secretion solely by enhancing the
rate of ASC commitment per division round. Furthermore, ASCs gen-
erated in culture are a plasmablast population that continues to divide
in concert with other cells. Together these observations provided an
opportunity to measure how cytokines and cell division simultaneously
regulate the development of cells of mixed phenotype from non-
switched naive precursors. Our results show that the two differentiation
events, isotype switching and ASC commitment, operate indepen-
dently within cells and that a simple probabilistic, division-based
framework can be used to accurately predict the generation of three
new cell types from the undifferentiated starting population. This sto-
chastic mode of generating cells of mixed phenotype has broad impli-
cations for our understanding of the source and control of cell
heterogeneity and, by logical extension, of immune complexity.
RESULTS
The number of ASCs increases with each division
Stimulation of B cells in culture by the ligand for CD40 (CD40L) and
cytokines induces the appearance of secreted immunoglobulin in the
supernatant16,17,21–23. The source of immunoglobulin in these cul-
tures is poorly understood. One possibility is that the differentiation of

1Centenary Institute of Cancer Medicine and Cell Biology, Locked Bag Number 6, Newtown, NSW 2042, Australia. 2The Walter and Eliza Hall Institute of Medical
Research, 1G Royal Parade, Parkville, Victoria 3050, Australia. 3Present address: The Walter and Eliza Hall Institute of Medical Research, 1G Royal Parade,
Parkville, Victoria 3050, Australia. Correspondence should be addressed to P.D.H. (hodgkin@wehi.edu.au).

Published online 30 November 2003; doi:10.1038/ni1016

NATURE IMMUNOLOGY VOLUME 5 NUMBER 1 JANUARY 2004 55

 ARTICLES

Figure 1 IL-4-induced immunoglobulin secretion

shows a cell division–related increase in ASC 6 4
a 20 25
b c

IgG1+ASC/Total IgG1+ (%)

numbers. (a) B cells were activated with CD40L IgM IgG1 IgM
in the absence or presence of IL-4 (1,000 U/ml IgG1 20

Ig (× 103 ng/ml)
3 15
as indicated on horizontal axis.). After 7 d, 4

ASC (%)
culture supernatants were collected and amounts 15
of total IgM and IgG1 were determined. 2 10
10
(b) CFSE-labeled B cells were stimulated with 2
CD40L in the presence of IL-4. On day 4, cells 1 5
5
were collected and an equal number of cells in
each division were sorted by flow cytometry 0 0 0 0
0 1,000 0 1,000
© 2004 Nature Publishing Group http://www.nature.com/natureimmunology

directly into ELISpot plates, and the proportion 0 1 2 3 4 5 6 7 8 0 1 2 3 4 5 6 7 8

of sorted cells secreting IgM and IgG1 was IL-4 (U/ml) Division number Division number
determined by ELISpot assay. (c) CFSE-labeled
B cells stimulated with CD40L and IL-4 for 4 d were collected and stained with anti-IgG1–phycoerythrin. IgG1+ cells from consecutive divisions were
sorted directly into ELISpot plates to determine the frequency of secreting cells within the IgG1+ population for each division. Data represent mean and
s.e.m. of triplicate cultures.

a separate subpopulation of ASCs, such as a terminally differentiated
plasma cell, is induced. Alternatively, immunoglobulin secretion may be
an intrinsic feature of activated B cell blasts and so all cells would be
secreting some antibody. We examined this question with an in vitro
culture system based on CD40L stimulation of small resting B cells. The
addition of IL-4 together with CD40L considerably increased the
amount of both immunoglobulin M (IgM) and IgG1 secreted into the
culture supernatant over 4 d (Fig. 1a). To ascertain whether develop-
ment of ASCs in these cultures altered with division, we labeled B cells
with 5- (and 6-)carboxyfluorescein diacetate succinimidyl ester (CFSE)
and stimulated them for 4 d with CD40L and IL-4, and then sorted them
directly, based on the number of cell divisions, into enzyme-linked
immunospot (ELISpot) assay plates. ASC numbers showed a clear
increase with successive divisions, with IgM-secreting cells appearing in
earlier divisions than IgG1 secretors (Fig. 1b). The delay in IgG1 ASC
generation compared with that of IgM could not be accounted for by the
requirement for division-linked isotype switching15, as sorted IgG1+-
switched cells showed a division-associated increase in ASC number
similar to that of IgM+ cells (Fig. 1c). These results showed that a dis-
tinct immunoglobulin-secreting cell is being generated within these cul-
tures and is changing in frequency with each division.
Flow cytometric detection of ASCs
To detect the development of ASCs by flow cytometry, we examined cul-
tured cells for expression of the surface proteoglycan molecule Synd1.
This marker correlates well with the generation of ASCs in vivo4,12. We
confirmed this correlation for CD40L- and IL-4-stimulated B cells
in vitro (Fig. 2a). Sorted Synd1+ cells secreted 10- to 50-fold more

a c d Synd1 Figure 2 Identification of immunoglobulin-

IgM IgG1 40
1,000 400 N H producing B cells by flow cytometry. (a) B cells
C – + – + RT stimulated with CD40L and IL-4 for 4 d were
300 30 sort purified into Synd1+ (+) and Synd1– (–)
Synd1 (%)

750 Actb
Ig (ng/ml)

populations. The sorted cells were restimulated

Igj
500 200 20 with CD40L and IL-4 for a further 24 h, and
Prdm1 immunoglobulin concentrations in culture
100 10 Xpb1 supernatants were assessed. Data represent mean
250
Aicda and s.e.m. of triplicate samples. (b) Stimulated
0 cells from 4-day cultures were stained for Synd1.
0 0 Bcl6
– + – + 0 1 2 3 4 5 6 7 8 Top, CFSE versus Synd1 expression. Bottom,
Synd1 sort Division number Pax5 CFSE intensity. (c) The percentage of cells in
each division expressing Synd1. (d) Synd1+
Division number (H) and Synd1– (N) cells were sorted, RNA
b 104
e 7 6 5 4 3 2 1 0 was extracted and cDNA was prepared.
104
9.01
Semiquantitative RT-PCR was done to detect
103 H transcripts of Actb (β-actin), Igj (J chain), Prdm1
103
(Blimp-1), Xbp1 (Xpb-1), Aicda (AID), Bcl6
102 (Bcl-6) and Pax5 (Pax-5). Their products were
102
detected by Southern blot of PCR products, with
Synd1

Synd1

101 101 gene-specific probes. C, no-DNA control; – and +,

N first-strand synthesis without (–) or with (+) added
100 100 reverse transcriptase (RT). (e) B cells stimulated
100 101 102 103 104
with CD40L and IL-4 were collected on day 4
90
200 of culture, and Synd1+ (H) and Synd1– cells (N)
H N were sorted by gating on division 5 (divisions,
60
above dot plot). Cells were recultured with CD40L
Events

100
Events

30
and IL-4 for 24 h and reanalyzed by flow
cytometry. Bottom, CFSE histogram positions
0 0 of the sorted cells. Overlayed histograms are
100 101 102 103 104 100 101 102 103 100 101 102 103 104 shown for the sorted cells at 0 hour (thin lines)
CFSE CFSE and after 24 h of reculture (solid lines).

56 VOLUME 5 NUMBER 1 JANUARY 2004 NATURE IMMUNOLOGY

 ARTICLES

Figure 3 Tracking simultaneous immunoglobulin

switching and development of ASCs. CFSE- a 40 c 15 e
1,000 U/ml IL-4 30
labeled B cells were stimulated with CD40L and IgM 1,000 U/ml IL-4

Ig+Synd1+ (%)
100 U/ml IL-4

IgM+Synd1+ (%)
IgG1 1,000 U/ml IL-4
10, 100 or 1,000 U/ml of IL-4. (a,b) After 4 d, 30 10 U/ml IL-4
10 IgM 100 U/ml IL-4

IgG1 (%)
the proportion of cells in each division expressing IgG1 100 U/ml IL-4
20
IgG1 (a) and Synd1 (b) was determined by 20
flow cytometry. (c,d) The proportion of dual- 5
expressing IgM+Synd1+ (c) and IgG1+Synd1+ 10
(d) cells in each division was also measured. 10
Broken lines (d), proportions of IgG1+Synd1+ 0 0
cells expected if switching and ASC commitment 0 1 2 3 4 5 6 7 8 0 1 2 3 4 5 6 7 8
Division number Division number
© 2004 Nature Publishing Group http://www.nature.com/natureimmunology

are independent. (e) IgM+Synd1+ cells plotted as 0

a percentage of the total number of IgM+ cells in b 40 d 15
0 1 2 3 4 5 6 7 8
Division number
each division and compared with the equivalent

IgG1+Synd1+ (%)
calculation for the IgG1+Synd1+ cells at two IL-4 30

Synd1 (%)
concentrations. Data represent mean and s.e.m. 10
of triplicate samples. 20
5
10

0 0
0 1 2 3 4 5 6 7 8 0 1 2 3 4 5 6 7 8
Division number Division number

immunoglobulin into culture supernatants than did Synd1– cells dur-
ing a 24-hour secondary culture. Synd1 staining, in combination with
CFSE division tracking, yielded data consistent with the ELISpot data
(Fig. 1), with the percentage of Synd1+ cells increasing with each suc-
cessive division (Fig. 2b,c). The percent increase per division was sim-
ilar between the two measurements, consistent with the conclusion
that most Synd1+ cells were secreting antibody (compare Fig. 2c and
the combined values for IgM and IgG1 in Fig. 1b). The total number
of Synd1+ cells in culture also showed close correlation with the total
number of ASCs from similar cultures detected by ELISpot assay
(data not shown).
To confirm the plasma cell nature of the Synd1+ population, we
sorted cells and prepared mRNA from them, and examined a series
of plasma cell markers by RT-PCR (Fig. 2d). Synd1+ cells displayed a
gene expression profile of plasma cells, with increased expression of J
chain, B lymphocyte–induced maturation protein 1 (Blimp-1) and
X-box-binding protein 1 (Xbp-1) mRNA and a reduction in activa-
tion-induced cytidine deaminase (AID), B cell lymphoma 6 (Bcl-6)
and paired box gene 5 (Pax-5) mRNA24. Thus, the development of
a c
10 5 100
10 4
75
IgM (ng/ml)
ASCs in culture could be followed and quantified by monitoring
Synd1+ expression in combination with other cellular changes with
flow cytometry.
Synd1+ cells are a rapidly dividing population
Although a traditional view holds that ASCs are nondividing, many
reports have identified proliferating ASCs in early responses in
vitro and in vivo25–29. We also found evidence that ASCs generated
in culture were proliferating. DNA staining showed that 53% of the
Synd1+ population was in the S and G2-M phases of the cell cycle
(data not shown). Furthermore, when we sorted Synd1+ and
Synd1– cells from the fifth division and recultured them for a fur-
ther 24 h (Fig. 2e), most of the Synd1+ cells had divided at least
once, indicating they are an actively dividing population. This divi-
sion rate seemed to be slightly faster than that of the undifferenti-
ated Synd1– cells (Fig. 2e).
Isotype switching and ASC development
The increase in the frequency of Synd1+ ASCs with consecutive divi-
sions in CD40L- and IL-4-stimulated B cell
cultures parallels the activity of cells under-
e going isotype switching to IgG1 (refs.
75
15,16). To explore the relationship between
isotype switching and ASC commitment, we

Synd1 (%)

50
IgM (%)

monitored simultaneously the expression of

10 3
50
10 2 1,000 U/ml IL-5
100 U/ml IL-5 25
10 U/ml IL-5 25 Figure 4 IL-5 reduces switching and increases
10 1 0 U/ml IL-5 the appearance of Synd1+ cells in each division.
10 0 0 0 (a,b) CFSE-labeled B cells were stimulated
b d 0 1 2 3 4 5 6 7 8 in culture with CD40L and IL-4 and various
Division number concentrations of IL-5 (key). Cumulative IgM
40
10 3 (a) and IgG1 (b) concentrations in culture
IgG1 (ng/ml)

30 supernatants over 7 d were determined.

IgG1 (%)

10 2 (c,d) B cells from cultures equivalent to those

20 in a and b collected at day 4 were fixed,
10 1 permeabilized, stained and analyzed by flow
10
cytometry for intracellular and cell surface
10 0 0 expression of IgM+ (c), IgG1+ (d) and Synd1+
0 1 2 3 4 5 6 7 8 0 1 2 3 4 5 6 7 8 (e). The percent positive cells found in each
Time (d) Division number division is shown.

NATURE IMMUNOLOGY VOLUME 5 NUMBER 1 JANUARY 2004 57

 ARTICLES

After sort 18-hour culture Figure 5 ASCs downregulate division-associated

104 104 104
isotype switching. B cells stimulated with CD40L
0.39 0.87 and IL-4 in the presence of IL-5 for 4 d were
90.1
103 103 103 stained for Synd1 and IgG1 and sorted into two
+ populations: IgG1–Synd1+ and IgG1–Synd1–.
Synd1
102 102 102 Left, contour plot of analyses for Synd1 and IgG1
expression after sorting. Right, sorted cells from
101 101 101
each group were then recultured in the same
stimulation conditions for a further 18 h and
100 100 100
100 101 102 103 104 100 101 102 103 104 100 101 102 103 104 re-examined for IgG1 expression. Numbers
inside the gates on each plot show the percent
104 104 104
© 2004 Nature Publishing Group http://www.nature.com/natureimmunology

1.34 8.27 of positive cells for Synd1 or IgG1.

0.85
103 103 103

Synd1– 102 102 102

Synd1

101 101 101

IgG1

IgG1
100 100 100
100 101 102 103 104 100 101 102 103 104 100 101 102 103 104

CFSE CFSE

immunoglobulin isotype and of Synd1. As the amount of surface We sought to determine whether isotype switching by the dividing
immunoglobulin is lower on Synd1+ cells than on Synd1– B lymphoblasts affected their likelihood of subsequently becoming
B cells4,12, we fixed and permeabilized the cells before staining for an ASC. For this, we determined the proportion of Synd1+ cells
immunoglobulin to more clearly show the isotype expressed by the among both IgM+ and IgG1+ cells in each division after culture with
Synd1+ cells. Both isotype switching (Fig. 3a) and the appearance of 1,000 and 100 U/ml of IL-4. The results show that the proportion of
Synd1+ cells (Fig. 3b) were sensitive to IL-4 concentration and IgM+ and IgG1+ cells that were ASCs in each division were similar
increased with cell division number. The rate of appearance of (Fig. 3e), strongly indicating that switching did not affect the likeli-
IgM+Synd1+ cells increased with higher concentrations of IL-4, hood of becoming an ASC.
approaching different plateaus at later divisions (Fig. 3c). The
appearance of IgG1+ ASCs was also notably affected by IL-4 concen- IL-5 increases differentiation to ASC
tration; however, the dose-response curve was different from that A second candidate for a soluble regulator of ASC commitment per
for IgM ASCs, as few IgG1+ ASCs cells were generated with a con- division was IL-5. This cytokine increases secreted immunoglobulin in
centration of 10 U/ml of IL-4, and no plateau was seen at later divi- culture21,30. The IL-5-induced increase in IgM was greater than that for
sions for higher concentrations of IL-4 (Fig. 3d). Thus, the IgG1 (approximately 30-fold and 6-fold, respectively, above control;
generation of IgM and IgG1+ ASC cells seemed to have a different Fig. 4a,b). The highest concentration (1,000 U/ml) of IL-5 seemed to
sensitivity to IL-4 concentration. reduce the rate of appearance of IgG1-switched cells in each division,

Sorted
divisions: 2 3 4 5 2 3 4 5
a 104
b 104

103 0.08 0.14 0.13 0.08 103 0.31 0.55 0.55 2

102
102 After
After
sort
101 101 sort
100 100

103 1.46 1.65 1.91 2.88 103 5.58 8.8 15.8 29.1
18-hour
102 18-hour
102 culture
culture
Synd1

Synd1

101 101

100 100
100 101 102 103 100 101 102 103 100 101 102 103 100 101 102 103 104 100 101 102 103 100 101 102 103 100 101 102 103 100 101 102 103 104

CFSE CFSE

c 104
0.15 1.02 0.94 1.14
103

102
AfterFigure 6 Confirmation of division-associated increases in differentiation.
101 sort (a,b) CFSE-labeled B cells were stimulated in culture with CD40L and IL-4
100 in the absence (a) or presence (b) of IL-5. After 4 d, cells were collected
2.48 2.95 11.7 14.7 and Synd1– B cells were sorted from consecutive divisions 2–5 (above
103
plots). Top rows, reanalysis for Synd1 expression after sorting. Bottom
18-hour
102 rows, sorted cells were placed back in stimulatory cultures for a further
culture
24 h before being restained and analysed for Synd1 expression. (c) An
IgG1

101

100
experiment similar to that in a and b, except that cells were sorted by IgG1
100 101 102 103 100 101 102 103 100 101 102 103 100 101 102 103 104 expression and placed back in culture. Numbers indicate the percent of
CFSE cells in the gates boxed.

58 VOLUME 5 NUMBER 1 JANUARY 2004 NATURE IMMUNOLOGY

 ARTICLES

a Probability of switching b Probability of ASC commitment

0.2
No IL-5 No IL-5 10 U/ml IL-5 100 U/ml IL-5 1,000 U/ml IL-5
IgM+/Synd1–
Probability

IgM+/Synd1+
0.1 × =
IgG1+/Synd1–
IgG1+/Synd1+
0
0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7
Division number (q)
© 2004 Nature Publishing Group http://www.nature.com/natureimmunology

Cellular heterogeneity with division

c 100
No IL-5 10 U/ml IL-5 100 U/ml IL-5 1,000 U/ml IL-5
Percent of population
in division number

75
Observed
50
Predicted
25

0
1 2 3 4 5 6 7 1 2 3 4 5 6 7 1 2 3 4 5 6 7 1 2 3 4 5 6 7

Division number (q)

Figure 7 The generation of cellular heterogeneity by two independent differentiation events. The data from the experiment in Figure 4 were analyzed
quantitatively to test for independence of sorting two differentiation events. CFSE-labeled B cells were stimulated in culture for 4 d with CD40L and
100 U/ml of IL-4 and 0, 10, 100 or 1,000 U/ml of IL-5. After being collected, cells were stained for IgM, IgG1 and Synd1 expression. (a) The calculated
probability of becoming IgG1+ per division round for cultures without IL-5. These probabilities are calculated from the frequency data presented in Figure 4d
using equation (15) in Methods. Values are plotted against the half division scale (q) as described in Methods. (b) Results for the same calculations for
the probability of expressing Synd1 in cultures for each concentration of IL-5 from Figure 4e. (c) The experimental data observed for the proportion of
each of the four cell types (IgM+Synd1–, IgM+Synd1+, IgG1+Synd1– and IgG1+Synd1+) found in each division in cultures with CD40L and IL-4, plus IL-5
(concentrations, top right corners). These data points are shown in the colored symbols (key) and are not connected by lines. The dashed lines show the
result of calculating the expected proportion of each cell type using the relevant probabilities shown in a and b and the rules of independent assortment
given in Methods (equations (11–14) using sq and gq). The predicted proportion of each cell type is given in the same color as the relevant data symbol.

and this was mirrored by a small increase in the number of unswitched
IgM+ cells (Fig. 4c,d). When we examined the percent Synd1+ cells per
division, we found a substantial dose-dependent increase in the pres-
ence of IL-5. High concentrations of IL-5 produced a 6- to 10-fold
increase in the number of ASCs in each division (Fig. 4e). Additional
experiments determined that IL-5 did not affect the rate of prolifera-
tion or the survival of either Synd1+ or Synd1– cells (data not shown).
Thus, the increase in immunoglobulin secretion induced by IL-5 is
solely attributable to the increase in ASCs generated per division.
Committed ASCs no longer undergo isotype switching
IL-5 increased the rate of commitment to Synd1+ ASCs per division
and also seemed to reduce the overall rate of isotype switching
(Fig. 4c,d). These two observations could be related if commitment
to the ASC lineage downregulated division-associated isotype
switching. We examined this issue directly by comparing the switch-
ing rate per division of sorted IgG1–Synd1+ cells and IgG1–Synd1–
cells. The results show that dividing ASCs generated only 10% of the
isotype-switched cells produced by Synd1– cells (Fig. 5). Thus, the
greater number of ASCs generated early in culture in the presence of
IL-5 decreases the pool of cells capable of undergoing isotype
switching in each subsequent division. Therefore, IL-5 specifically
and exclusively increases the rate of commitment of B cell blasts at
each division to ASC differentiation, without affecting proliferation,
survival or immunoglobulin secretion rate. This result indicates that
the probability of ASC commitment per division can be altered
independently of other cellular functions and presents a notable
contrast to the influence of IL-4, which simultaneously regulates
isotype switching and ASC commitment (Fig. 3) as well as prolifer-
ation and survival16,17,31.
Confirmation of division-linked differentiation
It remained a formal possibility that the observed increase in ASCs
found in consecutive divisions was due to a separate ASC precursor
population that specifically expanded in the cultures, particularly
given the observation that the ASCs were an actively dividing popula-
tion. To directly test this hypothesis, we stimulated CFSE-labeled B
cells in culture for 4 d with CD40L and IL-4 in the presence or absence
of IL-5, then collected the cells and sorted Synd1– cells from each of
divisions 2–5. We placed sorted cells back in the identical culture con-
ditions for a further 18 h and then re-examined them for Synd1
expression. The results showed a change in differentiation rate per
division that was considerably increased by IL-5 (Fig. 6a,b).
Therefore, differentiation from non-ASC in later divisions is a major
contributor to the observed division-based increase in ASC number
(Figs. 1b and c, 2c and 4c). A similar experiment confirmed the
increasing frequency of IgG1 switched B cells from unswitched pre-
cursors with progressive division (Fig. 6c).
Stochastic mechanisms generate isotype switched ASC
It is possible to estimate the probability of becoming an ASC or of iso-
type switching within each division by analyzing the observed change in

NATURE IMMUNOLOGY VOLUME 5 NUMBER 1 JANUARY 2004 59

 ARTICLES
frequency of the population18,32. We further analyzed the data from the
experiment presented in Figure 4 with quantitative methods (Fig. 7a).
From the changing frequency of IgG1+ cells per division, we determined
the underlying probability of switching that is required to generate the
experimental curve (Methods, equation (15)). Repeating this estimation
for commitment to Synd1 expression at each of the IL-5 concentrations
demonstrated the effect of this cytokine (Fig. 7b). With these underlying
probabilities, we can attempt to predict the net number of secreting,
switched cells in each division in the presence of IL-4 with and without
added IL-5. Thus, at each division the estimated probabilities (Fig. 7a,b)
© 2004 Nature Publishing Group http://www.nature.com/natureimmunology
can be used to calculate how many of the unswitched, nonsecreting cells
entering that division will switch, become an ASC or undergo both
events. For the calculations shown it is assumed those cells that attempt
to do both will become switched ASCs. It is also assumed that ASCs con-
tinue to divide but do not switch and that switched cells are subject to
the same likelihood of becoming ASCs as nonswitched cells. With these
assumptions, the proportion of each of the four possible cell phenotypes
(IgM+ASC–, IgM+ASC+, IgG1+ASC– and IgG1+ASC+) can be calculated
for consecutive divisions (Fig. 7c). Further details for these calculations
are given in Methods. The predicted values are shown as dashed lines in
Figure 7c and can be compared with the observed values. There is excel-
lent agreement between predicted and observed results confirming the
independent probabilistic assortment of the two division-linked differ-
entiation events. These calculations also show how IL-5 can paradoxi-
cally reduce the overall rate of switching to IgG1 while increasing the
amount of secreted IgG1 (Fig. 4b,d). Thus, in the presence of high con-
centrations of IL-5 there is a net loss of switched cells; however, a larger
proportion of these cells become ASCs.
We used the same calculation method to predict the expected num-
ber of IgG1+ ASCs generated at different IL-4 concentrations from the
data presented in Figure 3. This calculation produced the dashed
‘expected’ lines in Figure 3d, which show a close fit to the observed
data points for IgG1+ ASCs at each of the three IL-4 concentrations.
Thus, the very different IL-4 dose response curves for the generation
of IgG1+ ASCs (Fig. 3d) compared with that of IgM+ ASCs arise as the
result of the two interleaved independent events, isotype switching
and ASC differentiation, and are not due to different IL-4 sensitivities
for IgM+ and IgG1+ cells.
DISCUSSION
Many of the main activity changes of stimulated naive B cells, includ-
ing proliferation, isotype switching and development of ASCs, can be
reproduced in a relatively short period after stimulation in vitro.
Additional stimuli provided by B cell receptor engagement and
cytokines can modulate this activity in a predictable way17,21,23,33,34.
This in vitro system provides an opportunity to dissect in detail the
contribution of both external and internal influences to cell fate deci-
sions. We have extensively studied the combination of the T cell–
derived stimuli CD40L and IL-4 on proliferation and immunoglobu-
lin isotype switching and here have extended the study to quantify the
process of ASC differentiation. By labeling cells with CFSE and then
sorting cells from individual divisions, we found that ASCs appeared
early in culture and arose at a low frequency, usually <5%, in the ear-
liest divisions. The frequency of ASC development increased with
consecutive divisions in these cultures in a manner similar to that
noted for isotype switching15,16. This division-associated increase
could potentially be explained by the outgrowth of ASC precursors or
by a division-associated mechanism that alters the probability of
undergoing differentiation. To explore these possibilities in greater
detail, we used flow cytometry to simultaneously monitor proliferation
and commitment to the ASC phenotype, based on the characteristic
60
acquisition of Synd1. Synd1+ cells correlated strongly with ASCs and
displayed a gene expression profile consistent with that of ASCs.
Tracking the development of Synd1+ cells confirmed the increase in
their frequency within consecutive divisions collected on the same
day. This analysis also showed that Synd1+ ASCs continued to divide
in culture. Thus, these cells are best described as plasmablasts and
may be the in vitro equivalent of the extrafollicular ASC population
found in vivo2–4,27 but not the nondividing plasma cells found in
bone marrow9–11. These cells also have properties similar to those of
the actively dividing ASC population generated from human memory
B cells28. The lineage relationship between plasmablasts and plasma
cells is unclear, and the requirements for further terminal differentia-
tion may not be fulfilled until these plasmablast cells home to the
bone marrow35 and/or receive additional signals only available in the
germinal center4,6.
Several cytokines are known to increase the amount of immunoglob-
ulin secreted by activated B cells in culture. As the generation of anti-
body is the result of a series of complex processes, net increases in
immunoglobulin amounts could result from enhanced activation, pro-
liferation, survival, differentiation or secretion rate. The ability to relate
the rate of differentiation to division number allowed us to determine
how the cytokines IL-4 and IL-5 increased the amount of immunoglob-
ulin secreted into supernatants after stimulation with CD40L. IL-4 has
multiple effects on CD40L-stimulated B cells, increasing proliferation,
survival and switching rate16,17. In addition to these effects, IL-4
increased the frequency of ASC development per division in a dose-
dependent manner. IL-5, in contrast, operated much more specifically.
This cytokine increased the frequency of ASC development per division,
but did not affect the rates of survival or proliferation. This isolated
effect of IL-5 confirmed that the probability of undergoing differentia-
tion per division could be directly regulated within the cell and is not
necessarily linked to, or the result of, simultaneous changes to division
rates as seen with IL-4. The probability of commitment to ASC was
altered both by changes in division number and by regulatory cytokines.
Many of the molecular triggers for commitment to the plasma cell lin-
eage have been identified. Blimp-1 is considered a ‘master regulator’ for
plasma cell differentiation36, although it is not apparent whether it has a
similar function in plasmablast development. The probability of com-
mitment to becoming an ASC might be linked to the level of expression
of key triggering molecules. Further experiments can answer these ques-
tions. Irrespective of the mechanism, however, the highly predictable
and manipulable nature of commitment to ASC in vitro should provide
an excellent system for studying complex regulatory processes relating
stochastic molecular events to biological outcomes.
We noted an asymmetry in the ‘downstream’ consequences of iso-
type switching and commitment to the ASC lineage. When cells under-
went isotype switching, they continued to divide and remained subject
to the independent mechanism for commitment to ASC. The reverse
situation, however, did not occur. Cells that had committed to ASC as
IgM+ cells were much less likely to undergo isotype switching, despite
continued division. The effect of this ASC dominance on the net num-
ber of switched cells is small when a relatively low proportion of cells
commit to ASC; for example, in cultures with IL-4 only. However when
IL-5 was present, there was a noticeable reduction in overall switching,
because of the enhanced rate of ASC commitment at the expense of
switching. For this reason, the enhancing effect of IL-5 on secreted IgM
is much greater than that for IgG1. Analysis of expression of AID, an
enzyme essential for isotype switching37, by RT-PCR showed a marked
reduction in AID mRNA in the Synd1+ plasmablast population. Thus,
the reduced frequency of switching among dividing ASCs is consistent
with the recent report that AID is reduced in plasma cells36.
VOLUME 5 NUMBER 1 JANUARY 2004 NATURE IMMUNOLOGY

When stimulated, B cells rapidly generate phenotypic heterogene-
ity, exemplified here by the emergence of the three differentiated cell
types, IgM+Synd1+, IgG1+Synd1– and IgG1+Synd1+, from IgM+
Synd1– precursors. Our quantitative analysis showed an unexpect-
edly simple mechanism controlled the generation of each new phe-
notype. Immunoglobulin isotype switching and ASC commitment
acted as independent division-linked processes within each cell. The
probability of each differentiation event per division could be esti-
mated from the frequency change and could be presented as a ‘map’
showing the associated probability per division. The cytokine IL-4
© 2004 Nature Publishing Group http://www.nature.com/natureimmunology
can alter the probability map for switching and generating ASCs,
whereas IL-5 solely influences the probability of commitment to
ASC. When stimulated in the presence of cytokines, the cells begin
to divide, and as they do so they are subject to the probabilities of
differentiation dictated by the concentration of cytokines in culture.
The numbers of each cell type to be found in each division can be
calculated from the probabilities associated with the culture condi-
tions, assuming they are operating independently. The emergence of
cellular heterogeneity among immunoglobulin-secreting cells is
therefore an inevitable and predictable consequence of their divi-
sion-linked differentiation activity. A system of independently alter-
able, division-linked stochastic cellular processes can efficiently
enable the generation of a very large number of unique cell pheno-
types. Thus, three such processes could ‘code’ for eight different cell
phenotypes, and ten could predictably yield over 1,000 different cell
types. The extent of the contribution of independent division-
alterable stochastic differentiation events to the development of
cellular heterogeneity is unknown at present, although salient sto-
chastic features of T cell differentiation have been noted for some
time38. Our study emphasizes how considerable complexity can be
easily and efficiently ‘encoded’ through a few division-linked sto-
chastic control mechanisms. This model offers a powerful interpre-
tive framework for further investigation of complex processes in
immune activity.
It is implicit in many immunological analyses that differentiated
cells of different phenotype each have a unique history of stimula-
tion and exposure to signals. By this view, the generation of hetero-
geneity from our starting cells must be attributed to variants in the
starting population or to heterogeneity in the stimulation that each
cell receives in the microenvironment of culture. However, both
explanations have difficulty accounting for the data presented here,
in which the number of switched immunoglobulin-secreting cells is
regulated by cytokine concentration according to simple rules of
independent probability. Our experiments raise this question: How
much is cell fate determined by external signals and how much is
due to internal, stochastic events playing out within each cell? By
adopting the first view, the experimentalist is forced to pursue the
continual dissection of phenotype into identifiable differentiation
pathways. By the other view, heterogeneity is encoded to a large
extent intrinsically, with perhaps only modest regulation of differ-
entiation rates by external signals. The consequence of the paradigm
adopted is profound. On the one hand, every environmental contri-
bution, and cellular difference must be identified and measured
before we can understand the immune system, reinforcing the cur-
rent idea of immense complexity. Alternatively, enormous hetero-
geneity could be regulated subtly by signals evoking small changes
in the probability of events that were already occurring at a measur-
able rate. If the latter explanation is correct, the former view pre-
sents a substantial barrier to the progress of this subject, and a
change to a probabalistic perspective may lead to rapid advances in
our understanding.
NATURE IMMUNOLOGY VOLUME 5 NUMBER 1 JANUARY 2004
ARTICLES
METHODS
Mice. CBA/H inbred mice of both sexes, 8–12 week old, were used for most
experiments. These were either obtained from the Animal Resource Center
(Perth, Australia) or bred at the animal facility of The Walter and Eliza Hall
Institute and were maintained in specific pathogen–free conditions in accor-
dance with institutional animal ethics committee regulations.
Reagents and antibodies. Cell membranes expressing mouse CD40L were
prepared as described39 from the Sf21 insect cell line infected with a mouse
CD40L–expressing baculovirus vector. Recombinant mouse IL-4 was pro-
vided by R. Kastelein (DNAX Research Institute, Palo Alto, California) and
recombinant IL-5 was a gift from A. Hapel (John Curtin School of Medical
Research, Canberra, Australia). Biotinylated antibody to mouse IgG1 (anti-
mouse IgG1; A85.1) and phycoerythrin-conjugated Synd1 antibody were pur-
chased from BD Pharmingen. Streptavidin-Tricolor was purchased from
Caltag Laboratories.
Cell culture and flow cytometry. Single-cell suspensions of small resting
B cells were prepared from spleens of CBA/H mice by T cell depletion and
were purified on Percoll density gradients as described14,18. B cells were typi-
cally 90–95% IgM+IgD+B220+, as determined by flow cytometry. These cells
were labeled for division tracking with CFSE (Molecular Probes) according to
the originally published method40. B cells were cultured in B cell medium
comprising RPMI 1640 medium (Gibco BRL) with 2 mM L-glutamine, 1 mM
Na-pyruvate, 0.1 mM nonessential amino acids, 10 mM HEPES, pH 7.4,
100 µg/ml of streptomycin, 100 U/ml of penicillin, 50 µM β-mercap-
toethanol (all supplements were purchased from Sigma) and 10% heat-
inactivated FCS (CSL). Typically, 1 × 105 purified B cells were cultured in 1 ml
or 2 × 104 cells, in 0.2 ml cell culture medium.
The stimulated B cells were collected at various times and washed twice
with a solution of 0.1% BSA and 0.1% NaN3 in PBS (PBA solution). For
intracellular staining, approximately 5 × 105 cells were fixed for 10 min on
ice with 0.25 ml 2% paraformaldehyde (BDH). The fixed cells were then per-
meabilized overnight at room temperature by the addition of 0.25 ml PBS
and 0.5 ml 0.2% Tween-20 to yield a final concentration of 0.5%
paraformaldehyde with 0.1% Tween-20. Cells were then washed twice with
PBA and stained with primary monoclonal antibodies for 45 min and incu-
bated with Streptavidin-Tricolor for 30 min. All incubations were done on
ice, and after each step cells were washed twice with cold PBA. Samples were
analyzed on FacScan or LSR flow cytometers (Becton Dickinson) with
CellQuest (Becton Dickinson) or Flow-Jo (Tree Star) software. Cells were
sorted on a FacStar+ (Becton Dickinson) or MoFlo (Cytomation) cytometer.
For ELISpot assays, cells were directly deposited into 96-well filter plates
with an automated cell deposition unit.
CFSE peaks were fitted and individually gated with the proliferation feature
of FlowJo software (Treestar), and the proportion of immunoglobulin- or
Synd1-positive cells within each division round was calculated as described41.
In some experiments, the number of live cells per culture was estimated by a
reference to a known number of CaliBRITE beads (unlabeled; Becton
Dickinson), run simultaneously with the cells on the flow cytometer17,41.
Enzyme-linked immunosorbent assay (ELISA) and ELISpot assay. For
ELISA and ELISpot assays, antibodies purchased from Southern
Biotechnology Associates were goat anti-mouse IgM or IgG1 to coat the
plates, and biotinylated goat anti-IgM or IgG1 as a secondary detection step.
Purified mouse monoclonal IgM (TEPC 183) and IgG1 (MOPC 21) were
purchased from Sigma and were used as a standard for the quantification of
immunoglobulin concentrations. To develop the ELISAs, streptavidin–
horseradish peroxidase conjugate (Jackson ImmunoResearch Laboratories)
was used, followed by 2,2′-azino-bis(3-ethylbenz-thiazoline-6-sulfonic acid)
(Sigma) substrate reaction. ELISpot assays were done on MultiScreen-HA
filter plates (Millipore). Typically, 1 × 102 to 2 × 102 cells in 0.2 ml cell culture
medium were incubated on precoated plates for 4 h at 37 °C. Development
of spots with alkaline phosphatase was as described42.
RT-PCR analysis of plasma cell markers. CFSE-labeled cells that had been cul-
tured in CD40L and IL-4 for 4 d were stained for Synd1, and late divisions were
61
 ARTICLES
sorted into Synd1-low and Synd1-high fractions (Fig. 2b). Total RNA was pre-
pared with an Rneasy Mini Kit with RNase-Free DNase treatment (both prod-
ucts from Qiagen). First-strand cDNA was generated with the SuperScript
First-Strand Synthesis System (Invitrogen). Amounts of cDNA were titrated by
β-actin RT-PCR on serial dilutions of the first-strand reaction products. The
products were separated by gel electrophoresis, transferred to a nylon mem-
brane and probed with a β-actin probe. The specific signals were quantified on
a Molecular Dynamics PhosphorImager with the ImageQuant analysis pro-
gram. All subsequent PCR reactions were done on template dilutions giving
equivalent β-actin signals in the linear range of the assay. The RT-PCR products
were separated by gel electrophoresis, transferred and probed as described
© 2004 Nature Publishing Group http://www.nature.com/natureimmunology
above, to detect the specific products shown in Figure 2e. The probes used in
each case corresponded to the RT-PCR fragments themselves, which have all
been cloned and sequenced to confirm their identities. Primers used were:
β-actin, 5′-GTGGGCCGCTCTAGGCACCAA-3′ and 3′-CTCTTTGATGTC
ACGCACGATTTC-5′; J chain, 5′GTCTTCACTGGGGAGTCCTAGCC-3′ and
3′-GGG TGCAAATGGAGAGCCTCTAAGG-5′; Blimp-1, 5′-CATTCCTGTC
CCCAACGCATCAACTG-3′ and 3′-GGTGCCCAAGCACCAAAGTCATAGC-
5′; Xbp-1, 5′-GCTGGAGCAGCAAGTGGTGGATTTGG-3′ and 3′-GGCTTC
CAGCTTGGCTGATGAGGTCC-5′, AID, 5′-CCGGCACGTGGCTGAGTTT-3′
and 3′-GATGCGCCGAAGTTGTCTGGTTAG-5′; bcl-6, 5′-CAGCACCTTCC
TCTTCTCTGATGAGGAGCTCC-3′ and 3′-CTGGCGGAGAGCCAGAGG
CCTGAAGGATGC-5′; and Pax-5, 5′-CCTACCCTATTGTCACAGGCC-3′ and
3′-CCTCTGTCTGTCTCAGGGGGTT-5′.
Calculating differentiation probabilities per division. A method for calcu-
lating the proportion of cells of mixed phenotype expected in each division
was developed using rules based on experiments presented here. This method
assumes that any cell death or variation in the time to divide does not select for
a subpopulation of cells more or less likely to differentiate. These assumptions
are consistent with published experiments16–18. We also assume for these cal-
culations the differentiation event does not alter the division rate of the cell.
This is correct for isotype switching, but is a minor simplification for ASCs. A
starting population of cells when stimulated will undergo activation and a
series of divisions. It is assumed that the conditions initiate internal cell pro-
gramming that sets a probability of differentiation within each division round
that may change with each division. The consecutive sequence of probabilities
is referred to as the division-based differentiation ‘map’18. If the map is known
the cumulative proportion of differentiated cells at the end of each division
can be calculated. In general, the proportion of differentiated cells at the end of
the division round i (Qi) is given by:
Qi = pi(1 – Q(i – 1)) + Q(i – 1), (1)
where pi is the probability of an undifferentiated cell differentiating between
the beginning and end of division i.
When two events (that is, isotype switching and ASC commitment) can
occur, a mixture of cell phenotypes becomes possible. If the two events are
operating independently, simple probability calculations yield the propor-
tion of the cells that began the division round undifferentiated that will
undergo one or other event only, or initiate both events. For those cells that
would attempt both differentiation events, a rule must be given to resolve the
phenotypic outcome. The simplest rule, that both events occur in all cases, is
assumed here. For calculating the proportion of the four possible pheno-
types associated with isotype switching and commitment to Synd1 expres-
sion from individual differentiation maps, the following formulas can be
used, calculated in sequence from division 1 (gi and si are the probabilities of
switching to IgG1 in division i and committing to Synd1 expression in divi-
sion i, respectively). The formulas also take account that Synd1+ cells are no
longer subject to switching in subsequent divisions, whereas switched
Synd1– cells have the same probability of committing to Synd1 expression as
unswitched cells. Thus:
Q(M+S1–)i = Q(M+S–)(i – 1) – (new G+ + new M+S+) (2)
Q(M+S+)i = Q(M+S+)(i – 1) + (new M+S+) (3)
Q(G+S–)i = Q(G+S–)(i – 1) + (new G+S–) – (new G+S+ from G+ precursors) (4)
Q(G+S+)i = Q(G+S+)(i – 1) + (new G+S+), (5)
62
where M, G and S represent IgM, IgG and Synd1, respectively, and new cells
are the proportion generated between the beginning and end of the division
round i. The values of the new cells required for the above formulas can be cal-
culated as follows:
New G+ = giQ(M+S–)(i – 1) (6)
New M+S+ = siQ(M+S–)(i – 1) – (gisiQ(M+S–)(i – 1)) (7)
New G+S– = giQ(M+S–)(i – 1) – (gisiQ(M+S–)(i – 1)) (8)
New G+S+ from G+ precursors = siQ(G+S–)(i – 1) (9)
New G+S+ = gisiQ(M+S–)(i – 1) + siQ(G+S–)(i – 1) (10)
Substituting and simplifying yields the following for calculating the values of
Qi for each cell type:
Q(M+S–)i = (1 – gi)(1 – si)Q(M+S–)(i – 1) (11)
Q(M+S+)i = Q(M+S+)(i – 1) + si(1 – gi)Q(M+S–)(i – 1) (12)
Q(G+S–)i = (1 – si)Q(G+S–)(i – 1) + gi(1-si)Q(M+S–)(i – 1) (13)
Q(G+S+)i = Q(G+S+)(i – 1)+ gisiQ(M+S–)(i – 1) + siQ(G+S–)(i – 1) (14)
Typically the values of gi and si must be estimated from the changes in fre-
quency found in each division observed by CFSE experiment. This involves a
simplification. In general, cells are dividing asynchronously and the cohort of
cells found within a division are relatively evenly distributed between those
that have recently divided and those about to divide again. Thus, the fre-
quency change between, for example, cells making up the CFSE peak for divi-
sions 3 and 4 represents, on average, the transition from division 3.5 to 4.5
(where 0.5 represents progression through half the cell cycle and we assume
differentiation is equally likely at any phase of the cell cycle). Thus, when
using CFSE data for calculations it is convenient to use ‘q’ instead of i in the
above formulas, where q represents discrete division intervals starting at 0.5
and continuing 1.5, 2.5 and so on. The probability of differentiating between
division intervals q – 1 and q can be estimated by taking the proportional
change in differentiation frequency between the divisions as a fraction of the
undifferentiated frequency at q – 1. That is:
pq = (Q(d)q – Q(d)(q – 1))/(1 – Q(d)(q – 1)), (15)
where pq is the probability of differentiating between divisions q and q – 1, and
Q(d)q is the proportion of differentiated cells at the end of division q. Division
0 is a special case in which the difference from the time 0 phenotype is used.
Plots of pq versus division number yield the probability map required.
Equations (2–14) can then be used with gq and sq in place of gi and si to predict
the frequency of differentiated cells at CFSE division q.
ACKNOWLEDGMENTS
We thank M. Kehry and B. Castle for gifts of reagents used in this study. Supported
by the National Health and Medical Research Council of Australia.
COMPETING INTERESTS STATEMENT
The authors declare that they have no competing financial interests.
Received 26 June; accepted 21 October 2003
Published online at http://www.nature.com/natureimmunology/
1. Benner, R., Hijmans, W. & Haaijman, J.J. The bone marrow: the major source of
serum immunoglobulins, but still a neglected site of antibody formation. Clin. Exp.
Immunol. 46, 1–8 (1981).
2. van Rooijen, N., Claassen, E. & Eiekelenboom, P. Is there a single differentiation
pathway for all antibody-forming cells in the spleen? Immunol. Today 7, 193–195
(1986).
3. Jacob, J., Kassir, R. & Kelsoe, G. In situ studies of the primary immune response to
(4-hydroxy-3- nitrophenyl)acetyl. I. The architecture and dynamics of responding
cell populations. J. Exp. Med. 173, 1165–1175 (1991).
4. Smith, K.G., Hewitson, T.D., Nossal, G.J. & Tarlinton, D.M. The phenotype and fate
of the antibody-forming cells of the splenic foci. Eur. J. Immunol. 26, 444–448
(1996).
5. Smith, K.G., Light, A., Nossal, G.J. & Tarlinton, D.M. The extent of affinity matura-
tion differs between the memory and antibody-forming cell compartments in the pri-
mary immune response. EMBO J. 16, 2996–3006 (1997).
6. Takahashi, Y., Dutta, P.R., Cerasoli, D.M. & Kelsoe, G. In situ studies of the primary
immune response to (4-hydroxy-3- nitrophenyl)acetyl. V. Affinity maturation devel-
ops in two stages of clonal selection. J. Exp. Med. 187, 885–895 (1998).
VOLUME 5 NUMBER 1 JANUARY 2004 NATURE IMMUNOLOGY

7. Szakal, A.K., Kosco, M.H. & Tew, J.G. Microanatomy of lymphoid tissue during
humoral immune responses: structure function relationships. Annu. Rev. Immunol.
7, 91–109 (1989).
8. Ochsenbein, A.F. et al. Protective long-term antibody memory by antigen-driven and
T help-dependent differentiation of long-lived memory B cells to short-lived plasma
cells independent of secondary lymphoid organs. Proc. Natl. Acad. Sci. USA 97,
13263–13268 (2000).
9. Ho, F., Lortan, J.E., MacLennan, I.C. & Khan, M. Distinct short-lived and long-lived
antibody-producing cell populations. Eur. J. Immunol. 16, 1297–1301 (1986).
10. Slifka, M.K., Antia, R., Whitmire, J.K. & Ahmed, R. Humoral immunity due to long-
lived plasma cells. Immunity 8, 363–372 (1998).
11. Manz, R.A., Lohning, M., Cassese, G., Thiel, A. & Radbruch, A. Survival of long-lived
plasma cells is independent of antigen. Int. Immunol. 10, 1703–1711 (1998).
© 2004 Nature Publishing Group http://www.nature.com/natureimmunology
12. Lalor, P.A., Nossal, G.J., Sanderson, R.D. & McHeyzer-Williams, M.G. Functional
and molecular characterization of single, (4-hydroxy-3-nitrophenyl)acetyl (NP)-spe-
cific, IgG1+ B cells from antibody-secreting and memory B cell pathways in the
C57BL/6 immune response to NP. Eur. J. Immunol. 22, 3001–3011 (1992).
13. Kopper, L. & Sebestyen, A. Syndecans and the lymphoid system. Leuk. Lymphoma
38, 271–281 (2000).
14. Hodgkin, P.D. & Kehry, M.R. Methods for polyclonal B lymphocyte activation to pro-
liferation and Ig secretion in vitro. in Weir’s Handbook of Experimental Immunology,
Vol. 89, 89.1–89.13 (eds. Herzenberg, L.A., Weir, D.M., Herzenberg, L.A. &
Blackwell, C.) (Blackwell Science, Cambridge, Massachusetts, 1996).
15. Hodgkin, P.D., Lee, J.H. & Lyons, A.B. B cell differentiation and isotype switching is
related to division cycle number. J. Exp. Med. 184, 277–281 (1996).
16. Hasbold, J., Lyons, A.B., Kehry, M.R. & Hodgkin, P.D. Cell division number regulates
IgG1 and IgE switching of B cells following stimulation by CD40 ligand and IL-4.
Eur. J. Immunol. 28, 1040–1051 (1998).
17. Hasbold, J., Hong, J.S., Kehry, M.R. & Hodgkin, P.D. Integrating signals from IFN-γ
and IL-4 by B cells: positive and negative effects on CD40 ligand-induced prolifera-
tion, survival, and division-linked isotype switching to IgG1, IgE, and IgG2a.
J. Immunol. 163, 4175–4181 (1999).
18. Deenick, E.K., Hasbold, J. & Hodgkin, P.D. Switching to IgG3, IgG2b, and IgA is
division linked and independent, revealing a stochastic framework for describing dif-
ferentiation. J. Immunol. 163, 4707–4714 (1999).
19. Tangye, S.G., Ferguson, A., Avery, D.T., Ma, C.S. & Hodgkin, P.D. Isotype switching
by human B cells is division-associated and regulated by cytokines. J. Immunol.
169, 4298–4306 (2002).
20. Janas, M.L., Hodgkin, P., Hibbs, M. & Tarlinton, D. Genetic evidence for Lyn as a
negative regulator of IL-4 signaling. J. Immunol. 163, 4192–4198 (1999).
21. Maliszewski, C.R. et al. Recombinant CD40 ligand stimulation of murine B cell
growth and differentiation: cooperative effects of cytokines. Eur. J. Immunol. 23,
1044–1049 (1993).
22. Hollenbaugh, D., Ochs, H.D., Noelle, R.J., Ledbetter, J.A. & Aruffo, A. The role of
CD40 and its ligand in the regulation of the immune response. Immunol. Rev. 138,
23–37 (1994).
23. McIntyre, T.M., Kehry, M.R. & Snapper, C.M. Novel in vitro model for high-rate IgA
class switching. J. Immunol. 154, 3156–3161 (1995).
24. Calame, K.L. Plasma cells: finding new light at the end of B cell development. Nat.
NATURE IMMUNOLOGY VOLUME 5 NUMBER 1 JANUARY 2004
ARTICLES
Immunol. 2, 1103–1108 (2001).
25. Jelinek, D.F. & Lipsky, P.E. The role of B cell proliferation in the generation of
immunoglobulin-secreting cells in man. J. Immunol. 130, 2597–2604 (1983).
26. Jego, G. et al. Reactive plasmacytoses are expansions of plasmablasts retaining the
capacity to differentiate into plasma cells. Blood 94, 701–712 (1999).
27. Sze, D.M., Toellner, K.M., Garcia de Vinuesa, C., Taylor, D.R. & MacLennan, I.C.
Intrinsic constraint on plasmablast growth and extrinsic limits of plasma cell sur-
vival. J. Exp. Med. 192, 813–821 (2000).
28. Tangye, S., Avery, D. & Hodgkin, P. A division-linked mechanism for the rapid generation
of Ig-secreting cells from human memory B cells. J. Immunol. 170, 261–269 (2003).
29. Tangye, S., Avery, D., Deenick, E. & Hodgkin, P. Intrinsic differences in the prolifer-
ation of naive and memory human B cells as a mechanism for enhanced secondary
immune responses. J. Immunol. 170, 686–694 (2003).
30. Takatsu, K. Interleukin 5 and B cell differentiation. Cytokine Growth Factor Rev. 9,
25–35 (1998).
31. Rush, J.S. & Hodgkin, P.D. B cells activated via CD40 and IL-4 undergo a division
burst but require continued stimulation to maintain division, survival and differenti-
ation. Eur. J. Immunol. 31, 1150–1159 (2001).
32. Hasbold, J. et al. Quantitative analysis of lymphocyte differentiation and prolifera-
tion in vitro using carboxyfluorescein diacetate succinimidyl ester. Immunol. Cell.
Biol. 77, 516–522 (1999).
33. Snapper, C.M., Kehry, M.R., Castle, B.E. & Mond, J.J. Multivalent, but not divalent,
antigen receptor cross-linkers synergize with CD40 ligand for induction of Ig synthe-
sis and class switching in normal murine B cells. A redefinition of the TI-2 vs T cell-
dependent antigen dichotomy. J. Immunol. 154, 1177–1187 (1995).
34. Rush, J.S., Hasbold, J. & Hodgkin, P.D. Cross-linking surface Ig delays CD40 ligand-
and IL-4-induced B cell Ig class switching and reveals evidence for independent reg-
ulation of B cell proliferation and differentiation. J. Immunol. 168, 2676–2682
(2002).
35. Hargreaves, D.C. et al. A coordinated change in chemokine responsiveness guides
plasma cell movements. J. Exp. Med. 194, 45–56 (2001).
36. Shaffer, A.L. et al. Blimp-1 orchestrates plasma cell differentiation by extinguishing
the mature B cell gene expression program. Immunity 17, 51–62 (2002).
37. Muramatsu, M. et al. Class switch recombination and hypermutation require activa-
tion-induced cytidine deaminase (AID), a potential RNA editing enzyme. Cell 102,
553–563 (2000).
38. Kelso, A. et al. Heterogeneity in lymphokine profiles of CD4+ and CD8+ T cells and
clones activated in vivo and in vitro. Immunol. Rev. 123, 85–114 (1991).
39. Kehry, M.R. & Castle, B. Regulation of CD40 ligand expression and use of recombi-
nant CD40 ligand for studying B cell growth and differentiation. Semin. Immunol. 6,
287–294 (1994).
40. Lyons, A.B. & Parish, C.R. Determination of lymphocyte division by flow cytometry.
J. Immunol. Methods 171, 131–137 (1994).
41. Lyons, A.B., Hasbold, J. & Hodgkin, P.D. Flow cytometric analysis of cell division
history using dilution of carboxyfluorescein diacetate succinimidyl ester, a stably
integrated fluorescent probe. Methods Cell Biol. 63, 375–398 (2001).
42. Sedgwick, J.D. & Holt, P.G. The ELISA-plaque assay for the detection and enumera-
tion of antibody- secreting cells. An overview. J. Immunol. Methods 87, 37–44
(1986).
63