Download Quantitative analysis of lymphocyte differentiation and proliferation

Immunology and Cell Biology (1999) 77, 516–522
Special Feature
Quantitative analysis of lymphocyte differentiation and proliferation
in vitro using carboxyfluorescein diacetate succinimidyl ester
J H A S B O L D, AV G E T T, J S RU S H , E D E E N I C K , D AV E RY, J J U N a n d P D H O D G K I N
Immune Regulation Group, Medical Foundation, University of Sydney, The Centenary Institute of Cancer Medicine
and Cell Biology, Sydney, New South Wales, Australia
Abstract Mature T and B lymphocytes respond to receptor-delivered signals received during and following
activation. These signals regulate the rates of cell death, growth, differentiation and migration that ultimately
establish the behaviour patterns collectively referred to as immune regulation. We have been pursuing the philosophy that in vitro systems of lymphocyte stimulation, when analysed quantitatively, help reveal the logical attributes of lymphocyte behaviour. The development of carboxyfluorescein diacetate succinimidyl ester (CFSE) to
track division has enabled the variable of division number to be incorporated into these quantitative analyses. Our
studies with CFSE have established a fundamental link between differentiation and division number. Isotype
switching, expression of T cell cytokines, surface receptor alterations and changes to intracellular signalling components all display independent patterns of change with division number. The stochastic aspects of these changes
and the ability of external signals to independently regulate them argue for a probabilistic modelling framework
for describing and understanding immune regulation.
Key words: carboxyfluorescein diacetate succinimidyl ester, cell division, flow cytometry, immune regulation,
lymphocyte differentiation.
Introduction
The discovery by Lyons and Parish that cells labelled with the
dye (5- and 6-) carboxyfluorescein diacetate succinimidyl
ester (CFSE) accurately apportion fluorescence between
daughter cells upon mitosis made it possible for the division
history of a cell population to be monitored by flow cytometry.1 Importantly, cellular differentiation turned out to be
unaffected by the dye, because labelled lymphocytes were
shown to isotype switch, develop cytokine secretion potential
and alter their cell surface phenotype as normal.2–4 Thus, it
became possible to explore how lymphocytes integrate simultaneous proliferation- and differentiation-inducing signals
and compute a net outcome.
Methods
Methods used in our laboratory have been discussed in some detail
previously;5 however, a brief summary of important practical information follows.
Labelling method
Typically we use purified naïve lymphocytes for differentiation and
proliferation studies. For B cells, percoll is used to separate high
density cells,6 whereas T cells with the desired phenotype are sorted
from pooled lymph nodes.4 Naïve cells are excellent subjects for
Correspondence: Philip D Hodgkin, Centenary Institute of Cancer
Medicine and Cell Biology, Locked Bag 6, Newtown, NSW 2042,
Australia. Email: <[email protected]>
Received 2 September 1999; accepted 2 September 1999.
CFSE labelling, because they are relatively homogeneous in size and
the level of labelling with the dye is approximately proportional to
cell volume. The CFSE method is less well suited to labelling more
heterogeneous populations, such as dividing cell lines.
A stock solution of CFSE should be prepared by dissolving
CFSE (Molecular Probes, Eugene, OR, USA) in DMSO at a concentration of 5 mmol/L. This stock can be stored frozen in small
aliquots to avoid excessive freeze and thaw cycles.
To label cells with CFSE, they are suspended at 107 cells/mL in
PBS with 0.1% BSA (PBS/BSA). Sufficient CFSE stock solution is
added directly to make a final concentration of 10 mmol/L. The suspension should be vortexed immediately following CFSE addition, to
ensure rapid dispersal, and then left to incubate for 10 min in a 37°C
water bath. After labelling, a 10× volume of PBS/0.1% BSA is added
and cells are washed twice in culture medium. Labelled cells can
then be placed directly in culture under conditions where they are
stimulated to divide. After harvesting they are run on the flow
cytometer under standard conditions for monitoring fluorescein.
Label concentration
The fluorescence intensity of labelled cells is directly proportional to
the concentration of CFSE.5 However, at high concentrations the
CFSE load becomes toxic to cells (giving very bright, but very dead
cells). It is advisable to titrate the CFSE concentration and therefore
labelling intensity for each new cell type because the optimal conditions will vary. Curiously, T cells take up twice as much dye as B cells
and human cells are more brightly stained than mouse lymphocytes.
Controls for flow cytometry
When CFSE-labelled lymphocytes are simply left in culture at 37°C,
the intensity of fluorescence decays rapidly over the first 48 h.5
Lymphoctye differentiation analysis with CFSE
517
Figure 1 Monitoring lymphocyte responses in vitro using carboxyfluorescein diacetate succinimidyl ester (CFSE). (a) Basic features of
a CFSE tracking experiment for cell division using B cell stimulation by CD40L for 4 days. The dashed line at the right illustrates the CFSE
intensity of B cells cultured unstimulated with IL-4 alone for the same time period. These cells serve as the control to locate the undivided
cell position. The solid line shows the dividing B cell population revealing the typical asynchronous profile. Progressive divisions are apparent by the accurate two-fold dilutions of the fluorescence intensity, which on the log scale appear as even spacings. The dashed line to the
left illustrates the position of an equivalent population that was not labelled with CFSE, allowing the position of the autoflourescent intensity to be determined. (b) The same cells presented as a forward versus side scatter plot. The live cell gate set is illustrated, as is the position of CaliBRITE beads used to determine the total number of cells in culture. (c) Use of computer fitting to determine the best series of
Gaussian curves to accommodate the experimental data. The proportion of cells in each division can be determined by the relative area under
each normal curve which, when combined with the ratio of total live cells compared to the beads, can be used to determine the absolute
number of cells in each division in the culture (d). (e) Use of simultaneous surface staining with a different fluorochrome is illustrated by
IgG1 expression. Clearly, the most divided cells are more likely to express IgG1. This is further illustrated by ‘slicing’ the data and calculating the proportion of cells in each division that express IgG1 (f). In addition to cell surface labelling, CFSE-stained cells can be fixed
and permeabilized for concurrent monitoring of intracellular components. This is illustrated in (g), which shows labelling of bromodeoxyuridine (BrdU) incorporation following a 3 h pulse. (h) STAT-1 expression diminishes progressively with each division.
518
J Hasbold et al.
Presumably this is due to catabolism of CFSE-bound proteins in the
cell. This loss can be as high as 10-fold and occurs in the absence of
cell division. After 48 h cells will continue to lose CFSE, although
at a much reduced rate. For this reason it is always important to know
where the undivided peak would be when analysing a proliferating
group of cells. Fortunately, this is easily determined by controls,
because the rate of loss of fluorescence in unstimulated cells is identical to that observed for stimulated and activated cells.5 Therefore,
the CFSE intensity of unstimulated cells can be used to determine
the position of the undivided peak in stimulated cultures where no
undivided cells are apparent (Fig. 1a).
Another useful and important control is to set up cells activated
under the same conditions but not labelled with CFSE. This autofluorescence control is useful for determining the limits of resolution
of division tracking (Fig. 1a).
Monitoring the total cell number in culture
If it is important to measure the total number of cells in each division, a known number of beads can be added prior to harvesting cells
from culture wells. We typically add 5000 CaliBRITE beads
(Becton-Dickinson, San Jose, CA, USA) to a 1 mL culture. These
beads are easily distinguished from cells by the scatter profile
following acquisition on the flow cytometer (Fig. 1b). Because the
number of beads per well is known, a simple ratio determination can
be used to estimate final cell numbers accurately. For example, if
5000 beads were added to the culture well and, following analysis by
flow cytometry, the ratio of beads to live cells was 2:1 then there
were 2500 cells in the culture. This number can be combined with the
determination of the proportion of cells in each division to calculate
the total cell number in each division (see later).
How many divisions can be tracked?
The number of divisions that can be followed after CFSE labelling
is limited by the autofluorescent level of unlabelled cells. Thus, as
the cells divide and dilute CFSE accurately in two, the fluorescence
approaches the level of unlabelled cells. Typically, the intensity is
plotted on a log scale and therefore the geometric two-fold reductions will appear as even spacings, as shown in Fig. 1a and by Lyons
and Parish1 and Hodgkin et al.2 As the cells approach the autofluorescent level, the peaks appear to get closer together. For most
applications, seven to eight divisions is the practical limit for accurate resolution. The expected position of each peak for any division
(Di, where i is the division number) can be accurately calculated
using the formula:5
Di = ((D0 – A)/2i) + A
[1]
where D0 is the position of the undivided peak and A is the average
(geometric) autofluorescence intensity.
Computer-based fitting
We have used PROFIT (Quantumsoft, Zurich, Switzerland), or EXCEL
(Microsoft, Redmond, WA, USA) to fit a series of Gaussian curves
to CFSE data using the above formula to set the position of each division peak. Typically, the results are consistent with the expected features of sequential log-normal Gaussian distributions (Fig. 1c). By
fitting a series of curves, the area under each peak can be expressed
as the proportion of live cells in each division, which, by reference
to bead numbers, can then be used to calculate the total final cell
number per division (Fig. 1d). Examples of using this method for
monitoring the impact of cytokines on the rates of growth and death
of dividing cells are given in Hasbold et al.7
Analysing differentiation
The phenotype or behaviour of cells from different divisions is
often markedly different. We have used a number of methods to
explore this phenomenon. The expression of a cell surface marker
in comparison with CFSE intensity can be monitored by flow
cytometry (Fig. 1e). Although bright, CFSE-labelled cells can be
simultaneously examined with any conventional combination of
fluorochromes (except fluorescein isothiocyanate (FITC)), including
phycoerythrin (PE), tricolour, PerCP, Texas red and allophycocyanin
(APC). Immediately after labelling at standard CFSE concentrations,
lymphocytes are too bright to compensate for the use of PE. Therefore, a lower concentration of CFSE can be used, although this will
limit the number of divisions that can be monitored at later times.
Alternatively, the early phenotype of very bright CFSE-labelled cells
can be determined using a two-laser system employing APC as a
second colour.
To monitor differentiation, a two-dimensional plot of CFSE
versus the second fluorochrome is drawn as shown in Fig. 1e, using
CELLQUEST software (Becton Dickinson). The position of each division peak is determined and gates set denoting the interval along the
CFSE axis maximally occupied by each division. By appropriate
gating, the proportion of cells of interest in each division can then be
determined. This process we refer to as ‘division slicing’.3 By this
method the proportion of positive (or negative) cells per division can
be calculated and plotted, as in Fig. 1f, for expression of IgG1 by B
cells cultured with IL-4. This information conforms to a divisionbased map as discussed further below.
Because harvested CFSE labelled cells are viable, they can be
sorted and their behaviour monitored further in vitro. This method
has proved useful for monitoring the changing patterns of cytokine
production with cell division.4 It is also possible to extract mRNA or
DNA for PCR8 or even protein for western blotting from cells in
different divisions.
Intracellular staining
Cell fixation and permeabilization conditions can be used that
preserve CFSE profiles and allow simultaneous intracellular
labelling with antibodies. Cells harvested from culture are washed
and resuspended in 4% formaldehyde or 2% paraformaldehyde (in
PBS) for 20 min at room temperature. Fixed cells are then rinsed
with PBS and stored in PBS/0.1% BSA at 4°C, protected from light
until staining. To stain intracellular components, fixed cells must be
permeabilized. We have used two methods successfully:
1. Saponin at 0.5% with 0.1% BSA in PBS. For this method, all
antibody labelling and cell washing is performed in this solution.
All steps can be carried out at room temperature. This is the method
of choice for intracellular cytokine staining.4 For analysis by flow
cytometry, cells can be resuspended in PBS/0.1% BSA.
2. Tween 20.9 Cells are fixed for 10 min at room temperature
before suspension volume is increased four-fold by addition of PBS
and Tween 20 to a final concentration of Tween 20 of 0.1%. These
cells are left overnight at room temperature. The next day cells are
washed and resuspended in PBS/0.1% BSA and left on ice for 30
min. These cells can now be stained with antibodies according to
standard methods. This permeabilization method is optimal for staining bromodeoxyuridine (BrdU)-labelled cells,3,7 as illustrated in
Fig. 1g, or intracellular signalling molecules such as STAT-1
(Fig. 1h).
Lymphoctye differentiation analysis with CFSE
519
Differentiation and cell division
Isotype switching is linked to cell division
B lymphocytes can be activated by T-dependent (TD) or
T-independent (TI) protocols in vitro by the use of CD40L or
LPS, respectively. B lymphocytes initially express IgM and
IgD as antigen receptors, but may switch to another isotype
during an immune response. The nature of this isotype switch
is affected by the activating stimulus and by the presence of
cytokines.10 When B cells are stimulated with the TD stimulus CD40 L, CFSE labelling reveals that they proliferate
asynchronously but do not switch from IgM expression over
the first 7–8 division cycles.3 When IL-4 was included in
culture the CD40 L stimulated B cells switched to IgG1 and
IgE in a division related manner, with approximately 50% of
the cells switching to IgG1 between divisions 3–8, whereas
the switch to IgE occurred at the later divisions.3
The appearance of switched cells only at later divisions
suggested isotype switching was either under the regulation of
a form of counting mechanism or that a subpopulation of
switched cells or cells destined to switch divided faster, allowing these cells to dominate at the later divisions. This latter
possibility was ruled out by studying the kinetics of appearance of switched cells. B cells stimulated by CD40L divide
asynchronously and enter division at widely disparate times.3
If cells entering division at different times had different
switching potential, then the ratio of switched cells within
each division would be different when monitored at different
times. Remarkably, however, the proportion of switched cells
in each division remains constant, regardless of when the
cultures are harvested.3 Furthermore, the switched and unswitched cells at later divisions are dividing at the same rate,3
a result inconsistent with the hypothesis that the switched cells
are dominating the cultures due to a faster division rate.
To explain these results we have adopted a probabilistic
view of switching and differentiation.3,6 By this view, cells
vary in sensitivity to CD40L and therefore the time of entry
to division; however, that variation bears no relationship to
the probability of a cell switching, which is dictated by division number and not time spent in any division. Simultaneously, each cell also varies in the division number at which it
will switch under the influence of IL-4. Thus, there are two
independent probabilities encapsulated in each B lymphocyte
that direct its differentiation after stimulation.6
Further analysis has revealed that the switch to all isotypes displays a relation with division that is independent of
time. Stimulation by LPS alone is sufficient to induce cells to
switch to IgG3, whereas the addition of transforming growth
factor (TGF)-β induces the same cultures to additionally
switch to IgG2b.11 B cells stimulated by the TD stimulus
CD40L will switch to IgG2a when IFN-γ is added,7 to IgA if
TGF-β is present11 and as already noted, to IgG1 and IgE
if IL-4 is included.2,3
T cell differentiation to cytokine secretion is also division
related
Altering the B cell isotype is an example of class switching
by the immune system. The equivalent process for T cells is
the development of different cytokine repertoires. The T cell
can become a Th1- or Th2-like cell after culture under appro-
Figure 2 The source of asynchrony in the standard model of
cell division. Under a given set of stimulation conditions, individual
cells in a lymphoctye population will exhibit a variable time to division that conforms to a probability distribution (top panel). At the
time of harvest, cells that fall to the left of the harvest time within
the time division distribution will have entered division whereas
those to the right make up the undivided population. The dividing
cells are moving through at a similar average division time (b) such
that the number of cells in their first, second and later divisions can
be determined (lower panel). CFSE, carboxyfluorescein diacetate
succinimidyl ester.
priate conditions.12,13 Typically, culture in IL-4 generates cells
capable of secreting a range of cytokines, including IL-4
itself, IL-10 and IL-5. In contrast, culturing the same cells
with IL-12 leads to a Th1 phenotype, typified by the secretion of IFN-γ.14 To test whether induction of T cell differentiation is also division linked and time independent, purified
naïve CD4 T cells (CD44 lo, CD62Lhi and CD25–ve) were
incubated with immobilized anti-CD3 in the presence of
IL-4, the same cytokine that induced division-linked differentiation in B cells. To examine cytokine secretion by cells in
different divisions, cells were sorted from overlapping divisions taken from cultures at different times. The results of
these experiments showed clearly that the secretion of IL-4
increased with division.4 Furthermore, the production of
IL-4 strongly correlated with division number and not time
after stimulation.4 The pattern of other cytokines also varied,
with IL-2 increasing then decreasing and IL-10, IL-5 and
IFN-γ appearing at late divisions. Again, as for B cells, the
relationship between division and cytokine is highly influenced by the presence of exogenous cytokines. The addition
of IL-12 induces the appearance of IFN-γ secreting cells at
early divisions, whereas IL-4 generates IL-4 secreting
cells.4,15
520
J Hasbold et al.
Figure 3 Progressive cell division in the standard model. (a) Effect of different harvest times on the standard model presented in Fig. 2.
The time of harvest is shown by the dashed arrow. The corresponding carboxyfluorescein diacetate succinimidyl ester (CFSE) profiles
are presented at the right. (b) Stimulation dose also alters the time to first division profiles as illustrated progressively for weak (upper
panels) to strong (lower panels) stimulation. The effect on the CFSE profiles is similar to the time course in (a) and the CFSE patterns
could arise from either scenario.
Cell surface marker expression alters with division
Asynchronous division
Lymphocyte differentiation is also associated with alteration
to cell surface molecules that subsequently influence the
cell’s sensitivity to the environment and alter its migratory
behaviour. We have noted that many of these cell surface
changes proceed in a division-related manner, with each
marker displaying an independent and distinct relationship
with division.2 Therefore, division number represents a
useful reference framework for describing differentiation. By
plotting the probability of a change in expression versus division number, a differentiation ‘map’ can be generated.2–4,11
For many division-linked changes so far monitored, the corresponding maps are approximated by a normal distribution
upon the division scale.11 Curiously, these maps need not
involve all the cells in the dividing population.2–4
We were interested in the underlying cause of division asynchrony in lymphocyte culture. It was possible the cells were
dividing at different rates, with those having undertaken the
most divisions being the fastest dividing cells. Alternatively,
cells could be dividing at the same rate and vary in the time
they enter into the first division round. To distinguish
between these possibilities, we pulsed cells with BrdU prior
to harvesting and then simultaneously monitored BrdU
incorporation and CFSE intensity (as illustrated in Fig. 1g).
Because BrdU is incorporated into DNA, a brief pulse can
indicate the number of cells in S phase of the cell cycle. Our
analysis has revealed a range of outcomes. We refer to the
most common outcome as the standard model of proliferation. T cells stimulated by anti-CD3 or B cells stimulated by
Lymphoctye differentiation analysis with CFSE
521
Figure 4 Evidence for independence of time to division rate and
probability of undergoing differentiation. (a) The progeny of
cells from different parts of the
time to first division plot reach
each subsequent division at different times. Irrespective, the proportion that change phenotype (for
example, switch isotype) is not
altered. Therefore, the probability
of switching is an independent
property of the time to division.
The cumulative expression of the
marker with division can be converted to a ‘map’ by plotting the
change per division as illustrated
by the arrows. Maps often
conform to a normal distribution.
(b) The average time to division
plot is also altered by stimulation
strength; however, again the
differentiation map followed is
related to division number and
is independent of proliferation
kinetics.
Figure 5 Independent regulation
of proliferation and differentiation
by cytokines. Extrinsic signals can
alter both proliferation and differentiation. These processes may be
linked, as shown in (a), where
increasing amounts of IL-4 reduce
the time to division of CD40Lstimulated B cells and reduce the
divisions at which switching to
IgG1 takes place. In contrast,
IL-12 acting on T lymphocytes
does not alter the proliferation rate
but profoundly alters the divisions
at which acquisition of IFN-γ
secreting ability occurs.
CD40 L and IL-4 conform to this standard. By this model,
the cells in division 1 or greater take up BrdU at the same rate
and are dividing evenly. Therefore, most of the asynchrony in
division number derives from variation in the time of entry
into the first division cycle. Underlying this pattern must be
a distribution presenting the proportion of cells entering division at each time after stimulation. We have measured this
distribution and found it to be approximately Gaussian, consistent with the difference in times to first division between
cells being stochastic rather than being due to different cell
populations. Figures 2 and 3 illustrate how this standard
model of growth gives rise to the characteristic asynchronous
CFSE profiles.
In other examples of proliferation, the rate of BrdU incorporation does change and increase with progressive division
number. This can be observed with CD40L stimulation of B
cells without IL-4, or B cells stimulated with anti-Ig reagents
(J Hasbold et al., unpubl. obs., 1999).
Independence of cellular parameters
Together, these studies precipitate a novel understanding of
lymphocytes that requires the adoption of a probabilistic view
of cell behaviour. This view is illustrated in Figs 4,5. It is
convenient to imagine each lymphocyte as comprised of a
series of machines that control the rates of growth and
522
J Hasbold et al.
differentiation. Furthermore, these machines are capable of
being controlled independently. The independence of proliferation and differentiation is illustrated in Fig. 4a. That is, a
cell population will vary in the time taken to divide due to
variations in the construction of the cell. This can be presented as a variable time to first division plot, as shown
(Fig. 4a). At the same time, cells will possess an innate differentiation program for linking division number and some
differentiation change; however, not all cells are constructed
identically and therefore the division at which the change will
occur is spread around a mean division number. Because the
time of division program and the differentiation program are
not related in the cell, the differentiation map of change per
division is not altered by the timing of entry into the first
cycle.
A variation of this behaviour is illustrated in Fig. 4b. Here
the stimulation dose is varied, which has the effect of changing the average time to first division. Despite this change, the
division-linked differentiation pattern is unaffected. This is
the case for altering CD40L dose during B cell switching
induced by IL-4 to IgG1.3 Further evidence that the time to
division and the differentiation program are capable of independent regulation is revealed by studying the effect of
cytokines. Some cytokines can hasten or lessen the time to
first division and also alter the division-linked differentiation
response. Interleukin-4 has this effect on both T and B cells
(Fig. 5a). In contrast, IL-12 does not alter the timing of
division, but will alter the divisions at which T cells acquire
IFN-γ secreting ability (Fig. 5b and AV Gett and PD
Hodgkin, unpubl. data, 1999).
Conclusions
The discovery that lymphocyte differentiation displays a predictable stochastic relationship with division number, which
is altered by cytokines and independent of the time spent in
traversing a division, provides a simplifying paradigm for
further analyses. The discovery indicates that there are only
three components required to predict the biological properties
of a population of lymphocytes after activation. These are:
1. The distribution of cells in each division cycle. This is
complex, but can be determined empirically with CFSE.
2. Maps of division linked changes. Such maps can be prepared experimentally and include information on the effect of
engaging receptors from the cell surface, including cytokines.
3. The rules of integration for different signals. For instance,
how does the combination of IFN-γ and IL-4 affect the separate differentiation probabilities of each? This must also be
determined experimentally. Because division number provides
a constant reference point, development of differentiation
maps is cumulative and can incorporate an inventory of
components at the cell surface and inside the cell, as well as
describing the functional properties of cells on restimulation.
It is anticipated that development of these models will
facilitate further dissection of the lymphocyte response by
helping us to look inward to determine the mechanism of how
these predictable relationships are controlled by intracellular
processes. Furthermore, they will enable us to look outward
from the cell to develop models that predict how the programmed response of lymphocytes will dictate their behaviour during ongoing immune reactions in vivo.
Acknowlegements
This work was supported by the Medical Foundation of the
University of Sydney and the National Health and Medical
Research Council.
References
1 Lyons AB, Parish CR. Determination of lymphocyte division by
flow cytometry. J. Immunol. Meth. 1994; 171: 131–7.
2 Hodgkin PD, Lee J-H, Lyons AB. B cell differentiation and
isotype switching is related to division cycle number. J. Exp.
Med. 1996; 184: 277–81.
3 Hasbold J, Lyons AB, Kehry MR, Hodgkin PD. Cell division
number regulates IgG1 and IgE switching of B cells following
stimulation by CD40 ligand and IL-4. Eur. J. Immunol. 1998; 28:
1040–51.
4 Gett AV, Hodgkin PD. Cell division regulates the T cell cytokine
repertoire revealing a mechanism underlying immune class
regulation. Proc. Natl Acad. Sci. USA 1998; 95: 9488–93.
5 Lyons AB, Hasbold F, Hodgkin PD. Flow cytometric analysis of
cell division history using dilution of CFSE, a stably integrated
fluorescent probe. Meth. Cell Biol. 1999; in press.
6 Hodgkin PD, Kehry MR. Methods for polyclonal activation to
proliferation and Ig secretion in vitro. In: Herzenberg LA, Weir
DM, Herzenberg LA, Blackwell C (eds). Weir’s handbook of
experimental immunology, vol. 3. Oxford: Blackwell Science,
1996.
7 Hasbold J, Hong JS-Y, Kehry MR, Hodgkin PD. Integrating
signals from IFN-γ and IL-4 by B cells: Positive and negative
effects on CD40 ligand-induced proliferation, survival and
division-linked isotype switching to IgG1, IgE, and IgG2a. J.
Immunol. 1999; in press.
8 McCall MN, Hodgkin PD. Switch recombination and germ-line
transcription are division-regulated events in B lymphocytes.
Biochim. Biophys. Acta 1999; 1447: 43–50.
9 Tough DF, Sprent J. Turnover of naive- and memory-phenotype
T cells. J. Exp. Med. 1994; 179: 1127–35.
10 Snapper CM, Mond JJ. Towards a comprehensive view of immunoglobulin class switching. Immunol. Today 1993; 14: 15–17.
11 Deenick EK, Hasbold J, Hodgkin PD. Switching to IgG3, IgG2b
and IgA is division-linked and independent revealing a
stochastic framework for describing differentiation. J. Immunol.
1999; in press.
12 Mosmann TR, Coffman RL. TH1 and TH2 cells: Different
patterns of lymphokine secretion lead to different functional
properties. Annu. Rev. Immunol. 1989; 7: 145–73.
13 Kelso A. Cytokines: Principles and prospects. Immunol. Cell
Biol. 1998; 76: 300–17.
14 O’Garra A, Murphy K. Role of cytokines in determining
T-lymphocyte function. Curr. Opin. Immunol. 1994; 6: 458–66.
15 Bird JJ, Brown DR, Mullen AC et al. Helper T cell differentiation is controlled by the cell cycle. Immunity 1998; 9: 229–37.