You are on page 1of 14

REVIEWS

T-CELL DEVELOPMENT AND THE


CD4–CD8 LINEAGE DECISION
Ronald N. Germain
Cell-fate decisions are controlled typically by conserved receptors that interact with co-evolved
ligands. Therefore, the lineage-specific differentiation of immature CD4+CD8+ T cells into CD4+ or
CD8+ mature T cells is unusual in that it is regulated by clonally expressed, somatically generated
T-cell receptors (TCRs) of unpredictable fine specificity. Yet, each mature T cell generally retains
expression of the co-receptor molecule (CD4 or CD8) that has an MHC-binding property that
matches that of its TCR. Two models were proposed initially to explain this remarkable outcome
— ‘instruction’ of lineage choice by initial signalling events or ‘selection’ after a stochastic fate
decision that limits further development to cells with coordinated TCR and co-receptor
specificities. Aspects of both models now appear to be correct; mistake-prone instruction of
lineage choice precedes a subsequent selection step that filters out most incorrect decisions.

CO-RECEPTOR Cell-fate specification is a key developmental event. infections. In accordance with this dichotomy, mature
A CD4 or CD8 molecule, which The physiological function of a multicellular organism αβ T cells comprise two lineages, the antigen recogni-
cooperatively recognizes an depends on the generation of the proper number and tion of which is focused on either MHC class-I- or
MHC class-II or class-I ligand,
diversity of cell types. Signals from broadly expressed class-II-associated peptides and whose effector func-
respectively, together with the
antigen receptor (T-cell receptors that interact with co-evolved germ-line lig- tions are tuned to these recognition biases. Typically,
receptor) of a T cell. ands control most differentiation decisions1. But, adap- these subsets are defined by exclusive surface expression
tive immunity depends on the function of T and B of either the pan-MHC-class-II-binding protein CD4
cells, which express unique surface receptors that are (helper or regulatory CD4+ T cells) or the pan-MHC-
created by somatic DNA rearrangement and random class-I-binding protein CD8 (cytotoxic CD8+ T cells)3,4.
chain pairing. These clonal receptors help to determine CD4 and CD8 are non-clonally distributed proteins
which precursor lymphocytes will successfully mature; that, because they co-recognize the ligands of the TCR,
this raises the issue of whether these receptors dictate are often referred to as CO-RECEPTORS5. Maintaining the
cell fate in the same general manner as more broadly proper number and proportion of CD4+ and CD8+
expressed receptors or whether novel mechanisms T cells is essential for optimal host defence — the ques-
underlie their activity. tion is how this required outcome can be achieved,
Lymphocyte Biology Section, Such questions are particularly intriguing with given the unpredictable peptide–MHC specificity of
Laboratory of Immunology, respect to the main peripheral pool of T cells that each precursor cell’s TCR.
National Institute of Allergy express αβ T-cell receptors (TCRs). These cells This problem was defined more precisely as the pre-
and Infectious Diseases,
National Institutes of respond to antigen in the form of short peptides cursor–product relationships in αβ T-cell differentiation
Health, Building 10, bound to MHC class I or class II molecules2. MHC were unravelled6 (FIG. 1). An important advance was the
Room 11N311 MSC-1892, class-I ligands alert the immune system to active intra- recognition that CD4+ and CD8+ T cells are derived from
10 Center Drive, cellular infection and target infected cells for destruc- a common precursor pool, in which each cell expresses
Bethesda, Maryland
20892-1892, USA.
tion, whereas MHC class-II ligands guide intercellular both CD4 and CD8 proteins. Subsequent experiments7–10
e-mail: rgermain@nih.gov co-operation between haematopoietic cells in an indicated that the successful maturation of precursor
doi:10.1038/nri798 immune response and help to combat extracellular CD4+CD8+ thymocytes seems to be restricted to those

NATURE REVIEWS | IMMUNOLOGY VOLUME 2 | MAY 2002 | 3 0 9


© 2002 Nature Publishing Group
REVIEWS

Outer cortex
Inadequate
TCR/co-receptor Cortical
signalling epithelial
cell

MHC class I MHC class II


recognition recognition
Death by
Inner cortex

neglect

Lymphoid
progenitor
Cortico–medullary CD8-committed DP CD4-committed DP
junction

Negative Positive Negative


selection selection selection
Medulla

Haematopoietic
precursor
Medullary Death Death Medullary
epithelial cell/ epithelial cell/
dendritic cell dendritic cell

CLONAL EXPRESSION Emigration to periphery


The presence of a particular
somatically generated antigen
receptor on a maturing T cell Figure 1 | Overall scheme of T-cell development in the thymus. Committed lymphoid progenitors arise in the bone marrow and
that is not shared with other migrate to the thymus. Early committed T cells lack expression of T-cell receptor (TCR), CD4 and CD8, and are termed double-negative
independently developing (DN; no CD4 or CD8) thymocytes. DN thymocytes can be further subdivided into four stages of differentiation (DN1, CD44+CD25−; DN2,
precursors. This receptor is CD44+CD25+; DN3, CD44−CD25+; and DN4, CD44−CD25−)41. As cells progress through the DN2 to DN4 stages, they express the pre-
shared among the progeny of the TCR, which is composed of the non-rearranging pre-Tα chain and a rearranged TCR β-chain43. Successful pre-TCR expression leads
mature cell after activation and to substantial cell proliferation during the DN4 to double positive (DP) transition and replacement of the pre-TCR α-chain with a newly
cell division stimulated by rearranged TCR α-chain, which yields a complete αβ TCR. The αβ-TCR+CD4+CD8+ (DP) thymocytes then interact with cortical epithelial
foreign antigen. cells that express a high density of MHC class I and class II molecules associated with self-peptides. The fate of the DP thymocytes
depends on signalling that is mediated by interaction of the TCR with these self-peptide–MHC ligands6,51. Too little signalling results in
RECOMBINATION-ACTIVATING delayed apoptosis (death by neglect). Too much signalling can promote acute apoptosis (negative selection); this is most common in the
GENE medulla on encounter with strongly activating self-ligands on haematopoietic cells, particularly dendritic cells123. The appropriate,
(RAG). The product of this gene intermediate level of TCR signalling initiates effective maturation (positive selection). Thymocytes that express TCRs that bind self-
is involved in creating the
peptide–MHC-class-I complexes become CD8+ T cells, whereas those that express TCRs that bind self-peptide–MHC-class-II ligands
double-strand DNA breaks that
become CD4+ T cells; these cells are then ready for export from the medulla to peripheral lymphoid sites. SP, single positive.
are necessary for producing the
rearranged gene segments that
encode the complete protein
chains of T-cell and B-cell cells that maintain expression of the co-receptor molecule instruction12,13, then to selection14–19, and more recently,
receptors. (CD4 or CD8) that has a self-MHC-class specificity that back to instruction20–36. This review summarizes our
NEGATIVE SELECTION
matches the CLONALLY EXPRESSED TCR. There are two possi- current understanding of αβ T-cell differentiation,
Active cell loss in the thymus, ble explanations for this phenomenon: either the interac- antigen-receptor signalling and the TCR–co-receptor-
which is mediated by apoptosis tion of a TCR–co-receptor pair with an MHC ligand dependent events that guide the lineage-specific devel-
induced by strong T-cell stimuli, instructs the cell as to which path to take and which opment of immature T cells to yield pools of CD4+ and
particularly at later stages of
(inappropriate) co-receptor-encoding gene to stop tran- CD8+ mature T cells that are properly suited to their
maturation.
scribing; or, after random (stochastic) lineage choice and physiological roles.
POSITIVE SELECTION the extinction of CD4 or CD8 expression, further matu-
The maturation of immature ration depends on signals that require a match between The αβ T-cell differentiation pathway
CD4+CD8+ precursor the MHC-class specificity of the remaining co-receptor Generation of CD4+CD8+ immature thymocytes.
thymocytes induced by T-cell
receptor signals that result from
and the cell’s TCR (selection11,12) (FIG. 2). Committed lymphoid progenitors arise in the bone
binding to self-peptide–MHC Over the years, the prevailing view has swung back marrow and migrate via the blood to the thymus. In the
ligands on thymic epithelial cells. and forth between these two possibilities — first to thymus, these cells lose the potential for B-cell37–39 and

310 | MAY 2002 | VOLUME 2 www.nature.com/reviews/immunol


© 2002 Nature Publishing Group
REVIEWS

a Instruction b Stochastic/selection natural-killer-cell40 development. The result is a double-


negative (DN; no CD4 or CD8), committed T-cell pre-
TCR–CD8/ TCR–CD4/ Stochastic choice of cursor (FIG. 1). DN thymocytes can be subdivided further
MHC class I MHC class II CD4 or CD8 lineage
interaction interaction (before or at time into four sequential stages of differentiation, which are
of first TCR signal) identified by their surface expression of CD44 and
Cortical
epithelial cell CD25: DN1, CD44+CD25−; DN2, CD44+CD25+; DN3,
CD44−CD25+; and DN4, CD44−CD25− (REF. 41).
Correct
CD8
Incorrect Double-negative T cells can give rise to either γδ or
CD4
MHC CD8 MHC
choice
choice αβ TCR-expressing cells6. For cells that proceed along
class I class II the αβ TCR pathway, DN3-stage cells first express pre-
TCR CD4 TCR-α42–44, which is encoded by a non-rearranging
locus. Pre-TCR-α pairs with the TCR β-chain, which is
the product of a set of somatic DNA rearrangements
that require expression of RECOMBINATION-ACTIVATING GENE 1
(RAG1) (REF. 45) and RAG2 (REF. 46) proteins. At the cell
Unique signal Unique signal surface, the pre-TCR-αβ pair is associated with a collec-
for CD8 lineage for CD4 lineage tion of proteins (the CD3/ζ complex) that is involved in
proximal signal transduction47. Active signalling is
required for further T-cell maturation (TCR β-chain
selection), because mutations that prevent the expres-
sion of key enzymes or adaptors that mediate the gener-
ation of intracellular second messengers by the pre-TCR
complex cause maturational arrest at this DN3
stage48–50. This is the same stage of developmental arrest
that is seen with defective Rag gene expression, which
prevents β-locus rearrangement and hence, pre-TCR
assembly, expression and signalling.
T cells that emerge from β-selection (late DN3 and
TCR–CD8/MHC TCR/MHC class I DN4) undergo 6–8 cell divisions, after which recombi-
class I interaction interaction without
CD8 involvement
nation at the TCR-α locus produces the second com-
ponent chain of the mature αβ antigen receptor. The
expression of pre-TCR-α is lost during this stage,
which results in the cell-surface display of a low level of
MHC- MHC- αβ TCR assembled with CD3/ζ proteins. The thymo-
class-I- class-II- cytes also begin to express co-receptor proteins —
specific specific
TCR TCR most often CD8 first, followed by CD4 (REF. 6) — and,
eventually, form a large population of double-positive
CD8+ CD4+ (DP; CD4+CD8+) αβ-TCR-expressing immature cells
SP SP CD8+
SP that constitute 90% of the lymphoid compartment in
the thymus of young individuals.
Cell survival Cell death
The opposite would occur with a DP cell TCR-mediated selection of DP thymocytes. From this
that expresses a TCR that is specific for
MHC class II large number of DP thymocytes, the subset that is best
suited to function in the host environment is permitted
Figure 2 | ‘Instruction’ and ‘stochastic’ models for the development of mature T cells
to mature and migrate to peripheral lymphoid tissues.
with coordinated TCR and co-receptor specificities. a | The ‘instruction’ model proposes
that biochemically distinct intracellular signals are generated when a T-cell receptor (TCR) and Four distinguishable processes characterize this selec-
CD8 molecule co-engage an MHC class-I ligand compared to when a TCR and CD4 co- tion — death by neglect, NEGATIVE SELECTION, POSITIVE
6,51
receptor co-recognize an MHC class-II ligand. These different messages induce commitment SELECTION and lineage-specific development . Most
to the correct lineage (the CD8 or CD4 pathway, respectively)11,12. Among the genetic events (~90%) DP thymocytes express TCRs that interact so
that are involved in lineage-specific differentiation, the most obvious is loss of expression of the poorly with the available self-peptide–MHC ligands
co-receptor (CD4 or CD8, respectively) that was not involved in generating the signal that
that the intracellular signals that are required to sustain
induced lineage commitment. b | By contrast, the ‘selection’ model suggests that precursor
CD4+CD8+ thymocytes are either already (randomly) committed to a lineage, or, at the start of
viability are not generated, which leads to death
positive selection, make a stochastic lineage choice that is unrelated to the specificity of their by neglect. It is not surprising that so few TCRs
TCR. This results in some cells choosing incorrectly — that is, committing to a developmental bind adequately to self-ligands. The diverse set of
pathway that will extinguish expression of the co-receptor that is required for optimal mature- peptide–MHC-class-I and -class-II structures on corti-
cell antigen responses with the TCR of that T cell. To eliminate such undesired cells, a filter is cal epithelial cells that are involved in initiating T-cell
applied during later development. After one co-receptor is lost, continued differentiation is selection is but a small fraction of the total number of
postulated to be dependent on additional TCR signals that require cooperation with the correct
(MHC-class-specificity-matched) co-receptor. Cells that retain expression of the wrong co-
potential complexes. Although the unselected repertoire
receptor fail to produce this life-sustaining signal and die, so that successfully maturing cells of TCRs is biased towards the recognition of MHC in
have the expected correspondence between TCR and co-receptor specificity for MHC ligands. general52,53, it is unlikely that the random DNA
DP, double positive; SP, single positive. rearrangement and protein-chain pairing process that

NATURE REVIEWS | IMMUNOLOGY VOLUME 2 | MAY 2002 | 3 1 1


© 2002 Nature Publishing Group
REVIEWS

controls the fine specificity of TCRs54 would generate might facilitate optimal signalling, both through the
many receptors that are well-enough matched to the action of the additional LCK that is recruited by this
particular peptide–MHC ligands that are present in any event and the association of co-receptors with the
individual to result in efficient signalling. A small frac- ZAP70-substrate LAT64. The latter might promote
tion of immature T cells (~5%) bears TCRs that bind rapid elevation of intracellular MAPK1 (extracellular
very well to self-ligands; these lymphocytes could cause signal-regulated kinase; ERK) activity, which helps to
autoimmune pathology if they were permitted to leave sustain the function of LCK within the TCR complex60.
the thymus. Signalling on engagement of these TCRs As a result, highly effective signalling in response to for-
with self-peptide–MHC ligands promotes rapid apop- eign antigen will depend on the coordinate binding of
totic death (negative selection). Cells that express TCRs ligand by the TCR and a CD4 or CD8 molecule.
that recognize self-ligands and generate signals that Therefore, the correct class of co-receptor must be pre-
have an intensity between those resulting in neglect or sent on the mature T cell if it is to be sensitive to low
negative selection initiate a multi-step process known levels of foreign antigen.
as positive selection that ultimately results in lineage-
specific differentiation into either CD4+ or CD8+ Instruction versus selection
mature T cells. If optimal T-cell function depends on matching the
MHC-class specificity of co-receptor and TCR, how is
TCR–co-receptor coordination coordination between germ-line and somatic specifici-
The generation of intracellular messengers after TCR ties achieved? Thymocyte differentiation involves a
recognition of self-ligands has a crucial role in the pro- change from the DP to the SP (CD4+CD8− or CD4−CD8+)
duction of the mature T-cell pool. This involves not state by the silencing of transcription of one co-receptor
only life or death decisions, but also some mechanism locus65–75, which is accompanied by other genetic
that permits a differentiating DP thymocyte to coordi- events that determine the effector potential of the
nate its continued expression of CD4 or CD8 with the mature T cell (‘helper’ versus ‘cytotoxic’ cell). This
unpredictable specificity of the clonally expressed TCR. polarization could occur randomly due to stochastic
But, why is such coordination important? The answer processes before or at the initiation of positive selec-
lies in the need for sensitivity of peripheral T cells to tion. Alternatively, it could be regulated by the nature
foreign antigens. of the initial receptor signal, so that the extinguished
Antigen-receptor signalling begins with ligand co-receptor locus is (most often) the one with an
recognition, which leads to the activation of a SRC- MHC specificity that does not match that of the TCR.
family kinase, typically lymphocyte-specific protein- In the stochastic model, the remaining co-receptor
tyrosine kinase (LCK). This kinase is associated with would frequently be inappropriate for the TCR of that
the TCR–CD3/ζ complex, as well as with the cytoplas- cell. A later test of the MHC-class specificity of the
mic tails of CD4 and CD8 (REF. 55). Activated LCK clonal receptor and co-receptor molecule would be
tyrosine phosphorylates immunoreceptor tyrosine- required, so that only cells with the correct match are
based activation motifs (ITAMs) in CD3/ζ, which allowed to mature. This process of random choice fol-
produces a structure that is suitable for tight binding lowed by testing is known as the ‘stochastic’ or ‘selec-
by the paired SRC-homology 2 (SH2) domains of the tion’ model, to distinguish it from the case in which a
SYK-family kinase ZAP70 (ζ-chain-associated protein specific, unique biochemical event that arises from
kinase, 70 kDa)56,57. Recruited ZAP70 is phosphory- initial engagement of the TCR and one of the co-
lated and activated by the already functional LCK. receptors signals to the T cell to enter the proper lin-
Active ZAP70, in turn, tyrosine phosphorylates a eage (‘instruction’ model)11,12. FIGURE 2 illustrates these
major adaptor protein — linker for activation of two models.
T cells (LAT)58 — as well as several other enzymes and
adaptors. This leads to an increase in the concentration Rescue experiments support instruction. A series of
of intracellular Ca2+, which promotes the nuclear experiments in transgenic and gene-targeted (knock-
translocation of nuclear factor of activated T cells out) mice was the first to test the competing hypothe-
(NFAT) transcription factors and the activation of one ses. The selection model predicts that some developing
or more mitogen-activated protein kinase (MAPK) T cells die because they make the wrong stochastic
pathways. Together, these events help direct the gene choice of which co-receptor to maintain in relation-
expression that is necessary for the function of mature ship to their TCR specificity. Therefore, it would be
T cells59. expected that enforcing expression of the appropriate
The consequences of TCR interaction with pep- co-receptor, in addition to the randomly chosen co-
tide–MHC ligands depend strongly on the ‘quality’ of receptor, should rescue these cells at the developmental
the interaction. In particular, suboptimal binding does checkpoint that evaluates TCR–co-receptor matching.
not simply result in a smaller quantity of the same sig- The result would be mature T cells that express both
nals, but in a different quality of proximal signalling 60,61. the inappropriate endogenous co-receptor and
The capacity of a TCR–ligand pair to interact effec- the appropriate transgenic co-receptor (mature
tively with either the CD4 or CD8 co-receptor has a CD4+CD8+ T cells). Initial experiments with transgenic
crucial role in producing high-quality signals62,63. The mice that expressed either CD4 or CD8 throughout
association of an engaged TCR with a co-receptor T-cell development gave negative results — very few or

312 | MAY 2002 | VOLUME 2 www.nature.com/reviews/immunol


© 2002 Nature Publishing Group
REVIEWS

Lineage programme expression Transitional-cell studies argue for selection. The pen-
dulum then swung almost completely towards selec-
tion when several groups reported detailed phenotypic
CD4+CD8low analyses of thymocyte development in mice that were
TCRint/hi deficient in MHC class-I or class-II expression. Guidos
CD4+ SP and colleagues79 had described the transition of DP
TCRhi DP T cells to SP T cells in normal animals as comprising a
TCRlow
series of intermediate states in which the surface
expression of CD4 (on the path to SP CD8+ T cells) or
CD8 (on the path to SP CD4+ T cells) was gradually
CD4

CD4low lost. When thymocytes from MHC class-I-deficient


CD8low/+ CD4lowCD8+ mice were examined, they included substantial num-
TCRint TCRhi
bers of CD4mediumCD8hi cells, a transitional state that
would not be expected to exist if MHC class I mole-
cules were unavailable to ‘instruct’ development along
the CD8 pathway15,16. Likewise, CD4hiCD8medium T cells
were observed in MHC class-II-deficient mice, even
CD8+ SP
TCRhi though the DP T cells could not have been instructed to
DN
develop along the CD4 lineage in such animals15,80. On
the basis of these data, stochastic choice and selection
were invoked to explain TCR–co-receptor matching,
and various arguments were offered to explain the lim-
CD8 ited ability of co-receptor transgenes to rescue the devel-
opment of cells that were predicted to have chosen the
Uncommited CD4-commited CD8-commited wrong expression pattern.
A third model of development was proposed at this
Figure 3 | Complex pattern of CD4 and CD8 expression on positively selected time, known as the default model. Singer and col-
thymocytes. The discovery that CD4+ and CD8+ T cells are derived Nature Reviews
from | Immunology
CD4+CD8 +
double- leagues developed an assay (PRONASE STRIPPING) that
positive (DP) precursors led to the assumption that mature cells that express only one co- allowed analysis of the ongoing, rather than the past,
receptor arise from transcriptional silencing of the irrelevant co-receptor locus and gradual loss state of co-receptor synthesis by an individual T cell.
of surface expression of the corresponding protein. This widely accepted concept of progressive Using this method, they concluded that DP T cells
co-receptor loss contributed to the confusion concerning whether lineage commitment resulted
moved along the CD4 lineage pathway by ‘default’,
from instruction or stochastic choice. Because of the inconsistencies between experiments that
tested predictions of a stochastic/selection model12–14 and unanticipated findings on the lineage
even without TCR engagement, but were deviated
fate of isolated thymocyte subpopulations82, this basic assumption of linear co-receptor from this fate by engagement of the TCR with MHC
extinction was re-evaluated83. These studies revealed that DP cells become CD4+ or CD8+ class I (REF. 81).
single-positive (SP) cells by a complex, non-intuitive route. TCR signalling involving MHC class-I
or class-II recognition results in the partial loss of both CD4 and CD8 expression. These Back to instruction. The divergent conclusions that
CD4lowCD8low cells then begin to re-express CD4, followed by CD8, yielding CD4highCD8medium arose from rescue experiments, analysis of transitional
cells, regardless of whether the initiating signal involves MHC class I or MHC class II ligands. The
extent of CD4 and CD8 downregulation correlates generally with the strength of the initial TCR
cells in MHC-deficient animals and pronase-stripping
signal. So, for cells at the lower limit of signalling that is required for CD8 lineage commitment, studies hinted that some systematic error(s) were pre-
CD8 loss might not be great enough to generate cells with a CD4+CD8low phenotype that are venting the correct interpretation of experiments that
distinguishable from the broad range of co-receptor expression on DP cells. This potentially examined lineage commitment. Several studies helped
explains some reports that CD8-lineage development does not always include passage through to straighten matters out. Two groups showed that
this particular intermediate stage. CD4+CD8low cells that are committed to the CD4 lineage then CD4hiCD8medium T cells could develop into both CD4+
lose expression of CD8 to become CD4+ SP cells. Cells that are committed to the CD8 pathway
and CD8+ SP T cells, in contrast to the prediction of
lose expression of CD4 while gaining expression of CD8; they become CD4+CD8+ again, but
with higher levels of TCR expression than immature DP cells. Eventually, expression of all CD4 is the simple transitional-cell model81,82. Using both
lost and the cells emerge as CD8+ SP cells. This scheme has been verified using molecular kinetic and pronase-stripping analyses, Lucas and
genetic approaches124. Germain then found that this result, and much of the
confusing lineage data, could be explained by the previ-
ous acceptance of an incorrect model of thymocyte
no mature DP T cells were generated12,13. Because the phenotypic change after the DP stage83. It had been
expectation of the selection model was not met, these assumed that DP T cells simply extinguish transcrip-
data were taken as strong evidence for ‘instruction’. tion of the genes that encode CD4 or CD8 and that
Follow-up studies clouded this picture, however. With expression of the corresponding co-receptor disap-
PRONASE STRIPPING
sufficiently high levels of expression of the transgenic pears gradually. But, in reality, co-receptor extinction
The treatment of cells with a co-receptor or with different transgenic TCRs, signifi- occurs in a more complex manner (FIG. 3). TCR sig-
powerful protease that removes cant ‘rescue’ was observed14,76–78. The process was never nalling first leads to some loss of both co-receptors
surface protein molecules, as successful as would be predicted from a simple from the cell membrane, irrespective of whether MHC
followed by the analysis of
surface phenotype immediately
selection model, but the fact that a substantial number class-I or class-II ligands are recognized. The extent of
after such treatment and then of mature DP T cells was seen argued against a this loss seems to correlate with the strength of TCR
after further culture. ‘unique-signal’ instruction model. signalling. Therefore, weaker signals might not cause

NATURE REVIEWS | IMMUNOLOGY VOLUME 2 | MAY 2002 | 3 1 3


© 2002 Nature Publishing Group
REVIEWS

TCR–CD8/MHC class I interaction TCR–CD4/MHC class II interaction enough loss of co-receptor expression to easily distin-
guish these cells from the large pool of ‘neglected’ DP
T cells in the thymus, which explains some reports that
reached a different conclusion about the CD8 pathway
MHC class I MHC class II
of T-cell development (that some cells developing
CD8 CD4
along the CD8 pathway do not undergo this loss of
expression and asymmetric recovery of co-receptor
expression)84,85. Re-expression of CD4 and, with a
delay, of CD8 then occurs. This kinetic asymmetry in
re-expression results temporarily in CD4hiCD8medium
Rare Frequent Rare cells — regardless of whether they have TCRs that are
Long/moderate Short/low Long/moderate Short/low specific for MHC class I or class II — which explains
intensity signal intensity signal intensity signal intensity signal the cell-transfer data from REFS 81,82 and the pronase-
stripping results81. This stage is followed by the loss of
expression of CD4 on cells of the CD8 lineage —
which requires these cells to pass through a DP
CD4mediumCD8medium stage as they become CD8+ SP — or
the loss of expression of CD8 — which allows the cells to
move rapidly and directly from the CD4hiCD8medium state
CD4 committed CD8 committed CD4 committed CD8 committed to the CD4+ SP state.
Once this new scheme of thymocyte development
became clear, the strongest evidence for a
stochastic/selection model disappeared, because the
‘inappropriate’ transitional forms that were seen in
MHC-deficient mice no longer violated the precepts of
an instruction model. The CD4mediumCD8hi cells that
have been observed in MHC class-I-deficient mice can
be explained by the fact that on recognition of MHC
class II, the expression of CD4 is reduced somewhat
more than CD8 during the initial downregulation of
both co-receptors. Logic demanded that if CD4+CD8+
T cells become CD4+ or CD8+ SP T cells, then the DP
cells must lose expression of either CD4 or CD8. This
was a well-founded presumption. However, the extrap-
olation from this that co-receptor loss must occur in a
straightforward ‘linear’ manner was not, as it turned
Inadequate Persistent Inadequate out, correct.
signalling signalling signalling
Molecular mechanisms of lineage choice
‘Strength of signal’ model of lineage commitment. As
CD8+ CD4+ the problems with the transitional-cell analyses became
SP SP
clear, additional studies focused more on the possible
Cell death Cell death molecular mechanisms of ‘instructed’ lineage choice.
Three lines of evidence indicated that the extent of
Figure 4 | ‘Strength of signal’ model for the development of mature T cells with TCR–co-receptor-dependent intracellular signalling
coordinated TCR and co-receptor specificities. The ‘strength of signal’ model proposes had a decisive role in determining the fate of DP T cells
that commitment to the CD4 versus CD8 lineage depends on the intensity23,24,32,36 or
(FIG. 4). First, the cytoplasmic domains of the CD4 and
duration24,33,34,97 of signalling from the T-cell receptor (TCR). Short-duration signals lead to the
CD8 pathway; more prolonged signals lead to the CD4 pathway. Because CD4 is more CD8 co-receptors were exchanged by molecular
extensively associated with lymphocyte-specific protein-tyrosine kinase (LCK) in CD4+CD8+ cloning methods, and these recombinant proteins were
thymocytes30,32,88,89, MHC class-II binding by the TCR and CD4 will tend to produce a expressed in developing T cells22,86,87. The results of
prolonged signal more often than will TCR–CD8 co-recognition of an MHC class-I ligand, these experiments indicated that lineage fate was dic-
resulting in the ‘correct’ lineage choice by most immature T cells. Some of these cells will tated largely by the signalling end of the co-receptor,
express TCRs that only very weakly bind peptide–MHC-class-II ligands, which leads to short
rather than by the class of MHC that was bound. On
signals and an inappropriate CD8-lineage choice. Other cells will have TCRs that bind MHC
class-I ligands particularly well, producing long signals and an improper CD4-lineage choice.
the basis of these data, Itano et al.22 first proposed that
The latter cells will often be eliminated by means of negative selection. The former cells will ‘strength of signal’ dictated the CD4–CD8 decision.
often fail to complete maturation, because the signalling capacity of the TCR without CD4 is Second, Matechek et al.23 found that thymocytes with
inadequate to support subsequent development. This is the same late filtering step that is an MHC class-II-specific TCR became CD8+ SP T cells
postulated in the original ‘selection’ model. Enforced CD4 expression can rescue some MHC when CD4 was absent, rather than failing to mature at
class-II-recognizing T cells that incorrectly enter the CD8 lineage. This can explain the data of all. Given the idea that CD4 assists the TCR in sig-
Leung et al.75 on the effect of a CD4 silencer deletion that maintains CD4 expression in CD8-
lineage-committed cells, without the need to conclude from these findings that the lineage-
nalling when an MHC class-II ligand is recognized,
commitment process is stochastic. SP, single positive. these investigators suggested that weak signalling when

314 | MAY 2002 | VOLUME 2 www.nature.com/reviews/immunol


© 2002 Nature Publishing Group
REVIEWS

Box 1 | Analysis of lineage commitment and progression using a two-stage culture model
Jenkinson and Owen first introduced the use of re-aggregate cultures for the analysis of T-cell development134. They
appropriately emphasized the use of this method for the dissection of the role of different cell populations in this process.
In an adaptation of their pioneering method, Yasutomo et al.33,94 separated delivery of the signal that initiates positive
selection of double-positive (DP) T cells from the signals that are required for the completion of maturation to the
single-positive (SP) state. This was accomplished using DP T cells that were isolated from T-cell receptor (TCR)-
transgenic mice that had been bred onto an MHC background that was incapable of initiating either positive or negative
selection, which provided a pool of ‘virgin’ DP T cells. These DP cells were exposed to various peptide–MHC ligands in a
dispersed culture, and then cells with evidence of activation (surface expression of CD69) were isolated by flow
cytometry. Previous studies had shown that thymocytes that are undergoing selection in a normal thymus express CD69
(REF. 135). These activated cells, which have the CD4lowCD8lowTCRintermediate phenotype of early post-selection
thymocytes83,136, were then combined with isolated MHC-deficient thymic stromal cells and additional haematopoietic
dendritic cells of the desired MHC type to provide ligands for the completion of maturation. Previous work, as well as
other studies in this two-step model, had shown that successful maturation depended on such additional TCR
signalling90–94. By varying the quality of the peptide–MHC ligand that is available to the DP T cells in the first culture and
to the activated thymocytes in the second culture, it was possible to determine the stage at which lineage commitment
occurred and the rules that govern this process with respect to TCR and co-receptor engagement. It was also possible to
vary the amount and duration of exposure to the first signal, to investigate how ‘strength of signal’ might influence
lineage choice and/or progression.

DP TCR-transgenic Antigen-presenting Isolate


thymocytes from cells +/– specific CD69hi
non-selecting host peptide cells

Analyse by FACS
+
or function

CD69

First step = overnight Second step = 3 day


dispersed coculture; re-aggregate culture
next day, FACS sort with thymic stroma +/–
for activated CD69+ cells haematopoietic antigen-
presenting cells

CD4 was missing led to the CD8 fate. Third, Iwata and This high efficiency would not be expected if lineage
associates showed that by varying the concentration commitment is determined stochastically and an
and duration of exposure of pre-selection DP T cells to incorrect choice ‘dooms’ a cell35.
phorbol ester and ionomycin — which simulate down- The generation of CD4+ T cells by co-crosslinking
stream signalling by the TCR complex — one could of TCR and CD8 was problematic, however; this result
selectively produce CD4+ or CD8+ SP T cells24. Higher seemed to be in clear violation of an instructional
concentrations and longer exposures produced CD4+ model, in which the class of MHC that is co-engaged
SP T cells, whereas brief exposures to lower concentra- by TCR and co-receptor dictates cell fate when physio-
tions produced CD8+ SP T cells. A similar quantitative logical ligands, and not antibodies, are involved. Data
model was proposed on the basis of the effect of the on the association of Lck with CD4 and CD8 in imma-
level of Mapk activity on the generation of CD4+ versus ture thymocytes provided a possible solution. Veillette
CD8+ T cells25. Fourth, Zamoyska and associates et al.88 first reported, and Wiest et al.89 later confirmed,
showed in thymic-organ cultures that crosslinking the that CD4 molecules that are isolated from DP cells are
TCR with either CD4 or CD8, using hybrid antibodies, associated with more Lck than are CD8 molecules.
led to the emergence of CD4+ SP T cells, whereas Therefore, under physiological conditions, the co-
TCR–TCR crosslinking alone produced CD8+ SP engagement of CD4 and TCR by MHC class-II ligands
T cells26,29,30. The extent of Lck activation was proposed might promote greater LCK inclusion in the signalling
to control cell fate31; TCR–co-receptor crosslinking pro- complex than TCR–CD8 co-recognition of an MHC
duced CD4+ SP T cells because this provided more Lck class-I ligand. Signalling would, therefore, generally be
input than TCR crosslinking alone. The idea, from all of much stronger with MHC class-II ligands than with
these studies, that signal strength ‘instructed’ lineage MHC class-I ligands, providing a screen that ‘instructs’
choice was also supported by evidence that showed that most DP T cells into the correct lineage upon self-
by manipulating the parameters that affect TCR–co- ligand recognition. The finding that CD4+ SP T cells
receptor signalling, conditions can be achieved in which were generated by TCR–CD8 crosslinking could then be
almost 90% of DP precursors transit to the SP stage. explained by postulating that the antibodies markedly

NATURE REVIEWS | IMMUNOLOGY VOLUME 2 | MAY 2002 | 3 1 5


© 2002 Nature Publishing Group
REVIEWS

exceeded the usual physiological levels of engagement, co-receptor function failed to alter cell fate33. On the basis
and yielded signals that reached the threshold required of, in part, biochemical data24,29, signal duration was con-
for CD4 lineage commitment, even with the lower level sidered as an alternative candidate to ligand concentra-
of LCK that is associated with CD8. tion. This hypothesis was tested by limiting the length of
Further refinement of the ‘strength of signal’ lin- exposure of HY-specific-TCR-transgenic DP T cells to
eage-commitment model then came from experiments male antigen in the first culture. When the interaction
that involved a two-step thymic-organ culture system33 was limited to a few hours, the cells became CD8+ SP
(BOX 1). DP thymocytes from TCR-transgenic mice that T cells in the second culture, whereas if the DP cells were
lacked the MHC ligands required for positive selection exposed to HY antigen for more than 14 hours, they
were used as a source of ‘virgin’ precursors. These developed into CD4+ SP T cells. These data tied together
immature cells were exposed to MHC ligands for their the preceding observations that concerned signal
TCRs in a first culture step to initiate positive selection, strength in lineage choice by indicating that co-receptor
then recovered and re-aggregated with a mixture of binding contributes to the maintenance of signalling
stromal cells, after which the development of SP T cells above some threshold level, and that cell fate is respon-
was analysed. The amount and quality of the MHC sive to the length of this supra-threshold input.
ligands in each culture step could be independently Experiments in mature T cells have shown that the cou-
manipulated, which permitted an analysis of how pling of Lck to the co-receptor helps to dictate the quality
TCR and co-receptor engagement at each stage affected of receptor signalling62, which, in turn, affects the rate of
lineage-specific differentiation. Recognition of MHC by feedback desensitization and hence, the duration of sig-
the TCR was found to be necessary in both cultures, nalling61. This hypothesis is also consistent with the
which confirmed earlier studies that indicated a need for effects of altered Lck activity or amount on the CD4–
persistent receptor signalling in T-cell development90–94. CD8 choice31,32 and with data on the role of the duration
More importantly, the use of mutant MHC class II mol- of protein kinase C (Pkc) and Mapk1 activation in CD4
ecules that were unable to interact with CD4 but were versus CD8 lineage commitment24,97.
still recognized by the TCR95 revealed that CD4 lineage However, pronase treatment of cells that were recov-
fate is instructed by the initial signalling event that is ered from the first-step cultures in the re-aggregate
involved in positive selection, based on whether the CD4 model showed that both co-receptor loci were still
co-receptor is or is not productively co-engaged with the active, even though lineage fate was already fixed. This
TCR at that time. meant that it was possible to separate commitment to
What about CD8+ T-cell differentiation? Given the the CD4 versus CD8 lineages from progression along
TCR–CD8 crosslinking data and these new findings, it these differentiation pathways, at least in terms of silenc-
was proposed that very efficient co-engagement of CD8 ing of the relevant co-receptor loci. Differentiation
with a strong class-I-binding TCR could produce suffi- apparently requires additional signals that are, perhaps,
cient signal to direct DP T cells into the CD4 lineage. dependent on epithelial elements in the thymus. The
Such fate choices would normally be undetectable, evidence from these experiments that thymocytes are
because cells that expressed a TCR with such high affin- fully lineage-committed after a few hours of TCR sig-
ity for a self-MHC ligand would undergo negative nalling but don’t have lineage-specific co-receptor
selection (apoptosis) before completing maturation. silencing, as well as data that show that CD8-lineage
However, in the two-stage culture system, such late commitment precedes CD4-lineage commitment tem-
deletion could be avoided by replacing the potent lig- porally97,98, is inconsistent with a distinct ‘duration of
and with a less avid one that still supported receptor signal’ model of development, in which lineage choice is
signalling in the re-aggregate phase. This prediction proposed to default to the CD4 pathway on initiation of
was tested using HY-specific-TCR-transgenic T cells. TCR signalling and to be reversed to CD8-lineage devel-
Cells that carry this TCR usually die in the presence of opment in the case of T cells that recognize an MHC
the strongly binding, MHC class-I-associated male HY class-I ligand if TCR signalling is abrogated within a few
antigen96, but develop into CD8+ SP T cells in the pres- hours due to loss of CD8 expression34 (FIG. 4).
ence of female cells that lack this antigen8. Remarkably,
when first exposed to male cells and then re-aggregated Imperfections in the lineage commitment process. A
with female cells, HY-specific-TCR-expressing precur- fate-commitment mechanism that is based on the
sors developed into SP CD4+ T cells, rather than CD8+ quantitative parameter of signal duration cannot be
T cells33 — similar to the results of Bommhardt et al.26 perfect. The initial DP T-cell pool will contain TCRs of
using TCR–CD8 crosslinking with antibodies. MHC class-I and class-II specificity that have overlap-
Watanabe et al. reported similar results that showed ping ranges of affinities for their respective ligands. In
that varying the magnitude of stimulation of thymo- the context of the commitment model described, this
cytes with a transgenic TCR that is specific for an MHC must inevitably lead to ‘wrong’ choices, even with
class-I ligand resulted in the generation of CD8+ SP asymmetric co-receptor–LCK association. Some cells
T cells at low ligand levels and CD4+ SP T cells at with TCRs that bind weakly to MHC class-II ligands
increased concentrations of antigen36. will begin developing along the CD8 pathway and some
What does ‘strength of signal’ mean? In the re- that have strong MHC class-I binding will opt for the
aggregate system described, limiting the quantity of lig- CD4 fate. Such ‘incorrect’ lineage choice presumably
and that was available to the TCR without influencing accounts for the ‘rescue’ of some developing T cells in

316 | MAY 2002 | VOLUME 2 www.nature.com/reviews/immunol


© 2002 Nature Publishing Group
REVIEWS

P For this reason, mismatches between the class of


MHC that is involved in TCR stimulation and
SHP1 CD4–CD8 fate choice, as revealed in the gene-targeting
studies, do not necessarily argue against instruction and
support stochastic commitment, as some investigators
– have proposed75,100 (see legend to FIG. 2 for details).
MHC SHP1 Rather, these mismatches are merely inconsistent with
an instruction model that postulates the control of lin-
TCR
P Y394 P Y394 eage choice by qualitatively unique signals that arise
from CD4 versus CD8 co-engagement. The existence of
S59
p56 LCK p56 LCK p59 LCK
P
‘mismatched’ cells is fully in accord with the predicted
inability of a pathway that combines quantitative
instruction with post-commitment selection to ensure
+ an absolute match between TCR specificity and cell fate.
It seems that some inaccuracy in the system of cell-fate
Relative generation rates
P
choice among developing T cells is not sufficiently dam-
Antagonist peptide aging for evolutionary pressures to have selected against
MAPK1 MAPK1 P this occurrence.
P P

SHP1 MAPK1
Biochemistry and cell biology of lineage choice. How
Agonist peptide
do variations in the strength and/or duration of
TCR–co-receptor signalling dictate lineage choice?
P P
Unfortunately, the information that is available on
MAPK1 SHP1 this important issue is limited at present. In part, this
is because many of the relevant molecules also have
Figure 5 | Two different roles for MAPK1 in thymocyte development. A large body of literature roles in earlier stages of thymocyte differentiation. For
supports a key role for mitogen-activated protein kinase 1 (MAPK1) in thymocyte selection and example, genetic inactivation of the Zap70 (REF. 48),
lineage-specific differentiation of thymocytes. Active MAPK1 seems to be crucial to successful Lck and Fyn49, or Slp76 (SH2-domain-containing
positive selection in general125–130, and higher levels of MAPK1 activity promote thymocytes to protein of 76 kDa)50 loci results in the arrest of devel-
become CD4+ single-positive (SP) T cells, whereas interference with MAPK1 activity results in opment at the DN3 stage (β-checkpoint), before the
excess CD8+ SP T-cell generation25,101. This kinase was, until recently, presumed to mediate its
CD4–CD8 decision.
effect on these developmental processes through its action on nuclear transcription factors, and
there is strong evidence that this is the case. For example, MAPK1 has been implicated in
Despite these limitations, some informative experi-
induction of the transcription factor early growth response 1 (EGR1) and the subsequent ments exist. Expression of a dominant active Mek1
expression of inhibitor of DNA binding 3 (ID3), which is necessary to inactivate E2A and permit (MAPK kinase 1) transgene — which activates Mapk1
T-cell development beyond the double-negative (DN)1 stage131,132. But, two other sets of — promotes the development of CD4+ T cells at the
experiments raise another, non-mutually exclusive, possibility. First, the data from the two-stage expense of CD8+ T cells25. Conversely, blockade of the
culture model indicate that the duration of T-cell receptor (TCR)–co-receptor signalling controls Mapk1 pathway results in a reduction in the generation
lineage choice33, and the work of Wilkinson and Kaye implicates the persistence of MAPK1
activation as a key aspect of the duration-controlled lineage decision97. Second, as shown in the
of CD4+ T cells and an increase in CD8+ T-cell produc-
figure, MAPK1 has been found to be the central element in a positive-feedback loop that regulates tion25,101. In a tumour-cell model, active RAS promotes
the duration of TCR signalling in mature T cells60. Active MAPK1 phosphorylates serine at position the loss of CD8 expression102, which is consistent with
59 and activates lymphocyte-specific protein-tyrosine kinase (LCK). This phosphorylation limits the data that indicate a role for the MAPK pathway in lin-
ability of SRC-homology-2-domain-containing protein tyrosine phosphatase 1 (SHP1) to bind and eage choice. The same laboratory has shown very
dephosphorylate LCK (crossed arrow). When SHP1 is recruited to the TCR complex, it inactivates recently that the duration of activation of MAPK1
LCK and ZAP70 (ζ-chain associated protein kinase, 70 kDa), which terminates T-cell
seems to have a key role in CD4+ versus CD8+ T-cell
signalling103–110. So, MAPK1 activity leads to prolonged TCR signalling by preventing SHP1 from
limiting the duration of LCK and ZAP70 kinase function. This would favour the entry of cells into the development97,98; more than ten hours of elevated
CD4 pathway, whereas blockade of MAPK1 activation would result in truncated signalling that MAPK1 activity is necessary for cells to become CD4+
promotes a CD8-lineage decision — as seen in experiments that modify MAPK1 activity in these SP T cells, which agrees with the data from the two-
two directions. Furthermore, MAPK1 is selectively involved in this feedback loop and is non- stage culture experiments33. Interestingly, a short
redundant for efficient T-cell development133. So, in addition to any direct effect of MAPK1 on the period of MAPK1 activity is crucial for the appearance
transcriptional control of gene expression, MAPK1 might contribute to the control of thymocyte
of CD8+ SP T cells in this model, because such differen-
fate indirectly through its control of TCR signalling, which could then act through other downstream
mediators to regulate the differentiation process. Distinguishing between the roles of the TCR
tiation is prevented by the complete blockade of the
feedback and direct downstream effects of MAPK1 on thymocyte development will require the MAPK pathway. It was suggested that CD8-lineage
generation of animals with selective mutations in LCK that affect only the TCR-feedback function development is the default pathway that is selected
of MAPK1 and not its nuclear activities. when initial signalling exceeds the ‘neglect’ threshold.
Prolonged MAPK1 signalling presumably overrides
this default condition and results in the generation of
mice when persistent co-receptor expression is CD4-lineage-committed T cells. In this regard, this
enforced through transgenic14,76–78 or gene-targeting75 model is the polar opposite of the co-receptor reversal
strategies, and the ‘incorrect’ co-receptor expression hypothesis of Singer and associates34. These data on the
that is observed on cells exposed to particular duration of MAPK signalling and lineage-choice do
TCR–MHC combinations99. agree with the findings of Ohaka et al.24.

NATURE REVIEWS | IMMUNOLOGY VOLUME 2 | MAY 2002 | 3 1 7


© 2002 Nature Publishing Group
REVIEWS

Box 2 | NOTCH and T-cell development


NOTCH is a transmembrane protein that is cleaved during
biosynthesis by a furin-like activity to produce a non-
covalently associated extracellular fragment and a
transmembrane/cytoplasmic fragment. The interaction of
NOTCH with one of its ligands (jagged 1 (JAG1), JAG2 or
NOTCH ligand
δ-like 1 (DLK1) in mammals) induces a further cleavage of the
molecule just inside the plasma membrane, which is mediated
by presenilin/γ-secretase. The released intracellular fragment NOTCH1–4
(NOTCH-IC) regulates gene expression by its effects on
C-promoter-binding factor 1 (CBP)/Su(H) and Deltex.
The former activates the HES (hairy and enhancer of split)
and nuclear factor-κB2 (NF-κB2) family of transcription
factors, whereas the latter inhibits JUN N-terminal kinase NOTCH-IC
(JNK) pathway signalling113. The regulation of NOTCH
function by several co-modifiers, such as Lunatic fringe137, is
not shown. The general model of NOTCH function in cell-fate
specification is that NOTCH signalling leads to the adoption
of a secondary fate, whereas the absence of NOTCH signalling CBF/Su(H) Deltex
allows a cell to adopt a default or primary fate113.
On the basis of data that were obtained using
overexpression of a constitutively active Notch1 transgene,
HES, NF-κB2 and JNK
NOTCH signalling was first proposed to affect thymocyte other mediators pathway
development by promoting the secondary fate of CD8-
lineage development in concert with ongoing TCR
signalling, whereas TCR signalling alone would lead to the
primary fate of CD4-lineage development115. However,
Yasutomo et al. did not find evidence of a role for Notch1 in
the CD4–CD8 decision in their two-step culture model33.
Instead, they observed a role for Notch in the differentiation of already committed cells along the CD8, but not the
CD4, developmental pathway. Deftos et al. concluded that Notch signalling contributed to the upregulation of
expression of B-cell lymphoma protein 2 (Bcl2)117. Animals that were transgenic for an active Notch1 fragment —
similar, but not identical, to that used by Robey et al. — had an accumulation of CD8+ T cells that did not enter the
peripheral pool, which Deftos et al. likened to results obtained by the transgenic expression of Bcl2 in thymocytes138.
No loss of CD4+ T cells was seen in these studies, although more recent analysis by others of younger animals from this
transgenic line indicates that the phenotype is similar to that seen by Robey et al.115. Wolfer et al. eliminated expression
of Notch1 using a conditional gene-targeting strategy, in which the inactivation of the locus occurs only after
commitment to the T-cell lineage118 and hence, does not affect the T-cell–B-cell decision as does germ-line Notch1
deletion39. Under these conditions, these investigators found no effect on CD4+ and CD8+ T-cell generation, from
which it was concluded that Notch1 had no role in later stages of T-cell development. These divergent data cannot be
fully reconciled at present, although the main text suggests some possible explanations. Figure adapted, with
permission, from REF. 111 © Oxford University Press (1999).

It is not known how the MAPK1 pathway influences (FIG. 5). If active MAPK1 helps to delay receptor desensi-
lineage fate, because very few target genes have yet to be tization in thymocytes by SH2-domain-containing
defined as being expressed specifically by one or the protein tyrosine phosphatase 1 (SHP1), as it does in
other lineage, other than the co-receptor loci. Both co- mature T cells103–110, the manipulations of Mapk1 activ-
receptor genes have very complex sets of promoters ity in these various studies might not produce their
and enhancers; CD4 also uses a specific silencer ele- effects on lineage fate only by acting downstream on
ment for regulation65–75. Different regulatory elements gene transcription, as is commonly assumed. Rather,
control the initial expression of CD4 and CD8 by MAPK1 might contribute to lineage choice by prolong-
T cells that pass the β-checkpoint; the extinction of ing TCR signalling when LCK-coupled co-receptor
expression of a co-receptor during lineage progression recruitment is adequate, so as to reach the threshold
after the initiation of selection; and the maintenance of duration for commitment to the CD4 lineage.
expression of one co-receptor on cells of a given lin- Additional studies that separate the proximal feedback
eage. The biochemical events that link proximal TCR effects of MAPK1 from the gene-regulatory effects are
signals to the activity of these various co-receptor needed to clarify this issue.
genetic regulatory elements remain to be defined. In
addition, evidence that MAPK1 has a feedback regula- Notch and CD4–CD8 lineage choice and progression.
tory role in TCR signalling 60 complicates interpretation Another receptor–ligand system has also been pro-
of the data on the control of lineage fate by this kinase posed to have a central role in controlling thymocyte

318 | MAY 2002 | VOLUME 2 www.nature.com/reviews/immunol


© 2002 Nature Publishing Group
REVIEWS

development (BOX 2). NOTCH and its ligands are This still leaves the data by Radkte et al. on Notch1
expressed early in thymic development on both T-cell deletion to be explained. One can either conclude that
precursors and stromal elements111–114. Robey and the transgenic and in vitro studies are both unphysio-
associates reported that in the presence of TCR–MHC logical and irrelevant, or consider the possibility that
signalling, a constitutively active Notch1 transgene members of the Notch family other than Notch1 (par-
promotes the development of CD8+ SP T cells at the ticularly Notch2) might be responsible for the effects
expense of CD4+ SP T cells, and that CD8+ SP T cells seen by Yasutomo et al. and produce the physiological
are generated in these transgenic animals even when counterpart of the signalling that is induced using
only MHC class-II ligands are available to the TCR of active Notch1 transgenes. Immature thymocytes
the DP T cell115. Their interpretation of these data was express several NOTCH-family members111,113,114, so
that in the presence of a permissive TCR signal, Notch only additional studies will answer this question. One
dictates the lineage fate of thymocytes, as it does for so reason to continue to focus on NOTCH as a potential
many other developing tissues. Others have also player in post-signalling lineage progression comes
reported an effect of Notch on the CD4:CD8 ratio116. from the identification of sites in the regulatory regions
This conclusion has been challenged, however, by sub- of CD4 that bind HES1 (hairy and enhancer of split 1)73
sequent studies. Deftos et al. reported that a similar, — a key signalling molecule in the NOTCH pathway113.
active Notch1 transgene promotes thymocyte sur- Although evidence is lacking that such binding controls
vival117. They concluded that active Notch signalling the expression of CD4 in a manner that is consistent
results in the accumulation of the same immature with its silencing in cells that are developing in the CD8
CD8+CD4− thymocytes that are seen in transgenic pathway, this is a tempting possibility.
mice that express the anti-apoptotic protein B-cell
lymphoma protein 2 (Bcl2), and that Notch does not Conclusion
affect CD4- versus CD8-lineage choice. This picture is The past few years have seen substantial progress in
complicated by a more recent analysis of younger elucidating the molecular genetic basis for commit-
mice of this strain, which gives a picture that is more ment to the T-cell versus B-cell lineages and for effec-
similar to that of Robey and colleagues (B. J. Fowlkes, tive progression from immature to mature T cells, irre-
personal communication). Using the re-aggregate sys- spective of co-receptor-defined lineage122. Advances at
tem, Yasutomo et al. also found no evidence for this level of resolution are still lacking with respect to
Notch-mediated control of lineage commitment33. the commitment of precursor T cells to the CD4
However, in contrast to Deftos et al., these investiga- (helper) versus CD8 (cytotoxic) pathways. This is not
tors did see a selective contribution of Notch to post- to say that the field has not moved forward. Although
commitment progression along the CD8, but not not all workers interpret the available data in the same
CD4, pathway. Finally, any role for Notch1 in thymo- way, the bulk of evidence appears to favour a model in
cyte development after the β-checkpoint has been which the extent/duration of signalling that results
called into question by the absence of an effect of from TCR and co-receptor engagement of self-pep-
selective Notch1 gene deletion at the DN stage on tide–MHC ligands controls the lineage choice of bipo-
CD4+ or CD8+ SP T-cell differentiation118. tential DP αβ T cells. Because the decision is based on
At present, it is not possible to propose a fully a quantitative effect of ligand recognition, and the
coherent model that incorporates all of these data. range of TCR recognition affinities on precursor cells
One way to reconcile the active Notch1-transgene115 is broad and overlapping between those that preferen-
and reaggregate-culture33 data is to draw on the obser- tially engage self-peptide–MHC-class-I versus -class-II
vations of Pear and co-workers119. These investigators ligands, errors are made in the initial decision process.
found that the overexpression of active Notch1 down- These mistakes are largely, but not completely, cor-
regulated TCR signalling in developing thymocytes rected during the later stages of maturation, as co-
and blocked maturation. Several laboratories have receptor extinction takes place. So, elements from
reported that TCR signalling in response to the type both of the two competing hypotheses that concern
of weak ligands that promote positive selection thymocyte lineage commitment have been found to
becomes less efficient as a cell progresses past the DP contribute to the developmental process — ‘sloppy’
stage120,121. Therefore, it is possible that the overexpres- instruction at the first step and ‘leaky’ selection at the
sion of active Notch pushes early T cells along this second step (FIG. 2). The end result is a pool of mature
pathway of diminished TCR signalling capacity. This T cells that, for the most part, possesses the coordina-
would, in turn, put a larger fraction of the cells into tion of TCR and co-receptor specificity that is
the low/short signal-duration pool, which promotes required for effective immune responses to foreign
CD8- rather than CD4-lineage development, as antigens. Whether ‘mistakes’ that survive are simply
observed by Robey et al. This would not be due to a developmental ‘noise’ of no consequence, or whether
direct effect of Notch on lineage choice, but to an such cells might have particular roles in autoimmune
indirect effect on TCR signalling. The findings of disease, remains unknown. With some luck, it will be
Yasutomo et al. are consistent with this suggestion — only a few years before another review of this topic
loss of Notch expression did not affect fate choice, can provide a nearly complete map of CD4- and
only the development of already committed cells CD8-lineage development from the cell surface to the
along the CD8 maturation pathway. genome and back.

NATURE REVIEWS | IMMUNOLOGY VOLUME 2 | MAY 2002 | 3 1 9


© 2002 Nature Publishing Group
REVIEWS

1. Freeman, M. Feedback control of intercellular signalling in 22. Itano, A. et al. The cytoplasmic domain of CD4 promotes 43. von Boehmer, H. & Fehling, H. J. Structure and function of
development. Nature 408, 313–319 (2000). the development of CD4-lineage T cells. J. Exp. Med. the pre-T-cell receptor. Annu. Rev. Immunol. 15, 433–452
2. Germain, R. N. MHC-dependent antigen processing and 183, 731–741 (1996). (1997).
peptide presentation: providing ligands for T-lymphocyte 23. Matechak, E. O., Killeen, N., Hedrick, S. M. & Fowlkes, B. J. 44. Aifantis, I., Feinberg, J., Fehling, H. J., Di Santo, J. P. &
activation. Cell 76, 287–299 (1994). MHC class-II-specific T cells can develop in the CD8 von Boehmer, H. Early T-cell receptor-β gene expression
3. Biddison, W. E., Rao, P. E., Talle, M. A., Goldstein, G. & lineage when CD4 is absent. Immunity 4, 337–347 (1996). is regulated by the pre-T-cell-receptor–CD3 complex.
Shaw, S. Possible involvement of the OKT4 molecule in References 22 and 23 contain the initial descriptions J. Exp. Med. 190, 141–144 (1999).
T-cell recognition of class II HLA antigens. Evidence from of the ‘strength of signal’ model. 45. Mombaerts, P. et al. RAG-1-deficient mice have no
studies of cytotoxic T lymphocytes specific for SB 24. Ohoka, Y. et al. Regulation of thymocyte lineage mature B and T lymphocytes. Cell 68, 869–877 (1992).
antigens. J. Exp. Med. 156, 1065–1076 (1982). commitment by the level of classical protein kinase C 46. Shinkai, Y. et al. Restoration of T-cell development in
4. Swain, S. L. T-cell subsets and the recognition of MHC activity. J. Immunol. 158, 5707–5716 (1997). RAG-2-deficient mice by functional TCR transgenes.
class. Immunol. Rev. 74, 129–142 (1983). 25. Sharp, L. L., Schwarz, D. A., Bott, C. M., Marshall, C. J. & Science 259, 822–825 (1993).
5. Janeway, C. A. Jr. The T-cell receptor as a multicomponent Hedrick, S. M. The influence of the MAPK pathway on 47. van Oers, N. S., von Boehmer, H. & Weiss, A. The pre-
signalling machine: CD4/CD8 coreceptors and CD45 in T-cell lineage commitment. Immunity 7, 609–618 (1997). T-cell receptor (TCR) complex is functionally coupled to
T-cell activation. Annu. Rev. Immunol. 10, 645–674 (1992). This study provides clear evidence that high MAPK the TCR-ζ subunit. J. Exp. Med. 182, 1585–1590 (1995).
6. Robey, E. & Fowlkes, B. J. Selective events in T-cell activity favours the development of CD4+ single- 48. Negishi, I. et al. Essential role for ZAP-70 in both positive
development. Annu. Rev. Immunol. 12, 675–705 (1994). positive T cells and low MAPK activity favours the and negative selection of thymocytes. Nature 376,
7. Kruisbeek, A. M. et al. Absence of the Lyt-2-L3T4+ lineage development of CD8+ single-positive T cells. 435–438 (1995).
of T cells in mice treated neonatally with anti-I-A correlates 26. Bommhardt, U., Cole, M. S., Tso, J. Y. & Zamoyska, R. 49. van Oers, N. S., Lowin-Kropf, B., Finlay, D., Connolly, K. &
with absence of intrathymic I-A-bearing antigen-presenting- Signals through CD8 or CD4 can induce commitment to Weiss, A. αβ T-cell development is abolished in mice
cell function. J. Exp. Med. 161, 1029–1047 (1985). the CD4 lineage in the thymus. Eur. J. Immunol. 27, lacking both Lck and Fyn protein tyrosine kinases.
One of the very first papers to show that the class of 1152–1163 (1997). Immunity 5, 429–436 (1996).
MHC molecule that is recognized controls the 27. Goldrath, A. W., Hogquist, K. A. & Bevan, M. J. CD8- 50. Clements, J. L. et al. Requirement for the leukocyte-
development of a particular mature co-receptor- lineage commitment in the absence of CD8. Immunity specific adapter protein SLP-76 for normal T-cell
defined T-cell subset. 6, 633–642 (1997). development. Science 281, 416–419 (1998).
8. Teh, H. S. et al. Thymic major histocompatibility complex 28. Sebzda, E., Choi, M., Fung-Leung, W. P., Mak, T. W. & 51. von Boehmer, H., Teh, H. S. & Kisielow, P. The thymus
antigens and the αβ T-cell receptor determine the Ohashi, P. S. Peptide-induced positive selection of TCR- selects the useful, neglects the useless and destroys the
CD4/CD8 phenotype of T cells. Nature 335, 229–233 transgenic thymocytes in a coreceptor-independent harmful. Immunol. Today 10, 57–61 (1989).
(1988). manner. Immunity 6, 643–653 (1997). A seminal description of the TCR-dependent
The first use of TCR-transgenic mice to reveal the 29. Basson, M. A., Bommhardt, U., Cole, M. S., Tso, J. Y. & selection events that control thymocyte fate.
relationship between MHC class-specificity in Zamoyska, R. CD3 ligation on immature thymocytes 52. Merkenschlager, M. et al. How many thymocytes audition
foreign antigen recognition and co-receptor-defined generates antagonist-like signals appropriate for CD8 for selection? J. Exp. Med. 186, 1149–1158 (1997).
lineage development in the thymus. lineage commitment, independently of T-cell receptor 53. Zerrahn, J., Held, W. & Raulet, D. H. The MHC reactivity of
9. Marusic-Galesic, S., Longo, D. L. & Kruisbeek, A. M. specificity. J. Exp. Med. 187, 1249–1260 (1998). the T-cell repertoire prior to positive and negative
Preferential differentiation of T-cell receptor specificities 30. Basson, M. A., Bommhardt, U., Mee, P. J., Tybulewicz, V. L. selection. Cell 88, 627–636 (1997).
based on the MHC glycoproteins encountered during & Zamoyska, R. Molecular requirements for lineage 54. Davis, M. M. & Bjorkman, P. J. T-cell antigen receptor
development. Evidence for positive selection. J. Exp. commitment in the thymus — antibody-mediated genes and T-cell recognition. Nature 334, 395–402
Med. 169, 1619–1630 (1989). receptor engagements reveal a central role for Lck in (1988).
10. Kaye, J. et al. Selective development of CD4+ T cells in lineage decisions. Immunol. Rev. 165, 181–194 (1998). 55. Anderson, S. J., Levin, S. D. & Perlmutter, R. M.
transgenic mice expressing a class II MHC-restricted 31. Legname, G. et al. Inducible expression of a p56Lck Involvement of the protein tyrosine kinase p56lck in T-cell
antigen receptor. Nature 341, 746–749 (1989). transgene reveals a central role for Lck in the signaling and thymocyte development. Adv. Immunol. 56,
An extension of the results of reference 8 to MHC differentiation of CD4+ SP thymocytes. Immunity 12, 151–178 (1994).
class-II recognition and CD4+ single-positive T-cell 537–546 (2000). 56. Weiss, A. & Littman, D. R. Signal transduction by
development. 32. Hernandez-Hoyos, G., Sohn, S. J., Rothenberg, E. V. & lymphocyte antigen receptors. Cell 76, 263–274 (1994).
11. Robey, E. A., Fowlkes, B. J. & Pardoll, D. M. Molecular Alberola-Ila, J. Lck activity controls CD4/CD8 T-cell 57. Kane, L. P., Lin, J. & Weiss, A. Signal transduction by the
mechanisms for lineage commitment in T-cell lineage commitment. Immunity 12, 313–322 (2000). TCR for antigen. Curr. Opin. Immunol. 12, 242–249
development. Semin. Immunol. 2, 25–34 (1990). 33. Yasutomo, K., Doyle, C., Miele, L., Fuchs, C. & Germain, R. N. (2000).
12. Borgulya, P., Kishi, H., Müller, U., Kirberg, J. & The duration of antigen-receptor signalling determines 58. Zhang, W., Sloan-Lancaster, J., Kitchen, J., Trible, R. P. &
von Boehmer, H. Development of the CD4 and CD8 CD4+ versus CD8+ T-cell lineage fate. Nature 404, Samelson, L. E. LAT: the ZAP-70 tyrosine kinase
lineage of T cells: instruction versus selection. EMBO J. 506–510 (2000). substrate that links T-cell receptor to cellular activation.
10, 913–918 (1991). This study used a re-aggregate culture system to Cell 92, 83–92 (1998).
13. Robey, E. A. et al. Thymic selection in CD8-transgenic separate early and late signals through the TCR and 59. Kuo, C. T. & Leiden, J. M. Transcriptional regulation of
mice supports an instructive model for commitment to a co-receptor in controlling thymocyte development. T-lymphocyte development and function. Annu. Rev.
CD4 or CD8 lineage. Cell 64, 99–107 (1991). It showed that signal duration during the initiation of Immunol. 17, 149–187 (1999).
References 11–13 set out the original proposals for positive selection ‘instructs’ lineage choice and is 60. Germain, R. N. & Stefanova, I. The dynamics of T-cell
the ‘instruction’ versus ‘selection’ models of lineage controlled by co-operation between the TCR and receptor signaling: complex orchestration and the key
commitment and development. They also provided co-receptors. roles of tempo and cooperation. Annu. Rev. Immunol. 17,
the first experimental tests of the models, which 34. Brugnera, E. et al. Coreceptor reversal in the thymus: 467–522 (1999).
looked for ‘rescue’ with co-receptor transgenes. signaled CD4+8+ thymocytes initially terminate CD8 61. Germain, R. N. The T-cell receptor for antigen: signaling
14. Davis, C. B., Killeen, N., Crooks, M. E., Raulet, D. & transcription even when differentiating into CD8+ T cells. and ligand discrimination. J. Biol. Chem. 276,
Littman, D. R. Evidence for a stochastic mechanism in the Immunity 13, 59–71 (2000). 35222–35226 (2001).
differentiation of mature subsets of T lymphocytes. Cell 35. Itano, A. & Robey, E. Highly efficient selection of CD4- and 62. Madrenas, J., Chau, L. A., Smith, J., Bluestone, J. A. &
73, 237–247 (1993). CD8-lineage thymocytes supports an instructive model of Germain, R. N. The efficiency of CD4 recruitment to
15. Chan, S. H., Cosgrove, D., Waltzinger, C., Benoist, C. & lineage commitment. Immunity 12, 383–389 (2000). ligand-engaged TCR controls the agonist/partial agonist
Mathis, D. Another view of the selective model of 36. Watanabe, N., Arase, H., Onodera, M., Ohashi, P. S. & properties of peptide–MHC molecule ligands. J. Exp.
thymocyte selection. Cell 73, 225–236 (1993). Saito, T. The quantity of TCR signal determines positive Med. 185, 219–229 (1997).
16. van Meerwijk, J. P. & Germain, R. N. Development of selection and lineage commitment of T cells. J. Immunol. 63. Hampl, J., Chien, Y. H. & Davis, M. M. CD4 augments the
mature CD8+ thymocytes: selection rather than 165, 6252–6261 (2000). response of a T cell to agonist but not to antagonist
instruction? Science 261, 911–915 (1993). 37. Radtke, F. et al. Deficient T-cell fate specification in mice ligands. Immunity 7, 379–385 (1997).
References 15 and 16 conclude that stochastic with an induced inactivation of Notch1. Immunity 10, 64. Bosselut, R. et al. Association of the adaptor molecule
choice/selection is more probable than instruction, 547–558 (1999). LAT with CD4 and CD8 coreceptors identifies a new
based on the appearance of transitional-phenotype 38. Pui, J. C. et al. Notch1 expression in early lymphopoiesis coreceptor function in T-cell receptor signal transduction.
cells in various MHC-deficient mice. influences B- versus T-lineage determination. Immunity J. Exp. Med. 190, 1517–1526 (1999).
17. von Boehmer, H. & Kisielow, P. Lymphocyte lineage 11, 299–308 (1999). 65. Sawada, S. & Littman, D. R. Identification and
commitment: instruction versus selection. Cell 73, 39. Wilson, A., MacDonald, H. R. & Radtke, F. Notch-1- characterization of a T-cell-specific enhancer adjacent to
207–208 (1993). deficient common lymphoid precursors adopt a B-cell fate the murine CD4 gene. Mol. Cell. Biol. 11, 5506–5515
18. Chan, S. H., Benoist, C. & Mathis, D. In favor of the in the thymus. J. Exp. Med. 194, 1003–1012 (2001). (1991).
selective model of positive selection. Semin. Immunol. 40. Michie, A. M. et al. Clonal characterization of a bipotent 66. Sawada, S., Scarborough, J. D., Killeen, N. & Littman, D. R.
6, 241–248 (1994). T-cell and NK-cell progenitor in the mouse fetal thymus. A lineage-specific transcriptional silencer regulates CD4
19. van Meerwijk, J. P. M., O’Connell, E. M. & Germain, R. N. J. Immunol. 164, 1730–1733 (2000). gene expression during T-lymphocyte development. Cell
Evidence for lineage commitment and initiation of positive 41. Godfrey, D. I., Kennedy, J., Suda, T. & Zlotnik, A. 77, 917–929 (1994).
selection by thymocytes with intermediate surface A developmental pathway involving four phenotypically 67. Ellmeier, W., Sunshine, M. J., Losos, K., Hatam, F. &
phenotypes. J. Immunol. 154, 6314–6323 (1995). and functionally distinct subsets of CD3−CD4−CD8− triple- Littman, D. R. An enhancer that directs lineage-specific
20. Kersh, G. J. & Hedrick, S. M. Role of TCR specificity in negative adult mouse thymocytes defined by CD44 and expression of CD8 in positively selected thymocytes and
CD4 versus CD8 lineage commitment. J. Immunol. 154, CD25 expression. J. Immunol. 150, 4244–4252 (1993). mature T cells. Immunity 7, 537–547 (1997).
1057–1068 (1995). 42. Groettrup, M. et al. A novel disulfide-linked heterodimer 68. Ellmeier, W., Sawada, S. & Littman, D. R. The regulation of
21. von Boehmer, H. CD4/CD8 lineage commitment: back to on pre-T cells consists of the T-cell receptor β-chain and a CD4 and CD8 coreceptor gene expression during T-cell
instruction? J. Exp. Med. 183, 713–715 (1996). 33 kD glycoprotein. Cell 75, 283–294 (1993). development. Annu. Rev. Immunol. 17, 523–554 (1999).

320 | MAY 2002 | VOLUME 2 www.nature.com/reviews/immunol


© 2002 Nature Publishing Group
REVIEWS

69. Hostert, A. et al. A CD8 genomic fragment that directs 86. Seong, R. H., Chamberlain, J. W. & Parnes, J. R. Signal 107. Zhang, J. et al. Involvement of the SHP-1 tyrosine
subset-specific expression of CD8 in transgenic mice. for T-cell differentiation to a CD4 cell lineage is delivered phosphatase in regulation of T-cell selection. J. Immunol.
J. Immunol. 158, 4270–4281 (1997). by CD4 transmembrane region and/or cytoplasmic tail. 163, 3012–3021 (1999).
70. Hostert, A. et al. A region in the CD8 gene locus that Nature 356, 718–720 (1992). 108. Johnson, K. G., LeRoy, F. G., Borysiewicz, L. K. &
directs expression to the mature CD8 T-cell subset in 87. Itano, A., Cado, D., Chan, F. K. & Robey, E. A role for the Matthews, R. J. TCR signaling thresholds regulating T-cell
transgenic mice. Immunity 7, 525–536 (1997). cytoplasmic tail of the β-chain of CD8 in thymic selection. development and activation are dependent upon SHP-1.
71. Adlam, M., Duncan, D. D., Ng, D. K. & Siu, G. Positive Immunity 1, 287–290 (1994). J. Immunol. 162, 3802–3813 (1999).
selection induces CD4 promoter and enhancer function. 88. Veillette, A., Zuniga-Pflucker, J. C., Bolen, J. B. & 109. Plas, D. R. et al. Cutting edge: the tyrosine phosphatase
Int. Immunol. 9, 877–887 (1997). Kruisbeek, A. M. Engagement of CD4 and CD8 SHP-1 regulates thymocyte positive selection.
72. Hostert, A. et al. Hierarchical interactions of control expressed on immature thymocytes induces activation of J. Immunol. 162, 5680–5684 (1999).
elements determine CD8α gene expression in subsets of intracellular tyrosine phosphorylation pathways. J. Exp. 110. Chiang, G. G. & Sefton, B. M. Specific dephosphorylation
thymocytes and peripheral T cells. Immunity 9, 497–508 Med. 170, 1671–1680 (1989). of the Lck tyrosine-protein kinase at Tyr-394 by the SHP-1
(1998). The first report of marked asymmetry of Lck protein-tyrosine phosphatase. J. Biol. Chem. 276,
73. Kim, H. K. & Siu, G. The notch pathway intermediate association with CD4 versus CD8 in thymocytes. 23173–23178 (2001).
HES-1 silences CD4 gene expression. Mol. Cell. Biol. 18, 89. Wiest, D. L. et al. Regulation of T-cell receptor expression 111. Felli, M. P. et al. Expression pattern of notch1, -2 and -3
7166–7175 (1998). in immature CD4+CD8+ thymocytes by p56lck tyrosine and Jagged1 and -2 in lymphoid and stromal thymus
74. Zou, Y. R. et al. Epigenetic silencing of CD4 in T cells kinase: basis for differential signaling by CD4 and CD8 in components: distinct ligand–receptor interactions in
committed to the cytotoxic lineage. Nature Genet. 29, immature thymocytes expressing both coreceptor intrathymic T-cell development. Int. Immunol. 11,
332–336 (2001). molecules. J. Exp. Med. 178, 1701–1712 (1993). 1017–1025 (1999).
75. Leung, R. K. et al. Deletion of the CD4 silencer element 90. Pircher, H., Ohashi, P. S., Boyd, R. L., Hengartner, H. & 112. Robey, E. Notch in vertebrates. Curr. Opin. Genet. Dev.
supports a stochastic mechanism of thymocyte lineage Brduscha, K. Evidence for a selective and multi-step 7, 551–557 (1997).
commitment. Nature Immunol. 2, 1167–1173 (2001). model of T-cell differentiation: CD4+CD8low thymocytes 113. Osborne, B. & Miele, L. Notch and the immune system.
This report argues for a stochastic/selection model, selected by a transgenic T-cell receptor on major Immunity 11, 653–663 (1999).
on the basis of MHC class-II-dependent production histocompatibility complex class I molecules. Eur. J. 114. Anderson, G., Pongracz, J., Parnell, S. & Jenkinson, E. J.
of CD8-lineage cells after the elimination of CD4 Immunol. 24, 1982–1987 (1994). Notch-ligand-bearing thymic epithelial cells initiate and
silencer function. See the text and legend to FIG. 4 91. Kisielow, P. & Miazek, A. Positive selection of T cells: sustain Notch signaling in thymocytes independently of
for an alternative interpretation of these data that is rescue from programmed cell death and differentiation T-cell receptor signaling. Eur. J. Immunol. 31, 3349–3354
consistent with an instruction model. require continual engagement of the T-cell receptor. (2001).
76. Davis, C. B. & Littman, D. R. Thymocyte lineage J. Exp. Med. 181, 1975–1984 (1995). 115. Robey, E. et al. An activated form of Notch influences the
commitment: is it instructed or stochastic? Curr. Opin. 92. Wilkinson, R. W., Anderson, G., Owen, J. J. & Jenkinson, E. J. choice between CD4 and CD8 T-cell lineages. Cell 87,
Immunol. 6, 266–272 (1994). Positive selection of thymocytes involves sustained 483–492 (1996).
77. Itano, A., Kioussis, D. & Robey, E. Stochastic component interactions with the thymic microenvironment. This study concludes, from the phenotype of mice
to development of class I major-histocompatibility- J. Immunol. 155, 5234–5240 (1995). that express an active Notch1 transgene, that Notch
complex-specific T cells. Proc. Natl Acad. Sci. USA 91, 93. Kirberg, J., Berns, A. & von Boehmer, H. Peripheral T-cell signaling has a central role in CD4–CD8 lineage
220–224 (1994). survival requires continual ligation of the T-cell receptor to choice.
78. Robey, E., Itano, A., Fanslow, W. C. & Fowlkes, B. J. major-histocompatibility-complex-encoded molecules. 116. Jimenez, E. et al. Distinct mechanisms contribute to
Constitutive CD8 expression allows inefficient maturation J. Exp. Med. 186, 1269–1275 (1997). generate and change the CD4:CD8 cell ratio during
of CD4+ helper T cells in class II major histocompatibility 94. Yasutomo, K., Lucas, B. & Germain, R. N. TCR signaling thymus development: a role for the Notch ligand,
complex mutant mice. J. Exp. Med. 179, 1997–2004 for initiation and completion of thymocyte positive Jagged1. J. Immunol. 166, 5898–5908 (2001).
(1994). selection has distinct requirements for ligand quality and 117. Deftos, M. L., Huang, E., Ojala, E. W., Forbush, K. A. &
79. Guidos, C. J., Danska, J. S., Fathman, C. G. & presenting-cell type. J. Immunol. 165, 3015–3022 (2000). Bevan, M. J. Notch1 signaling promotes the maturation of
Weissman, I. L. T-cell-receptor-mediated negative 95. Riberdy, J. M., Mostaghel, E. & Doyle, C. Disruption of the CD4+ and CD8+ SP thymocytes. Immunity 13, 73–84
selection of autoreactive T-lymphocyte precursors occurs CD4–major-histocompatibility-complex-class-II interaction (2000).
after commitment to the CD4 or CD8 lineages. J. Exp. blocks the development of CD4+ T cells in vivo. Proc. Natl This study presents evidence from a different
Med. 172, 835–845 (1990). Acad. Sci. USA 95, 4493–4498 (1998). active Notch1-transgenic model that Notch
An initial description of intermediate surface- 96. Kisielow, P., Blüthmann, H., Staerz, U. D., Steinmetz, M. & contributes to the survival of both developing
antigen phenotypes that accompany the double- von Boehmer, H. Tolerance in T-cell-receptor transgenic CD4- and CD8-lineage cells; it argues against the
positive to single-positive transition. The concept of mice involves deletion of nonmature CD4+CD8+ concept that Notch has a selective role in the CD8
‘linear’ co-receptor loss that arose from this paper thymocytes. Nature 333, 742–746 (1988). pathway.
was responsible for the difficulty in correctly 97. Wilkinson, B. & Kaye, J. Requirement for sustained 118. Wolfer, A. et al. Inactivation of Notch 1 in immature
interpreting later studies of intermediate-phenotype MAPK signaling in both CD4 and CD8 lineage thymocytes does not perturb CD4+ or CD8+ T-cell
cells in MHC-deficient mice. commitment: a threshold model. Cell. Immunol. 211, development. Nature Immunol. 2, 235–241 (2001).
80. Crump, A. L., Grusby, M. J., Glimcher, L. H. & Cantor, H. 86–95 (2001). Using a conditional deletion strategy, it is shown
Thymocyte development in major-histocompatibility- 98. Shao, H., Wilkinson, B., Lee, B., Han, P. C. & Kaye, J. that Notch1 expression beyond the double-negative
complex-deficient mice: evidence for stochastic Slow accumulation of active mitogen-activated protein stage is not important in normal CD4+ and CD8+
commitment to the CD4 and CD8 lineages. Proc. Natl kinase during thymocyte differentiation regulates the single-positive T-cell development.
Acad. Sci. USA 90, 10739–10743 (1993). temporal pattern of transcription-factor gene expression. 119. Izon, D. J. et al. Notch1 regulates maturation of CD4+ and
81. Suzuki, H., Punt, J. A., Granger, L. G. & Singer, A. J. Immunol. 163, 603–610 (1999). CD8+ thymocytes by modulating TCR signal strength.
Asymmetric signaling requirements for thymocyte 99. Chan, S., Correia-Neves, M., Dierich, A., Benoist, C. & Immunity 14, 253–264 (2001).
commitment to the CD4+ versus CD8+ T-cell lineages: a Mathis, D. Visualization of CD4/CD8 T-cell commitment. 120. Davey, G. M. et al. Preselection thymocytes are more
new perspective on thymic commitment and selection. J. Exp. Med. 188, 2321–2333 (1998). sensitive to T-cell receptor stimulation than mature T cells.
Immunity 2, 413–425 (1995). 100. Chan, S., Correia-Neves, M., Benoist, C. & Mathis, D. J. Exp. Med. 188, 1867–1874 (1998).
82. Lundberg, K., Heath, W., Köntgen, F., Carbone, F. & CD4/CD8 lineage commitment: matching fate with 121. Lucas, B., Stefanova, I., Yasutomo, K., Dautigny, N. &
Shortman, K. Intermediate steps in positive selection: competence. Immunol. Rev. 165, 195–207 (1998). Germain, R. N. Divergent changes in the sensitivity of
differentiation of CD4+CD8intTCRint thymocytes into CD4− 101. Bommhardt, U., Basson, M. A., Krummrei, U. & maturing T cells to structurally related ligands underlies
CD8+TCRhi thymocytes. J. Exp. Med. 181, 1643–1651 Zamoyska, R. Activation of the extracellular signal-related formation of a useful T-cell repertoire. Immunity 10,
(1995). kinase/mitogen-activated protein kinase pathway 367–376 (1999).
83. Lucas, B. & Germain, R. N. Unexpectedly complex discriminates CD4- versus CD8-lineage commitment in 122. Rothenberg, E. V. Stepwise specification of lymphocyte
regulation of CD4/CD8 coreceptor expression supports a the thymus. J. Immunol. 163, 715–722 (1999). developmental lineages. Curr. Opin. Genet. Dev. 10,
revised model for CD4+CD8+ thymocyte differentiation. 102. Shao, H., Rubin, E. M., Chen, L. Y. & Kaye, J. A role for 370–379 (2000).
Immunity 5, 461–477 (1996). Ras signaling in coreceptor regulation during 123. Matzinger, P. & Guerder, S. Does T-cell tolerance require a
This paper proposes a complex, non-linear pattern differentiation of a double-positive thymocyte cell line. dedicated antigen-presenting cell? Nature 338, 74–76
of changes in co-receptor expression during J. Immunol. 159, 5773–5776 (1997). (1989).
thymocyte maturation that accounts for the 103. Lorenz, U., Ravichandran, K. S., Burakoff, S. J. & Neel, B. G. 124. Sant’Angelo, D. B. et al. A molecular map of T-cell
existence of transitional cells in MHC-deficient Lack of SHPTP1 results in src-family kinase development. Immunity 9, 179–186 (1998).
animals without postulating stochastic choice. hyperactivation and thymocyte hyperresponsiveness. 125. Alberola-Ila, J., Forbush, K. A., Seger, R., Krebs, E. G. &
This model also explains the data in references 81 Proc. Natl Acad. Sci. USA 93, 9624–9629 (1996). Perlmutter, R. M. Selective requirement for MAP kinase
and 82, which show that CD4+CD8low thymocytes 104. Pani, G., Fischer, K. D., Mlinaric-Rascan, I. & Siminovitch, K. A. activation in thymocyte differentiation. Nature 373,
can generate both CD4+ and CD8+ single-positive Signaling capacity of the T-cell antigen receptor is 620–623 (1995).
T cells. negatively regulated by the PTP1C tyrosine phosphatase. 126. Alberola-Ila, J., Hogquist, K. A., Swan, K. A., Bevan, M. J.
84. Barthlott, T., Kohler, H., Pircher, H. & Eichmann, K. J. Exp. Med. 184, 839–852 (1996). & Perlmutter, R. M. Positive and negative selection invoke
Differentiation of CD4highCD8low coreceptor-skewed 105. Plas, D. R. et al. Direct regulation of ZAP-70 by SHP-1 in distinct signaling pathways. J. Exp. Med. 184, 9–18
thymocytes into mature CD8 single-positive cells T-cell antigen-receptor signaling. Science 272, (1996).
independent of MHC class-I recognition. Eur. J. Immunol. 1173–1176 (1996). 127. O’Shea, C. C., Crompton, T., Rosewell, I. R., Hayday, A. C.
27, 2024–2032 (1997). 106. Carter, J. D., Neel, B. G. & Lorenz, U. The tyrosine & Owen, M. J. Raf regulates positive selection. Eur. J.
85. Correia-Neves, M., Mathis, D. & Benoist, C. A molecular phosphatase SHP-1 influences thymocyte selection by Immunol. 26, 2350–2355 (1996).
chart of thymocyte positive selection. Eur. J. Immunol. 31, setting TCR signaling thresholds. Int. Immunol. 11, 128. Mariathasan, S., Ho, S. S., Zakarian, A. & Ohashi, P. S.
2583–2592 (2001). 1999–2014 (1999). Degree of ERK activation influences both positive and

NATURE REVIEWS | IMMUNOLOGY VOLUME 2 | MAY 2002 | 3 2 1


© 2002 Nature Publishing Group
REVIEWS

negative thymocyte selection. Eur. J. Immunol. 30, are both required for T-cell development in the thymus. the ideas in this review. I also thank B.J. Fowlkes for reading a draft
1060–1068 (2000). Nature 362, 70–73 (1993). of this manuscript and making many very helpful suggestions for
129. Gong, Q. et al. Disruption of T-cell signaling networks and 135. Bendelac, A., Matzinger, P., Seder, R. A., Paul, W. E. & corrections, as well as improvements in presentation. Any errors
development by Grb2 haploid insufficiency. Nature Schwartz, R. H. Activation events during thymic selection. that remain are my own.
Immunol. 2, 29–36 (2001). J. Exp. Med. 175, 731–742 (1992).
130. Mariathasan, S. et al. Duration and strength of 136. Swat, W., Dessing, M., Baron, A., Kisielow, P. & von
extracellular signal-regulated kinase signals are altered Boehmer, H. Phenotypic changes accompanying positive Online links
during positive versus negative thymocyte selection. selection of CD4+CD8+ thymocytes. Eur. J. Immunol. 22,
J. Immunol. 167, 4966–4973 (2001). 2367–2372 (1992). DATABASES
131. Bain, G. et al. Regulation of the helix-loop-helix proteins, 137. Koch, U. et al. Subversion of the T/B lineage decision in The following terms in this article are linked online to:
E2A and Id3, by the Ras–ERK MAPK cascade. Nature the thymus by lunatic-fringe-mediated inhibition of InterPro: http://www.ebi.ac.uk/interpro/
Immunol. 2, 165–171 (2001). Notch-1. Immunity 15, 225–236 (2001). ITAM | SH2 domain
132. Engel, I., Johns, C., Bain, G., Rivera, R. R. & Murre, C. 138. Linette, G. P. et al. Bcl-2 is upregulated at the CD4+CD8+ LocusLink: http://.ncbi.nlm.nih.gov/LocusLink/
Early thymocyte development is regulated by modulation stage during positive selection and promotes thymocyte Bcl2 | CD3/ζ | CD4 | CD8 | CD25 | CD44 | CD69 | Deltex | DLK1
of E2A protein activity. J. Exp. Med. 194, 733–745 (2001). differentiation at several control points. Immunity 1, | E2A | EGR1 | Fyn | HES | HES1 | HY antigen | ID3 | JAG1 |
133. Pages, G. et al. Defective thymocyte maturation in p44 197–205 (1994). JAG2 | JNK | LAT | LCK | Lck | Lunatic fringe | MAPK1 | Mapk |
MAP kinase (Erk-1)-knockout mice. Science 286, Mek1 | MHC class I | MHC class II | NFAT | NF-κB2 | Notch1 |
1374–1377 (1999). Acknowledgements Notch2 | Pkc | presenilin | Rag | RAG1 | RAG2 | SHP1 | Slp76 |
134. Anderson, G., Jenkinson, E. J., Moore, N. C. & Owen, J. J. I wish to thank the colleagues in my own laboratory and around the ZAP70 | Zap70
MHC class-II-positive epithelium and mesenchyme cells world whose experiments and discussions have helped to shape Access to this interactive links box is free online.

322 | MAY 2002 | VOLUME 2 www.nature.com/reviews/immunol


© 2002 Nature Publishing Group

You might also like