<?xml version="1.0" ?>
<tei>
	<teiHeader>
		<fileDesc xml:id="0"/>
	</teiHeader>
	<text xml:lang="en">

		<p>The immune system has evolved to defend the host organism from<lb/> infection by
			exogenous pathogens. Most pathogens replicate much<lb/> faster than mammalian cells. To
			face this challenge, lymphocytes have<lb/> developed unique capacities for division.
			Indeed, although most mature<lb/> T cells are quiescent, they maintain a high capacity
			to self-renew <ref type="biblio">1</ref> ; within<lb/> a few days after antigen
			stimulation, each naive antigen-specific cells can<lb/> produce 10,000 progeny.
			Moreover, this extensive cell division is not<lb/> necessarily associated with terminal
			differentiation. After antigen elimi-<lb/>nation, a fraction of primed cells survive to
			constitute a quiescent mem-<lb/>ory pool that may persist throughout life <ref
				type="biblio">2,3</ref> . In the absence of antigen,<lb/> most memory cells divide
			slowly, with only 0.5% of the cells synthesiz-<lb/>ing DNA. The remaining cells are in
			the first gap phase (G0/G1) of the<lb/> cell cycle <ref type="biblio">4–6</ref> . After
			antigen rechallenge, however, memory cells improve<lb/> their capacity to divide.
			Compared with their naive counterparts, they<lb/> have a reduced G1 phase and they enter
			into cell division rapidly <ref type="biblio">6</ref> .<lb/> Therefore, T cells are
			unique because extensive cell division does not<lb/> necessarily lead to terminal
			differentiation, and more differentiated<lb/> memory cells improve their proliferative
			capacity. These properties con-<lb/>trast with those of other self-renewal systems in
			the body (such as other<lb/> hematopoietic lineages or the epithelia), in which
			progressive cell divi-<lb/>sion eventually leads to terminal differentiation, and cell
			maturation is<lb/> associated with a progressive inability to divide.<lb/></p>

		<p>The ability of memory T cells to divide rapidly during the secondary<lb/> immune response
			indicates they may have acquired different strategies<lb/> to control cell cycle entry
			and/or progression. In mammalian cells,<lb/> progression through the G1 and S phases
			involves several classes of<lb/> cyclin and cyclin-dependent kinases (CDKs). Cell
			activation induces<lb/> the expression of D-type cyclins that positively regulate and
			associate<lb/> with CDK4 or CDK6 <ref type="biblio">(ref. 7)</ref>. These complexes
			migrate to the nucleus<lb/> where they initiate retinoblastoma protein (Rb)
			phosphorylation <ref type="biblio">8</ref> .<lb/> Later in G1, cyclin E–CDK2
			phosphorylates Rb on additional sites.<lb/></p>

		<p>Hyperphosphorylated Rb dissociates from E2F transcription factors<lb/> family members,
			allowing the transcription of several genes<lb/> required for S-phase entry <ref
				type="biblio">8,9</ref> . Cell cycle progression is also con-<lb/>trolled by two
			types of CDK inhibitors. INK4 inhibitors bind<lb/> directly to CDK4 or CDK6. This
			binding inactivates CDKs and pre-<lb/>vents their association with D cyclins <ref
				type="biblio">8</ref> . Because D cyclins are unsta-<lb/>ble when not bound to CDK4
			or CDK6, the formation of<lb/> INK4-CDK4 or INK4-CDK6 complexes indirectly leads to
			degrada-<lb/>tion of D cyclins <ref type="biblio">8</ref> . Cip-Kip inhibitors form
			ternary complexes with<lb/> different types of cyclin-CDK complexes. When associated
			with<lb/> cyclin E–CDK2 complexes, Cip-Kip molecules (either p21 Cip1 or<lb/> p27 Kip1 )
			block CDK2 kinase activity and cell cycle progression <ref type="biblio">8</ref> .
			In<lb/> contrast, cyclin D–CDK complexes may sequester Cip-Kip molecules<lb/> and yet
			retain their kinase activity <ref type="biblio">8,10–14</ref> . Moreover,
			preferential<lb/> binding of Cip-Kip molecules to cyclin D–CDK complexes favors
			cell<lb/> cycle progression by a noncatalytic mechanism <ref type="biblio">8</ref> .
			Indeed, the high<lb/> expression of cyclin D–CDK6 complexes induced after cell
			activation<lb/> sequesters p27 Kip1 . The cyclin E–CDK2 complexes are then
			liberated<lb/> from p27 Kip1 constraint, favoring cell cycle progression.<lb/></p>

		<p>To elucidate the mechanisms underlying the unusual characteris-<lb/>tics of memory CD8 +
			T cells, we compared the expression of different<lb/> proteins and protein complexes
			involved in the S-phase transition in<lb/> naive and memory CD8 + T cells ex vivo. Our
			results show that G0/G1<lb/> memory T cells have high expression of preactivated CDK6 in
			the<lb/> cytoplasm, favoring rapid division after antigen restimulation.<lb/></p>

		<head>RESULTS<lb/></head>

		<head>Cycle status of naive and memory transgenic cells<lb/></head>

		<p>We compared naive and memory CD8 + T cells with the same T cell<lb/> receptor (TCR)
			specificity. We obtained naive cells from female mice<lb/> deficient in recombination
			activation gene 2 (Rag2 –/– ) that express a<lb/> transgenic αβ TCR specific for a
			peptide of the male antigen HY 15 .<lb/> Naive cells were all CD44 – (ref. 15) and had a
			single peak of 2N DNA<lb/> content, indicating that these cells were in G0/G1 <ref
				type="figure">(Fig. 1a)</ref>. Memory<lb/> transgenic cells were obtained as
			described before 6 . Naive transgenic<lb/> cells were immunized with male cells in vivo,
			in the presence of CD4 +<lb/> help. During this primary response, all naive cells divide
			extensively,<lb/> eliminating the male cells. After antigen clearance, all memory
			trans-<lb/>genic cells show improved functional capacities and persist for long<lb/>
			time periods in vivo, even in the absence of antigen 6 . Like memory<lb/> CD8 + T cells
			of other T cell specificities <ref type="biblio">4,5</ref> , over 99% of these
			trans-<lb/>genic memory cells were in G0/G1 <ref type="figure">(Fig. 1a)</ref>. After
			antigen stimula-<lb/>tion, however, memory cells progressed through division much
			faster<lb/> than naive T cells <ref type="figure">(Fig. 1b)</ref>.<lb/></p>

		<p>To investigate why memory cells showed an enhanced capacity to<lb/> divide, we studied
			the expression of D cyclins as well as CDK4 and<lb/> CDK6 in G0/G1-sorted memory <ref
				type="figure">(Fig. 1a)</ref> and naive transgenic<lb/> cells. D-type cyclin family
			members as well as their catalytic part-<lb/>ners, CDK4 and CDK6, have different tissue
			distribution, but are<lb/> thought to have redundant functions <ref type="biblio"
				>8,9</ref> . Cyclin D1 protein expres-<lb/>sion was similar in naive and memory
			cells <ref type="figure">(Fig. 1c)</ref>. Cyclin D2 was<lb/> twofold higher in memory
			cells than in naive cells. In several exper-<lb/>iments, cyclin D3 was undetectable in
			naive cells, but was always<lb/> evident in memory cells. In the experiments in which
			cyclin D3<lb/> could be detected in naive cells, the amount of cyclin D3 in mem-<lb/>ory
			cells was sixfold higher <ref type="figure">(Fig. 1c)</ref>. Thus, naive cells seem to
			have<lb/> very low expression of cyclin D3, whereas this protein was upregu-<lb/>lated
			considerably in memory cells. The different D cyclins also act<lb/> differently after
			naive T cell stimulation. D1 and D2 cyclins are<lb/> upregulated shortly after T cell
			stimulation, whereas D3 upregula-<lb/>tion only occurs much later <ref type="figure"
				>(Supplementary Fig.1 online)</ref>. T lym-<lb/>phocytes express mostly CDK6 rather
			than CDK4 <ref type="biblio">(ref. 16)</ref>, but<lb/> both these kinases were twice as
			abundant in memory transgenic<lb/> cells than in naive cells <ref type="figure">(Fig.
				1c)</ref>. High cyclin D and CDK6 expres-<lb/>sion is usually associated with Rb
			phosphorylation <ref type="biblio">17,18</ref> . Like naive<lb/> cells, however, memory
			cells expressed only hypophosphorylated<lb/> Rb forms <ref type="figure">(Fig.
			1d)</ref>. We were unable to detect Rb mobility shifts. Rb<lb/> Ser780 residues, known
			to be specifically phosphorylated by cyclin<lb/> D–CDK4 or cyclin D–CDK6 complexes 19 ,
			were not phosphory-<lb/>lated. Furthermore, Rb Ser795 residues, which can be
			phosphory-<lb/>lated by kinases other than CDK4 and CDK6, were also not<lb/>
			phosphorylated <ref type="figure">(Fig. 1d)</ref>. In contrast, Rb phosphorylation
			was<lb/> induced efficiently after in vitro activation <ref type="figure">(Fig.
				1e)</ref>. Thus, memory<lb/> transgenic cells were at a stage of G1 arrest, in which
			high expres-<lb/>sion of D cyclins and CDK6 coexist with hypophosphorylated Rb.<lb/>
			This cell cycle state has never been described to our knowledge in<lb/> resting
			cells.<lb/></p>

		<head>Expression of INK4 inhibitors<lb/></head>

		<p>To investigate whether memory transgenic cells remain in G0/G1<lb/> arrest because of
			preferential expression of CDK inhibitors, we com-<lb/>pared the expression of all the
			different INK4 inhibitor family mem-<lb/>bers in naive and memory T cells. These T cells
			expressed both<lb/> p19 INK4d and p18 INK4c , but did not express other INK4 family
			mem-<lb/>bers. There was similar expression of p19 INK4d in both populations,<lb/>
			whereas p18 INK4c was increased in memory transgenic cells compared<lb/> with naive
			cells <ref type="figure">(Fig. 2, left)</ref>.<lb/></p>

		<p>INK4 family members inhibit cell cycle progression by forming sta-<lb/>ble complexes with
			CDK6, which prevents cyclin D–CDK6 associa-<lb/>tion. Unbound cyclin D is rapidly
			degraded <ref type="biblio">8</ref> . Because INK4<lb/> inhibition is associated with
			low expression of D cyclins, it should not<lb/> be responsible for memory cell cycle
			arrest, as memory cells have high<lb/> expression of D cyclins <ref type="figure">(Fig.
				1c)</ref>. However, memory transgenic cells<lb/> may form INK4-CDK6 complexes with a
			different configuration that<lb/> associate with and stabilize cyclin D. To exclude this
			possibility, we<lb/></p>

		<figure>a<lb/> c<lb/> d<lb/> e<lb/> b<lb/> Figure 1 Cell cycle characteristics of naive and
			memory transgenic cells.(a) DNA content. Profiles show area (FL4-A) versus width (FL4-W)
			of Hoechst 33352<lb/> labeling, and the gate used to sort G0/G1 memory transgenic cells.
			Boxed cells are in G0/G1. (b) CFSE-labeled naive and memory transgenic cells were<lb/>
			transferred to female mice immunized with male cells and analyzed 40 h later. Bracketed
			numbers indicate percentages of nondividing cells. (c,d) Lysates<lb/> of G0/G1 naive (N)
			and memory (M) transgenic cells analyzed by immunoblot. Right margins (R), protein
			signal ratios between memory and naive cells.<lb/> Similar quantitative differences
			(ratios between memory and naive cells) were obtained in two other experiments: cyclin
			D2: 2.1 and 2.5; cyclin D3: 6 and<lb/> 5.3; CDK4: 1.8 and 2.1; CDK6: 2 and 2.2. (d) Rb
			content and Rb phosphorylation. (e) Transgenic cells were stimulated with specific
			peptide. Top, increased<lb/> molecular weight of hyperphosphorylated forms of Rb at
			different times after activation. Bottom, phosphorylation of specific Rb Ser780 and
			Ser795 residues<lb/> 24 h after stimulation. Data are from one of at least three
			independent experiments. pRb, phosphorylated Rb; S780, Ser780; S795,
			Ser795.<lb/></figure>

		<p>studied the type of complexes formed by p18 INK4c . We found that<lb/> p18 INK4c –CDK6
			complexes did not contain cyclin D <ref type="figure">(Fig. 2, right)</ref>.<lb/>
			Therefore, p18 INK4c upregulation cannot by itself be responsible for<lb/> memory cell
			G0/G1 cycle arrest. However, increased p18 INK4c expres-<lb/>sion may have other
			functions. By reducing the pool of available<lb/> CDK6, it may establish a threshold of
			activation. High expression of<lb/> p18 INK4c may also be required for CD8
			differentiation into effector<lb/> functions, as seen in plasma cells <ref type="biblio"
				>20,21</ref> .<lb/></p>

		<head>Expression of Cip-Kip inhibitors<lb/></head>

		<p>Next we studied Cip-Kip expression and complex formation.<lb/> Memory cell G0/G1 arrest
			could be due to Cip-Kip inhibition, as<lb/> Cip-Kip molecules associate with but do not
			disrupt cyclin-CDK<lb/> complexes <ref type="biblio">8</ref> . Therefore, Cip-Kip
			inhibition can occur even when<lb/> expression of cyclin D and CDK6 is high, as in
			memory transgenic<lb/> cells. As described before, p57 Kip2 <ref type="biblio">(ref.
				22)</ref> and p21 Cip1 <ref type="biblio">(ref. 23,24)</ref><lb/> were not detected
			in G0/G1 T cells, whereas p27 Kip1 was present<lb/>
			<ref type="figure">(Fig. 3a)</ref>. If p27 Kip1 had a major function in the arrest of
			memory<lb/> cells in G0/G1, it should have high expression in these cells. G0/G1<lb/>
			memory cells, however, expressed only half the p27 Kip1 expressed by<lb/> naive cells
				<ref type="figure">(Fig. 3a)</ref>. These results indicated that the inhibition
			of<lb/> cell cycle progression in memory T cells could involve mechanisms<lb/> other
			than p27 Kip1 expression.<lb/></p>

		<p>The function of p27 Kip1 also depends on the complexes with<lb/> which it is associated
				<ref type="biblio">8</ref> . When associated with cyclin E–CDK2,<lb/> p27 Kip1
			strongly inhibits CDK2 kinase activity, and has a major<lb/> function in G0/G1 arrest.
			In contrast, association of p27 Kip1 with<lb/> cyclin D–CDK promotes stability <ref
				type="biblio">11–13</ref> , is required for the migration<lb/> of these complexes
			into the nucleus <ref type="biblio">8</ref> and does not necessarily inhibit<lb/> CDK4
			and CDK6 kinase activity <ref type="biblio">8,10–14</ref> . We therefore analyzed
			the<lb/> cyclin-CDK-p27 Kip1 complex composition of naive and memory<lb/> cells <ref
				type="figure">(Fig. 3b)</ref>. T lymphocytes express mainly CDK6 rather than<lb/>
			CDK4 <ref type="biblio">(ref. 6)</ref>. Consistent with those results, we failed to
			detect<lb/> CDK4 in association with D cyclins in naive or memory<lb/></p>

		<figure>Figure 2 Expression of INK4 inhibitors in naive and memory transgenic<lb/> cells.
			Lysates of G0/G1 naive (N) and memory (M) transgenic cells<lb/> were analyzed by
			immunoblot (left) or were immunoprecipitated with<lb/> anti-p18 INK4c and analyzed by
			immunoblot (right). Left margins,<lb/> immunoblot antibodies. Right margins (R), protein
			expression ratios<lb/> between memory and naive cells. Similar quantitative differences
			in<lb/> p18 INK4c expression were obtained in the other two experiments (5 and<lb/>
			4.8). Cyclin D2 was not associated with p18 INK4c –CDK6 complexes in<lb/> these
			experiments. Data are from one of three independent experiments.<lb/></figure>

		<figure>Figure 3 Cip-Kip molecules, cyclin D<lb/> and cyclin E complexes in naive (N)
			and<lb/> sorted G0/G1 memory transgenic cells (M).<lb/> (a) Amounts of p27 Kip1 and p21
			Cip1 in cell<lb/> lysates. Right margin (R), protein expression<lb/> ratios between
			memory and naive cells. Similar<lb/> quantitative differences of p27 kip1
			expression<lb/> were obtained in the other two experiments<lb/> (0.5 and 0.48). (b,c)
			Amount, composition<lb/> (b) and kinase activity (c) of cyclin D3 and<lb/> cyclin E
			complexes. (b) Lysates were<lb/> immunoprecipitated with antibodies at top<lb/> (IP).
			Left margins, immunoblot antibodies.<lb/> Right margins (R), protein expression
			ratios<lb/> between memory and naive cells. (c) Lysates<lb/> (100 µg) were precipitated
			(IP) with anti-cyclin<lb/> D3 and tested for their capacity to<lb/> phosphorylate Rb-GST
			in vitro. Below blots,<lb/> kinase activity ratios. The kinase activity of<lb/> cyclin
			D3 complexes should rely on associated<lb/> CDK6, because these complexes do not
			contain<lb/> CDK4. Ac, transgenic cells recovered 24 h<lb/> after activation with
			HY-specific peptide.<lb/> (d) Nonsaturating conditions of the kinase<lb/> assay.
			Different amounts of cyclin D3<lb/> complexes were tested for their capacity to<lb/>
			phosphorylate Rb-GST. Linear regression curve<lb/> of the kinase activity of naive T
			cells and<lb/> the kinase activity of cyclin D3 complexes<lb/> precipitated from memory
			(closed square) and activated cells lysates (open circle) are shown. (e) Memory
			transgenic (Tg) cells were plated on<lb/> poly-L-lysine-coated coverslips; cyclin D3
			expression was measured by means of Alexa Fluor 488 (green) and nuclei were stained with
			propidium<lb/> iodide (red), and cells were examined with a confocal laser microscope.
			Images are of overlapping sections; the cyclin D3 location was confirmed<lb/> in
			individual cell sections. Data are from one of three independent experiments.<lb/>
			a<lb/> b<lb/> c<lb/> e<lb/> d<lb/></figure>

		<p>T cells (data not shown) . In naive cells, the amount of cyclin<lb/> D2–CDK6 complexes
			(data not shown) and cyclin D3–CDK6 com-<lb/>plexes was very low, these complexes
			contained low amounts of<lb/> p27 Kip1 and p27 Kip1 was preferentially associated with
			cyclin<lb/> E–CDK2 complexes <ref type="figure">(Fig. 3b)</ref>. The kinase activity of
			both CDK6 and<lb/> CDK2 was also very low <ref type="figure">(Fig. 3c)</ref>. These
			results demonstrate that<lb/> naive cells are in a classical state of early G0/G1 arrest
			in which<lb/> cyclin D–CDK6 complex expression is very low, p27 Kip1 expression<lb/> is
			high and p27 Kip1 remains bound to cyclin E–CDK2 complexes,<lb/> thereby blocking the
			kinase activity of these complexes. In contrast,<lb/> cyclin D–CDK6 complexes were
			abundant in memory T cells: cyclin<lb/> D2–CDK6 complexes were increased twofold (data
			not shown) and<lb/> cyclin D3–CDK6 complexes were increased sixfold. The complex<lb/>
			composition also differed: CDK6 and p27 Kip1 expression was<lb/> increased 11-fold in
			cyclin D3 precipitates of memory T cells, com-<lb/>pared with that of cyclin D3
			precipitates of naive cells <ref type="figure">(Fig. 3b)</ref>.<lb/> Therefore, after
			adjustment for increased cyclin D3, the cyclin<lb/> D3–CDK6 complexes from memory
			transgenic cells contained, on<lb/> average, twice the amount of CDK6 and p27 Kip1
			contained by the<lb/> cyclin D3–CDK6 complexes from naive transgenic cells. Cyclin
			D<lb/> complexes containing high ratios of CDK6 and p27 Kip1 were<lb/> thought to be
			present only in dividing cells <ref type="biblio">8</ref> , but we also found them<lb/>
			in memory transgenic cells in the G0/G1 state.<lb/></p>

		<p>The induction and maintenance of stable CDK6 kinase activity<lb/> requires CDK6
			association with both D cyclins and p27 Kip1 <ref type="biblio">(refs.<lb/>
				8,10–14)</ref>. Because memory transgenic cells had high expression of<lb/> cyclin
			D–CDK6–p27 Kip1 complexes, the CDK6 in these complexes<lb/> should have been able to
			phosphorylate Rb. Consistently, cyclin<lb/> D3–CDK6–p27 Kip1 complexes from memory cells
			phosphorylated Rb<lb/> in vitro as efficiently as complexes of cycling T cells <ref
				type="figure">(Fig. 3c,d)</ref>.<lb/> Induction of CDK6 kinase activity is normally
			followed by CDK2<lb/> activation, Rb phosphorylation and S-phase entry <ref
				type="biblio">8</ref> . The cyclin<lb/> E–CDK2 complexes from memory T cells were
			different from those of<lb/> naive cells <ref type="figure">(Fig. 3b,c)</ref>. They
			contained half the amount of associated<lb/> p27 Kip1 , and their CDK2 Rb kinase
			activity was higher. However, the<lb/> CDK2 kinase activity of memory T cells was much
			lower than the<lb/> CDK2 kinase activity of cycling cells <ref type="figure">(Fig.
				3b,c)</ref>. Therefore, memory<lb/> transgenic cells developed a unique state of
			cell cycle arrest in which<lb/> large amounts of cyclin D–CDK6 complexes with high
			kinase activity<lb/> are dissociated from CDK2 activation and Rb
			phosphorylation.<lb/></p>

		<p>Active cyclin D–CDK6 is thought to translocate to the nucleus,<lb/> where it can access
			and phosphorylate nuclear Rb. To determine<lb/> why the active CDK6 from G0/G1 memory
			cells would not phos-<lb/>phorylate Rb, we analyzed the location of cyclin D3–CDK6
			com-<lb/>plexes. We found that these active complexes remained in the<lb/> cytoplasm
				<ref type="figure">(Fig. 3e)</ref>. The cytoplasmic location of these active
			cyclin<lb/> D3–CDK6 complexes prevents access to nuclear Rb and Rb phos-<lb/>phorylation
			in intact memory T cells.<lb/></p>

		<head>Naive and memory T cells from normal mice<lb/></head>

		<p>We analyzed other T cells with different TCR specificities to deter-<lb/>mine if they
			shared the properties of these transgenic cells. P14<lb/> (specific for a Gp-33
			lymphocytic choriomeningitis virus peptide)<lb/> and OT-1 (anti-ovalbumin) monoclonal
			CD8 + naive populations<lb/> were reported to have a much higher affinity for self major
			histo-<lb/>compatibility complex (MHC) than CD8 HY-specific cells. All<lb/> these
			monoclonal naive transgenic cells have undetectable to low<lb/> expression of cyclin D,
			similar to that of HY-specific transgenic<lb/> naive cells <ref type="figure">(Fig.
				4a)</ref>. Thus high-affinity interactions with self MHC<lb/> are not sufficient to
			induce cyclin D3 upregulation. Next, we<lb/> sorted CD8 + CD44 – naive and CD8 + CD44 hi
			memory T cells from<lb/> normal mice, and compared these cell populations to naive
			and<lb/> memory transgenic cells, respectively <ref type="figure">(Fig. 4)</ref>. CD8 +
			CD44 – naive<lb/> cells resembled naive transgenic cells, whereas CD44 hi memory<lb/> T
			cells from normal mice resembled memory transgenic cells: they<lb/> upregulated cyclin
			D3 and contained preactivated CDK6, associ-<lb/>ated into cyclin D3–CDK6–p27 Kip1
			complexes <ref type="figure">(Fig. 4b–d)</ref>. The<lb/> kinase activity of CDK6 in the
			cyclin D3–CDK6 complexes of<lb/> CD8 + CD44 hi memory cells was similar to that of
			transgenic mem-<lb/>ory cells <ref type="figure">(Fig. 4d)</ref>. The active cyclin
			D3–CDK6 complexes in<lb/> CD8 + CD44 hi memory cells were exclusively located in the
			cyto-<lb/>plasm and were thus unable to phosphorylate nuclear Rb in the<lb/> intact cell
				<ref type="figure">(Fig. 4e)</ref>.<lb/></p>

		<head>Preactive CDK6 in cell cycle progression<lb/></head>

		<p>Although the cytoplasmic location of active cyclin D3–CDK6 com-<lb/>plexes prevents Rb
			phosphorylation in memory T cells, the existence<lb/> of such preactivated complexes
			could favor their rapid cell cycle pro-<lb/>gression after antigen stimulation. Cyclin
			D3–deficient mice have a<lb/> block in thymocyte differentiation (P. Sicinski, personal
			communica-<lb/>tion). Therefore, to investigate the function of cyclin D3–CDK6
			com-<lb/>plexes in the cell cycle of memory T cells, we studied the effect of<lb/>
			blocking cyclin D3 activity by intracellular delivery of a monoclonal<lb/> antibody to
			cyclin D3 (anti-cyclin D3).<lb/></p>

		<figure>a<lb/> b<lb/> c<lb/> e<lb/> d<lb/> Figure 4 Comparison of cell cycle parameters in
			naive and memory CD8<lb/> T cells from normal mice. Naive CD8 + CD44 – and memory CD8 +
			CD44 hi<lb/> polyclonal populations from normal mice were sorted. (a) Cyclin D3<lb/>
			expression on different monoclonal naive T cell populations (HY, P14<lb/> and OT1) and
			in memory transgenic cells (M). (b) Cyclin D3 expression in<lb/> polyclonal naive and
			memory lysates, compared with transgenic naive (N)<lb/> and memory (M) populations.
			Protein ratios are shown below lanes<lb/> (c,d) Lysates were precipitated (IP) with
			anti-cyclin D3 and analyzed for D3,<lb/> CDK6 and p27 kip1 content (c) and their
			capacity to phosphorylate Rb-GST<lb/> in vitro compared with memory transgenic cells
			(d), studied simultaneously.<lb/> Right margins (R), protein expression ratios between
			memory and naive<lb/> cells. (e) Cellular location of cyclin D3, in CD8 CD44 hi
			polyclonal memory<lb/> T cells, evaluated as described in Figure 3e.<lb/></figure>

		<p>As controls for these experiments, we determined if the anti-<lb/>cyclin D3 we used was
			able to precipitate all cyclin D3 from cell<lb/> lysates and was strictly specific for
			cyclin D3. We studied lysates<lb/> from a CD8 T cell line containing large amounts of
			all D cyclin fam-<lb/>ily members. Anti-cyclin D3 precipitated a single band and
			elimi-<lb/>nated all cyclin D3 from these lysates; no further cyclin D3 could be<lb/>
			identified in the supernatants <ref type="figure">(Fig. 5a)</ref>. Moreover, anti-cyclin
			D3<lb/> did not recognize D2 or D1 cyclins, as the amounts of these proteins<lb/> in the
			cell lysates remained unchanged after anti-cyclin D3 immuno-<lb/>precipitation <ref
				type="figure">(Fig. 5a).</ref> To control for any nonspecific effects of<lb/>
			intracellular antibody delivery, we treated T cells with anti-cyclin D3<lb/> or with the
			same concentration of an isotype antibody control.<lb/></p>

		<p>We determined the effects of these antibodies on memory CD8 + T cells<lb/> cycle
			progression after anti-CD3 stimulation <ref type="figure">(Supplementary Fig. 1<lb/>
				online)</ref>. Initially, we used a T cell line that contains active cyclin<lb/>
			D3–CDK6 complexes with a high capacity to phosphorylate Rb in vitro<lb/> (data not
			shown). When this line is cycling, these cyclin D3–CDK6 com-<lb/>plexes are detected
			both in the cytoplasm, where these proteins are syn-<lb/>thesized and assembled, and in
			the nucleus <ref type="figure">(Fig. 5b, left)</ref>. When this line<lb/> was
			synchronized in G0/G1, the cyclin D3 complexes remained in the<lb/> cytoplasm, as in
			memory transgenic cells <ref type="figure">(Fig. 5b, right)</ref>. Intracellular<lb/>
			delivery of anti-cyclin D3 to G0/G1 cells strongly reduced cell cycle<lb/> entry, as
			evaluated by early division, with loss of carboxyfluorescein<lb/> diacetate succinimidyl
			diester (CFSE) and 5-bromodeoxyuridine<lb/> (BrdU) incorporation after anti-CD3
			stimulation <ref type="figure">(Fig. 5c)</ref>. The effects<lb/> of anti-cyclin D3 in
			polyclonal CD8 + CD44 hi memory T cells from nor-<lb/>mal mice were similar. Anti-cyclin
			D3 also blocked the rapid cell cycle<lb/> entry and cell division after anti-CD3
			stimulation of these cells <ref type="figure">(Fig. 5d)</ref>.<lb/></p>

		<head>DISCUSSION<lb/></head>

		<p>Here we have shown that resting naive and memory CD8 T cells stud-<lb/>ied ex vivo are in
			very different states of G0/G1 arrest. Naive G0/G1<lb/> T cells have low amounts of D
			cyclins, high amounts of p27 Kip1 and<lb/> low CDK6 and CDK2 kinase activity. Memory
			G0/G1 T cells have high<lb/> amounts of cyclin D, low amounts of p27 Kip1 and very high
			CDK6<lb/> kinase activity. These results show that T lymphocytes in physiological<lb/>
			conditions use multiple strategies for G0/G1 arrest.<lb/></p>

		<p>Many mechanisms can induce cell cycle arrest in response to<lb/> stress. In physiological
			conditions, however, cell cycle arrest after<lb/> division was thought to be induced by
			p27 Kip1 upregulation and<lb/> redistribution. Once cell division is over, p27 Kip1
			expression is<lb/> upregulated, and p27 Kip1 dissociates from cyclin D–CDK6
			com-<lb/>plexes, associating with cyclin E–CDK2. Consequently, cyclin<lb/> D–CDK6
			complexes become unstable and are degraded, the kinase<lb/> activity of cyclin E–CDK2
			complexes is blocked, extinguishing Rb<lb/> phosphorylation and inducing G0/G1 arrest
				<ref type="biblio">8</ref> . Our results show<lb/> that naive T cells exist in this
			classical state of early G1 arrest. They<lb/> have little expression of cyclin D–CDK
			complexes and high expres-<lb/>sion of p27 Kip1 preferentially bound to cyclin E–CDK2
			complexes.<lb/> The kinase activity of both CDK6 and CDK2 is very low. In
			con-<lb/>trast, memory T cells are in a completely different stage of G1<lb/> arrest.
			They resemble cycling cells, as they have little expression of<lb/> p27 Kip1 and high
			expression of cyclin D3–CDK6–p27 Kip1 com-<lb/>plexes. Furthermore, when in the presence
			of a Rb substrate in<lb/> vitro, the CDK6 present in these complexes has Rb kinase
			activity<lb/> similar to that of cycling cells. Yet these active complexes cannot<lb/>
			phosphorylate Rb in an intact memory T cell. They are localized in<lb/></p>

		<figure>Figure 5 Effect of anti-cyclin D3 on the cell cycle<lb/> entry and progression of
			polyclonal memory<lb/> T cells. (a) Specificity of anti-cyclin D3. Cell<lb/> lysates
			from a CD8 T cell line were precipitated<lb/> with anti-cyclin D3. Left, cyclin D3
			content of<lb/> precipitate (IP) and supernatant (SN). Right, D1<lb/> and D2 content in
			lysates with (D3) or without (C)<lb/> cyclin D3 precipitation. Far right margin
			(R),<lb/> protein expression ratios between D3 and C cells.<lb/> (b) Location of cyclin
			D3 in a T cell line in cycling<lb/> (left) and G0/G1-synchronized (right) cells<lb/>
			evaluated as described in Figure 3e. (c) Effect<lb/> of anti-cyclin D3 on cell cycle
			entry and cell<lb/> division. Anti-cyclin D3 (bold line) or the same<lb/> amount of an
			isotype control antibody (thin lines)<lb/> were delivered intracellularly into this<lb/>
			synchronized T cell line with liposomes. Cells<lb/> were activated with anti-CD3, and
			CFSE loss<lb/> and BrdU incorporation was studied 24 h later.<lb/> Bracketed numbers
			indicate percentages of<lb/> BrdU-positive cells. (d) CD8 + CD44 hi polyclonal<lb/>
			memory T cells received anti-cyclin D3 (bold line)<lb/> or an isotype control (thin
			line) and were<lb/> stimulated with anti-CD3. Results show CFSE<lb/> loss and BrDU
			incorporation in one of two<lb/> experiments 1.5 d after stimulation. Bracketed<lb/>
			numbers indicate percentages of BrdU-positive<lb/> cells. In c and d, CFSE loss was
			studied shortly<lb/> after activation and thus only shows the first<lb/> division. For
			the effects of anti-cyclin D3 on<lb/> naive cell division see Supplementary Fig. 1
			online.<lb/> a<lb/> b<lb/> c<lb/> d<lb/></figure>

		<p>the cytoplasm and thus have no access to nuclear Rb. In this context,<lb/> G0/G1 memory
			cells retain other features that are characteristic of<lb/> resting cells. The cyclin
			E–CDK2 complexes have low kinase activity,<lb/> and Rb is not phosphorylated. It was
			believed that cell cycle arrest is<lb/> always associated with high p27 Kip1 , whereas
			cell cycle progression is<lb/> associated with p27 Kip1 degradation. Our results show
			that this is<lb/> not always the case. Resting memory cells maintain small amounts<lb/>
			of p27 Kip1 that are stable for long time periods.<lb/></p>

		<p>Several strategies may be involved in ensuring the cell cycle arrest of<lb/> memory T
			cells. The cytoplasmic location of active cyclin D3–CDK6-<lb/>p27 Kip1 complexes should
			contribute to the maintenance of memory<lb/> T cells in a G1 resting state, because
			these complexes are prevented from<lb/> reaching and phosphorylating nuclear Rb. Such
			retention could be<lb/> mediated by cytoplasmic signals that predominate over nuclear
			localiza-<lb/>tion signals, or by a cytoplasmic retention protein. Similar
			strategies<lb/> retain other proteins in the cytoplasm, like cyclin B <ref type="biblio"
				>25</ref> or Jun kinase <ref type="biblio">26</ref> .<lb/> However, the cytoplasmic
			location of cyclin D3–CDK6–p27 Kip1 com-<lb/>plexes does not explain all the features of
			resting G1 memory cells.<lb/> Indeed, cyclin E–CDK2 complexes are inactive, despite
			their low p27 Kip1<lb/> content. It is possible that the low amount of p27 Kip1 is
			sufficient to inac-<lb/>tivate the cyclin E–CDK2 complexes of memory cells.
			Alternatively,<lb/> p27 Kip1 degradation may not be sufficient to induce CDK2 kinase
			activ-<lb/>ity. The activation of CDK2 may require additional signals, including<lb/>
			mitogen-TCR-dependent activation. This latter hypothesis is supported<lb/> by the
			presence of resting cells in p27 Kip1 -deficient mice <ref type="biblio">27–29</ref> .
			Also, most<lb/> studies of cell cycle arrest mechanisms were done with cell lines
			that<lb/> were induced to stop division by growth factors or serum starvation.<lb/> In
			contrast, our study was done with ex vivo isolated memory T cells<lb/> that had stopped
			dividing in physiological conditions, after antigen<lb/> elimination. It is possible
			that other types of ex vivo isolated G0/G1<lb/> cells use multiple strategies to stop
			cell cycle progression that have not<lb/> been detected with cell lines in
			vitro.<lb/></p>

		<p>If the cytoplasmic location of active CDK6 is not solely responsible<lb/> for the arrest
			of memory cells in G0/G1, could active CDK6 be<lb/> involved in other memory T cell
			properties? One of the main features<lb/> of CD8 memory cells is their efficient
			division on rechallenge. We<lb/> propose that the presence of large amounts of active
			CDK6 in mem-<lb/>ory cells could guarantee their very rapid cell cycle progression
			after<lb/> reactivation. Indeed, D cyclins and CDK6 are already highly<lb/> expressed
			and form stable complexes with p27 Kip1 , and the CDK6 in<lb/> these complexes already
			has Rb kinase activity. These complexes<lb/> sequester p27 Kip1 , which partially
			relieves CDK2 from p27 Kip1 con-<lb/>straint, favoring cell cycle progression. The
			pre-existence of these<lb/> active complexes may also explain the lower activation
			threshold<lb/> characteristic of memory T cells. The function of pre-existent
			active<lb/> complexes in rapid cell cycle progression is supported by the major<lb/>
			effect of anti-cyclin D3 in cell cycle entry and progression.<lb/></p>

		<p>In summary, we have shown here that resting memory CD8 T cells,<lb/> either transgenic or
			CD8 + CD44 hi cells from normal mice, are in a more<lb/> advanced G1 phase than their
			naive counterparts. We propose a previ-<lb/>ously unrecognized stage of cell cycle
			arrest that maintains memory<lb/> T cell in a preactivated state and may facilitate cell
			cycle progression.<lb/> Most pathogens have a very high rate of self-replication, and
			therefore<lb/> the cell cycle properties of memory cells described herein may ensure
			a<lb/> fast progression in the cell cycle, allowing the rapid control of
			reinfec-<lb/>tion, a property characteristic of secondary immune responses.<lb/></p>

		<head>METHODS<lb/></head>

		<p>Mice. C57BL/6 mice were either Ly 5.1, Rag2 –/– , Cd3e –/– or Rag2 –/– transgenic<lb/>
			for the HY-specific TCRαβ 6 . P14 and OT1 Rag2 –/– mice were a gift from A.<lb/> Freitas
			(Institut Pasteur, Paris, France). All mice were bred at the Center for<lb/> Development
			of Advanced Experimentation Techniques (Orleans, France).<lb/> Live animal experiments
			were approved by Ministère de l&apos;Agriculture de la<lb/> Pèche et de
			l&apos;Alimentation (France).<lb/></p>

		<p>T cells. Naive transgenic cells were recovered from Rag2 –/– transgenic female<lb/> mice.
			Memory transgenic cells were obtained as described <ref type="biblio">6</ref> . Naive
			anti-HY<lb/> transgenic lymph node cells (0.5 × 10 6 ) were injected intravenously
			into<lb/> Rag2 –/– female mice together with 1 × 10 6 CD4 T cells, and mice were
			immu-<lb/>nized with 0.5 × 10 6 Cd3e –/– bone marrow cells from male mice. Primed
			cells<lb/> were recovered from a pool of lymph nodes and spleens 2 months after<lb/>
			immunization. These cells are all antigen experienced and functionally com-<lb/>petent,
			and persist in vivo in the absence of antigen <ref type="biblio">15</ref> . In this
			system, the<lb/> antigen is eliminated 2 weeks after immunization, but memory T cells
			main-<lb/>tain the same characteristics 2 months through 6 months after priming. To<lb/>
			obtain pure monoclonal populations, cell suspensions from monoclonal<lb/> TCR transgenic
			mice were depleted of macrophages, granulocytes, erythro-<lb/>cytes and B cells with a
			&apos;cocktail&apos; of monoclonal antibodies (Mac1, 8C5,<lb/> Ter119 and 6B2) and
			magnetic sorting with coated Dynabeads (Dynal), fol-<lb/>lowed by positive selection
			with a biotin CD8α monoclonal antibody (53-<lb/>6.7) and the Macs sorting system
			(Miltenyi Biotec). Memory anti-HY<lb/> transgenic cells were also stained with the vital
			dye Hoechst 33352<lb/> (Molecular Probes), and G0/G1 memory cells were sorted with
			a<lb/> FACSVantage flow cytometer (Becton Dickinson). CD8 + CD44 – and<lb/> CD8 + CD44
			hi cells from normal mice were also isolated by cell sorting. For<lb/> activation of
			transgenic cells, 1 × 10 6 transgenic T cells were incubated with<lb/> 2 × 10 6 Cd3e –/–
			spleen cells from female mice and 0.1 µM specific peptide <ref type="biblio">6</ref>
			.<lb/> The CD8 T cell line was a gift from J. di Santo (Institut Pasteur, Paris,<lb/>
			France). This T cell line was synchronized in vitro by serum starvation and<lb/> was
			activated in vitro with 2.5 µg/ml of anti-CD3. BrdU incorporation and<lb/> CFSE staining
			were evaluated and quantified as described <ref type="biblio">6,15</ref> .<lb/></p>

		<p>Immunoblotting, immunoprecipitation and in vitro kinase reactions. Cells<lb/> were lysed
			in Nonidet-P40 buffer containing protease inhibitors. Lysates were<lb/> prepared, and
			equal amounts of protein (20 µg/sample) were separated by<lb/> SDS-PAGE. Resolved gels
			were transferred to nitrocellulose membranes and<lb/> immunoblotted with monoclonal
			antibodies or antisera. Cyclin D2, cyclin E,<lb/> CDK4, CDK6, p15 INK4b , p18 INK4c ,
			p19 INK4d , pRb Ser 780 and pRb Ser 795 antis-<lb/>era were from Santa Cruz
			Biotechnology; monoclonal antibodies to cyclin D1,<lb/> cyclin D3, CDK2, p16 INK4a , p21
			Cip1 and p27 Kip1 were from Santa Cruz<lb/> Biotechnology; monoclonal antibody to actin
			was from Sigma; and mono-<lb/>clonal antibody to Rb was from Pharmingen. Immunodetection
			was done by<lb/> incubation with a horseradish peroxidase–conjugated antibody to
			mouse<lb/> immunoglobulin G (IgG; Amersham), rabbit IgG (Southern Biotechnology)<lb/>
			and goat IgG (Zymed), followed by ECL (Amersham). Immunoprecipitation<lb/> was done as
			described <ref type="biblio">17</ref> , with 50 µg of protein per sample. In vitro
			kinase reac-<lb/>tions were done as described <ref type="biblio">17</ref> , with 100 µg
			protein per sample. Rb–glutathione<lb/> S-transferase (GST) phosphorylation was measured
			with the Bio-Rad Quantity<lb/> One software system.<lb/></p>

		<p>Fluorescence confocal microscopy. Cells were plated on coverslips coated<lb/> with
			poly-L-lysine (Sigma), fixed in 2% paraformaldehyde and permeabilized<lb/> with 0.15%
			Triton X-100 (Sigma). Slides were incubated with anti-cyclin D3.<lb/> The secondary
			antibody was Alexa Fluor 488 goat anti-mouse IgG (Molecular<lb/> Probes). Controls
			included a mouse isotype control IgG as the primary anti-<lb/>body. Slides were mounted
			in mounting medium for fluorescence with pro-<lb/>pidium iodide from Vector Laboratories
			and were examined with a confocal<lb/> laser microscope (LSM 510; Carl Zeiss).<lb/></p>

		<p>Intracellular delivery of monoclonal antibodies. Liposomes (Imgenex) were<lb/> hydrated
			in serum-free medium containing 150 µg/ml of monoclonal IgG1<lb/> anti-cyclin D3 (mouse
			IgG1;Santa Cruz Biotechnology) or IgG1 isotype con-<lb/>trol (Pharmingen). T cells were
			then seeded at a density of 2 × 10 5 cells/well in<lb/> a 24-well plate with serum-free
			medium. Liposome-antibody mixtures were<lb/> added to the cells and incubated for 2.5 h
			at 37 °C in a 5% CO 2 incubator. T<lb/> cells were then washed in PBS, transferred into
			medium containing 5% FCS<lb/> and stimulated with specific peptide <ref type="biblio"
				>6</ref> or anti-CD3.</p>

	</text>
</tei>