Costimulation of naive human CD4⁺ T cells through intercellular adhesion molecule-1 promotes differentiation to a memory phenotype that is not strictly the result of multiple rounds of cell division

Abstract

The process by which naive T cells become activated, differentiate into effector cells and ultimately generate long-lived memory cells is dependent upon a number of factors, including the costimulatory signals received by the T cell. To best understand the multiple events involved, it is important to understand the potential contributions by individual signalling proteins using both in vitro and in vivo studies. Here, the potential for costimulation through intercellular adhesion molecule-1 (ICAM-1; CD54), resident on the surface of naive human T cells, to influence differentiation was investigated. Costimulation of naive T cells through ICAM-1 resulted in expansive cell division, high interleukin-2 production, and protection from apoptosis. Prolonged culture led to outgrowth of a subpopulation of cells with a highly differentiated CD45RA^– CD11a^hi CD27^– phenotype. In this respect, costimulation through ICAM-1 was similar to costimulation through CD28 and different from costimulation through leucocyte function-associated antigen-1. The CD45RA^– CD11a^hi CD27^– cells responded to suboptimal stimulation through the T-cell receptor alone with a more robust proliferative response compared with naive cells from the same subject. These cells also secreted higher levels of T helper type 1 cytokines in response to lower levels of stimulation than their naive counterparts. The surface phenotype and more sensitive response characteristics suggest the creation of a memory T-cell subpopulation as a result of costimulation through ICAM-1. Finally, generation of this memory population was the result of specific costimulatory signals, and not merely because of a high number of cell divisions. These data reveal a new role for resident ICAM-1 to influence the differentiation of naive T cells.

Introduction

Following antigen stimulation in vivo, naive T cells undergo clonal expansion and differentiate into a population of effector cells, which eventually contract in number leaving a small population of long-lived memory cells. The consequence is a significantly greater frequency of resident cells specific for a particular antigen.^1–4 Memory T cells are qualitatively superior to naive cells should that particular antigen be encountered again. Memory T cells respond to lower concentrations of antigen, are less dependent on costimulatory signals, and display effector functions more rapidly following activation than do naive T cells.^5–7 Progression of naive T cells to the effector and memory T-cell states is dependent upon signals received during activation and differentiation. It is not yet clear how many sets of signalling conditions can contribute to the varied differentiation outcomes available to naive T cells.

Early activation of naive T cells requires two distinct signals.^8,9 The first is mediated by interaction of the T-cell receptor (TCR) with its cognate antigen in the context of major histocompatibility complex. A naive T cell that receives only signal one, will enter a state of anergy or undergo apoptosis.^10,11 The second, or costimulatory, signal is delivered through accessory molecules on the T-cell surface and is antigen-independent. The TCR and various costimulatory signals differentially contribute to T-cell proliferation^12,13 and the acquisition of effector functions and the onset of memory marker expression correlate with increased cell division.^14,15

CD28 is the prototypic costimulatory molecule, and its ligands, B7.1 (CD80) and B7.2 (CD86), are expressed on professional antigen-presenting cells.^16–20 Costimulation through CD28 synergizes with signalling through the TCR, driving T-cell activation by enhancing gene expression, increasing proliferation and interleukin-2 (IL-2) production, providing protection from signal-1-induced apoptosis, and effectively promoting the progression of T cells from naive to effector and memory populations of both the Th1 and Th2 phenotypes. In addition to its role in T-cell to antigen-presenting cell adhesion,^21,22 stimulation through leucocyte function-associated antigen-1 (LFA-1, CD11a, CD18) also provides a costimulatory signal for T-cell activation.^23–25 Costimulation of naive human CD4⁺ T cells through LFA-1 can provide an initial burst of proliferation and IL-2 production^26,27 but fails to enhance the expansion of cell numbers or to promote cell viability.^28,29 The inability of costimulation through LFA-1 to cause human T cells to function in the same manner as CD28, makes LFA-1 an excellent control for the present study.

The role of resident intercellular adhesion molecule-1 (ICAM-1, CD54) in T-cell activation is considerably less well characterized. ICAM-1 is expressed at low levels on resting and naive T cells, and is up-regulated following activation or in response to pro-inflammatory mediators.³⁰ We demonstrated previously that stimulation through the low constitutive level of ICAM-1 on the T-cell surface is sufficient to activate some signalling processes within 5 min.³¹ Early reports suggested a role for ICAM-1 interactions in augmenting TCR-driven proliferation in cord blood T cells, and in protecting highly differentiated T-cell clones from antigen-induced apoptosis.^32,33 Our laboratory recently demonstrated that costimulation of mixed human peripheral blood lymphocyte (PBL) T cells through TCR (CD3) plus ICAM-1 could provide a functional second signal, leading to phosphatidylinositol-3 kinase activity, increased IL-2 and interferon-γ (IFN-γ) gene expression [a T helper type 1 (Th1) response] but not that of IL-5 (Th2 response), and to substantial cell division in a mixed T-cell population.³⁴ We also demonstrated that costimulation of mixed peripheral T cells through ICAM-1 could induce the appearance of an elevated number of CD4⁺ T cells with a memory-like phenotype (CD45RA^– CD11a^hi CD27^lo) by 7 days.³⁵ From that work, we proposed that ICAM-1 costimulation might favour the outgrowth of existing cells with a memory phenotype, induce some naive T cells contained within the mixed PBL-T population to differentiate into a more specialized cell, or perform both functions.

In the present work, costimulation of purified human naive CD4⁺ T cells through CD3 plus ICAM-1 promoted sustained activation and protection from apoptosis, culminating in the generation of a highly differentiated CD4⁺ CD45RA^– CD11a^hi CD27^– cell population capable of a robust response to single stimulation with suboptimal concentrations of anti-CD3. The appearance of these cells was dependent upon distinct costimulatory signals, and not merely the result of extensive cell division. In these studies, ICAM-1 costimulation was more similar to costimulation through CD28 than through LFA-1 and the resultant cells exhibited characteristics consonant with a memory phenotype.

Materials and methods

Antibodies and reagents

Hybridomas expressing anti-CD3 (OKT3), anti-CD11a (HB202), and anti-CD54 (R6·5D6) were purchased from the American Type Culture Collection (ATCC, Rockville, MD) and purified from serum-free culture medium with protein G-sepharose (Amersham Pharmacia, Piscataway, NJ). Anti-CD28 from two different suppliers (clone 28.2, BD Pharmingen, San Diego, CA, and clone ANC28.1, Ancell, Minneapolis, MN) were used interchangeably with similar results. Anti-CD11a–fluorescein isothiocyanate, anti-CD27–phycoerythrin, anti-CD45RA–TriColor and anti-CD69–fluorescein isothiocyanate were purchased from Caltag Laboratories (Burlingame, CA). Annexin V–phycoerythrin was purchased from Pharmingen. 5-(and-6) Carboxyfluorescein diacetate, succinimidyl ester (CFSE) was purchased from Molecular Probes (Eugene, OR) and used at 2·5 μm in labelling experiments. Propidium iodide was purchased from Sigma (St Louis, MO).

Cell purification

Naive human CD4⁺ T cells were isolated from the peripheral blood of healthy volunteers by negative selection using a naive T-cell enrichment kit (Stem Cell Technologies, Vancouver, BC, Canada). Blood collections were made with informed consent, and with the approval of the University's human subjects committee. All magnetically selected naive CD4⁺ T-cell populations were of >98% purity as assessed by flow cytometry. In the present work, purified naive human T cells represent cells that are CD4⁺ CD45RA⁺RO^– CD11a^lo CD27⁺. Isolation of T cells with a differentiated phenotype was accomplished by negative selection of stimulated populations with antibodies against CD45RA and CD27 (Stem Cell Technologies), resulting in T cells with a CD4⁺ CD45RA^– CD11a^hi CD27^– phenotype.

T-cell stimulation

Antibodies in phosphate-buffered saline were attached to tissue culture-treated plates (Midwest Scientific, St Louis, MO) by incubation at 37° for 2 hr, and wells were washed three times with phosphate-buffered saline to remove unbound antibody. Naive CD4⁺ T cells were added at a concentration of 2 × 10⁶/ml, and stimulated with 1 μg/ml anti-CD3 plus 10 μg/ml anti-ICAM-1, or 10 μg/ml anti-LFA-1, or 2 μg/ml anti-CD28. Optimal antibody concentrations were determined based on the minimum dose that led to maximum T-cell proliferation (not shown). In experiments involving secondary stimulations, naive CD4⁺ T cells were stimulated as indicated for 10 days. Cells were washed and incubated in fresh medium for 2 days to allow cells to return to a resting phenotype. Cultures were spun over Ficoll before secondary stimulation to remove dead cells.

Flow cytometry

Flow cytometry was performed using the FACScan (Becton Dickenson, San Jose CA). The instrument was calibrated using CaliBRITE beads (Becton Dickinson), and this was supplemented by compensation using singly stained cells. Data were analysed using Cell Quest (Becton Dickenson) and WinMDI software (Joe Trotter).

Proliferation assays

T-cell proliferation was measured by radioactive thymidine incorporation as previously described.³⁶ Briefly, cells were stimulated in 96-well plates for 72 hr, and pulsed with 1 μCi/well (67 Ci/mmol) of [³H]thymidine (New England Nuclear, Boston, MA) for the final 6 hr of culture. Samples were harvested using a PHD Cell Harvester (Cambridge Technology, Watertown, MA) and incorporated [³H]thymidine was measured by liquid scintillation counting. Analysis of cell division by CFSE dilution was performed as previously described by others³⁷ and by ourselves.³⁴ Data are displayed as the percentage of cells having undergone the indicated number of divisions (generations) relative to the entire population.

Cytokine measurements

Cell culture supernatants were harvested at the indicated times and stored at −80° until analysed. IL-2 and IFN-γ concentrations were measured using Quantikine kits (R & D Systems, Minneapolis, MN) according to the manufacturer's instructions.

Measurement of cell expansion and apoptosis

After stimulation, naive CD4⁺ T cells were harvested on the days indicated and stained with AnnexinV–phycoerythrin and propidium iodide (PI) at 2 μg/ml before the flow cytometric analysis. Viable cell number was quantified by comparing the number of non-apoptotic or non-necrotic cells (AnnexinV^– and PI^–) acquired on the flow cytometer during 1 min at a constant flow rate. Calibration standards of known cell concentrations were used to calculate the number of viable cells within each population. To measure the induction of apoptosis, CFSE-labelled cells were stained as indicated above and the numbers of apoptotic cells were determined (apoptotic cells are designated as AnnexinV⁺ and PI^–). The data shown are the percentage of apoptotic cells within each cell generation.

Identification of and functional testing of differentiated T cells

The generation of cells with a differentiated phenotype was assessed 14 days following the stimulation of naive CD4⁺ cells. CD4⁺ cells were gated on the CD45RA^– population and assessed for expression of CD11a and CD27 by flow cytometry. Analysis of the non-stimulated naive population was performed on day 1 without gating on CD45RA. To test the ability of these cells to respond to secondary stimulation, cells matching a memory phenotype were magnetically selected as described above. On the same day, fresh naive CD4⁺ T cells were isolated from the same individual and stimulated, as indicated, alongside the purified memory cells for comparison.

Results

Assignment of T-cell populations

Extensive multicolour analyses^38–40 have suggested that accurate description of naive human CD4⁺ cells requires the use of several markers. As a prime example, an accuracy of 99% was attained using CD45RA, CD11a and either CD62L or CD27 and functionality was determined by lack of expression of IFN-γ.³⁹ Thus, naive CD4⁺ human T cells can be accurately described as CD45RA⁺RO^– CD11a^lo CD27⁺. Of particular note, down-regulation of the tumour necrosis factor receptor family member CD27 from the T-cell surface has been described by several investigators as an index for the acquisition of T-cell memory function.^39,41–45 Recently, the most highly differentiated CD4⁺ memory subset was shown to be contained in the CD45RA^–RO⁺ CD27^– subpopulation, where the strongest response to antigen resides.⁴⁶ This was in contrast to CD27⁺ cells, which do not display the traditional recall response and retain the need for costimulation to become activated.⁴⁶ Thus, although discordant populations exist,³⁹ the CD4⁺CD45RA^–RO⁺ CD11a^hi CD27^– T-cell subpopulation reasonably represents a highly differentiated memory T-cell subset and expresses characteristics in sufficient contrast with a starting naive CD4⁺ CD45RA⁺RO^– CD11a^loCD27⁺ parental subset to allow the conclusion that legitimate differentiation has occurred. These parameters were used in the present work.

Naive human CD4⁺ T cells costimulated through ICAM-1 underwent marked cell division and were protected from apoptosis

We examined the ability of costimulation through ICAM-1, resident on the T-cell surface, to influence the division of purified naive human CD4⁺ T cells following activation over a 7-day period, as compared to CD28 or LFA-1 (Fig. 1a). CFSE dilution was used to determine the number of cell divisions induced by each set of costimuli and to estimate the percentage of each population that had undergone a specific number of divisions. Similar to CD28 (closed bars), costimulation through ICAM-1 (open bars) led to cell division that was markedly greater than that observed following costimulation through LFA-1 (hatched bars). The bulk of cells that divided in response to costimulation through CD3 + ICAM-1 or CD3 + CD28 had divided at least five times. This was in contrast to costimulation through LFA-1, which consistently induced two or three cell divisions during the 7 days. In accord with Fig. 1(a), costimulation through ICAM-1 induced an increase in viable cell number up to 7 days in culture (Fig. 1b), albeit not to the same level as CD28. The onset of cell expansion was delayed compared to costimulation through CD28 (data not shown), but this was not unexpected because we have previously demonstrated that mixed PBL-T-cell division in response to costimulation through ICAM-1 is slightly delayed compared to the response to LFA-1 or CD28.³⁵ In contrast, overall cell numbers were not increased following LFA-1 costimulation despite robust cell division, suggesting that these cell cultures were unable to maintain overall viability. This suggested a difference in the ability of the three costimulatory regimens to protect naive T cells from apoptosis.

Figure 1.

Sustained and efficient activation of naive CD4⁺ T cells following costimulation through ICAM-1. (a) Purified human naïve CD4⁺ T cells (2 × 10⁵/well) were loaded with CFSE and stimulated with anti-CD3 in combination with anti-ICAM-1, anti-LFA-1, or anti-CD28. Cell division was assessed at day 7 and the data (mean ± SEM) are presented as the percentage of cells within each generation relative to the entire population. Representative of >10 experiments. (b) Naive CD4⁺ T cells (2 × 10⁵/well) were stimulated with anti-CD3 in combination with anti-ICAM-1, anti-LFA-1, or anti-CD28. Cells were harvested on the days indicated and the total number of viable cells (mean ± SEM) was measured by flow cytometry. Representative of four experiments. (c) To directly measure apoptosis, naive CD4⁺ T cells were labelled with CFSE and stimulated for 7 days with anti-CD3 in combination with anti-ICAM-1, anti-LFA-1, or anti-CD28. The percentage of apoptotic cells (AnnexinV⁺ and PI^–) within each generation is displayed. Representative of eight experiments.

A hallmark of costimulation through CD28 is the ability to protect naive T cells from apoptosis induced by stimulation through the TCR alone. To determine the propensity of naive CD4⁺ T cells to undergo apoptosis in response to the three costimulatory signals, CFSE-labelled cells were analysed 7 days following stimulation and cells were grouped according to generation number based on the dilution of CFSE (Fig. 1c). Results are presented as the percentage of cells during each cell division that were undergoing apoptosis. Similar to published information,²⁹ cells costimulated through LFA-1 were highly prone to apoptosis regardless of the number of cell divisions and, on average, 50% of cells costimulated through LFA-1 were undergoing apoptosis within each generation, even within the comparatively few cells that had divided five times as seen in Fig. 1(a). In contrast, cells costimulated through ICAM-1 or CD28 were largely protected from apoptosis within each generation. These data indicate that costimulation through ICAM-1 protected naive CD4⁺ T cells from activation-induced apoptosis in a manner similar to CD28 and in contrast to LFA-1.

ICAM-1 costimulation resulted in strong IL-2 production

The ability to produce IL-2 is directly linked to the survival of naive T cells following activation. Naive T cells costimulated through TCR + CD28 are known to produce high levels of IL-2, whereas naive T cells receiving stimulation through the TCR alone produce very little IL-2, and naive cells being costimulated through CD2, CD5, CD9, CD11a, or CD44 also produce very little IL-2.⁴⁷ In addition, the transition from naive to effector cells is largely reliant on cytokine-dependent differentiation, which occurs following antigen stimulation.⁴⁸ To investigate the effect of costimulation through ICAM-1 on IL-2 production, cytokine release was measured in response to stimulation over 7 days (Fig. 2). The onset of IL-2 secretion was most rapid in response to CD28 costimulation, peaking at 3 days. LFA-1 costimulation displayed similar kinetics, albeit with lower concentrations of IL-2. ICAM-1 also was able to induce high amounts of IL-2 secretion, with increasing amounts detectable from days 3 to 7. Interestingly, the difference in the kinetics is similar to that which we observed in the onset of cell expansion. Together with Fig. 1, these data suggest that costimulation through ICAM-1 is capable of efficient and sustained activation of naive human CD4⁺ T cells.

Figure 2.

Costimulation of naive CD4⁺ T cells through ICAM-1 induced high IL-2 production. Triplicate cultures of naive CD4⁺ T cells were left not stimulated or stimulated with anti-CD3 alone or in combination with anti-ICAM-1, anti-LFA-1, or anti-CD28. Cell culture supernatants were harvested on days 1, 2, 3, 5 and 7, and IL-2 secretion was measured by ELISA. Values for all stimuli were plotted but some were too low to be visible. Data are presented as the mean of triplicate samples. Representative of four experiments.

Costimulation of naive T cells through ICAM-1 generated T cells with a phenotype characteristic of highly differentiated memory cells

Having confirmed that costimulation of naive CD4⁺ T cells through ICAM-1 is capable of promoting expansive cell division and providing protection from apoptosis, we sought to determine whether this population could ultimately differentiate into defined subpopulations corresponding to known T-cell phenotypes. As described in the Introduction, one well-characterized human CD4⁺ phenotype that is indicative of cell differentiation is the CD45RA^– CD11a^hi CD27^– subpopulation, which describes a most highly differentiated memory T-cell subpopulation. In Fig. 3, naive CD4⁺ T cells were isolated magnetically according to the phenotype CD4⁺ CD45RA⁺ CD11a^dimCD27⁺, as shown in the upper left quadrant. These cells were stimulated for 14 days in culture before analysis for changes in cell phenotype. The upper right quadrant (CD3 + ICAM-1) and both lower quadrants represent cells that had become CD45RA^– during stimulation. Costimulation through either ICAM-1 or CD28 consistently produced a respectable percentage (8% and 6%, respectively) of CD45RA^– CD11a^hi CD27^– T cells under these in vitro conditions. This demonstrated that costimulation through ICAM-1 is sufficient to drive naive CD4⁺ cell differentiation. Interestingly, even though costimulation through LFA-1 could induce the appearance of CD45RA^– cells with elevated expression of CD11a, these cells did not down-regulate CD27 in any of our multiple experiments, suggesting that although this mode of costimulation could change the phenotype of naive human T cells it did not drive them to the final step(s) of differentiation.

Figure 3.

T cells with a differentiated phenotype were generated from naive CD4⁺ cells following costimulation through ICAM-1. Naive CD4⁺ T cells were left not stimulated or stimulated for 14 days with anti-CD3 in combination with anti-ICAM-1, anti-LFA-1, or anti-CD28. Purity of the starting naive population (CD45RA⁺ CD11a^dim CD27⁺) was confirmed by staining non-stimulated cells on day 1 (upper left panel, Ø). Results shown for stimulated cells were gated on CD45RA^– cells before analysis (remaining three panels). The percentage of cells matching a stringent memory phenotype (CD4⁺ CD45RA^– CD11a^hi CD27^–) in response to costimulation is depicted in the lower right quadrant of each appropriate panel. Representative of eight experiments.

Costimulation of naive T cells through ICAM-1 generated T cells that were capable of responding to a second stimulation through the TCR alone

Differentiated memory T cells are capable of responding to lower concentrations of antigen, are less dependent on costimulation, and display effector functions, such as cytokine production, more readily following activation than their naive counterparts.^49,50 To determine whether the CD4⁺ CD45RA^– CD11a^hi CD27^– population that we observed was functionally analogous to antigen-experienced memory T cells, these cells were negatively selected to yield a purified (98%) population (Fig. 4a). These resting (CD69^–, data not shown) CD4⁺ CD45RA^– CD11a^hiCD27^– T cells generated from costimulation through ICAM-1 were compared with CD4⁺ naive T cells (CD45RA⁺ CD11a^lo CD27⁺) from the same individual for their ability to respond to stimulation through the TCR (anti-CD3) in the absence of costimulatory signals. As expected, both populations were capable of increased activation marker expression (CD69) following stimulation through the TCR alone (Fig. 4b), confirming that each was capable of responding to stimulation.

Figure 4.

Memory-like cells generated following costimulation through ICAM-1 were pure and capable of responding to stimulation. (a) Cells costimulated through ICAM-1 were isolated by negative magnetic sorting using antibodies against CD45RA and CD27, and this routinely yielded a highly purified population (98%) of cells with the CD45RA^– CD11a^hi CD27^– phenotype. (b) Purified memory cells (CD45RA^– CD11a^hi CD27^–) and freshly isolated naive cells (CD45RA⁺ CD11a^lo CD27⁺) from the same subject were stimulated for 8 hr with anti-CD3 alone (1 μg/ml) and assessed for expression of the activation marker CD69 by flow cytometry. Representative of four experiments.

As mentioned, memory cells respond more readily to stimulation through the TCR alone than do naive T cells. Naive T cells and induced memory-like cells from the same patient were compared in Fig. 5. In Fig. 5(a), the CD4⁺ CD45RA^– CD11a^hi CD27^– T cells generated by costimulation through ICAM-1 (open squares) proliferated at lower concentrations of anti-CD3 stimulation, and responded more vigorously at higher concentrations of anti-CD3, than did naive (CD4⁺ CD45RA⁺ CD11a^loCD27⁺) T cells (open triangles) from the same subjects.

Figure 5.

Memory-like cells generated following costimulation through ICAM-1 were more responsive than corresponding naive cells. (a) As used in Fig. 4, purified, ICAM-1-induced memory-like cells and freshly isolated naive cells from the same subject were stimulated with increasing concentrations (0·005–5·0 μg/ml) of anti-CD3 alone for 72 hr. Proliferation was measured by [³H]thymidine incorporation and data are presented as the mean c.p.m. ± SEM. Representative of four experiments. Purified memory cells and freshly isolated naive cells were tested for their capacity to produce (b) IL-2 following stimulation with PHA for 6 and 24 hr or (c) IFN-γ in response to stimulation through the TCR (antibody at 1 μg/ml) or PHA at 24 hr. Cytokine concentrations are listed as the mean ± SEM. Representative of three experiments.

The ability to elicit effector functions following activation also was investigated. In Fig. 5(b), IL-2 production by the memory population (closed bars) was more rapid and more robust within the first 24 hr than in the naive population (open bars) in response to PHA stimulation. Naive T cells are generally considered incapable of secreting IFN-γ. In Fig. 5(c), the memory-like cells (closed bars) were able to secrete high levels of the effector cytokine IFN-γ in response to stimulation through the TCR alone or to PHA. In contrast, the naive population (open bars) secreted barely detectable IFN-γ in response to PHA and none in response to stimulation through the TCR alone, as expected. These results suggested that highly differentiated CD45RA^– CD11a^hi CD27^– T cells generated following costimulation of naive CD4⁺ T cells through ICAM-1 are functionally adept cells capable of responding rapidly and vigorously to secondary stimulation through the TCR alone. In keeping with the analyses of this T-cell phenotype by several other laboratories, these cells appear to be highly differentiated, and seem to fill several criteria attributed to memory T cells.

The generation of CD4⁺ CD45RA⁻ CD11a^hi CD27⁻ T cells was dependent on specific costimulatory signalling rather than only increased cell division

It has been widely discussed that the appearance of the memory T-cell subset is a consequence of repeated antigen stimulation and multiple rounds of cell division. It was possible therefore, that the absence of this population following costimulation through LFA-1 was simply the result of fewer numbers of cell divisions compared to costimulation through ICAM-1 or CD28. To address this issue, CFSE-labelled naive CD4⁺ T cells were subjected to successive stimulations using either ICAM-1 or LFA-1 as a costimulatory signal. Cells were costimulated for 10 days, washed, rested for 2 days, purified over Ficoll to remove dead cells, and then costimulated a second time for an additional 10 days. Following two cycles of costimulation of CD4⁺ naive T cells through CD3 + ICAM-1, a sizeable percentage (39%) of cells had acquired the CD45RA^– CD11a^hi CD27^– memory phenotype (Fig. 6a). In contrast, very few CD45RA^– CD11a^hi CD27^– T cells were generated (1·4%) following two successive costimulation series through CD3 + LFA-1 (Fig. 6b). The absence of cells bearing the CD4⁺ CD45RA^– CD11a^hi CD27^– phenotype in response to costimulation through LFA-1 was not the result of a lack of cell division (Fig. 6c). During the first cycle of costimulation, ICAM-1 costimulation (open bars) produced the expected more than five cell divisions and LFA-1 costimulation (closed bars) elicited an average of three divisions. Following the second series of costimulation, the majority of LFA-1-costimulated cells (hatched bars) had undergone five or more divisions, and represented a much greater level of activity than that observed following two cycles of costimulation through ICAM-1 (shaded bars). Nevertheless, the lesser ICAM-1-induced cell division still gave rise to elevated numbers of the CD4⁺ CD45RA^– CD11a^hi CD27^– population (Fig. 6a). Most interestingly, the ability to generate this highly differentiated population was not lost by the naive cells during the initial rounds of cell division induced by costimulation through CD3 + LFA-1. For Fig. 6(d), ICAM-1 was used as the second cycle of costimulation for cells that had first been costimulated for 10 days through CD3 + LFA-1 and then rested for 2 days. High levels (26%) of the CD4⁺ CD45RA^– CD11a^hi CD27^– T-cell population were observed in response to costimulation of this LFA-1-induced population through ICAM-1. Thus, the generation of this highly differentiated effector population was not merely the result of extensive cell division. This supports the earlier work of others based on similar mutation rates of the HPRT gene in CD27⁺ as compared with CD27^– cells from the CD45RO⁺ compartment.⁵¹ Thus, the generation of CD4⁺ CD45RA^– CD11a^hi CD27^– T cells seems to be governed, at least in part, by the nature of the costimulatory signal delivered to the cell. Taken together, these data suggest a new role for ICAM-1 in promoting efficient activation of naive human CD4⁺ T cells, and their differentiation into memory-like T-cell populations.

Figure 6.

Generation of CD45RA^– CD11a^hi CD27^– cells was dependent on costimulatory signalling rather than number of cell divisions. Naive CD4⁺ T cells were costimulated for 10 days with anti-CD3 plus either (a) anti-ICAM-1 or (b) anti-LFA-1 as a costimulatory signal, washed and rested for 2 days, purified over Ficoll to remove dead cells and costimulated a second time with the same costimulatory regimen. The generation of CD45RA^– CD11a^hi CD27^– effector cells was assessed by flow cytometry 10 days following the second stimulation and the percentages are shown in the lower right quadrant. (c) Parallel cultures were labelled with CFSE before the first stimulation and cell division was assessed at the conclusion of both the first costimulation (ICAM-1, open bars and LFA-1 closed bars) and the second costimulation (ICAM-1, shaded bars, and LFA-1, hatched bars). (d) Naive CD4⁺ T cells were costimulated first for 10 days with LFA-1 as the costimulatory signal, rested and re-purified as described above, and costimulated a second time with ICAM-1 as the costimulatory signal. The generation of CD45RA^– CD11a^hi CD27^– effector cells was assessed by flow cytometry 10 days following the second stimulation and the percentage is noted. Representative of three experiments.

Discussion

During an immune response, it is thought to be most likely that the differentiation of antigen-specific T cells from naive to effector and memory populations is driven by multiple costimulatory interactions and cytokines, rather than by a single signal. Using in vivo studies alone, it can be difficult to discern all the effects attributable to an individual costimulus because of the complexity of the in vivo environment, the tendency of various costimulatory signals to partially overlap in their effects on T-cell activation, and to outright redundancy of function. Therefore, to better understand the aggregate contribution of the several alternative costimulatory molecules it is useful to complement in vivo studies by employing in vitro systems that allow for individual contributions to be assessed, and then to examine the effects of combined signalling. In the present study, we report that costimulation through ICAM-1 can drive naive human CD4⁺ T-cell activation, leading to the generation of T cells with a highly differentiated cell surface phenotype evocative of memory T cells. This transition was accompanied by sustained cell division, Th1 cytokine production in response to suboptimal stimulation through the TCR, and the cells were largely protected from apoptosis.

It is becoming apparent that in the context of T-cell costimulation, resident ICAM-1 shares several features with CD28 but is not completely redundant with CD28. Others have shown that blockade of ICAM-1/LFA-1 interactions fosters greater appearance of Th2 cytokines during activation.⁵² We have previously reported that ICAM-1 costimulation induces secretion of Th1 but not Th2 cytokines from mixed PBL-T cells, whereas CD28 costimulation induced secretion of both Th1 and Th2 cytokines,³⁴ as was already known. In addition, the dynamics of expression of these two surface proteins are markedly different. Naive T cells express low levels of ICAM-1, and high levels of CD28. ICAM-1 expression is induced by a number of signalling mechanisms, including inflammatory cytokines and stimulation through the TCR.³⁰ In contrast, CD28 is often down-regulated following activation, and in chronic disease states and during aging a large percentage of the T-cell population is CD28^–.⁵³ It is therefore possible that these molecules exert their primary influence on the T-cell response at different times or under different circumstances, although the data presented here suggest that this is not necessarily always the case.

Several additional proteins expressed by T cells exert costimulatory function, generally as a result of their ability to augment proliferation and cytokine production in conjunction with suboptimal stimulation through the TCR (reviewed in ref. 54). Thus far, none of these costimulatory molecules is completely redundant with CD28. Despite supporting initial naive T-cell activation, costimulation through CD3 plus any of CD2, CD5, CD9, CD11a or CD44 did not sustain proliferation and IL-2 production, prevent apoptosis or promote differentiation.⁴⁷ Some costimulatory molecules appear to act only at later stages of T-cell activation, to effect the acquisition of distinct effector functions and promote memory cell generation. These include 4-1BB,^55,56 OX-40^57,58 and LIGHT.⁵⁹ It seems that ICAM-1 is not redundant with any of the potential costimulatory molecules reported thus far, and seems to more closely approximate the effects of CD28 under some circumstances.

The present work also supports the argument that in certain instances, specific signals are required for differentiation to memory T cells and that merely dividing for several generations is insufficient to trigger the changes in gene expression necessary to achieve specific phenotypes. It will be interesting to define the different panels of genes expressed under the conditions defined herein.

In summary, using purified naive human CD4⁺ T cells (CD4⁺ CD45RA⁺ CD11a^lo CD27⁺) we have demonstrated that costimulation through ICAM-1 can promote sustained activation and drive T-cell differentiation into cells (CD4⁺ CD45RA^– CD11a^hi CD27^–) capable of responding to secondary stimulation in the manner that is predicted for memory T cells. This differentiated memory-like population was not generated simply in response to multiple rounds of cell division, but was a consequence of specific costimulatory signalling. Studies addressing the signalling pathways involved in ICAM-1-mediated costimulation, as well as work with ICAM-deficient animals, will allow us to better understand the role of ICAM-1 in modulating T-cell responses.

Abbreviations

CFSE

5-(and-6) carboxyfluorescein diacetate, succinimidyl ester

ICAM-1

intercellular adhesion molecule-1

IFN-γ

interferon-γ

IL-2

interleukin-2

LFA-1

leucocyte function-associated antigen-1

TCR

T-cell receptor

Th1

T helper type 1