Control of Memory CD4 T Cell Recall by the CD28/B7 Costimulatory Pathway¹

Memory T cells are known to exhibit enhanced activation properties relative to naive T cell counterparts manifested by robust responses to low doses of Ag (1), rapid production of potent effector cytokines (2), and reduced activation requirements (3, 4, 5). These features allow memory T cells to mediate protective immunity to pathogens; however, their robust responses are detrimental when directed to self Ags in autoimmunity and alloantigens in transplantation (6, 7). In addition, memory T cells are heterogeneous in expression of lymph node homing receptors and tissue distribution delineating central memory T (T_CM;⁴ CD62L^high/CCR7⁺) cells in lymphoid tissue and effector memory (T_EM; CD62L^low/CCR7⁻) cells in peripheral tissues (8, 9). Memory CD4 T cell subsets exhibit functional diversity in cytokine production and expansion (8, 10, 11, 12) and play distinct roles in pathogenic processes (13), whereas memory CD8 T cell subsets can differ in their protective responses to virus infection or tumors in vivo (14, 15, 16, 17). These results suggest that differential modulation of memory subsets and/or migratory properties can have profound effects on controlling recall responses, although strategies to achieve this level of regulation have not been defined.

Costimulatory pathways regulate T cell activation and differentiation, and serve as therapeutic targets for manipulation of T cell responses. Of particular importance is the CD28/B7 costimulatory pathway, as interaction of the CD28 costimulatory receptor with its ligands B7-1/B7-2 (CD80/CD86) on APCs is required for activation of naive CD4 T cells. Blocking this pathway with the well-characterized CTLA4Ig fusion molecule that binds B7 ligands (18) inhibits naive CD4 T cell activation and proliferation (19), and can prevent the development of autoimmunity and inhibit graft rejection in animal models (20, 21). By contrast, it is generally accepted that memory T cells do not require CD28/B7-derived costimulation for recall responses (22), based on previous studies showing that memory CD4 T cells are fully activated by B7-deficient APC in vitro (5, 23), and that CD28^−/− mice are not impaired in the generation or recall of memory CD8 T cell responses to lymphocytic choriomeningitis virus infection (24). However, CTLA4Ig treatment can alleviate ongoing autoimmunity in animal models (25), and shows clinical efficacy in treating rheumatoid arthritis (26) and psoriasis (27), diseases known to be driven by memory CD4 T cells (28). These results suggest that the CD28/B7 pathway may regulate memory CD4 T cell-mediated responses; however, the in vivo requirements for CD28/B7 costimulation and the CTLA4Ig susceptibility of Ag-specific memory CD4 T cells have not been examined.

In this study, we used CTLA4Ig to investigate the CD28/B7 costimulatory requirements of heterogeneous memory CD4 T cells specific for influenza hemagglutinin (HA) and chicken OVA peptides, as well as polyclonal influenza virus-specific memory cells generated from flu infection in vivo. In vitro activation of HA- and OVA-specific memory CD4 T cells in the presence of CTLA4Ig induces selective inhibition of IL-2 and TNF-α production, without affecting early IFN-γ production and up-regulation of activation markers. Likewise, IL-2 production of influenza virus-specific memory CD4 T cells was profoundly inhibited by CTLA4Ig treatment, whereas the IFN-γ production was partially down-regulated. When administered in vivo, CTLA4Ig did not affect very early activation of HA- and OVA-specific memory T cells, although it profoundly inhibited in vivo Ag-driven proliferation, and triggered a loss of Ag-stimulated memory CD4 T cells that was most striking in the effector memory (CD62L^low) subset. Importantly, the responding memory T cells in CTLA4Ig-treated mice exhibited a predominant T_CM phenotype compared with a T_EM phenotype in control-treated mice. These results reveal a novel role for CD28 costimulation in regulating the expansion and resultant homing capacities of Ag-recalled memory CD4 T cells, and have important clinical implications for the use of CTLA4Ig in controlling the balance of pathologic and protective immune responses in autoimmunity and infectious diseases.

BALB/c mice (8–16 wk of age) were obtained from the National Cancer Institute Biological Testing Branch. Influenza HA-TCR-transgenic mice (29) and DO11.10 mice (30) bred as heterozygotes onto BALB/c (Thy1.2) or BALB/c (Thy1.1) hosts, and RAG2^−/− mice (31) on BALB/c backgrounds (Taconic Farms) were maintained under specific pathogen-free conditions. BALB/c mice used for influenza infections were transferred into the biocontainment facility. All mice were maintained in the Animal Facility at the University of Maryland Medical Center, and the experimental protocols were approved by the Institutional Animal Care and Use Committee at the University of Maryland.

The following Abs were purified from bulk culture supernatants and purchased from BioExpress: anti-CD8 (TIB 105), anti-CD4 (GK1.5), anti-I-A^d (212.A1), and anti-Thy-1 (TIB 238). The 6.5 anti-clonotype Ab directed against the HA-TCR (29) was purified and conjugated to biotin (Pierce). The clonotype-specific Ab KJ1-26 specific for the DO11.10 CD4 TCR was purchased conjugated to PE (BD Pharmingen) or allophycocyanin (Caltag Laboratories). Allophycocyanin- or PE-conjugated anti-IFN-γ, and -IL-2, -CD25, -CD44, -CD62L, and PerCP-conjugated anti-CD4 were purchased from BD Pharmingen. Murine and human CTLA4Ig were obtained from Bristol-Myers Squibb. Single-chain anti-mouse CD28-blocking Ab was provided by R. O’Hara (Wyeth Laboratories, Madison, NJ).

HPLC-purified influenza HA peptide (110–120, SFERFEIFPKE) and OVA peptide (323–339, ISQAVHAAHAEINEAGR) were synthesized by the Biopolymer Laboratory at the University of Maryland School of Medicine.

For generation of HA- and OVA-specific effector CD4 T cells, CD4 T cells were purified from spleens of HA-TCR and DO11.10 mice, as described (12), and cultured with 5.0 μg/ml HA peptide or 1.0 μg/ml OVA peptide, respectively, and mitomycin C-treated APCs were prepared from BALB/c splenocytes, as described (32), in complete Clicks medium (Irvine Scientific) (12) for 3 days at 37°C. The resultant effector cells were purified by Ficoll density gradient centrifugation (ICN/Cappel), resulting in 90–95% CD4⁺6.5⁺ or KJ1-26⁺ T cells. Memory generation from these effector cells was achieved by i.v. transfer of 5 × 10⁶ purified effector cells into RAG2^−/− adoptive hosts, as done previously (12, 33, 34, 35). Persisting memory CD4 T cells were harvested from spleen and mesenteric lymph node 2–5 mo posttransfer, as described (35), for subsequent in vitro and in vivo analysis (see below). In some cases, spleen-derived memory cells were further fractionated into CD62L^high and CD62L^low subsets with anti-CD62L-coupled magnetic beads (Miltenyi Biotec) using the autoMACS (Miltenyi Biotec), as described (35).

Intracellular cytokine staining analysis of IFN-γ and IL-2 was performed, as previously described (33, 35). Briefly, memory CD4 T cells were cultured with APC and 5 μg/ml HA or 1 μg/ml OVA peptide ± 50 μg/ml murine CTLA4Ig or isotype control IgG2a for 18–36 h, and monensin (Golgistop; BD Pharmingen) was added 6 h before harvesting. Specific timepoints for each experiment are indicated in the figure legends. Cells were stained with Abs for surface CD4, CD25, CD44, KJ1-26, or 6.5, fixed (Cytofix buffer; BD Pharmingen), permeabilized, and stained intracellularly with fluorochrome-conjugated anti-IFN-γ, anti-IL-2, or their respective isotype controls in permeabilization buffer before washing and FACS analysis. Stained cells were analyzed using either FACSCalibur or LSRII and CellQuest or FACSdiva software, respectively (BD Biosciences).

The relative frequencies of IFN-γ-, IL-2-, TNF-α-, IL-4-, IL-5-, and IL-10-secreting memory CD4 T cells in response to stimulation with HA/OVA peptide were determined using ELISPOT, as previously described (34, 35), using 5 × 10⁴-1 × 10⁵ purified memory CD4 T cells, 1.5 × 10⁵-3 × 10⁵ BALB/c APC, and HA or OVA peptide (5 or 1 μg/ml, respectively) per well. Plates were incubated at 37°C for 12–18 h, washed, and developed, as described (34, 35), and spots were counted using the Immunospot ELISPOT reader (CTL; BD Biosciences).

Polyclonal memory CD4 T cells specific for influenza were generated by infecting BALB/c mice intranasally with a sublethal dose (10² PFU in 20 μl of PBS) of influenza A virus (A/PR/8/34, PR8) obtained from American Type Culture Collection and grown in the allantoic fluid of embryonated chicken eggs in the laboratory of D. Perez (University of Maryland, College Park, MD). Infection was monitored in mice by weight loss measured daily for the first week, after which mice recovered. Splenic CD4 T cells containing influenza virus-specific memory CD4 T cells were harvested from primed hosts 12–16 wk postinfection. The relative frequency of IL-2- and IFN-γ-secreting cells was determined using the ELISPOT assay, as previously described (34, 35), using 5 × 10⁴-1 × 10⁵ purified CD4 T cells and 1.5 × 10⁵-3 × 10⁵ BALB/c APC preincubated with 1 multiplicity of infection of PR8 influenza for 1 h at 37°C, and 50 μg/ml CTLA4Ig or IgG2a was added to respective wells. Plates were incubated for 12–18 h, washed, and developed, as described (34, 35); spots were counted using the Immunospot ELISPOT reader (CTL; BD Biosciences).

To monitor in vivo recall of Ag-specific memory CD4 T cells, HA- and OVA-specific memory CD4 T cells were harvested from spleens of three to six RAG2^−/− adoptive hosts (see above) (33, 34), purified as described (34), labeled with 5 μM CFSE (Molecular Probes) (36), and injected into four groups of BALB/c (Thy 1.1) hosts (1.5–2.5 × 10⁶ cells per mouse). These secondary hosts were treated with 250 μg of murine CTLA4Ig or IgG2a on days 1, 3, and 5 with respect to memory transfer, and challenged with 100 μg of HA or 25 μg of OVA peptide in PBS or PBS alone as a negative control on day 5, or CD4 T cells were recovered from spleen 4 h after peptide administration, and from spleen and mesenteric lymph nodes 60 h after peptide administration (Fig. 4 A). Cells were analyzed for expression of CD4, CD62L, CD25, CD69, and 6.5TCR or KJ1-26 by flow cytometry, as described above. Analysis of proliferation was calculated from the percentage divided at each cell division, as previously described (34, 37). Statistical comparison between treatment groups was done using Student’s t test.

To study the role of CD28/B7 costimulation in memory CD4 T cell function, we generated Ag-specific memory CD4 T cells using an adoptive transfer system extensively characterized and validated by our laboratory and others (5, 12, 23, 33, 34, 35, 38, 39, 40), and analyzed their antigenic recall functions in vitro and in vivo in the presence of CTLA4Ig to inhibit CD28/B7 costimulation. For memory generation, we obtained CD4 T cells specific for influenza HA from HA-TCR-transgenic mice (29), primed them in vitro with HA peptide and APC, transferred the resultant effector cells into RAG2^−/− or BALB/c (Thy1.1) hosts, and recovered persisting memory CD4 T cells after 2–6 mo in vivo. We previously showed that HA-specific memory CD4 T cells generated in either lymphocyte-deficient or intact mouse hosts exhibit the phenotypic and functional attributes of endogenous memory CD4 T cells (12, 32, 33, 35, 41). The phenotype of Ag-specific naive and memory CD4 T cells used in these experiments was consistent with our previous results (12, 32, 33, 42); HA-specific naive CD4 T cells are CD25^low, CD44^low, and CD62L^high, whereas memory CD4 T cells are CD25^low, CD44^high, and heterogeneous for CD62L expression (Fig. 1 A). Functionally, HA-specific memory CD4 T cells exhibit rapid production of IFN-γ and IL-2 (12, 33, 34, 35).

We initially compared the ability of CTLA4Ig to inhibit activation of HA-specific naive and memory CD4 T cells in vitro. As expected, in vitro activation of HA-specific naive CD4 T cells was significantly impaired by CTLA4Ig, as manifested by greatly reduced up-regulation of IL-2Rα (CD25) (Fig. 1,B), a marker of early activation. By contrast, activation of HA-specific memory CD4 T cells in the presence of CTLA4Ig did not affect CD25 up-regulation at any timepoint (Fig. 1 B and data not shown). Thus, CTLA4Ig appeared to have selective inhibitory effects on naive CD4 T cell activation, while not affecting memory T cell activation, consistent with previous in vitro results comparing naive and memory CD4 T cell responses in the absence of CD28/B7 costimulation (5, 23).

We have shown previously that CD25 up-regulation on Ag-stimulated memory CD4 T cells does not necessarily reflect underlying modulations in cytokine production that can occur by altering recall parameters (33, 34). When additional memory functions were assessed, we found that IL-2 production was substantially inhibited (average 65 ± 14% inhibition; n = 4) from Ag-stimulated memory CD4 T cells in the presence of CTLA4Ig compared with isotype controls (Fig. 1,C, second row), whereas IFN-γ production, a hallmark of memory T cell recall, was not affected (third row). To determine whether this selective IL-2 inhibition by CTLA4Ig on unfractionated memory CD4 T cells was due to differential effects on a particular memory T cell subset, we sorted splenic HA-specific memory CD4 T cells into CD62L^low (T_EM) and CD62L^high (T_CM) subsets and assessed Ag-driven cytokine production in the presence of CTLA4Ig. We have determined previously that HA-specific T_EM and T_CM subsets both produced IL-2 as well as the effector cytokine IFN-γ (35), similar to findings that lymphocytic choriomeningitis virus-specific T_CM and T_EM CD8 T cell subsets both produced effector cytokines (17). We found that CTLA4Ig selectively inhibited IL-2 production by 50–70% for both CD62L^low and CD62L^high memory CD4 T cell subsets without affecting IFN-γ production, similar to results obtained with unfractionated memory CD4 T cells (Fig. 1 D). These results demonstrate a selective requirement for CD28/B7 costimulation for optimal IL-2 production from heterogeneous subsets of memory T cells.

Previous studies showed that memory CD4 T cells generated using similar adoptive transfer approaches from DO11.10 (OVA-specific) or AND (cytochrome c-specific) TCR-transgenic CD4 T cells produced high levels of IL-2 and effector cytokines when stimulated independent of CD28 costimulation with B7-deficient APC (5, 23). To establish that CTLA4Ig-mediated inhibition in our system was not peculiar to HA-TCR-derived memory T cells, we generated OVA-specific memory CD4 T cells by adoptive transfer of in vitro primed DO11.10 CD4 T cells, as previously described (3), and tested their requirement for CD28/B7 costimulation in vitro. Similar to HA-specific memory cells, IL-2 production by OVA-specific memory CD4 T cells was selectively inhibited (70%) in the presence of CTLA4Ig compared with controls (Fig. 2 A, second row), whereas CD25 up-regulation and IFN-γ production were not suppressed. These results clearly demonstrate that CTLA4Ig selectively inhibits IL-2 production by memory CD4 T cells in two different peptide Ag systems.

To establish that CTLA4Ig-driven inhibition of IL-2 production was due to direct blockade of the CD28 pathway, we assessed memory CD4 T cell responses in the presence of a blocking anti-CD28 Ab (see Materials and Methods). We found similar selective inhibition of IL-2, but not IFN-γ production by OVA-stimulated memory CD4 T cells in the presence of anti-CD28-blocking Ab or CTLA4Ig (Fig. 2,B). These results establish an intrinsic requirement for CD28/B7 costimulation for optimal IL-2 production by Ag-specific memory CD4 T cells, and that CTLA4Ig functions primarily to block positive CD28-mediated signals in memory CD4 T cells. We further assessed the role of CTLA4Ig-mediated blockade on the production of other cytokines, and observed an inhibition (50–70%) of early TNF-α in the presence of CTLA4Ig (Fig. 2 C), whereas late TNF-α was unaffected and IL-4, IL-5, and IL-10 were produced at low levels and were unaffected by CTLA4Ig treatment (data not shown). Therefore, CTLA4Ig-mediated blockade selectively down-regulates IL-2 and early TNF-α production from memory CD4 T cells.

The above results demonstrate that CTLA4Ig and anti-CD28 can inhibit IL-2 production from peptide-stimulated memory CD4 T cells generated from TCR-transgenic mice. To establish that CD28 costimulation likewise plays a role in the recall function of memory CD4 T cells generated under physiological conditions in vivo, we used an influenza mouse infection system to generate polyclonal virus-specific memory CD4 T cells. For these experiments, we infected BALB/c hosts intranasally with a sublethal dose of influenza A virus (A/PR/8/34), and 12–16 wk postinfection harvested splenic CD4 T cells containing a polyclonal population of influenza-primed memory CD4 T cells. To quantitate the flu-specific memory CD4 T cell response in the presence of CTLA4Ig or control IgG2a, we stimulated total CD4 T cells from naive or influenza-infected mice with splenic APC incubated with live influenza virions and analyzed cytokine production by ELISPOT. As shown in Fig. 3, CD4 T cells harvested from flu-infected hosts produced significant levels of IL-2 and IFN-γ within 12–18 h of virus stimulation characteristic of a memory response, whereas CD4 T cells from uninfected controls lacked flu-specific responses. In the presence of CTLA4Ig, however, there was almost complete inhibition of IL-2 and partial down-regulation (50%) of IFN-γ production from flu-specific memory CD4 T cells (Fig. 3). The greater inhibition of IL-2 compared with IFN-γ production by CTLA4Ig in polyclonal virus-specific memory CD4 T cells is similar to our results with TCR-transgenic peptide-specific memory CD4 T cells, establishing that CTLA4Ig can exert its effects on memory responses to multiple Ags and in the presence of pathogen-associated signals. Interestingly, the extent of CTLA4Ig-mediated inhibition was higher for recall of flu-specific compared with TCR-transgenic memory CD4 T cells, which is most likely due to the 20-fold difference in precursor frequency in these two systems. These results further validate that the Ag-specific memory cells generated in this study reflect the activation and functional and costimulatory requirements of memory cells generated under physiological conditions such as virus infection.

It is well established that IL-2 is an important growth/survival factor for T cells (43, 44), and recently it has been shown to be important for the expansion of memory CD8 T cells during secondary responses (45). Based on our in vitro results showing a decrease in memory CD4 T cell IL-2 production, we hypothesized that the CD28/B7 pathway may be important for recall expansion in vivo. To examine the effect of CTLA4Ig-mediated costimulation blockade on a specific population of Ag-specific memory CD4 T cells in vivo, we used a secondary transfer system previously optimized in the laboratory to follow the fate of Ag-recalled memory CD4 T cells in an intact mouse host (34). We transferred CFSE-labeled HA- or OVA-specific memory CD4 T cells into BALB/c (Thy1.2 or Thy1.1) hosts that were administered 250 μg of CTLA4Ig (weight equivalent for 10 mg/kg dose used in patients (26)) or murine IgG2a isotype control for three successive treatments (Fig. 4,A). Each group of treated mice was subsequently boosted with HA or OVA peptide Ag (or PBS as a control), as we have shown previously robust in vivo secondary responses by memory CD4 T cells boosted with antigenic peptide (34). Recalled memory CD4 T cells were recovered from spleen and lymph nodes at early times after boosting (4 h), and at a later timepoint (60 h) (Fig. 4 A), which we previously determined was the optimal timepoint for measuring peak in vivo proliferation of Ag-stimulated memory CD4 T cells (34). Visualization of Ag-specific memory CD4 T cells in vivo was accomplished based on Thy-1 allelic differences for HA-specific memory CD4 T cells, which likewise represented 6.5⁺ T cells, and based on KJ1-26 clonotype expression for OVA-specific memory CD4 T cells (data not shown).

For immediate recall in vivo, we found that OVA-specific memory CD4 T cells exhibited rapid up-regulation of the early activation markers CD25 and CD69 4 h after peptide boosting that was similar in IgG2a- and CTLA4Ig-treated mice (Fig. 4,B). CTLA4Ig treatment likewise did not affect rapid CD25 and CD69 up-regulation on HA-specific memory CD4 T cells, although a lower proportion of HA-specific cells was CD25⁺ or CD69⁺ at this timepoint (Fig. 4 C). These results establish that the CD28 pathway is not required for rapid activation of memory CD4 T cells in vivo, and that the majority of memory T cells have encountered the Ag.

Despite the lack of CTLA4Ig-mediated effects at this early timepoint, we found that in vivo proliferation and expansion of Ag-specific memory CD4 T cells were significantly inhibited by CTLA4Ig treatment that was apparent at later times (60 h). In control-treated mice, Ag-specific memory CD4 T cells underwent significant proliferation in response to peptide boosting compared with PBS boosting (Fig. 5,A, first row), with OVA-specific memory CD4 T cells proliferating more extensively than HA-specific memory CD4 T cells. In the presence of CTLA4Ig, however, recall proliferation of HA- and OVA-specific memory CD4 T cells was dramatically reduced compared with isotype controls (Fig. 5 A, second row), manifested by a significant increase (p < 0.05) in the percentage of undivided (CFSE^high) HA- and OVA-specific memory CD4 T cells (right quadrant), and substantially fewer number of cells undergoing >2 divisions (area outlined by box). These results show striking decreases in recall proliferation of memory CD4 T cells in CTLA4Ig-treated hosts in two distinct Ag systems with different magnitudes of proliferation.

To obtain a quantitative assessment of CTLA4Ig-mediated inhibition of memory CD4 T cell proliferation in vivo, we calculated the percentage of undivided precursors at each division cycle from multiple experiments, according to Lyons (37). There was a significant (p < 0.05, p < 0.01) decrease in HA- and OVA-specific memory CD4 T cells, respectively, undergoing more than two division cycles in the hosts treated with CTLA4Ig in comparison with IgG2a (Fig. 5,B). From this analysis, one can clearly discern the higher magnitude of the OVA-driven proliferative response compared with HA-specific memory CD4 T cells, with an increased proportion of OVA-specific cells undergoing three and four division cycles (Fig. 5,B). CTLA4Ig-mediated effects can be observed as a significant increase in nondividing cells for both HA- and OVA-specific T cells (Fig. 5,B), and striking decreases in the proportion of cells undergoing two or more divisions for HA-specific memory CD4 T cells, and three and four divisions for OVA-specific memory CD4 T cells (Fig. 5 B). These results demonstrate substantial inhibition of the progression of Ag-driven memory CD4 T cell proliferation by CTLA4Ig independent of the magnitude of the recall response.

To determine whether CTLA4Ig-mediated inhibition of memory CD4 T cell proliferation affected the overall yield of boosted memory CD4 T cells, we calculated the absolute numbers of Ag-specific memory CD4 T cells present in CTLA4Ig vs control-treated mice. There was a striking loss of Ag-recalled HA- and OVA-specific memory CD4 T cells in the hosts that received CTLA4Ig compared with IgG2a isotype control expressed either as a percentage of isotype control (Fig. 6,A) or as absolute numbers of Ag-specific memory T cells (Fig. 6,B). For OVA-specific memory CD4 T cells, a dramatic expansion of OVA-specific memory T cells was observed in spleen and lymph nodes after boosting with OVA, and this expansion was dramatically inhibited by CTLA4Ig treatment (Fig. 6,B, right). For HA-specific memory T cells, there was no net expansion upon recall consistent with our previous findings (34), probably due to the lower proliferative capacity of this system. However, there was a significant loss of HA-boosted memory CD4 T cells in CTLA4Ig-treated mice (Fig. 6 B, left). These results demonstrate that CTLA4Ig treatment results in a dramatic decrease in the yield of Ag-activated memory CD4 T cells, consistent in multiple mice studied and in two diverse antigenic systems.

We asked whether the loss of Ag-specific memory CD4 T cells seen in the presence of CTLA4Ig was specific to a particular memory subset, by analyzing CD62L expression of in vivo boosted memory CD4 T cells as a function of CFSE dilution in CTLA4Ig- and control-treated mice. We found that in control-treated mice, all of the OVA-stimulated memory T cells underwent division, and these OVA-stimulated memory CD4 T cells exhibited heterogeneous CD62L expression with 30–40% exhibiting a CD62L^low phenotype (Fig. 7,A, upper left). By contrast, in CTLA4Ig-treated mice, although the majority of OVA-stimulated cells underwent division (although not as extensively as controls; see Fig. 5), these responding OVA-specific CD4 T cells exhibited a predominant CD62L^high (80–90%) phenotype, with a very low proportion of CD62L^low phenotype cells even among cells undergoing extensive division (Fig. 7,A, lower left). We obtained similar results with HA-specific memory CD4 T cells in that the responding (dividing) memory CD4 T cells in control-treated mice exhibited a substantially higher proportion of CD62L^low phenotype cells, compared with memory CD4 T cells activated in the presence of CTLA4Ig (Fig. 7 A, right).

The CD62L profile of Ag-specific memory CD4 T cells from multiple PBS- and Ag-boosted mice is shown in Fig. 7,B. These results show that the CD62L profile of unactivated HA- and OVA-specific memory T cells was not significantly altered in CTLA4Ig-treated mice, and that the CTLA4Ig-induced loss of CD62L^low phenotype memory CD4 T cells was only observed among the Ag-stimulated memory CD4 T cells (Fig. 7,B). Furthermore, the biased CD62L^high profile of Ag-stimulated memory CD4 T cells in CTLA4Ig-treated mice occurred in both spleen and lymph node (Fig. 7 C) with a substantial fraction of CD62L^low phenotype cells in spleen and a predominance of CD62L^low memory CD4 T cells in lymph node of control-treated mice, compared with a preponderance of CD62L^high phenotype memory CD4 T cells in both spleen and lymph node of CTLA4Ig-treated mice. These results indicate a pronounced reduction of CD62L^low T_EM phenotype cells during antigenic recall in the presence of CTLA4Ig.

We demonstrate in this study that inhibition of the CD28/B7 pathway using CTLA4Ig profoundly alters the recall function of memory CD4 T cells generated by priming with Ag or during infection with influenza virus. In vitro, we observed a biased inhibition of IL-2 production by CTLA4Ig from TCR-transgenic or polyclonal influenza-specific memory CD4 T cells. In vivo, although CTLA4Ig does not inhibit early activation events by Ag-stimulated memory CD4 T cells, as shown in two Ag systems, there are striking defects in proliferative expansion in vivo coincident with a biased loss of the CD62L^low T_EM subset in the responding population. Our results identify a novel role for the CD28/B7 pathway as a regulator of memory T cell responses and as a potential target for modulation of memory CD4 T cell expansion and homing capacities.

Our findings challenge the generally accepted view that memory T cells are minimally dependent on CD28/B7 costimulation for function based on in vitro recall studies and in vivo studies in CD28-deficient mouse models (22, 28). In vitro studies established a CD28-independent mechanism for memory T cell activation by showing that B7-deficient APCs could activate memory, but not naive CD4 T cells, using memory CD4 T cells generated by adoptive transfer, as accomplished in this study (5, 23). However, B7 deficiency on APC eliminates both positive signaling through CD28 and negative regulation through CTLA4, which is known to be constitutively up-regulated in memory T cells (46). In our system, we observed a consistent and selective inhibition of IL-2 production in the presence of CTLA4Ig or an anti-CD28 Ab, consistent with previous findings showing that CTLA4Ig-mediated inhibition preferentially blocks the lower avidity CD28/B7 interaction, leaving the higher avidity CTLA4-B7 interaction intact (47).

Results showing unimpaired recall responses by memory CD4 and CD8 T cells in CD28-deficient (CD28^−/−) mice (24, 48) likewise supported a negligible role for CD28 costimulation and memory. However, CD28^−/− mice exhibit generalized immune abnormalities, including a lack of regulatory T cells (49), and naive CD28^−/− CD4 T cells can also be activated in vitro and in vivo (despite the potent ability of CTLA4Ig to down-regulate naive T cell activation (21)), suggesting that additional pathways are operable in CD28^−/− mice (50) that may or may not be operable in wild-type hosts in vivo. The tracking and analysis of an Ag-specific population of memory CD4 T cells in vivo in the presence of CD28/B7 inhibition, as shown in this study, have not been previously reported.

Our results indicate that whereas specific recall functions of memory CD4 T cells, such as up-regulation of activation marker expression, are independent of CD28 costimulation, optimal IL-2 production and proliferative expansion of memory CD4 T cells require CD28 costimulation. It is known that CD28/B7 costimulation promotes naive T cell activation by enhancing IL-2 production (51, 52, 53, 54), and that CTLA4Ig-mediated inhibition of the CD28/B7 pathway in naive T cells blocks progression, but not the induction of proliferation in vitro (19). Our findings, that CD28 costimulation is likewise required for optimal IL-2 production (Figs. 1 and 3) and proliferation (Fig. 5) of memory CD4 T cells, suggest similar control of IL-2 production and cell cycle progression by CD28 in naive and memory CD4 T cells in vivo. The production of TNF-α by naive T cells is also stabilized by CD28 costimulation (51, 55, 56), and we found inhibition of early TNF-α production by memory CD4 T cells in the presence of CTLA4Ig (Fig. 2 C), suggesting that CD28 signals are also required for optimal TNF-α production in memory T cells. This selective regulation of memory T cell cytokine function by CD28 costimulation indicates that multiple pathways may be triggered in memory T cells upon TCR ligation, and that each contributes to the varied functional outcomes of these multifaceted cells.

A striking and unexpected effect of CTLA4Ig treatment on memory T cell recall in vivo was the lack of the CD62L^low T_EM subset among the Ag-responsive memory CD4 T cells. CD62L expression on T_CM and T_EM subsets can be altered by activation, homing, and homeostasis (17, 35, 57, 58), and has been shown to direct memory T cell migration to lymphoid (CD62L^high) or nonlymphoid (CD62L^low) tissue (35, 57, 59, 60). The pronounced loss of the CD62L^low subset that we observed in the Ag-stimulated memory T cells in the presence of CTLA4Ig could be due to a block in activation-induced conversion of T_CM into a T_EM phenotype cell that is known to occur in vivo (17) (data not shown); biased attrition of the T_EM subset activated in the presence of CTLA4Ig; or a combination of both mechanisms. We propose a model whereby memory T cell recall in the presence of adequate CD28 costimulation favors a T_EM-driven response, as T_EM phenotype cells are generated by activation of both resting T_CM and T_EM cells that undergo rapid expansion, resulting in T_EM cells that can readily migrate to peripheral tissues to mediate site-specific recall responses (Fig. 8). When CD28 costimulation is limited, T_EM phenotype cells are not generated due to lack of substantial conversion and expansion of resting T_CM cells, and possible attrition of activated T_EM cells, resulting in a predominant T_CM profile and greatly reduced migration of activated memory CD4 T cells to nonlymphoid compartments, inhibiting peripheral responses (Fig. 8). Our results showing that influenza-primed memory CD4 T cells are susceptible to CTLA4Ig-mediated inhibition suggest that protective response to influenza challenge in vivo may likewise be affected by CTLA4Ig, a possibility that we are currently investigating.

This susceptibility of memory CD4 T cells to CTLA4Ig-mediated inhibition of CD28 costimulation has important clinical implications. Currently, use of human CTLA4Ig (Abatacept) is efficacious in treating ongoing rheumatoid arthritis (26), which is an autoimmune disease likely driven by memory T cells (28). Although CTLA4Ig is considered exclusively to inhibit naive T cell activation (22, 28), our findings together with clinical results strongly imply that CTLA4Ig can be highly effective in inhibiting immune responses driven by memory T cells, by repressing the propagation of pathogenic memory T cells. The prevalent T_CM phenotype of Ag-recalled memory CD4 T cells in the presence of CTLA4Ig suggests a mechanism whereby CTLA4Ig may inhibit the propagation of pathogenic T_EM cells that could potentially mediate pathology at peripheral tissue sites. Our results therefore identify a novel role for the CD28/B7 pathway as a key regulator of memory T cell responses and as a new target for modulation of pathogenic and productive memory CD4 T cell responses.

We extend gratitude to Wendy Lai and Elizabeth Kadavil for mouse colony maintenance; to Laureanne Lorenzo for help with the influenza ELISPOT; to Dr. Daniel Perez for growing influenza virus; and to Dr. Anita Tang for experimental support.

The authors have no financial conflict of interest.