CD73 and Ly-6A/E Distinguish In Vivo Primed but Uncommitted Mouse CD4 T Cells from Type 1 or Type 2 Effector Cells1

Primed CD4 T cells may develop into effector T cells such as Th1 and Th2, or remain uncommitted as Th primed precursor (Thpp) cells that can subsequently differentiate into Th1 and Th2 cells. Although mouse Thpp-like cells have also been identified among spleen and particularly lymph node cells, further characterization of these cells has been difficult without a defining cell surface marker. Using Affymetrix GeneChips followed by FACS analysis, we found that in vitro-derived Thpp cells expressed CD73 but not Ly-6A/E, whereas Th1 and Th2 cells showed the reciprocal pattern. CD73+ Ly6A/E− memory CD4 T cells were identified in normal C57BL/6 mice, and the proportion of these cells was highest in lymph nodes, lower in spleens, and lowest in the lungs. These cells produced IL-2 and MIP-1α, but much less IL-4 and IFN-γ than CD73− Ly6A/E+ cells. Similar results were obtained with additional Ly-6.2 mouse strains, but not Ly-6.1 strains. Restimulation of Thpp-like CD73+ Ly-6A/E− cells in Th1- or Th2-polarizing conditions induced differentiation into populations producing mainly IFN-γ or mainly IL-4, respectively. In contrast, the effector-like CD73− Ly-6A/E+ population was more committed, and continued to produce both IL-4 and IFN-γ in both conditions. CD73 and Ly-6A/E expression therefore identify a population of Thpp-like cells in C57BL/6 mice and at least some other Ly-6.2 mice.

Similar cells exist in vivo. Mouse CD4 T cells in lymphoid organs express a higher ratio of IL-2 to IFN-␥ than cells in peripheral tissues (5). A Thpp-like population of mouse CD44 high CD4 ϩ Ag-specific T cells was identified by cytokine production and split cloning in lymph nodes: these cells produced IL-2 but not IFN-␥ or IL-4; and individual cells could subsequently become either Th1 or Th2 cells (6). These cells were enriched in lymph nodes relative to spleen (X. Wang and T. R. Mosmann, unpublished observations) and in vitro-derived Thpp cells homed preferentially to lymph nodes rather than spleen, lungs, or peritoneal cavity (7).
It is critical for the immune system to choose the right effector functions for particular infections, otherwise the reciprocal inhibi-tion of Th1 and Th2 responses (8) may hinder subsequent induction of the correct response. Thus the uncommitted Thpp cells might represent an expanded pool of memory cells with the flexibility to become either Th1 or Th2 cells later in the same or subsequent responses. The chemokines produced by Thpp cells may attract other immune cells into the site of a Thpp reaction, without biasing the type of response.
Similar uncommitted CD4 T cells may exist among human memory (CD45RA Ϫ ) T cells. Central memory T cells expressed higher levels of CCR7 (which should enhance lymph node homing) and produced mainly IL-2 on activation (9). The effector memory cell population expressed lower levels of CCR7 and produced IL-2, IFN-␥, IL-4, and IL-5. Central memory T cell populations contained precursors of cells that could differentiate into effector cells that produce IL-4, IL-5, or IFN-␥. Although in vitroderived mouse Thpp cells selectively express CCR7 (4), CCR7 expression is not always associated with nonpolarized memory CD4 T cells (10,11).
Thus the further characterization of the mouse and human Thpp/ central memory T cell subsets has been held back by the lack of well-defined surface markers. During the analysis of T cell gene expression (4) we found that in vitro-derived Thpp expressed CD73 but not Ly-6A/E, in contrast to the reciprocal pattern expressed by Th1 and Th2 effector subtypes. These two markers also define a Thpp-like subpopulation of in vivo memory CD4 T cells: CD73 ϩ Ly-6A/E Ϫ cells are found preferentially in lymph nodes, secrete IL-2 and MIP-1␣ but not IL-4 or IFN-␥, and show flexibility of differentiation. Although the Ly-6A/E marker may be useful only in Ly-6.2 mouse strains, nevertheless the combination of CD73 and Ly-6A/E should have considerable utility for ex vivo isolation and further analysis of the primed, uncommitted CD4 T cell subset.

Mice
Female C57BL/6, C57L, PL, BALB/c, and C3H mice, 6-to 8-wk-old, were obtained from The Jackson Laboratory. All animal experiments were approved by the University of Rochester University Committee on Animal Resources.

T cell purification and differentiation
Thpp, Th1, and Th2 cell populations were prepared and characterized as previously described (4). Briefly, C57BL/6 CD4 T cells were enriched by magnetic bead depletion, and naive cells (CD44 low , CD4 high , CD62L high ) were purified by cell sorting. These cells were stimulated for 6 days by alloantigenic TA3 cells in Th1 conditions (IL-2 plus IL-12), Th2 conditions (IL-2 plus IL-4), or Thpp conditions (IL-2 plus TGF-␤ with anti-IFN-␥ Abs). Cytokine and chemokine levels in the 24-h supernatants were measured before the cells were used for other assays.

In vitro CFSE labeling
Naive cells (CD44 low , CD4 high , CD62L high ) were labeled with CFSE (Molecular Probes) by incubation with 2 M CFSE in PBS containing 5% FBS for 5 min at room temperature. Cells were then washed twice with PBS.

Isolation and analysis of lymphocytes from mouse lymphoid and nonlymphoid (lung) tissues
C57BL/6 mice were sacrificed, and after cardiac perfusion with PBS, lymphocytes were prepared from lymph nodes (a pool of cervical, axillary, brachial, inguinal, and popliteal nodes), spleen, and lungs. Single-cell suspensions of the tissues were prepared by passage through 40-m nylon mesh, followed by depleting erythrocytes by Fico/Lite-LM (Atlanta Biologicals) for lymph node and spleen, or Histopaque 1083 (Sigma-Aldrich) for lung. Lymphocytes at the interface were collected and washed before staining with PerCp anti-CD4, allophycocyanin anti-CD44, FITC anti-Ly-6A/E, and PE anti-CD73. The stained cells were analyzed by flow cytometry to obtain the frequency of CD73 ϩ Ly-6A/E Ϫ and CD73 Ϫ Ly-6A/E ϩ memory CD4 cells.

Immunization and T cell subset analysis
Mice were immunized i.p. with 100 g of keyhole limpet hemocyanin (KLH, endotoxin-free; Calbiochem) in 100 l of PBS. After 14 days, cells from both spleen and lymph nodes (cervical, axillary, brachial, inguinal, and popliteal) were enriched for CD4 T cells by depleting CD8 and B cells on MACS columns. After staining with PerCp anti-CD4, allophycocyanin anti-CD44, FITC anti-Ly-6A/E, and PE anti-CD73, these cells were analyzed by flow cytometry, and subpopulations were sorted as indicated in individual experiments. Differentiation of sorted populations was conducted as described for naive T cells.

T cell subset gene expression evaluation by GeneChip
Gene expression analysis of Thpp, Th1, and Th2 cells has been previously described (4). Briefly, mRNA was extracted from stimulated or unstimulated naive, Thpp, Th1, and Th2 populations, and these mRNA samples were analyzed on Affymetrix mouse U74 GeneChips. The Affymetrix Microarray Suite software was used to assess significant expression levels.

In vitro-derived Thpp cells express CD73 mRNA, whereas Th1 and Th2 cells express Ly-6A/E mRNA
In a previous Affymetrix GeneChip analysis of cytokine and chemokine expression in activated and resting Thpp, naive, Th1, and Th2 cells (4), we also examined these data sets for selective expression of cell surface proteins, to identify markers that might be used to isolate Thpp cells from mixed in vivo populations.
Two genes encoding cell surface proteins were selectively expressed in Thpp vs naive, Th1, or Th2 cells. CD73, a GPI-linked surface protein with ecto-5Ј-nucleotidase activity (13), was significantly ( p Ͻ 0.05) expressed by Thpp cells (Fig. 1), but not naive, Th1, or Th2 cells. Conversely, Ly-6A/E was strongly expressed in both Th1 and Th2 cells as expected (14), but not in Thpp cells (Fig.  1). As controls for the fidelity of the Genechip analysis, two additional genes are shown. The TCR subunit CD3␦ was expressed in all nine T cell samples and T1/ST2L (15) was expressed only in resting and activated Th2 cells (Fig. 1).

In vitro-derived Thpp cells express higher levels of CD73, and lower levels of Ly-6A/E proteins
To test whether these mRNA differences resulted in selective protein expression, we examined surface expression on in vitro-derived Thpp, Th1, and Th2 cells by flow cytometry (Fig. 2). Unstimulated Thpp cells expressed high levels of CD73, but very low levels of Ly-6A/E. In contrast, Th1 and Th2 cells expressed low levels of CD73, but uniformly high levels of Ly-6A/E. Activation with anti-CD3 Abs did not alter the expression of CD73 on any of the three T cell populations, but induced slightly higher levels of Ly-6A/E on Thpp cells. The FACS analysis therefore confirmed that in vitro Thpp cells express higher CD73 and lower Ly-6A/E levels than Th1 or Th2 effector cells.
The kinetics of the differentiation of naive cells in Thpp-, Th1-, or Th2-promoting conditions were analyzed by labeling sorted naive (CD4 ϩ CD44 low CD62L high ) cells with CFSE, stimulating with alloantigen and analyzing cell surface marker expression at different time points. Results were gated on the CD4 ϩ CFSE low 7-ami-noactinomycin D Ϫ population to analyze only live cells that had divided in response to the alloantigen. Expression of CD73 was limited to cells differentiating to the Thpp phenotype (Fig. 2B). From 5 days onward, the majority of the cells in Thpp conditions expressed CD73, whereas the frequency of CD73-expressing cells in Th1 or Th2 cultures remained low at all time points.
Ly6A/E was expressed at moderate levels under all three differentiation conditions, consistent with previous data showing that Ly-6A/E is induced on activation of T cells. However, only in the Th1 and Th2 cultures did the expression of Ly-6A/E increase to high levels (Fig. 2B). Almost all of the proliferating cells in the Th1 cultures expressed high Ly-6A/E levels by day 4, the earliest point at which proliferating cells were detected, whereas Th2 cultures showed similarly high levels of expression only after 9 days.

CD73 and Ly-6A/E expression identify primed, non-Th1, non-Th2 cells in vivo in C57BL/6 mice
To test whether CD73 ϩ Ly-6A/E Ϫ cells expressed the cytokine pattern of in vitro-defined Thpp cells, Thpp-like (CD4 ϩ , CD44 ϩ , CD73 ϩ , Ly-6A/E Ϫ ) and effector-like (CD4 ϩ , CD44 ϩ , CD73 Ϫ , Ly-6A/E ϩ ) cell populations were purified by cell sorting from C57BL/6 mice that had been immunized for 2 wk with KLH. After restimulation with KLH and allophycocyanin, Thpp-like (CD73 ϩ , Ly-6A/E Ϫ ) T cells from both spleen and lymph nodes contained a high frequency of KLH-specific cells that produced IL-2 and/or MIP-1␣, but not IL-4 or IFN-␥ (Fig. 4). In contrast, substantial numbers of the effector-like cells (CD73 Ϫ Ly-6A/E ϩ ) secreted IL-4 and IFN-␥, in addition to IL-2 and MIP-1␣. CD4 ϩ CD44 ϩ memory T cells of C57BL/6 mice also contained low numbers of CD73 Ϫ Ly-6A/E Ϫ cells, which may have been contaminating naive cells because the separation of naive and memory cells by CD44 expression is not complete. More substantial numbers of CD73 ϩ Ly-6A/E ϩ cells were present. When analyzed by cell sorting followed by ELISPOT assay, these cells were more similar to the CD73 Ϫ Ly-6A/E ϩ effector cells than Thpp-like cells, as they produced IL-2, IL-4, IFN-␥, and MIP-1␣ (data not shown). Although we cannot exclude the possibility that there could be additional Thpp-like cells in the double-positive population, the in vivo CD73 ϩ Ly-6A/E Ϫ population expressed both the cell surface markers and cytokine patterns of the in vitro-defined Thpp cells.
These in vivo Thpp-like cells are present in a range of conditions. In addition to the cytokine patterns produced after KLH immunization, we also tested the cytokines produced by these two populations in response to alloantigens, which should stimulate a representative sampling of memory CD4 T cell responses, and anti-CD3 plus anti-CD28, to stimulate all T cells. Analyzing populations from both KLH-primed or unprimed mice, these two additional stimuli also induced mainly IL-2 and MIP-1␣ responses from CD73 ϩ Ly-6A/E Ϫ cells, but higher levels of the effector cytokines IFN-␥ and IL-4 from CD73 Ϫ Ly-6A/E ϩ cells (data not shown). Also, following administration of Ag in CFA, the CD73 ϩ Ly-6A/E Ϫ cells still produced a substantially higher ratio of IL-2 to IFN-␥ than the reciprocal CD73 Ϫ Ly-6A/E ϩ population (data not shown).

Relationship to other cell surface markers
We compared CD62L expression on the CD73 ϩ Ly-6A/E ϩ and CD73 Ϫ Ly-6A/E Ϫ cell populations, and found that CD62L expression was heterogeneous on both populations (data not shown), in agreement with previous results (6,7). We also found that CCR7 mRNA was expressed by both populations, and that there was no striking difference in expression between Thpp-like (CD73 ϩ Ly-6A/E Ϫ ) and effector-like (CD73 Ϫ Ly-6A/E ϩ ) CD4 memory cells (data not shown). This is consistent with reports (10, 11) that CCR7 expression alone does not distinguish uncommitted (nonpolarized) CD4 memory cells from effector cells.

cells of C57BL/6 mice contain uncommitted precursor cells
One of the important characteristics of in vitro Thpp cells is that they can further differentiate into either Th1 or Th2 cells. To test the differentiation flexibility of the in vivo populations, we stimulated sorted cell populations with irradiated TA3 cells (alloantigen) in the presence of IL-12 (Th1 conditions) or IL-4 (Th2 conditions). Six days later, the T cells were washed and restimulated with anti-CD3 plus anti-CD28 Abs and tested for cytokine and chemokine secretion. Thpp-like CD73 ϩ Ly-6A/E Ϫ cells, from either spleen or lymph nodes, showed flexibility in their differentiation patterns: they produced more IFN-␥ or more IL-4 under Th1 or Th2 conditions, respectively (Fig. 5). In contrast, the spleen and lymph node CD73 Ϫ Ly-6A/E ϩ effector-like populations produced  both IL-4 and IFN-␥ before in vitro differentiation, and the ratios of IFN-␥ to IL-4 were not substantially changed after culture in either Th1-or Th2-polarizing conditions. This suggests that the CD73 ϩ Ly-6A/E Ϫ cell population comprised mainly uncommitted precursor cells, whereas the CD73 Ϫ Ly-6A/E ϩ population contained many committed effector cells. Thus the CD73 ϩ Ly-6A/E Ϫ population isolated from C57BL/6 mice were in vivo Thpp-like cells based on their preferential localization in lymph nodes, production of IL-2 and MIP-1␣ but not IL-4 or IFN-␥, and flexible in vitro differentiation.

CD73 and Ly-6A/E also identify Thpp-like cells in C57L and PL mice, but not in BALB/c mice
In Ly-6.2 (Ly-6A.2) mouse strains, Ly-6A/E is expressed on a subpopulation of resting memory T cells, as well as activated T cells (16), whereas Ly-6.1 (Ly-6E.1) mouse strains express Ly-6A/E mainly on activated, not resting T cells. We therefore tested whether the CD73 ϩ Ly-6A/E Ϫ population could be identified in additional mouse strains.
Subpopulations of CD73 ϩ Ly-6A/E Ϫ CD4 ϩ CD44 high cells were present in additional Ly-6.2 strains (C57L, PL, C58, SJL, and AKR), particularly in lymph nodes (Fig. 6A and data not shown). Sorted CD73 ϩ Ly-6A/E Ϫ spleen or lymph node cells from C57L and PL mice produced high levels of IL-2 and MIP-1␣, but low levels of IL-4 or IFN-␥ (Fig. 6B). However, CD4 T cells from Ly-6.1 mice (BALB/c and C3H/He) expressed low levels of Ly-6A/E (Fig. 6A and data not shown). Sorting BALB/c lymph node cells using CD73 alone did not allow separation into Thpp-like and effector populations (data not shown), even though Thpp-like cells are present in BALB/c mice (3,6). Thus reciprocal expression of FIGURE 6. CD73 and Ly-6A/E expression identify Thpp-like cells in Ly-6.2 but not Ly-6.1 mouse strains. A, Lymph node and spleen cell populations from C57BL/6, C57L, PL, and BALB/c mice were stained and analyzed as Fig. 3A. B, Thpp-like (CD73 ϩ Ly-6A/E Ϫ ) and effector-like cells (CD73 Ϫ Ly-6A/E ϩ ) from C57L and PL lymph nodes were isolated and analyzed as described in Fig. 4. These results are representative of three experiments. CD73 and Ly-6A/E resolves a population of resting Thpp from effector memory cells in some but not all mouse strains.

Discussion
CD73 is a 69-kd GPI-linked protein with four N-terminal glycosylation sites (13). CD73 functions as an ecto-5Ј-nucleotidase, which converts extracellular pyrimidine and particularly purine monophosphates to the corresponding nucleosides, which can pass through the plasma membrane and are thus available for cell metabolism. Dephosphorylation of AMP by CD73 produces adenosine, which binds to and activates adenosine receptors (17). There are also suggestions that CD73 could costimulate T cell activation (18), or mediate lymphocyte adhesion to cultured endothelial cells (19).
By expressing CD73, Thpp cells might contribute to the extracellular pool of adenosine, which regulates a number of physiological processes including inflammation. Adenosine inhibited IL-12 production but increased IL-10 production by dendritic cells and macrophages and selectively decreased some cytotoxic functions of neutrophils (20). A protective role for adenosine in Con Aor exotoxin-induced inflammatory liver damage was indicated by treatment with adenosine agonists and antagonists, or in mice deficient for A2aR, the major adenosine receptor expressed on lymphocytes, dendritic cells, and neutrophils (21). Although Thpp cells may induce a mild inflammatory response by secreting chemokines (4), at least in some mouse strains this response may be dampened by the concomitant action of adenosine produced by CD73 on the Thpp cell surface.
Ly-6A/E (Sca-1) is a small (10 -18 kD) GPI-linked surface protein, mainly expressed on hemopoietic stem cells, subsets of lymphocytes (T, B, NK) and some nonlymphoid tissues (16). Ly-6E.1 and Ly-6A.2 are two alleles of the same gene that differ by only one amino acid in the mature proteins, but more significantly, these two alleles show striking differences in their patterns of expression (22). In Ly-6.1 (Ly-6E.1) mouse strains, such as BALB/c, C3H, CBA, and DBA/1, New Zealand B, and New Zealand A, few resting peripheral lymphocytes express high levels of Ly-6A/E; whereas in Ly-6.2 (Ly-6A.2) mouse strains, including C57BL/6, C57L, C58, PL, SJL, DBA/2, SWR, AKR, 129, and C57BR, Ly-6A/E molecules are expressed on substantial numbers of resting lymphocytes (Ͼ5-fold higher than Ly-6.1 strains). The nature of this regulation may be related to the extra requirement for TNF signaling to induce Ly-6A/E expression on Ly-6.1 strains (23). Thus, even though Thpp-like cells are present in BALB/c mice (3,6), we found that Ly-6A/E could not be used as a marker to separate Thpp from effector memory cells in Ly-6.1 mouse strains. In Ly-6.2 mice, the Ly-6A.2 marker distinguishes uncommitted from effector T cells in resting T cell populations, but both Ly-6.1 and Ly-6.2 strains express higher Ly-6A/E levels on all CD4 T cells after activation (16) (Fig. 2).
Ly-6A/E may play roles in the immune responses of T cells (24) and B cells (25) but the major function remains unknown. In T cells, anti-Ly-6A/E induces an intracellular Ca 2ϩ flux and enhances IL-2 secretion (26). This effect requires expression of the TCR-CD3 complex, but anti-Ly-6A/E can also inhibit anti-CD3⑀induced IL-2 production (27), possibly due to differences in the extent of cross-linking or the expression levels of Ly-6A/E. The lack of expression of Ly-6A/E on all subtypes of resting T cells in Ly-6.1 mice suggests that this molecule does not have an important role in resting T cells, and so the significance of the selective lack of Ly-6A/E on Thpp of Ly-6.2 strains is not clear.
The in vivo Thpp-like CD73 ϩ Ly-6A/E Ϫ population may be similar or identical with the CD4 "central memory" T cells described in human PBMC populations (9, 28). Both cell types pro-duce IL-2 but not IL-4 or IFN-␥, both are found preferentially in lymph nodes rather than nonlymphoid tissue, and both show plasticity of differentiation. It should be noted, however, that the term central memory cells has also been applied to cells with lymphoid homing patterns, even if they show committed cytokine patterns (29). Both mouse and human Thpp-like (i.e., IL-2-secreting, uncommitted) cells are a substantial part of the response to immunization with protein Ags, with or without adjuvants such as alum or Ribi adjuvant (6,9)  Independently from the functional roles of CD73 and Ly-6A/E expression on different T cell subsets, these two markers provide valuable tools for identifying and enriching Thpp cells from normal Ly-6.2 lymphocyte populations. For example, we have used these markers to confirm that Thpp-like CD73 ϩ Ly-6A/E Ϫ cells are mainly localized in lymphoid tissue rather than peripheral tissue (lung). The ability to separate Thpp from Th1-and Th2-like effectors should greatly aid the analysis of the induction, distribution and roles of the uncommitted, nonpolarized Thpp cells during in vivo immune responses.