Memory T Cells Are Enriched in Lymph Nodes of Selectin-Ligand–Deficient Mice

As part of immune surveillance, T cells continuously recirculate through lymphoid and nonlymphoid organs via the blood and lymph. The vast majority of T cells enter lymph nodes (LNs) from the blood via high endothelial venules (HEVs), whereas a smaller fraction enters LNs through the afferent lymphatics, which drain the interstitial fluids of organs and MALT. In general, it is thought that naive T cell populations preferentially home to lymphoid organs, whereas memory T cells preferentially home to either bone marrow and lymphoid organs (T central memory), or nonlymphoid organs (T effector memory) (1–3). However, several groups have also observed that naive T cells regularly circulate through nonlymphoid organs such as lung, and alternative models have been proposed to account for these observations (4–7).

Selectin-mediated tethering represents the earliest step in T cell entry to the LNs via the HEVs. For naive and central memory T cells, this step is mediated through the interaction of L-selectin expressed on the T cell, with selectin ligands expressed on HEVs. L-selectin ligands are sialylated, fucosylated, and sulfated structures whose synthesis depends on the activity of several glycosyltransferases, including α1,3 fucosyltransferases, which are absolutely required for the synthesis of sialyl Lewis X (sLe^x) (8, 9). Mice deficient in both fucosyltransferase-IV and -VIII (fucosyltransferase-IV and -VII double knockout [FtDKO]) are profoundly deficient in T cell trafficking to the LNs due to loss of expression of sLe^x by HEVs (10, 11). Additionally, L-selectin ligands are further modified by the activity of two sulfotransferases, and addition of these sulfated sLe^X structures to glycoproteins, such as peripheral LN addressin, glycosylation dependent cell adhesion molecule-1, and mucosal addressin cell adhesion molecule-1, allows naive and central memory T cells to tether on HEVs, initiate chemokine-dependent signaling through CCR7, and initiate firm arrest and transendothelial migration via activated integrin receptors, such as LFA-1 (2, 12, 13).

In contrast to naive T cells, activated T cells downregulate L-selectin, but induce expression of fucosyltransferase-IV or -VII and begin to express selectin ligands on modified T cell glycoproteins, such as P-selectin glycoprotein ligand-1 (14–16). As a consequence, activated T cells are able to interact with E- and P-selectin expressed on inflamed endothelial cells to initiate tethering and slow rolling under shear forces. Additionally, activated T cells increase expression of the adhesive proteoglycan CD44 and several integrins, including VLA-4, that also contribute to slow rolling (17, 18). Moreover, studies have shown that effector/memory T cells are enriched in the afferent lymph and peripheral organs, allowing these T cells to bypass the requirement for selectin-mediated trafficking by entering LNs via the afferent lymph (19, 20).

Recently, we discovered that naive T cells preferentially redistributed and accumulated in nonlymphoid organs, such as the lung in FtDKO mice (21). In the current study, we have further examined FtDKO mice to understand how alterations in T cell trafficking to LNs affect memory T cell localization and homeostasis. In this work, we present our observations that FtDKO mice show a significant increase in the frequency of memory T cells in LNs, although total T cell numbers are reduced in LNs. Ab treatment of wild-type (WT) mice using anti-CD62L (L-selectin) recapitulated the increase in the ratio of memory T cells in LNs. We tested whether Ag-specific CD44^high effector or memory CD8 T cells showed alterations in trafficking to FtDKO LNs following lymphocytic choriomeningitis virus (LCMV) infection. Our results demonstrated that Ag-specific memory, but not effector T cells, were significantly enriched in the LNs of FtDKO mice, even though memory T cells remain severely impaired in homing to LNs of FtDKO mice. Furthermore, although substantial numbers of memory T cells homed to the bone marrow and spleen in both WT and FtDKO mice, T cells transferred into FtDKO mice show significantly increased homeostatic proliferation. We suggest that longer residency times of memory T cells in LNs with few lymphocytes, and increased homeostatic expansion together contribute to the preferential accumulation of Ag-specific memory T cell populations in LNs of FtDKO mice.

FtDKO mice backcrossed at least nine generations to C57BL/6 were originally generated by J.B. Lowe (Genentech, South San Franciso, CA) (10). Mice were obtained from the Consortium for Functional Glycomics (Scripps Research Institute, La Jolla, CA) and originally bred in-house as homozygous double knockouts, but breed poorly, and are currently maintained as heterozygous breeding pairs and genotyped by PCR. Transgenic (Tg) P14 × Thy1.1 mice were provided by R. Ahmed (Emory Vaccine Center, Atlanta, GA) and bred in-house. C57BL/6 (WT) mice were purchased from The Jackson Laboratory. Memory P14 × Thy1.1 chimeric mice were generated by adoptive transfer of 5 × 10⁵ total splenocytes from Tg P14 × Thy1.1 mice into C57BL/6 recipients and infected with 2 × 10⁵ PFU LCMV-Armstrong. FtDKO and C57BL/6 WT control mice were immunized i.p. with 2 × 10⁵ PFU LCMV-Armstrong. All mice were maintained under specific pathogen-free conditions at the University of Tennessee in accordance with Institutional Animal Care and Use Committee guidelines and used at age 2–6 mo.

Where indicated, MACS (Miltenyi Biotec, Auburn, CA)-purified naive and memory P14 × Thy1.1 CD8 T cells (1–2 × 10⁶) from the spleen were adoptively transferred into WT or FtDKO recipient mice (Thy1.2). Naive or memory cells were labeled with CFSE to differentiate transferred naive and memory P14 × Thy1.1 CD8 T cell populations, or to examine homeostatic proliferation. Equal numbers of Ag-specific CD8 T cells were injected i.v. via tail vein into age-matched recipient mice. For anti-CD62L experiments to measure exit rates, equal numbers of naive and memory T cells were adoptively transferred into recipient mice, and 1 d later 100 μg MEL-14 (anti-CD62L) blocking Ab (BioXcell, West Lebanon, NH) was injected i.v. Organs were collected before treatment and 8, 24, and 48 h after treatment, and exit rates were calculated as indicated below.

At indicated time points, mice were sacrificed, organs were perfused with 5 ml cold PBS, and all organs were harvested for processing, as previously described (22, 23). Briefly, for lung and liver, perfused organs were minced, incubated with HBSS with 1.3 mM EDTA for 30 min at 37°C, and resuspended in 225 U/ml type I collagenase for 60 min at 37°C, and cells were centrifuged using a Percoll density gradient to isolate lymphocytes from the parenchyma. Isolated lymphocytes were counted from all organs, and single-cell suspensions were stained with D^b nucleoprotein (NP)396–404 or D^bgp33–41 tetramers, provided by the National Institutes of Health Tetramer Facility (Atlanta, GA), and mAbs (CD8, CD4, CD44, CD62L, Thy1.1, Thy1.2) (24). Total number of cells in bone marrow was calculated by multiplying the number of cells from two femurs × 7.9 (25). mAbs were purchased from BD Pharmingen (La Jolla, CA). All samples were run on a FACSCalibur (BD Biosciences, San Jose, CA) and analyzed with FlowJo software (Tree Star, Ashland, OR).

For transfer studies, to calculate the proportions of transferred cells in different organs, total, naive, or memory Thy1.1⁺ CD8⁺ T cells in nine indicated organs were enumerated. The number of transferred memory cells in each organ was then divided by the memory cell populations in all examined organs as depicted by a pie graph. Statistical significance was determined using unpaired Student t tests where *p < 0.05, **p < 0.01, and ***p < 0.001.

Our analysis is based on the following, simple model of T cell inflow into a LN:

where T(t) is the density of a given CD8⁺ cell type, naive or memory, at time t, λ(t) is the immigration rate of these cells into a given LN, and ε is the egress rate. Assuming that the drug treatment results in λ(t) being reduced to near zero values, Equation 1 implies that T(t) will undergo an exponential decay process, i.e., given a cell density at some time t_i, T(t)_i, the density of cells at time t + τ is simply as follows:

Taking the log of both sides of Equation 2 yields the following:

Thus, we can estimate ε at different time points by fitting a linear regression model to the log-transformed data at times t_i and t_i + τ. Statistical tests for differences in ε as a function cell type, tissue, and time period were evaluated using an F test. Specifically, we set t_i = 0h and τ = 8h for the early treatment and t_i = 24h and τ = 24h for the late treatment. Regression fits to the log-transformed data, and F statistic calculations and comparisons, were done in R 2.11 using the “lm” and “ANOVA” functions, which are part of the R's standard statistical package stats.

We isolated lymphocytes from organs of age-matched naive WT and FtDKO mice. As has been shown previously, total cell numbers in the LNs of FtDKO mice were significantly reduced (10). We observed a significant decrease in CD8 T cell numbers in the LNs of FtDKO mice, with concomitant increases in spleen and blood (Fig. 1B). However, analysis of the activation/memory marker CD44 on T cell populations revealed that whereas frequencies of CD44^high T cells were similar in most organs, in the LNs and O-NALT, the frequency of CD44^high T cells was significantly higher in FtDKO mice (Fig. 1A, 1C). Analysis of the number of CD44^highCD8 T cells in organs revealed that whereas these effector/memory CD8 T cells were generally reduced in the LNs of FtDKO mice, the impairment was not as great as the reduction in total T cells. These data suggested that the impairment of trafficking of T cells to some secondary lymphoid organs (SLOs) in the absence of selectin ligands may not be as severe for effector/memory T cell populations.

We then investigated expansion and trafficking of CD44^high Ag-specific effector CD8 T in LNs following viral infection. WT and FtDKO mice were infected with 2 × 10⁵ PFUs of LCMV-Armstrong i.p., and Ag-specific CD8 T cells were enumerated in blood and SLOs (Fig. 2). As expected, CD8 T cells in WT and FtDKO mice similarly downregulate expression of CD62L by day 8 postinfection (p.i.), although expression levels of CD62L were slightly elevated on naive T cells in FtDKO mice (Fig. 2A). Both WT and FtDKO mice show similarly high frequencies of Ag-specific CD44^highCD8 T cells and total CD44^highCD8 T cells in PBMCs and LNs on day 8 p.i. (Fig. 2B, 2D, 2E). WT and FtDKO mice showed no differences in viral control by plaque assay similar to earlier reports (data not shown). However, we observed a significant reduction in total CD8 T cell numbers in LNs of FtDKO mice compared with WT mice, and as a consequence we also observed significantly reduced Ag-specific CD8 T cells in the LNs (Fig. 2C). Thus, whereas the frequencies of effector CD8 T cell populations were comparable between WT and FtDKO mice, expression of selectin ligands remains critically important in achieving normal numbers of Ag-specific CD44^high effector CD8 T cells in the LNs following viral infection. These data extend the findings of two previous studies using contact hypersensitivity challenge of skin or intranasal mycobacterial infection of FtDKO mice, showing loss of selectin ligands significantly impairs activated T cell migration to LNs (11, 26).

We concluded from our analysis of effector CD44^highCD8 T cells in LNs that despite studies suggesting cooperative interaction of CD44 and VLA-4 to enable slow rolling of activated T cell populations, T cells require expression of fucosyltransferases and synthesis of selectin ligands for activated T cell entry to the LNs, as described in earlier studies (14, 15, 18). In contrast to our findings with Ag-specific effector CD8 T cells at day 8, our analysis of Ag-specific memory CD8 T cell populations in LNs of immune FtDKO mice revealed significant increases in the frequencies of memory cells compared with WT mice at day 60 (Fig. 3A, 3C). Similar to our observations in uninfected FtDKO mice, in day 60 immune mice, the frequency of CD44^highCD8 T cell populations was increased in LNs of FtDKO mice (Fig. 3E). Importantly, despite our continued observation that total CD8 T cells remained substantially reduced in LNs of FtDKO mice, the significantly higher frequency of Ag-specific memory CD8 T cells in the LNs resulted in comparable numbers of Ag-specific memory CD8 T cells in LNs of WT and FtDKO mice (Fig. 3B, 3D). Thus, despite the loss of selectin ligands, Ag-specific memory T cells were similarly abundant in the LNs of FtDKO mice.

There were several possible explanations for our results, demonstrating increased frequencies of memory T cells in the LNs of FtDKO mice. T cell density has been associated with T cell survival, and it was possible that there was increased survival of memory T cells in LNs of FtDKO mice. However, studies that have examined T cell survival and density have found that high T cell density favors survival, whereas low T cell density results in increased apoptosis (27). Therefore, we initially considered two explanations to test experimentally. One possible explanation was that in contrast to naive or effector CD8 T cell populations, Ag-specific memory CD8 T cells were being preferentially retained in LNs of FtDKO mice. A second hypothesis was that memory T cells were preferentially homing to the LNs of FtDKO mice via selectin ligand-independent mechanisms.

To test whether inhibition of T cell entry changed the ratio of naive versus memory T cell populations, we treated normal mice with anti-CD62L. As expected, treatment of WT mice with anti-CD62L resulted in a significant reduction in total CD8 T cell numbers in the LNs, but not spleen (Fig. 4A). However, this reduction in total CD8 T cell numbers was almost entirely due to a decrease in naive CD44^lowCD8 T cells (Fig. 4A). In contrast, the memory/effector CD44^high population was relatively stable, and we observed a significant increase in the percentage of CD44^highCD8 T cells in LNs, similar to our observations in FtDKO mice (Fig. 4). Similar results were observed with the CD4 T cell population (data not shown). Thus, blockade of T cell entry via the L-selectin–mediated HEV-dependent pathway results in the increased frequency of endogenous effector/memory CD44^high T cell populations in LNs similar to the observed increases in FtDKO mice (Fig. 1). This observation was consistent with data presented in an elegant recent study that suggested that memory T cells were retained for longer periods in LNs (28). However, in both studies, T cells were activated by an unknown, endogenous Ag, and we cannot determine what proportion of the CD44^high memory/effector population was proliferating in response to endogenous Ags present in the LNs. Thus, whether memory T cells in the absence of specific Ags show differences in retention in LNs was unclear.

To directly test the hypothesis that memory T cells were retained longer in the LNs compared with naive T cells, we transferred bone fide Thy1.1⁺ naive or memory CD8 T cells with the same TCR specificity into WT Thy1.2⁺ recipient mice. We used the P14 Tg TCR specific for the gp33–41 peptide of LCMV. Both populations of Thy1.1⁺ CD8 T cells were CD62L^high and either CD44^low (naive, Ag inexperienced) or CD44^high (memory, Ag experienced). Following transfer into recipient mice, further T cell entry was blocked with anti-CD62L treatment of mice to measure the decay in transferred T cell populations in different organs over time. We then calculated the exit rate of these T cell populations. This evaluation allowed us to determine whether memory T cells show differences in residency time in LNs compared with naive T cells.

There was one zero cell count for both the memory and naive cell populations. Both zeros occurred in the mediastinal LNs (MedLNs) at 48 h, but in separate mice. Because we were working with log-transformed data, these zero counts were replaced with the lowest nonzero count for the entire dataset (13.2 cells). The best fit model was the one that assumed that egress rates differed between the early, 0- to 8-h, and the late, 24- to 48-h, time periods (F = 21.035, df = 1, p < 10⁻⁴), but where the exit rate was the same for both memory and naive cell types within each time frame (F = 0.0682, df = 2, p > 0.9) for all LNs except for the late iliac LN (ILN) treatment (F = 9.1651, df = 1, p < 0.005). The egress parameter estimates are summarized in Table I.

Our analysis revealed that naive and memory CD8 T cells showed no statistical differences in exit rates from the LNs (Fig. 5). Initially, both naive and memory CD8 T cell populations show residency times of ~5–6 h in LNs, with memory T cells showing a slight trend toward faster exit. However, our data also revealed that as the density of lymphocyte populations decreased in the LNs over the treatment time (due to anti-CD62L blockade of new lymphocyte entry to LNs with continued cell exit), the rate of T cell egress significantly slowed for both naive and memory T cell populations to ~15–16 h in some LNs, to as much as 84 h (3.5 d) in others (Table I). Several studies examining T cell recirculation have suggested that T cell migration through the T cell area of the LNs is slowed when T cell density is reduced, resulting in reduced exit and increased T cell retention (29–31). Although undoubtedly in our dataset the relative contribution from non-HEV–dependent T cell entry increased as the overall HEV-dependent contribution decreased, our data are consistent with this earlier work measuring thoracic duct T cells in irradiated and nude animals, which showed that density of T cells in LNs influences T cell retention (29–31). Furthermore, these data suggested that because endogenous T cell populations are quite low in LNs of FtDKO mice, T cells that enter LNs of FtDKO mice would be retained longer relative to T cells that enter LNs of WT mice because T cell density is up to 95% higher in LNs of WT mice.

Because exit rates of naive and memory T cells were similar in our study, we tested our second hypothesis that memory T cells were selectively homing to the LNs of FtDKO mice. We transferred bone fide Thy1.1⁺ naive or memory CD8 T cells that were specific for the gp33–41 peptide of LCMV into WT or FtDKO Thy1.2⁺ recipient mice. As before, both populations of Thy1.1⁺ CD8 T cells were CD62L^high and either CD44^low (naive, Ag inexperienced) or CD44^high (memory, Ag experienced). Additionally, because these T cells were derived from WT mice, they retained the ability to express fucosyltransferases following activation and make sLe^x-bearing selectin ligands, and could therefore potentially interact with E- or P-selectin expressed on HEVs of both WT and FtDKO mice. Contrary to our hypothesis, transfer of memory CD8 T cells into FtDKO mice also revealed a profound impairment in homing to LNs (Fig. 6). Whereas in both WT and FtDKO mice the majority of transferred memory T cells migrated to the bone marrow and spleen, the remaining memory CD8 T cells preferentially homed to the LNs in WT mice with very few cells homing to the LNs in FtDKO mice (51% WT versus 2% FtDKO). In contrast, in the FtDKO mice, the memory T cells that did not home to bone marrow or spleen preferentially homed to the lung and liver (89% FtDKO).

These data strongly argue that overall, memory T cells remain significantly impaired in homing to the LNs of FtDKO mice compared with WT mice (Fig. 6B, 6C). Furthermore, they suggest memory T cells also require selectin ligand expression by HEVs for efficient homing to LNs. However, despite significant reductions in the total number of T cells that homed to the LNs of FtDKO mice, our analysis also revealed that comparing memory versus naive T cell populations, the small population of transferred T cells that did migrate to LNs in FtDKO mice was more likely to be memory T cells compared with WT mice (Fig. 6E). Specifically, in WT mice we observed a ∼3:1 preferential homing of naive to memory T cells in LNs, whereas in FtDKO mice we observed no preferential homing of naive T cells (Fig. 6D, 6E). In fact, the frequency of memory T cells was slightly increased in the FtDKO mice (Fig. 6E). Stated another way, these data suggest that entry of T cells into the LNs via the non-HEV–dependent pathway (i.e., the afferent lymph) may be similar for naive and memory T cells, whereas HEV–dependent entry strongly favors naive T cells, even when memory T cells are CD62L^high.

Our data indicated that memory T cells were not preferentially retained in LNs compared with naive T cells, and also showed memory T cells did not preferentially migrate to LNs of FtDKO mice. However, despite significant impairment in migration of both naive and memory T cell populations to LNs, the ratio of transferred memory T cells was higher in FtDKO mice compared with WT mice. Because T cell populations are substantially reduced in FtDKO LNs, our studies examining T cell exit suggested that the few T cells that entered the low T cell density LNs of FtDKO mice would be retained longer at these sites. We therefore proposed an alternate hypothesis that memory T cell populations that entered and exited LNs of FtDKO mice would show increased homeostatic proliferation, and memory T cells would preferentially accumulate in LNs over time.

To test this, we labeled P14 × Thy1.1⁺ memory CD8 T cells with CFSE and transferred them into WT and FtDKO mice. Three weeks later, we isolated organs from recipient mice and assessed proliferation by CFSE dilution. As shown in Fig. 7, memory T cells transferred into FtDKO mice show greater homeostatic proliferation compared with memory T cells transferred into WT mice (Fig. 7A). Moreover, we also observed that significantly more memory T cells were found in LNs of FtDKO mice at day 21 compared with day 1. As before, the majority of transferred memory T cells were located in bone marrow and spleen of WT and FtDKO recipient mice (∼88%). Of the remaining transferred cells, a greater fraction of memory T cells was found in the LNs of FtDKO mice at day 21 compared with day 1 after transfer (Fig. 7B). At day 1 after transfer into recipient mice, only 2% of transferred memory T cells were found in LNs of FtDKO mice, but by day 21 the percentage of memory T cells observed in LNs increased >15-fold, to 35% (Figs. 6B, 7B). Taken together, these data suggest that memory T cells are enriched and remain abundant in the LNs of FtDKO mice, despite severely impaired trafficking to LNs, as a consequence of several mechanisms of T cell homeostatic maintenance, including a higher ratio of memory T cells entering the LNs when HEV-dependent entry is inhibited, longer retention of T cells in low cell density LNs, and increased homeostatic proliferation (Fig. 8).

The generation of effective immune responses is critically dependent on encounters between recirculating T cells and their specific Ags, which primarily occur in secondary lymphoid organs. One of the defining characteristics of the immune system is the constant movement of constituent cell populations, including T cells and dendritic cells through SLOs, where a highly efficient filtration and surveillance system are optimized for the capture of Ags and appropriate Ag presentation (32–34). Furthermore, in contrast to naive T cells, memory T cells respond more rapidly to specific Ag, based on both quantitative increases in their starting population and qualitative differences in effector responses, including rapid cytokine secretion (35).

In this study, we present evidence that FtDKO mice, which are incapable of synthesizing sLe^x-bearing selectin ligands, show significant enrichment of endogenous CD44^high effector/memory T cell populations in LNs and MALTs, such as organized nasal-associated lymphoid tissue (Fig. 1). When we infected FtDKO mice with LCMV to track Ag-specific CD44^highCD8 T cell responses in LNs, we show similar responses in blood and spleen of mice at day 8, but Ag-specific CD44^highCD8 T cells were reduced in LNs (Fig. 2). Thus, although Ag-specific CD8 T cells increased significantly in LNs of FtDKO mice at day 8 compared with naive FtDKO mice, selectin ligands were required for efficient Ag-specific CD8 T cell trafficking and expansion in LNs. To our knowledge, these data show for the first time that Ag-specific T cell responses remain compromised in LNs following viral infection of these mice, even though these mice show similar kinetics of viral clearance (36). However, our study does not rule out that there are potentially additional defects in priming or proliferation of T cells in LNs of these mice.

In contrast to our observations at day 8 following viral infection, we observed that at day 60 p.i., FtDKO mice demonstrated abundant Ag-specific CD8 T cell responses both in spleen and LNs (Fig. 3). Whereas total CD8 T cells remained substantially reduced in LNs of FtDKO mice, the higher frequencies of Ag-specific memory CD8 T cells in the LNs compensated for the overall low T cell numbers and particularly low CD44^low naive T cell population (Fig. 3). These data suggest that selectin ligands are not required for efficient localization of Ag-specific memory CD8 T cells in LNs of immune mice.

We can recapitulate a similar increase in the percentage of CD44^high T cells in LNs by treating WT mice with anti-CD62L, which substantially reduced LN T cell numbers by blocking entry of new lymphocytes (Fig. 4). Our analysis of T cell exit rates from LNs revealed no significant differences comparing naive and memory T cells (Fig. 5). Whereas these data contrast with a recent study examining in vivo CD44^high endogenous CD4 T cell populations using Kaede-Tg mice, several differences in the two studies must be considered (28). First, we examined memory T cells of known Ag specificity, while the Tomura et al. (28) study examined endogenous CD44^high T cells. When we examined endogenous CD44^high T cell populations, we also observed stable numbers of this population, whereas naive CD44^low T cell numbers plummeted (Fig. 4). In our experiment, this could be due to continued entry of CD62L^lowCD44^high T cells, slower exit of the CD44^high population, or conversion of CD44^low to CD44^high T cells. However, because we cannot determine whether these CD44^high cells have recently encountered Ag, we cannot conclude that activated/memory T cells not exposed to specific Ags were retained for longer periods based on these data. We therefore examined bona fide Ag-specific memory T cells, in which analysis of egress experiments suggested that naive T cells do not exit faster than memory T cells (Fig. 5). Also, whereas our analysis focused on CD8 T cells, the Tomura et al. (28) study focused on CD4 T cells, which interact with a limited population of APCs expressing MHC class II. Because memory CD4 T cells also show differences in survival and maintenance compared with memory CD8 T cells, it is possible that memory CD4 T cells are indeed retained longer in LNs, as this study concluded, whereas memory CD8 T cell retention is regulated differently (28, 37).

Importantly, our study revealed that T cell density influences T cell exit rates from the LNs. Our data suggested that exit rates slowed as T cell density decreased (Fig. 5). A previous study suggested that T cells that enter LNs with many cells exit faster (i.e., shorter residence time), whereas T cells that enter LNs with fewer cells exit more slowly (i.e., longer residency times), and our study examining Ag-specific naive and memory T cells is consistent with this study (31). Furthermore, when we examined memory T cell trafficking into the LNs of FtDKO mice, we discovered that both naive and memory T cells were significantly impaired in homing to the LNs of FtDKO mice following adoptive transfer (Fig. 6). Thus, the enrichment of memory T cells in the LNs of immune FtDKO mice is unlikely to be due to significantly increased selectin-independent entry of memory T cell populations into LNs.

The distribution of memory T cells in different organs will be different at different time points because T cell migration is a very dynamic process dependent on the expression of several adhesion molecules and chemotactic responses (38, 39). Our experiments examining day 1 posttransfer showed that for both WT and FtDKO mice, a significant fraction of memory T cells homed to the bone marrow and spleen (Fig. 6). These results are consistent with several studies that have demonstrated that memory T cells preferentially reside in the bone marrow where they undergo homeostatic proliferation or proliferative renewal due to IL-15 maintenance signals (6, 25, 40, 41). In contrast to WT mice, in which the majority of the remaining CD62L^high memory T cells homed to the LNs, most of the transferred memory T cells were located in nonlymphoid organs in FtDKO mice (Fig. 6B). We have recently demonstrated that naive T cells also preferentially redistribute to nonlymphoid organs such as lung in FtDKO mice (21). We conclude from these studies that naive, effector, and memory CD8 T cell populations all show dependence on the expression of selectin ligands for homing to LNs. However, we also observed that frequencies of effector/memory (CD44^high) T cells were consistently enriched in FtDKO mice, except at day 8, when both WT and FtDKO showed similarly high proportions of CD44^highCD8 T cells of 60–90% (Fig. 2).

Even though entry of memory T cells into LNs was significantly impaired, the few T cells that entered LNs of FtDKO mice were more likely to be memory T cells compared with LNs of WT mice, in which naive T cells preferentially homed (Fig. 6). Because these memory T cells entered LNs with very low T cell density in FtDKO mice, we considered that longer retention in these organs could trigger greater homeostatic proliferation, and our observations confirm this hypothesis (Fig. 7). In lymphoid organs, dendritic cells and stromal cells, such as fibroblastic reticular cells, form a complex three-dimensional meshwork through which T cells and other lymphocytes must travel (33). T cells that enter the depleted LNs of FtDKO mice could undergo lymphopenia-induced proliferation in the T cell zone if retained for a sufficient length of time at these locations. However, Ag-independent T cell proliferation occurs more slowly than Ag-driven proliferation, and we were uncertain whether the retention times in LNs would be sufficient for lymphopenia-induced proliferation to occur. Additionally, most memory T cells were located in bone marrow and spleen, major sites of memory T cell proliferation, and we observed no defects in migration to these locations following transfer into FtDKO mice (25). Nonetheless, our data clearly demonstrated that memory T cells divided more when transferred into fucosyltransferase-deficient mice (Fig. 7). Thus, we suggest that the enrichment of memory CD8 T cells in LNs of FtDKO mice compared with WT mice is due to several homeostatic maintenance mechanisms, as shown in our proposed model (Fig. 8). These mechanisms include a higher ratio of memory T cells that enter LNs via non-HEV–dependent pathways (afferent lymphatics) compared with HEV-dependent entry, which favors naive T cells, a significantly slower exit rate of T cells from LNs with low T cell density, and increased homeostatic proliferation of T cells. Together, these mechanisms enable abundant populations of memory T cells to reside in LNs and mucosal-associated tissues of mice lacking a critical glycosylation-dependent determinant of cell adhesion molecules, even though migration of naive, effector, and memory T cells to these critical secondary lymphoid organs is severely compromised.

In summary, our studies suggest that memory T cell homeostasis is strongly influenced by T cell density in LNs. We observed that despite the presence of significant numbers of memory T cells in the spleen and bone marrow of FtDKO mice, the severe deficit of T cells in LNs triggered greater proliferation of memory T cells transferred into FtDKO mice. Ultimately, this led to an accumulation and enrichment of memory T cells in the LNs of FtDKO mice, in which we observed significantly higher frequencies of memory T cells in LNs over several weeks. Additionally, we also observed that LCMV-immune FtDKO mice at day 60 showed an enrichment of Ag-specific memory T cells in LNs, despite significantly fewer Ag-specific T cells in LNs at the peak of expansion (Fig. 3). This occurred despite continual depressed numbers of T cell populations in the LNs. Taken together, these results suggest that compensatory mechanisms may exist to specifically maintain memory T cell populations at high ratios in the LNs following antigenic exposure (Fig. 8). How compensatory mechanisms are instigated, such as longer T cell retention, remains undetermined. Several studies have demonstrated that elevated common γ-chain cytokines and more self-pMHC availability in lymphopenic hosts are critical homeostatic signals that regulate naive T cell proliferation, whereas IL-7 and IL-15 are critical for memory T cell homeostasis (42). However, LN homing also plays a key role in homeostasis, and receptors that are involved in lymphocyte egress from the LNs, such as the sphingosine 1-phosphate receptor 1 (S1P₁), may be influenced by receptors that sense T cell density. In this regard, recent work has shown that CD69, a type II transmembrane protein that antagonizes S1P₁ expression, also interacts to form a complex with S1P₁ that promotes high-affinity ligand binding of S1P (43). Interestingly, CD69 is a C-type lectin and has been shown to bind fucoidan, a high fucose- and sulfate-containing polysaccharide previously shown to bind selectins (44). Potentially, the reduction of fucose-containing ligands in FtDKO LNs may alter CD69–S1P₁ interactions, slowing T cell egress from the LNs.

We thank Junwei Zeng, Pratima Krishna Suvas, and Heather Dech for technical assistance, as well as Scott Mueller, Mark Sangster, and Russell Zaretzki for helpful discussions.

Disclosures The authors have no financial conflicts of interest.