PMCCPMCCPMCC

Search tips
Search criteria 

Advanced

 
Logo of nihpaAbout Author manuscriptsSubmit a manuscriptHHS Public Access; Author Manuscript; Accepted for publication in peer reviewed journal;
 
J Immunol. Author manuscript; available in PMC 2011 January 1.
Published in final edited form as:
PMCID: PMC2957822
NIHMSID: NIHMS241434

IL-15 regulates both quantitative and qualitative features of the memory CD8 T cell pool1

Abstract

Memory T cells are critical for immunity to various intracellular pathogens. Recent studies have indicated that CD8 secondary memory cells, induced by prime-boost approaches, show enhanced protective function compared to primary memory cells and exhibit phenotypic and functional characteristics which distinguish them from primary memory cells. However, little is known about the cytokine requirements for generation and maintenance of boosted memory CD8 T cells. We studied the role of IL-15 in determining the size and composition of the secondary (2°) memory CD8 T cell pool induced by Listeria monocytogenes infection in mice. Following boosting, IL-15 deficient animals failed to generate a subset of CD8 effector memory cells, including a population of IL-7Rα low cells which were prominent among secondary memory cells in normal mice. IL-15 deficiency also resulted in changes within the IL-7Rαhigh CD62Llow subset of 2° memory CD8 T cells, which expressed high levels of CD27 but minimal granzyme B. In addition to these qualitative changes, IL-15 deficiency resulted in reduced cell cycle and impaired Bcl-2 expression by 2° memory CD8 T cells, suggesting a role for IL-15 in supporting both basal proliferation and survival of the pool. Analogous qualitative differences in memory CD8 T cell populations were observed following a primary response to Sendai virus in IL-15-/- animals. Collectively these findings demonstrate that IL-15 plays an important role in dictating the composition rather than simply the maintenance of the CD8 memory pool.

Introduction

Immunological memory results in a more rapid and effective immune response to antigens which have been encountered previously (1-3). A number of studies on memory CD8 T cells have shown that they differ from naïve cells in their speed and efficiency with which they enter cell cycle and differentiate into effector cells (2, 4-6). In the context of infection, these factors translate into a more rapid and robust clearance of pathogens by memory versus naïve CD8 T cells (1, 2, 7).

There is considerable data on the role of cytokines in the homeostasis of naïve and memory CD8 T cells, especially with respect to the function of IL-7 and IL-15 (8-10). The receptors for IL-7 and IL-15 both utilize the common gamma chain (γC) but these cytokines have distinct effects on T cell maintenance. IL-7 signaling is mediated through γC and the unique alpha chain IL-7Rα (CD127). IL-7 is critical for the development and survival of naïve T cells (11, 12), and is also important for differentiation of memory CD8 T cells following the effector stage (13, 14). Similar to its impact on naïve CD8 T cells, IL-7 promotes maintenance of memory cells, in part, through upregulation of Bcl-2 (15). IL-15 utilizes γC, IL-2Rβ (CD122) and the unique chain, IL-15Rα. Several reports have demonstrated that IL-15 plays a critical role in maintenance of memory CD8 T cells, as illustrated by the decay of this pool in IL-15 and IL-15Rα deficient animals (16-18). Although IL-15 can enhance expression of pro-survival Bcl-2 family member proteins in naïve and memory T cells (19), its best defined role in T cell homeostasis is to promote proliferation of CD8 memory T cells (16, 20, 21). Hence, IL-15 driven basal proliferation of CD8 memory cells allows for the indefinite maintenance of this pool, while the memory pool declines in IL-15-/- (and in IL-15Rα-/-) animals (8, 15, 16, 20-22).

Most current studies, however, have focused on the properties of CD8 memory T cells produced following a single round of immunization, while vaccination protocols advocate repeated boosting in order to produce the most effective memory response (2, 7, 23, 24). Interestingly, recent studies have shown differences in the differentiation state and functional potential of primary (1°) versus secondary (2°) (i.e. boosted) memory CD8 T cells (25, 26). These reports indicate that 2° memory CD8 T cells take much longer than 1° memory cells to lose expression of effector molecules (such as Granzyme B) and to acquire expression of CD62L (L-selectin) and the ability to make IL-2 upon re-stimulation (25, 26). Presumably as a result of their decreased expression of CD62L, boosted memory CD8 cells exhibit differential tissue localization, being less prevalent in lymph nodes compared to 1° memory CD8 T cells (25, 26). Finally the functional capacity of 2° memory CD8 cells was found to be enhanced compared to their 1° memory counterparts, exhibiting increased cytolytic potential and providing more efficient protection against the intracellular pathogen Listeria monocytogenes (25). Together, these data indicate that there are qualitative as well as quantitative differences in the CD8 memory pool following boosting.

Very little is known about the homeostasis of the 2° memory pool, in particular whether the established role for IL-15 in maintenance of the 1° memory CD8 T cell pool (8, 15, 16, 20-22) also applies to boosted CD8 memory T cells. Indeed, recent studies suggest IL-15 may be less significant in supporting the 2° memory CD8 T cell pool. Secondary memory CD8 T cells were found to be much less sensitive to IL-15 driven in vitro proliferation compared to their 1° memory counterparts (25). This finding was also correlated with the delayed acquisition of CD62L by 2° memory CD8 T cells, since CD62L was shown to be upregulated only after cell division of the 2° memory pool (25). Moreover, in vivo basal proliferation in both normal and lymphopenic animals was impaired in the 2° versus 1° memory CD8 pool (25). Together, these data suggest that 2° memory CD8 T cells may show reduced sensitivity to endogenous levels of IL-15 and hence undergo slower basal proliferation compared to the 1° memory CD8 T cell pool (25).

We sought to further investigate the role of IL-15 in the maintenance, phenotype and function of 1° and 2° memory CD8 T cells by examining these populations in normal versus IL-15-/- animals. In agreement with others, we find that 2° memory CD8 T cells exhibit reduced basal proliferation compared to 1° memory counterparts (25), but that the presence of IL-15 still impacts the maintenance of the 2° memory CD8 T cell pool. Surprisingly, while IL-15 has not been reported to influence the phenotype or subset distribution of the 1° memory CD8 T cell pool, we found that IL-15 has a qualitative effect on the 2° memory CD8 T cell population: In the absence of IL-15, 2° memory CD8 T cells expression of Granzyme B was reduced, and an IL-7Rαlow population (found prominently among 2° memory CD8 T cells pool in wild type animals) was completely absent. IL-15 deficiency also resulted in impaired basal proliferation and reduced expression of the pro-survival protein Bcl-2. In addition, we found that IL-15 deficiency lead to similar changes in the subsets of primary memory CD8 T cells induced by intranasal immunization with Sendai virus. Hence, our findings suggest a novel role for IL-15 in supporting the maintenance, subset composition and phenotype of the memory CD8 T cell pool.

Results

Maintenance of the secondary memory CD8 T cell pool requires IL-15

Previous studies indicated that certain features of the 2° memory CD8 T cell pool (such as their low basal proliferation and slow re-acquisition of CD62Lhigh phenotype) might be explained by insensitivity to IL-15 (25), leading to the hypothesis that physiological levels of IL-15 may be less relevant for maintenance of 2° memory CD8 T cells compared to their 1° memory counterparts. To test this idea, we investigated the characteristics of 2° memory CD8 T cells produced by priming and boosting wildtype B6 mice and IL-15-/- mice with Listeria monocytogenes (LM-Ova). In these studies we examined the response of both polyclonal (endogenous) CD8 T cells and of OT-I TCR transgenic (tg) T cells adoptively transferred (at low numbers) into B6 and IL-15-/- hosts. The latter approach was designed to avoid complications related to potential changes in the antigen specific TCR repertoire in 1° versus 2° responses of B6 versus IL-15-/- strains, and also controlled for the potential influence of IL-15 during CD8 T cell development (17, 18)(19).

Earlier studies showed that, following priming, CD8 memory cells decline in an IL-15-/- host (16, 20). However, if the hypothesis described above were correct, we might expect that the 2° CD8 memory pool may be less influenced by IL-15. Both 1° and 2° memory Ova/Kb specific CD8 T cells were analyzed in IL-15-/- and B6 mice infected with LM-Ova, at a minimum of 30 days post (re)infection. As expected (16, 20), the number of antigen specific memory CD8 T cells in the IL-15-/- hosts was lower than in B6 following the 1° infection (Fig 1A), but at 5 days after boosting we observed similar numbers of Ova/Kb+ CD8+ cells in B6 and IL-15-/- animals, indicating memory cells in both populations were competent to respond and expand to similar total numbers (Fig 1A). At >30 days following boosting, the antigen specific memory pool had contracted in both strains, but the frequency of Ova/Kb+ CD8+ cells was elevated compared to the primary memory pool, suggesting effective boosting. However, similar to the pattern observed in the 1° response, we found that the number of 2° memory cells was decreased in IL-15-/- compared to B6 animals (Fig 1A). Similar results were obtained when tracking TCR tg OT-I cells (adoptively transferred at 104 cells before priming) (data not shown), which argues that these results are not due to differences in TCR repertoire or recruitment of new naïve responder cells during the secondary response. These data suggested that, even in the secondary response, IL-15 had an impact on the maintenance of the CD8 memory pool.

Figure 1
IL-15 influences the maintenance and basal proliferation of 2° memory CD8 T cells

Next we sought to test if IL-15 was required for the generation and/or maintenance of 2° memory cells. Therefore we employed a secondary adoptive transfer of OT-I T cells that were primed and boosted in B6 mice; on day 5 following the boost the OT-I T cells were re-transferred into uninfected B6 or IL-15-/- hosts then memory cells were enumerated >30 days following the boost. If IL-15 was required for maintaining 2° memory cells we would expect to see a decline in numbers of OT-I cells recovered from the IL-15-/- hosts. Our data supported this hypothesis, and the number of OT-I 2° memory cells recovered from the IL-15-/- hosts was much lower than the number recovered from the B6 hosts (Fig 1B). These data further suggest that deprivation of IL-15 after 2° infection results in a dramatic decrease in cell numbers.

IL-15 is thought to contribute to maintenance of 1° memory CD8 T cells chiefly by induction of slow basal proliferation, which maintains the CD8 memory pool indefinitely: Hence a hallmark feature of IL-15 deficient animals is severely reduced memory CD8 T cell proliferation (8, 10, 15). Since 2° memory CD8 T cells already show a reduced turnover compared to the 1° memory pool (25), it was unclear whether a similar mechanism could account for the impaired maintenance of 2° memory CD8 T cells we observed in IL-15-/- hosts (Fig 1A & B). Thus, we assayed the proliferative capacity of 1° and 2° memory OT-I CD8 T cells in B6 and IL-15-/- animals by assessing their in vivo BrdU incorporation over a 14 day period (Fig 1C). Basal proliferation within the 1° memory OT-I CD8 T cell pool was reduced in IL-15-/- compared to B6 hosts, as expected (16, 20), and we also observed reduced turnover of 2° memory CD8 T cells compared to 1° memory CD8 T cells in B6 hosts, as previously reported (25) (Fig 1C). However, the proliferation of 2° memory CD8 T cells was reduced still further in IL-15-/- animals (Fig 1C). These data suggest that, even though the turnover of 2° memory cells in B6 animals is reduced (compared to 1° memory CD8 T cells), this proliferation still depends, at least in part, on IL-15.

IL-15 deficiency alters the phenotype of the 2° memory CD8 T cell pool

While these data suggested IL-15 may influence maintenance similarly in both 1° and 2° memory CD8 T cell homeostasis, other data suggest 2° memory CD8 T cells differ from their 1° memory counterparts by multiple phenotypic and functional properties (26). Hence, we next sought to determine whether IL-15 influences the composition of the 2° memory CD8 population.

A feature of the 2° CD8 T cell response is that there is a delay in contraction of the activated population and a prolonged representation of cells that display low levels of CD62L expression compared to the 1° memory pool (25-27). Also, it has been suggested that this phenotype is linked to the reduced proliferative response of 2° memory CD8 T cells in response to IL-15 (25). Such a model would predict that the presence or absence of IL-15 in the host would not impact the 2° memory CD62L phenotype. Indeed, when we assessed CD62L expression on either endogenous Ova/Kb tetramer+ (data not shown) or OT-I CD8 T cells 30+ days after boosting with LM-Ova in various tissues (spleen, lymph node (LN), bone marrow (BM) and liver), we found that the prominent CD62Llow phenotype of the splenic 2° memory CD8 pool was unaffected by IL-15 deficiency (Fig 2A). In addition, we found that the IL-15 status of the host did not influence the relative frequencies of CD62Lhigh and CD62Llow cells in BM and liver (data not shown). However, CD62L expression was high on 2° memory CD8 T cells recovered from the LN of boosted animals, regardless of their host (data not shown). Such data argued that the presence or absence of IL-15 did not impact re-expression of CD62L in 2° memory cells. Likewise, we observed a similar (high) frequency of CD62Lhigh memory cells in the 1° memory OT-I population in both B6 and IL-15-/- hosts (data not shown), in keeping with previous reports that the presence of IL-15 in vivo does not influence memory CD8 T cell phenotype (8, 15). However, the CD62Llow subset which persists long term in 2° CD8 memory pools have been reported to exhibit phenotypic features, including low expression of CD27 (26). Low expression of both CD27 and CD62L has been correlated with low expression of CCR7 (28), which is reminiscent of “effector memory” (TEM) phenotype memory cells (29, 30). Such a population was readily detected in the 2° memory pool of B6 animals (Fig 2B), yet we found markedly reduced frequencies of CD27low or CCR7low cells in the IL-15-/- 2° memory pool, even among the CD62Llow subset (Fig 2B). Expression of other markers, including CD122 (a critical signaling component of the receptor for IL-15) and CD44 were similar between the 2° memory from wildtype and IL-15-/- mice (data not shown). Together these data indicate that, while some characteristics of 2° memory CD8 population (such as prominent representation of CD62Llow cells) were maintained in IL-15-/- animals, other features (such as low expression of CD27 and CCR7) were altered in the absence of IL-15.

Figure 2
IL-15 deficiency affects the expression of CD27 and CCR7, but not CD62L

Absence of CD127low memory CD8 T cells in IL-15-/- animals

In addition to the features discussed above, it has been reported that the boosted CD8 T cell memory pool includes a sizeable and relatively long-lived population of CD127low cells (26). This contrasts with 1° responses, where there is compelling evidence that IL-7 is important for long term survival of the 1° memory CD8 T cell pool (13, 14). Since there may be overlap between the CD8 T cell responses induced by IL-15 and IL-7 receptors (10, 15), we were particularly interested in whether the prominent pool of CD127low 2° memory CD8 T cells would be IL-15 dependent.

Hence, we analyzed 1° and 2° memory OT-I CD8 T cells in B6 and IL-15-/- hosts. The vast majority of 1° memory CD8 T cells were CD127high in both B6 and IL-15-/- animals yet, as expected based on previous studies (26), at least 25% of 2° memory CD8 T cells in B6 hosts were CD127low cells (Fig 3A), and virtually all these cells were CD62Llow (Fig 3B). In contrast to this finding however, the 2° memory CD8 T cells recovered from the IL-15-/- animals were uniformly CD127high (Fig 3A,B). This pattern of expression was also observed for 2° memory OT-I cells recovered from the BM and liver of both B6 and IL-15-/-, while in the LN, 2° memory CD8 T cells were CD62Lhigh and CD127high in both B6 and IL-15-/- mice (data not shown). In addition, these cells lack expression of activation markers, such as CD69 (data not shown), indicating they were not recently activated effector cells. These data suggest that the generation and/or maintenance of CD127low 2° memory CD8 T cells was compromised in IL-15-/- animals.

Figure 3
IL-15 maintains CD127 low 2° memory cells

We also wanted to determine when IL-15 was necessary for the emergence of the CD127low 2° memory CD8 T cells in B6 hosts. To address this issue, we performed the re-transfer experiments as described in figure 1B. If IL-15 during priming or boosting was sufficient to program the appearance of CD127low 2° memory CD8 T cells, then we would expect to see CD127low cells appear in both B6 and IL-15-/- secondary hosts. In contrast, if the CD127low cells required IL-15 for their maintenance we would only see them emerge in the secondary B6 recipient. Our results supported the second model, since we observed the adoptively transferred OT-I 2° effector cells was uniformly high for CD127 in IL-15-/- hosts, but included both CD127high and CD127low subsets in B6 hosts (Fig 3A right panel).

Since most CD127low cells were found in the CD62Llow subset (Fig 3B), it was possible that the aberrant expression of other TEM markers discussed in Figure 2B related simply to the selective absence of the CD127low pool. Indeed, closer analysis revealed that even within the CD127high subset of CD62Llow cells, there was increased expression of CD27 in the IL-15-/- animals (Fig 3C), although CCR7 staining was similar in IL-15-/- and B6 hosts (data not shown). The expression of these markers (CD27 and CCR7) on the CD62Lhigh, CD127high subset was similar in B6 and IL-15-/- animals, suggesting that the IL-15 deficiency had little effect on the phenotype of the TCM–like pool.

Hence, IL-15 deficiency has numerous effects on the composition of the 2° memory CD8 T cell pool, including the loss of CD127low cells and changes in the phenotype of the CD62Llow, CD127high subset.

IL-15 deficiency alters the composition of the primary memory CD8 T cell pool following infection with Sendai virus

In our studies, a CD127low subset was prominent in the secondary memory CD8 T cell pool, but was rare among primary memory CD8 T cells induced in B6 mice following LM-OVA immunization (Fig. 3A), and similar observations were reported by others using LM infection (26). However, previous studies on the primary response to respiratory infection with Sendai virus observed a detectable population of CD127low cells which persisted for months following intranasal infection (31). To determine whether IL-15 was involved in the composition of this primary memory CD8 T cell pool, B6 and IL-15-/- mice were subjected to respiratory infection with Sendai virus and antigen specific CD8 T cell populations studied 65 days later. Antigen specific (SenNP324-332Kb tetramer binding) primary memory CD8 T cells were detected in both hosts at this time point (Fig 4A). Sendai-specific CD8 T cells in IL-15-/- animals were only slightly reduced in frequency compared to B6 animals (average ~40% lower percentage, data not shown), which was a less severe defect than we had observed in the LM response (Fig 1). Yet there were substantial changes in representation of Sendai-specific memory CD8 T cell subsets in the two mouse strains. While a CD127low population was readily observed among antigen specific primary memory CD8 T cells in B6 animals (Fig. 4A), this population was virtually absent in IL-15-/- hosts. In this model, CD127low cells lack expression of CXCR3 and CD27 (Fig. 4A)(31), and we observed a corresponding decrease in representation of CXCR3low and CD27low populations in the Sendai-specific memory CD8 T cell pool in IL-15-/- compared to B6 animals (Fig. 4A,B). These data suggest that perturbations of CD8 memory subsets due to IL-15 deficiency can be observed in both primary and secondary memory populations following infection with distinct pathogens.

Figure 4
IL-15 deficiency alters the primary memory CD8 T cell pool following respiratory infection with Sendai virus

IL-15 deficiency compromises Granzyme B expression in the 2° memory CD8 T cell pool

In addition to phenotypic differences, 1° and 2° memory CD8 T cells exhibit significant changes in functional properties. Compared to 1° memory cells, the boosted memory CD8 T pool shows a diminished capacity to produce IL-2, a similar ability to make IFN-γ and TNFα, and an increased cytolytic capability upon antigen encounter. This latter feature is exemplified by the ex vivo expression of Granzyme B in 2° (but not 1°) memory CD8 T cells (25, 26), a feature which may correlate with improved control of pathogens by 2° memory cells (25).

We observed no significant differences in the capacity of splenic 2° memory cells to produce IFNγ and TNFα whether the cells were generated in B6 or IL-15-/- environments, and there was little evidence for enhanced cytokine production by 2° versus 1° memory pools (Fig. 5A and B). Also, the ability of 2° memory cells to simultaneously produce both TNFα and IFNγ was not impaired in the IL-15-/- host (data not shown). Similarly we tested the ability of the cells to produce IL-2. In contrast to previous reports (25, 26), we found little difference in the ability of the 2° memory cells to produce IL-2 in comparison to 1° memory cells (Fig. 5C). In addition, the lack of IL-15 did not impact the ability of the cells to produce IL-2 (Fig. 5C). The basis for the difference between our findings and those of published reports (25, 26) is unclear but may relate to the length of time after boosting that memory function was assessed. In any case, these data suggest that the presence or absence of IL-15 has a minimal impact on the capacity of 2° (or 1°) memory CD8 T cells to produce IFNγ, TNFα and IL-2 after restimulation.

Figure 5
IL-15 does not regulate cytokines production from either 1° or 2° memory cells

Next, we examined the ex vivo expression of granzyme B and determined that 1° memory cells express minimal granzyme B, regardless of whether they are recovered from B6 or IL-15-/- hosts (Figure 6A). However, we observed an increase in granzyme B expression in 2° memory cells in the B6 host when compared to 1° memory cells (Fig 6A), as previously reported (25, 26). Interestingly, granzyme B expression was considerably diminished in 2° memory cells from IL-15-/- hosts (Fig. 6A and B). In contrast to previous studies (25), we observed that the majority of granzyme expressing cells were in the CD62Llow pool (Fig 6B, top row). This raised the question of whether the reduced granzyme B expression simply correlated with the absence of the CD127low subset in IL-15-/- animals. However, we observed a difference in granzyme B expression even within the CD127high (Fig 6B, bottom row) suggesting that, similar to the phenotypic differences discussed above, the IL-15 deficit impacts the differentiation state of the surviving cells. We observed similar vigorous cytolytic activity of both B6 and IL-15-/- derived populations of 2° memory CD8 T cells using the 51Cr-release assay (data not shown): this assay chiefly depends on the effector cell expression of perforin rather than granzymes (32) but does suggest cytolytic potential is normal (or quickly acquired, in vitro) by 2° memory CD8 T cells from IL-15-/- animals.

Figure 6
Sustained expression of granzyme B depends on IL-15

Together these data suggested that basal expression of Granzyme B, but not potential for cytokine production or cytolysis, is influenced by IL-15.

Maintenance of CD127low memory cells through an IL-15 dependent mechanism

While our earlier data (Fig 1) indicated that IL-15 enhanced proliferation in the 2° memory pool, the fact that the composition of the 2° memory pool differed in B6 and IL-15-/- animals demanded reassessment of the function of IL-15 in 2° memory CD8 T cell maintenance. In particular, the presence of CD127low cells in the 2° memory pool from B6 but not IL-15-/- animals suggested a key role for IL-15 in their maintenance. Potentially, the differences between basal proliferation of 2° memory cells in B6 and IL-15-/- animals might arise by IL-15 driven proliferation of the CD127low subset. To test this we divided the OT-I cells into either CD127high or CD127low subsets and determined the percentage of those cells that incorporated BrdU over a 14 day labeling period. In contrast with this model, we observed that the CD127low subset exhibits relatively poor basal proliferation (Fig. 7A). Instead, the majority of cells undergoing basal proliferation in the B6 hosts are CD127high (Fig. 7A). In addition, the proliferation of the corresponding CD127high population in the IL-15-/- host was slightly reduced and similar to the data in Fig 1C.

Figure 7
Bcl-2 expression is increased in the presence of IL-15

We showed earlier that total numbers of 2° memory CD8 T cells were consistently lower in IL-15-/- versus B6 hosts (Figs. 1A, B). While this is partially explained by the lack of CD127low cells, the number of CD127high cells was also reduced in IL-15-/- hosts compared to B6 hosts (data not shown). Furthermore, while 2° memory CD8 T cells showed lower basal proliferation in IL-15-/- rather than B6 hosts (Figs. 1C & 7A), the magnitude of this difference appeared to be insufficient to account for the decreased absolute numbers of 2° memory CD8 T cells in IL-15-/- hosts (Figs. 1A, B). Therefore, we considered whether IL-15 also enhances survival of 2° memory CD8 T cells. Bcl-2 is an important cell survival protein that is increased as a result of IL-7R signaling and is upregulated in memory CD8 T cells stimulated with IL-15 (19). Hence we analyzed intracellular Bcl-2 levels in 1° and 2° memory CD8 T cells recovered from B6 and IL-15-/- hosts. While there were minimal differences between expression levels of Bcl-2 in 1° versus 2° memory cells from the individual hosts, there was greater expression of Bcl-2 in memory cells recovered from B6 hosts compared with IL-15-/- hosts (Fig. 7B). We also determined Bcl-2 expression by CD127low and CD127high cells in 2° memory cells recovered from B6 hosts. The 2° memory CD127high CD8 T cells in B6 hosts expressed higher levels of Bcl-2 compared to the CD127low pool (Fig. 7C and D). Similar gating of 2° memory CD8 from IL-15-/- hosts reveals that the CD127high cells, which persist in the IL-15-/-, express less Bcl-2 than the corresponding population in the B6 host (Fig. 7C and D). Together these data would suggest that optimal expression of Bcl-2 requires recognition of both IL-7 and IL-15. Furthermore, these data indicate that the absence of CD127low cells in the 2° memory cells from IL-15-/- animals likely arises from a deprivation of survival signals provided by both cytokines rather than reduced IL-15 driven proliferation.

Discussion

As various pathogens (such as HIV and malaria) are poorly protected by antibody responses, there has been renewed interest in developing vaccines that elicit efficient T cell responses. As with humoral immunity, boosting can augment cellular immune response (2, 7, 23, 24). However, recent studies have revealed that boosting the CD8 T cell response not only increases the number of antigen specific T cells but also results in qualitative changes in the memory CD8 T cell pool, which impact the phenotype, trafficking and function of 2° memory cells compared to their 1° memory progenitors (25, 26). These changes in the frequency, subset distribution and phenotype of 2° memory subsets raise the question of whether the homeostasis of this population is regulated in the same way as 1° memory cells. Current models suggest that, for 1° memory CD8 T cells, IL-7 plays an important role in dictating differentiation and survival of 1° memory CD8 T cells, while IL-15 is critical for their basal proliferation, and supporting long-term maintenance of the pool (1, 10, 33). Accordingly, there is a severe decline in production/maintenance of 1° memory CD8 T cells if they are deprived of IL-7, while deprivation of IL-15 results in a severe reduction of 1° memory CD8 T cell basal proliferation. In contrast, 2° memory CD8 T cells exhibit reduced basal turnover (and impaired lymphopenia driven proliferation) in vivo, and show inefficient IL-15 induced proliferation in vitro (25). Hence, it was possible that IL-15 becomes a less relevant cytokine for maintenance of the 2° memory CD8 T cell pool. Instead, our findings indicate that IL-15 has a significant impact on the composition, maintenance and function of 2° memory CD8 T cells.

Specifically, we find that IL-15 impacts the 2° memory pool via three distinct mechanisms. First, IL-15 is evidently critical for the differentiation and/or survival of the CD127low subset of 2° memory cells. The CD127low pool survives well into the memory stage of the 2° response, yet cells of this phenotype gradually disappear over the course of many months (26), though it is not yet clear whether they die or change in phenotype. Also, while responsiveness to IL-7 appears important for long term persistence of memory cells, CD127low CD8 memory T cells are found for several months following some primary responses (31, 34), and such cells might play a role in acute protective immunity. Given the potential overlap between IL-7 and IL-15, it is reasonable to conclude that cells lacking CD127 might be more dependent on IL-15 for their survival. Indeed, cells with this phenotype have been shown to die more rapidly when deprived of IL-15 (34). An alternative hypothesis is that IL-15 itself induces downregulation of CD127. Naïve T cells have been shown to downregulate expression of CD127 in response to various cytokines, including IL-15 (35). Furthermore, recent studies show that treatment with multivalent IL-15/IL-15Ra complex can promote appearance of CD127low cells during the contraction phase of an immune response (36). It is worth reiterating that the CD127low subset is usually less prominent in the 1° memory CD8 T cell pool, therefore despite access to physiological levels of IL-15, the majority of 1° memory cells re-express IL-7Rα, which seems critical for their differentiation from the effector stage (13, 14). What regulates the capacity of some 2° memory cells to “wean” themselves off IL-7 and onto IL-15 as a survival cytokine is currently unclear. However, we also observed an impact of IL-15 deficiency on primary CD8 memory subsets, following respiratory infection with Sendai virus. Hikono et al. demonstrated phenotypic and functional heterogeneity among the memory populations produced by this infection, including the appearance of a CD127low memory CD8 T cell population which persists for months (but not long term) (31). The majority of cells in this CD127low subset was also found to lack expression of CD27 and CXCR3. We observed that the Sendai virus specific memory CD8 T cell pool in IL-15 deficient mice selectively lacked CD127low, CXCR3low and CD27low subsets (Fig. 4). Hence these data indicate that the effect of IL-15 on CD8 memory subset composition occurs in response to divergent pathogens (Sendai virus and Listeria bacteria), routes of infection (mucosal and systemic infection) and phases of the immune response (primary and secondary responses).

Second, the lack of IL-15 resulted in altered phenotypic and functional changes in the cells which persist into the 2° memory pool. A subset of CD127high CD62Llow 2° memory CD8 T cells was found in both B6 and IL-15-/- hosts. However, while in IL-15-sufficient mice this subset contained CD27low and Granzyme B+ cells, in the IL-15-/- animals, these cells were CD27high and Granzyme B-. A CD27low and Granzyme B+ memory CD8 T pool has been previously observed following the primary respiratory response to Sendai virus (31). CD27 is a member of the TNF-receptor superfamily and can undergo functionally relevant interactions with CD70. CD27 signals on CD8 T cells has been shown to promote the generation and survival of memory cells (37-39), therefore the retained expression of CD27 on cells in the IL-15-/- may be a compensatory mechanism of survival. Alternatively, the presence of CD27 on the IL-15-/- 2° TEM subset may allow these cells to participate in additional responses denied to their B6 counterparts. The lack of Granzyme B in the IL-15-/- 2° memory CD8 T cells is similar to previous studies which showed reduced expression of Granzyme B in IL-15-/- effector cells (40). Although Granzyme B is an important component of cytolytic granules, Granzyme B deficient CD8 T cells are typically competent for cytolysis, an assay that chiefly depends on the ability of the effector cells to secrete perforin (32), and we observed similar cytolytic potential of 2° memory CD8 T cells from both B6 and IL-15-/- animals (data not shown). On the other hand, Granzyme B deficient animals have a severely reduced ability to control ectromelia (41) and show a delayed control of LCMV (42). Thus, the lack of steady state Granzyme B in IL-15-/- 2° memory CD8 T cells may potentially influence their ability to rapidly control certain pathogens. Regardless, these data together argue that IL-15 influences the expression of functionally relevant markers (CD27 and Granzyme B) in the 2° memory pool. This may be direct in the case of Granzyme B, which appears to be induced by IL-15R signals (40). A recent study reported that treatment with IL-15/IL-15Ra complex following priming lead to enhanced representation of CD127low cells but no significant changes in expression of CD27 or CXCR3 (among other markers) following the treatment. In contrast, we observed that IL-15 deficiency lead to changes in expression of both CD27 (Figs. 2,,33,,4)4) and CXCR3 (Fig. 4). This may reflect the distinct phases of the immune response studied, since Rubenstein et al studied the contraction phase of the response (through to day 17 post-infection) while our studies focused on the memory phase.

Third, IL-15 deficiency leads to decreased frequency of 2° memory CD8 T cells. This was probably due to a combination of decreased basal proliferation and impaired survival of 2° memory CD8 T cell pool in IL-15-/- animals. While we and others (25) found that 2° memory CD8 T cells show reduced turnover compared to their 1° memory counterparts, this proliferative rate was further reduced in IL-15-/- hosts (even when analysis was focused on the CD127high subset found in both strains Fig 1C and and7A).7A). This is somewhat surprising given that the in vitro response of 2° memory T cells to IL-15 was impaired (25), however, this assay may have underestimated the cells sensitivity to the cytokine. Since the expression of the anti-apoptotic protein Bcl-2 has been linked to IL-15 signals (19), we then explored the survival role of IL-15. While we observed a reduction in Bcl-2 in IL-15 deprived 1° memory cells, this was even more marked for the 2° memory CD8 T cell pool (Fig. 7). Together these data suggest that IL-15 plays an important role in directing expression of pro-survival factors as well as supporting basal proliferation of the 2° memory pool.

In these studies, we chiefly examined responses of adoptively transferred OT-I T cells following priming and boosting with LM-Ova, and focused on relatively early stages of the memory phase (typically 4 weeks post immunization). This was partly driven by the low numbers and poor maintenance of antigen specific memory CD8 T cells generated in IL-15-/- animals. However we observed similar general features of 2° memory CD8 T cells in B6 and IL-15-/- hosts when endogenous Ova/Kb specific CD8 T cells were tracked as well as when heterologous prime/boost approaches were employed (using LM- for priming and boosting with vaccinia-Ova, data not shown). Furthermore, the characteristic features of 2° memory CD8 T cells discussed above have been reported in different mouse strains, responding to various viral and bacterial pathogens, and appears to hold true for responses by both the endogenous polyclonal CD8 T cell pool and for TCR transgenic T cells (25, 26). Hence, while there are certainly limitations with use of adoptive transfer approaches (27), the differences between 1° and 2° memory CD8 T cell responses appear to transcend these concerns.

IL-15-/- mice clear 1° and 2° LM infection similarly to B6 mice (our unpublished observations), and this was part of our rationale for using LM in this study. However IL-15-/- animals have been reported to have defects in clearance of vaccinia virus (17) as well as defects in the response to heat killed Propionibacterium acnes (43). Our data on the multiple effects of IL-15 deprivation on the 2° memory CD8 T cell pool suggest that boosting does not compensate for the lack of IL-15, and it will be interesting to see whether defective immune control of these pathogens by IL-15-/- animals is compounded still further in secondary immune responses.

In summary, our data suggest IL-15 impacts both the maintenance and, more surprisingly, the composition and functional potential of the 2° memory CD8 T cell pool. Hence, despite reported decreased sensitivity of 2° memory CD8 T cells for IL-15, this cytokine still plays a prominent role in dictating the boosted memory CD8 T cells. Hence, its proposed inclusion as a component of vaccines for cell-mediated immunity (44) may be justified, even when a prime boost strategy is employed. It will be interesting to determine whether the role of other homeostatic cytokines (for example IL-7) changes in the 2° memory CD8 T cell pool.

Material and Methods

Mice

C57BL/6 (B6) mice were purchased from Jackson Labs and C57BL/6 IL-15-/- were purchased from Taconic Farms. OT-I TCR transgenic (tg) mice (45) were bred to B6.PL (Thy1.1) mice (Jackson labs) to generate the OT-I.PL strain. Enrichment of CD44low OT-I.PL cells was performed by negative selection purification, as previously described (46) and 1×104 OT-I cells adoptively transferred into B6 and IL-15-/- mice via the tail vein one day prior to infection. All mice were maintained under SPF conditions at the University of Minnesota. All animal handling and experiments were carried out with approval from the University of Minnesota or Trudeau Institute IACUCs.

Infections

Recombinant Listeria monocytogenes expressing secreted ovalbumin protein (LM-ova) 10403s were provided by H. Shen (University of Pennsylvania School of Medicine, Philadelphia, Pennsylvania) and J.T. Harty (University of Iowa, Iowa City, Iowa). Bacteria were grown in tryptic soy broth (TSB) with 50 g/ml of streptomycin to an absorbance at 600 nm of about 0.1. For primary infections 1×104 to 5×104 CFU were injected intravenously, while for secondary infections ~ 1×105 were injected intravenously. Boosting was performed at least 4 weeks after the primary infection. The number of bacteria injected was confirmed by dilution and growth on tryptic soy agar plates containing streptomycin. For Sendai virus (Enders strain) infections, 8-12 week old mice were anesthetized with 2,2,2-tribromoethanol (200 mg/kg) and 250 50% egg infectious doses (EID50) were delivered intranasally in a volume of 30 μl.

Flow cytometry

Single cell suspensions from spleen and pooled lymph nodes, bone marrow and liver were analyzed by flow cytometry. Endogenous Ova specific CD8 T cells were identified by co-staining with CD8α and H-2Kb-Ova tetramer produced as described (47). Tetramers specific for the Sendai virus nucleoprotein epitope (SenNP324-332Kb) were produced by the Trudeau Institute Molecular Biology Core Facility. Co-staining with CD8α and Thy1.1 identified transferred OT-I.PL cells. Phenotypic analysis was carried out with additional staining with antibodies to CD122, CD62L, CD127, CD27, CD43 (1B11) (purchased from BD Biosciences and eBiosciences), CXCR3 (purchased from R&D Systems), and CCL19-Fc (generously provided by J. Cyster, UCSF). Flow cytometry was performed on a FACSCalibur and LSRII (BD Biosciences) and data analyzed using FlowJo software (Tree Star).

CD8 T cell enrichment for secondary adoptive transfers

B6 mice receiving adoptive transfer of 104 OT-I T cells were primed and boosted with LM-OVA as described above. At day 5 of the recall response, splenocytes were enriched for CD8+ T cells by negative selection, using a cocktail of FITC labeled antibodies against CD4, CD19, NK1.1 and I-Ab, followed by staining with anti-FITC paramagnetic microbeads (Miltenyi Biotech) and removal over magnetic columns (Miltenyi Biotech). OT-I cells in the flow through were enumerated as described above before secondary transfer into uninfected B6 or IL-15-/- hosts, and were analyzed 23 days later (4 weeks from the booster infection).

Intracellular staining

For intracellular staining cells were surface stained first, then fixed and permeabilized (BD Pharmingen) followed by anti-mouse Bcl-2, or hamster IgG isotype control (BD Pharmingen); or anti-human granzyme B or mouse IgG1 isotype control (Caltag).

Intracellular cytokine staining

Splenocytes were stimulated with our without SIINFEKL peptide (250nM for IFNγ and TNFα production 1uM for IL-2 production), in the presence of Brefeldin A, for 4 hours at 37°C. Cells were fixed and permeabilized (BD Pharmingen) according to manufacturers protocol and then stained with the appropriate antibodies. An unstimulated sample was generated for each animal and the frequency of cytokine staining in these samples subtracted from the frequency in antigen stimulated samples to quantitate the specific response.

BrdU labeling

Two weeks after the last infection, mice were administered with freshly made BrdU (0.8mg/mL administered with 2% Sucrose) in the drinking water, for 14 days. Cells were harvested and surface stained with CD8α, Thy1.1 and IL-7Rα. Cells were then fixed, permeabilized and DNAse treated according to manufacturers protocol. Cells were then stained with anti-BrdU (BD Pharmingen).

Statistical analysis

An unpaired, two-tailed Student's t-test was used to determine significance with Prism software (GraphPad Software).

Acknowledgments

We thank Kris Hogquist, Troy Baldwin, and Lisa McNeil for their critical input as well as various lab members for their suggestions and Tom McCaughtry and Kris Hogquist for comments on the manuscript. We also thank Shannon Miller for technical assistance.

Footnotes

1This work was supported by NIH grants R37AI38903 to S.C.J., an Immunology Pre-doctoral Training Grant T32 AI07313 to M.M.S., F32 AI71478 to J.E.K., and AI67967, AI76499, and T32 AI49823 to D.L.W.)

References

1. Rocha B, Tanchot C. CD8 T cell memory. Semin Immunol. 2004;16:305–314. [PubMed]
2. Kaech SM, Wherry EJ, Ahmed R. Effector and memory T-cell differentiation: implications for vaccine development. Nat Rev Immunol. 2002;2:251–262. [PubMed]
3. Sprent J, Surh CD. Generation and maintenance of memory T cells. Current opinion in immunology. 2001;13:248–254. [PubMed]
4. Veiga-Fernandes H, Walter U, Bourgeois C, McLean A, Rocha B. Response of naive and memory CD8+ T cells to antigen stimulation in vivo. Nat Immunol. 2000;1:47–53. [PubMed]
5. Veiga-Fernandes H, Rocha B. High expression of active CDK6 in the cytoplasm of CD8 memory cells favors rapid division. Nat Immunol. 2004;5:31–37. [PubMed]
6. Seder RA, Ahmed R. Similarities and differences in CD4+ and CD8+ effector and memory T cell generation. Nat Immunol. 2003;4:835–842. [PubMed]
7. Harty JT, Badovinac VP. Shaping and reshaping CD8+ T-cell memory. Nat Rev Immunol. 2008;8:107–119. [PubMed]
8. Schluns KS, Lefrancois L. Cytokine control of memory T-cell development and survival. Nat Rev Immunol. 2003;3:269–279. [PubMed]
9. Jameson SC. T cell homeostasis: Keeping useful T cells alive and live T cells useful. Semin Immunol. 2005;17:231–237. [PubMed]
10. Surh CD, Boyman O, Purton JF, Sprent J. Homeostasis of memory T cells. Immunol Rev. 2006;211:154–163. [PubMed]
11. Maraskovsky E, Teepe M, Morrissey PJ, Braddy S, Miller RE, Lynch DH, Peschon JJ. Impaired survival and proliferation in IL-7 receptor-deficient peripheral T cells. J Immunol. 1996;157:5315–5323. [PubMed]
12. Sudo T, Nishikawa S, Ohno N, Akiyama N, Tamakoshi M, Yoshida H, Nishikawa SI. Expression and function of the interleukin 7 receptor in murine lymphocytes. Proc Natl Acad Sci U S A. 1993;90:9125–9129. [PubMed]
13. Schluns KS, Kieper WC, Jameson SC, Lefrancois L. Interleukin-7 mediates the homeostasis of naive and memory CD8 T cells in vivo. Nature Immunology. 2000;1:426–432. [PubMed]
14. Kaech SM, Tan JT, Wherry EJ, Konieczny BT, Surh CD, Ahmed R. Selective expression of the interleukin 7 receptor identifies effector CD8 T cells that give rise to long-lived memory cells. Nat Immunol. 2003;4:1191–1198. [PubMed]
15. Ma A, Koka R, Burkett P. Diverse functions of IL-2, IL-15, and IL-7 in lymphoid homeostasis. Annu Rev Immunol. 2006;24:657–679. [PubMed]
16. Schluns KS, Williams K, Ma A, Zheng XX, Lefrancois L. Cutting Edge: Requirement for IL-15 in the Generation of Primary and Memory Antigen-Specific CD8 T Cells. J Immunol. 2002;168:4827–4831. [PubMed]
17. Kennedy MK, Glaccum M, Brown SN, Butz EA, Viney JL, Embers M, Matsuki N, Charrier K, Sedger L, Willis CR, Brasel K, Morrissey PJ, Stocking K, Schuh JC, Joyce S, Peschon JJ. Reversible defects in natural killer and memory CD8 T cell lineages in interleukin 15-deficient mice. J Exp Med. 2000;191:771–780. [PMC free article] [PubMed]
18. Lodolce JP, Boone DL, Chai S, Swain RE, Dassopoulos T, Trettin S, Ma A. IL-15 receptor maintains lymphoid homeostasis by supporting lymphocyte homing and proliferation. Immunity. 1998;9:669–676. [PubMed]
19. Berard M, Brandt K, Bulfone-Paus S, Tough DF. IL-15 promotes the survival of naive and memory phenotype CD8+ T cells. J Immunol. 2003;170:5018–5026. [PubMed]
20. Becker TC, Wherry EJ, Boone D, Murali-Krishna K, Antia R, Ma A, Ahmed R. IL-15 is required for proliferative renewal of virus-specific memory CD8 T cells. J Exp Med. 2002;195:1541–1548. [PMC free article] [PubMed]
21. Burkett PR, Koka R, Chien M, Chai S, Chan F, Ma A, Boone DL. IL-15R alpha expression on CD8+ T cells is dispensable for T cell memory. Proc Natl Acad Sci U S A. 2003;100:4724–4729. [PubMed]
22. Goldrath AW, Sivakumar PV, Glaccum M, Kennedy MK, Bevan MJ, Benoist C, Mathis D, Butz EA. Cytokine requirements for acute and Basal homeostatic proliferation of naive and memory CD8+ T cells. J Exp Med. 2002;195:1515–1522. [PMC free article] [PubMed]
23. Woodland DL. Jump-starting the immune system: prime-boosting comes of age. Trends in immunology. 2004;25:98–104. [PubMed]
24. Robinson HL, Amara RR. T cell vaccines for microbial infections. Nature medicine. 2005;11:S25–32. [PubMed]
25. Jabbari A, Harty JT. Secondary memory CD8+ T cells are more protective but slower to acquire a central-memory phenotype. J Exp Med. 2006;203:919–932. [PMC free article] [PubMed]
26. Masopust D, Ha SJ, Vezys V, Ahmed R. Stimulation history dictates memory CD8 T cell phenotype: implications for prime-boost vaccination. J Immunol. 2006;177:831–839. [PubMed]
27. Marzo AL, Klonowski KD, Bon AL, Borrow P, Tough DF, Lefrancois L. Initial T cell frequency dictates memory CD8(+) T cell lineage commitment. Nat Immunol. 2005;6:793–799. [PMC free article] [PubMed]
28. Wherry EJ, Teichgraber V, Becker TC, Masopust D, Kaech SM, Antia R, von Andrian UH, Ahmed R. Lineage relationship and protective immunity of memory CD8 T cell subsets. Nat Immunol. 2003;4:225–234. [PubMed]
29. Lanzavecchia A, Sallusto F. Understanding the generation and function of memory T cell subsets. Current opinion in immunology. 2005;17:326–332. [PubMed]
30. Lefrancois L, Marzo AL. The descent of memory T-cell subsets. Nat Rev Immunol. 2006;6:618–623. [PubMed]
31. Hikono H, Kohlmeier JE, Takamura S, Wittmer ST, Roberts AD, Woodland DL. Activation phenotype, rather than central- or effector-memory phenotype, predicts the recall efficacy of memory CD8+ T cells. J Exp Med. 2007;204:1625–1636. [PMC free article] [PubMed]
32. Pardo J, Balkow S, Anel A, Simon MM. The differential contribution of granzyme A and granzyme B in cytotoxic T lymphocyte-mediated apoptosis is determined by the quality of target cells. European journal of immunology. 2002;32:1980–1985. [PubMed]
33. Williams MA, Holmes BJ, Sun JC, Bevan MJ. Developing and maintaining protective CD8+ memory T cells. Immunol Rev. 2006;211:146–153. [PubMed]
34. Joshi NS, Cui W, Chandele A, Lee HK, Urso DR, Hagman J, Gapin L, Kaech SM. Inflammation directs a gradient of T-bet expression that specifies memory precursor and short-lived effector CD8 T cell fates. Immunity 2007 [PMC free article] [PubMed]
35. Park JH, Yu Q, Erman B, Appelbaum JS, Montoya-Durango D, Grimes HL, Singer A. Suppression of IL7Ralpha transcription by IL-7 and other prosurvival cytokines: a novel mechanism for maximizing IL-7-dependent T cell survival. Immunity. 2004;21:289–302. [PubMed]
36. Rubinstein MP, Lind NA, Purton JF, Filippou P, Best JA, McGhee PA, Surh CD, Goldrath AW. IL-7 and IL-15 differentially regulate CD8+ T cell subsets during contraction of the immune response. Blood 2008 [PubMed]
37. Hendriks J, Xiao Y, Rossen JW, van der Sluijs KF, Sugamura K, Ishii N, Borst J. During viral infection of the respiratory tract, CD27, 4-1BB, and OX40 collectively determine formation of CD8+ memory T cells and their capacity for secondary expansion. J Immunol. 2005;175:1665–1676. [PubMed]
38. Hendriks J, Gravestein LA, Tesselaar K, van Lier RA, Schumacher TN, Borst J. CD27 is required for generation and long-term maintenance of T cell immunity. Nat Immunol. 2000;1:433–440. [PubMed]
39. Borst J, Hendriks J, Xiao Y. CD27 and CD70 in T cell and B cell activation. Current opinion in immunology. 2005;17:275–281. [PubMed]
40. Yajima T, Yoshihara K, Nakazato K, Kumabe S, Koyasu S, Sad S, Shen H, Kuwano H, Yoshikai Y. IL-15 regulates CD8+ T cell contraction during primary infection. J Immunol. 2006;176:507–515. [PubMed]
41. Mullbacher A, Waring P, Tha Hla R, Tran T, Chin S, Stehle T, Museteanu C, Simon MM. Granzymes are the essential downstream effector molecules for the control of primary virus infections by cytolytic leukocytes. Proc Natl Acad Sci U S A. 1999;96:13950–13955. [PubMed]
42. Zajac AJ, Dye JM, Quinn DG. Control of lymphocytic choriomeningitis virus infection in granzyme B deficient mice. Virology. 2003;305:1–9. [PubMed]
43. Ohteki T, Tada H, Ishida K, Sato T, Maki C, Yamada T, Hamuro J, Koyasu S. Essential roles of DC-derived IL-15 as a mediator of inflammatory responses in vivo. J Exp Med. 2006;203:2329–2338. [PMC free article] [PubMed]
44. Waldmann TA. The biology of interleukin-2 and interleukin-15: implications for cancer therapy and vaccine design. Nat Rev Immunol. 2006;6:595–601. [PubMed]
45. Hogquist KA, Jameson SC, Heath WR, Howard JL, Bevan MJ, Carbone FR. T cell receptor antagonist peptides induce positive selection. Cell. 1994;76:17–27. [PubMed]
46. Kieper WC, Prlic M, Schmidt CS, Mescher MF, Jameson SC. Il-12 enhances CD8 T cell homeostatic expansion. J Immunol. 2001;166:5515–5521. [PubMed]
47. Daniels MA, Jameson SC. Critical Role for CD8 in T Cell Receptor Binding and Activation by Peptide/Major Histocompatibility Complex Multimers. Journal of Experimental Medicine. 2000;191:335–346. [PMC free article] [PubMed]