- Top of page
- Materials and methods
To study the role of T cells in gram-negative sepsis, we developed a mouse model in which i.v. injection of Escherichia coli results in severe systemic illness, with high mortality rates after day 5. A large proportion of both CD4+ and CD8+ T cells are activated within 1 day after infection, as evidenced by up-regulation of CD69 and down-regulation of CD62L. Even more surprisingly, T cell-deficient mice exhibit markedly decreased disease severity compared to WT mice, indicating a pathogenic role of T cells. Mice lacking IFN-γ also show diminished disease, and exhibit reduced T cell activation. Therefore, the pathogenic role of T cells may be mediated by IFN-γ. Both T cell- and IFN-γ-deficient mice have reduced serum IL-6 levels compared to WT mice, suggesting that T cells may stimulate innate immune responses, resulting in enhancement of disease. These data indicate an important role for T cells in a mouse model of E. coli sepsis, and reveal an unexpected early and pathogenic T cell response to this bacterial infection.
- Top of page
- Materials and methods
The experiments described here were designed to study the role of T cells in a mouse model of gram-negative sepsis. Surprisingly, activation of a large percentage of T cells occurs very early after i.v. injection of E. coli, and absence of T cells is associated with decreased disease severity, indicating a pathogenic role for T cells in this model. Mice lacking IFN-γ also develop less severe disease than WT mice upon E. coli infection, which appears to be associated with decreased activation of T cells. Amelioration of disease in both IFN-γ–/– mice and TCRα–/– mice is associated with decreased levels of IL-6.
A role for T cells in sepsis was first suggested by Hotchkiss and colleagues 5, who demonstrated rapid apoptosis of CD4+ T cells in the murine colonic ligation and puncture model of sepsis and showed that prevention of apoptosis resulted in improved outcome. These results suggest that T cells are protective, although the exact mechanism behind this phenomenon remains unclear. Although we did not study apoptosis in our model, we did not observe depletion of CD4+ T cells in the spleens of E. coli-infected mice. Rather, our data clearly show that the absence of T cells reduces disease severity, i.e. activated T cells are pathogenic. These differences may reflect difference in the models: colonic ligation and puncture results in peritonitis and polymicrobial bacteremia, whereas in our model sepsis does not arise from a nidus of infection and involves one microbe only.
In a study of humans with severe gram-negative sepsis, a decreased percentage of CD4+ T cells was noted in the peripheral blood of septic patients as compared to healthy controls 6. Patients in this study were selected on criteria that included a known (pulmonary or intra-abdominal) nidus of infection, which, as noted above, may result in different immune responses. In addition, it is unclear whether changes in peripheral blood cell counts reflect those in spleen and lymph nodes; because of small cell numbers, quantification of CD4+ T cells in peripheral blood is difficult in a mouse model. Nevertheless, our results are the first to show a pathogenic role of T cells in gram-negative sepsis.
Activation of T cells occurs in response to antigens presented by antigen-presenting cells together with a co-stimulatory signal; this process is thought to take several days. The percentage of antigen-specific T cells for any given antigen is low; therefore only a small percentage of T cells are expected to become activated in response to a microbial infection. Surprisingly, T cell activation was detected within 24 h after i.v. injection of E. coli, and involved a large percentage of T cells (up to 80%).
It has been postulated that T cells can respond directly to microbial products such as LPS via direct engagement of TLR on T cells 7, 8. E. coli infection of mice lacking MyD88, a central adaptor protein for signal transduction of TLR, resulted in similar early and extensive T cell activation compared to WT mice (data not shown), making this explanation less likely, although signaling through MyD88-independent pathways could still be involved. Early and pronounced up-regulation of CD69 on CD4+ and CD8+ T cells in response to injection of LPS and other microbial mitogens into mice has been demonstrated previously 9, 10. In subsequent studies it was shown that the effect of LPS on T cells is mediated by dendritic cells, in part via secretion of IFN-αβ, which stimulates T cells directly, and in part by stimulating IFN-γ production by NK cells 11. In concordance with these observations, we demonstrated reduced T cell activation in E. coli-infected IFN-γ–/– mice. Of note, we did not find evidence for increased IFN-γ or IL-2 production by T cells, indicating that the observed activation of T cells may be incomplete and not lead to a full effector phenotype with resulting cytokine secretion.
T cell-deficient mice developed significantly less severe disease upon E. coli infection than WT mice, indicating that T cells have a detrimental effect on disease outcome. We were unable to identify with certainty the specific subset of T cells responsible for this effect. Results from antibody depletion experiments pointed in the direction of CD8+ T lymphocytes. In the colonic ligation and puncture model of murine sepsis, mice deficient in CD8+ T cells were found to have longer survival compared to WT mice, although mortality was still 100% after 72 h 12. Considering the observation that both CD4+ and CD8+ T cells show up-regulation of activation markers in our model, we believe it is unlikely that either subset of T cells is solely responsible for the pathogenic effect. T cells involved in modulation of disease severity in our model could have an effector phenotype, or, alternatively, be regulatory cells that have suppressive effects on innate or adaptive immune responses leading to enhanced disease in the presence of these cells.
We initially postulated that the pathogenic effect of T cells was mediated by IFN-γ, since neutralization of IFN-γ has been demonstrated to reduce mortality in mice after LPS injection 13 as well as E. coli infection 14. Indeed, mice lacking IFN-γ demonstrated diminished disease severity after E. coli infection compared to WT mice. Serum IFN-γ levels in WT mice were, however, not different from those in TCRα–/– mice. In addition, we were not able to detect increased IFN-γ production by intracellular staining of lymphocytes isolated from spleen or peripheral lymph nodes after restimulation with either plate-bound anti-CD3, PMA and ionomycin, or heat-killed E. coli lysates (data not shown).
Although this still leaves the possibility of the presence of IFN-γ-secreting T cells in other tissues and a direct pathogenic role for IFN-γ in this disease, it seems more likely that in the absence of IFN-γ, a T cell-mediated pathogenic process is not fully activated. As noted above, T cells are known to express TLR, but the function of these TLR is far from clear 15. In addition, other, yet to be identified pathogen recognition receptors may be implicated. Activation of such receptors and/or the downstream signaling events may be dependent on IFN-γ.
Comparison of serum cytokine levels in E. coli-infected TCRα–/–, IFN-γ–/– and WT mice revealed a marked difference in IL-6 levels, with much higher levels noted in WT mice at all time points studied. High serum levels of IL-6 have been demonstrated in human subjects with sepsis, and correlate with an increased risk of mortality 16–19. It is unclear whether IL-6 is merely a marker of disease severity or whether it is involved in the development of the disease. Increased levels of IL-6 have also been detected in several mouse models of sepsis, and neutralization of IL-6 with monoclonal antibody treatment improved survival 20. Studies in IL-6-knockout mice subjected to colonic ligation and puncture, however, have found no difference in survival 21, 22, whereas mortality in IL-6-deficient mice compared to WT mice was increased upon systemic E. coli infection 23.
IL-6 is synthesized by both immune cells (including activated T cells) and non-immune cells, and has a wide range of biological effects. These include stimulation of acute-phase protein production by hepatocytes, activation of T cells, regulation of neutrophil and T cell trafficking 24, and stimulation of IL-17 production 25. In particular, IL-6 is thought to decrease neutrophil recruitment and induce neutrophil apoptosis while increasing T cell trafficking to inflamed tissues and rescuing T cells from apoptosis, thereby shifting the immune response from innate to acquired immunity. In addition, IL-6 has been shown to block the suppressor activity of CD4+CD25+ regulatory T cells in the context of TLR signaling 26. The results of our studies imply that in the presence of T cells, IL-6 secretion in response to E. coli infection is enhanced. This suggests the presence of a positive feedback mechanism that results in ongoing activation of T cells.
The precise nature of this postulated feedback mechanism remains to be elucidated. It may be cytokine-mediated, and dependent on T cell activation. A possible candidate is IL-17, which is produced by both CD4+ and CD8+ T cells in response to microbial infection and IL-6 secretion, and augments neutrophil chemoattraction as well as granulopoiesis 27. Serum IL-17 levels were modestly higher in WT mice than in IFN-γ–/– or TCRα–/– mice on day 4 after infection; however, this difference did not reach statistical significance (p=0.08). Although this observation fits with increased IL-6 levels in WT mice, it is at odds with findings in other mouse models, in which IFN-γ–/– mice were noted to have increased numbers of IL-17-producing T cells 28, 29. These reports describe findings in models of chronic disease, which may not extrapolate to an acute infectious process. The possible involvement of IL-17 in sepsis deserves further study.
Of note, E. coli-infected BALB/c mice exhibited an increase in neutrophilic granulocytes among splenocytes as compared to PBS-treated animals, which was remarkably pronounced on day 7 after infection. Only 40% of E. coli-infected WT mice where still alive on day 7; it is therefore possible that only mice capable of mounting an extensive neutrophil response are able to survive until this time point, possibly through control of further bacterial proliferation. On the other hand, neutrophils have been implicated in sepsis-associated multi-organ dysfunction, both in humans with the disease as well as in animal models, although the exact role of neutrophils in sepsis is far from clear 30. The role of neutrophils in our model of E. coli sepsis warrants further investigation in the future.
In IFN-γ–/– mice, T cell activation after E. coli infection was significantly decreased compared to WT mice, but still pronounced compared to PBS-injected mice. In addition, IFN-γ–/– mice, but not TCRα–/– mice, had significantly decreased levels of TNF-α on day 1 after E. coli infection. This suggests that an alternate mechanism may be (in part) responsible for the decreased levels of IL-6 observed in E. coli-infected IFN-γ–/– mice. Others have shown that IFN-γ–/– mice have delayed recruitment and diminished clearance of polymorphonuclear cells in response to an inflammatory stimulus, which is associated with decreased IL-6 levels 31.
Results of our studies do not demonstrate a clear relationship between disease severity and bacterial elimination. Both TCRα–/– and IFN-γ–/– mice had markedly less severe disease than WT mice, but only TCR-α–/– mice cleared bacteria faster from the blood stream whereas bacterial load in the liver was not different between the different strains at any of the time points analyzed. Bacteria could not be cultured from the blood of any mouse past day 3, regardless of disease severity. This is not unlike the situation in humans with sepsis, in whom negative blood cultures, even without prior antibiotic therapy, are not uncommon 32. Conversely, many animals that appeared to have recovered completely from disease in terms of weight loss and disease score were found to have significant bacterial counts in the liver. Taken together we believe these data to indicate that it is not the extent of the bacterial infection itself, but rather the nature of the hosts’ response that determines disease severity.
The pathogenesis of sepsis is complex, and our understanding of the immune mechanisms involved remains incomplete. The development of therapeutic modalities for this highly lethal disease is impeded by this lack of insight. The experiments described in this report add to a growing body of evidence that pathogenic immune responses are not limited to the innate immune system. T cells do get activated early after a systemic bacterial infection, and this appears to be detrimental to disease outcome. Our results challenge the idea that adaptive immune responses are slow and protective against infection, and emphasize the need for further delineation of the interplay between innate and adaptive immune cells and the mediators involved.