Sepsis syndrome remains the most common cause of mortality in ICUs, causing about 750,000 deaths annually in the USA 12.

Despite this, our knowledge of the underlying process of sepsis is still unclear, and this is particularly obvious from an immunological standpoint. Indeed, after having considered for years the host response to severe infection only as a stormy inflammatory reaction, it appears now that anti-inflammatory components are released at the same time as pro-inflammatory mediators and the net result of these two arms of immune response determines the patient's phenotype: overwhelming inflammation with early death, delicate balance between pro- and anti-inflammation with progressive recovering, or exaggerated anti-inflammation predisposing to superinfections 34.

The later phenotype has been called 'immune paralysis' or 'leukocyte reprogramming' and is characterized by a lymphocytes' anergy, apoptosis, and a reduced capacity to present antigens 5. Several mechanisms may partly explain this phenomenon such as an increased production of anti-inflammatory cytokines IL-10 and transforming growth factor (TGF)-β 3.

More recently, several groups have implicated an active lymphoid suppressor cell population: the naturally occurring regulatory T cells (Tregs) 56. Although these CD4⁺CD25⁺ cells make up only a small fraction of the T-lymphocyte population, they are potent inhibitors of immune response 7, and are now widely regarded as the primary mediators of peripheral tolerance, able to maintain immune homeostasis, to prevent autoimmunity, and to modulate inflammation induced by pathogens and environmental insults 5.

From a functional perspective, the suppressive mechanisms used by Tregs can be divided into four groups: suppression by inhibitory cytokines (such as IL-10 and TGF-β) and cell-to-cell contact, suppression by cytolysis through the secretion of granzymes, suppression by metabolic disruption, and suppression by the modulation of dendritic cells maturation or function (through cell-surface molecules such as cytotoxic T lymphocyte antigen-4, IL-10 and TGF-β) 8.

Although Tregs have a pivotal role in preventing autoimmune diseases and limiting chronic inflammatory disorders, they may also block beneficial immune responses by preventing sterilizing immunity to certain pathogens 39. As an elevated percentage of Tregs was observed during septic shock, a role of this cell subset has been suspected during this syndrome 10. Indeed, data are scarce and conflicting: whereas Heuer and colleagues demonstrated that adoptive transfer treatment with ex vivo-stimulated Tregs improved sepsis mortality 11, Scumpia and colleagues found that endogenous Tregs did not play a role in sepsis mortality in mice 12.

In humans, Venet and colleagues showed that, although the percentage of Tregs increased during sepsis, absolute cell count was decreased 10. Moreover, these authors recently found an inverse correlation between Tregs and lymphocyte proliferation capacity 13.

In the present study we report that the relative increase in Tregs appears lately during septic shock and show that this phenomenon is not specific to sepsis but also occurs during non-septic shock. We also observe that the percentage of Tregs is inversely correlated with severity at admission, although with no relation to cytokine plasma concentrations. Finally, we demonstrate that Tregs do not influence survival or cytokine production in a polymicrobial model of septic shock in mice.

From January to June 2007, 43 consecutive patients suffering from shock were included. There were 26 septic shocks (10 of pulmonary origin, 4 peritonitis, 2 endocarditis, 3 urinary tract infections and 7 bacteremia of unknown origin) and 17 non-septic shocks (10 cardiogenic, 3 hemorrhagic, and 4 toxic). Characteristics of the patient population are shown in Table 1. The high mortality (42%) was expected regarding the SOFA and SAPS II severity scores.

The septic shock patients had lower neutrophils and lymphocytes counts, higher plasma concentrations of C-reactive protein and procalcitonin as compared with non-septic patients.

There is now enough evidence that morbidity and mortality occurring during septic, as well as non-septic, shock may be due to the effect of distinct mechanisms over time: in addition to an overwhelming pro-inflammatory immune response, concomitant counter-inflammatory mechanisms develop that may be responsible for 'immune paralysis' 3. Although naturally occurring Tregs have been found to play major roles in a wide range of disorders, such as tumour progression, inflammation or transplantation tolerance 479, their implication during shock states is unclear.

In agreement with the findings of Monneret and colleagues 1019, we observed that the percentage of Tregs increased as early as three days after the onset of shock, while their absolute number remained lower than in healthy volunteers. This phenomenon has been explained by Venet and colleagues by a decrease of the CD4⁺CD25^- lymphocytes subset 10, most probably through apoptosis. Adding to the similarities that exist between septic and non-septic shock in terms of immune derangement, we found similar patterns of Treg kinetics in infected and non-infected patients. Indeed, immune paralysis is also observed during trauma/haemorrhage or cardiogenic shock as demonstrated by a reduced ability of antigen presentation or ex vivo pro-inflammatory cytokine production 20. Nevertheless, this apparent absence of difference between septic and non-septic patients may also, at least in part, be explained by the small sample size and the variability in each of the measurements.

Although we observed an inverse correlation between severity, assessed by SOFA score or arterial lactate concentration, and percentage of Tregs, the time course of the percentage or absolute number of Tregs was similar between survivors and non-survivors. Most probably, this finding may also be explained by the negative correlation between severity and CD4⁺CD25^- lymphocyte number (data not shown). Nevertheless, only considering the group of septic shock patients, we observed that survivors had higher percentage and absolute numbers of Tregs by day five than non-survivors. Although this finding may suggest a specific effect of Tregs in sepsis outcome, the low number of non-survivors studied at days five and seven preclude any definite conclusion.

The influence of cytokines on Treg proliferation or activation is unknown during shock states, as is the effect of Tregs on cytokine production. Here we were unable to show any correlation between percentage or absolute count of Tregs and plasma cytokine concentration (IL-2, IL-4, IL-5, IL-10, INFγ, TNFα and TGF-β), taken individually or expressed as a Th1/Th2 balance. Therefore, Tregs do not seem to modulate the inflammatory response during shock in vivo. Of note, this absence of relation was found in septic as well as in non-septic patients.

As NK cells have recently been advocated to play a role in sepsis outcome, we also investigated the kinetics of this cell's subset. In contrast to the findings of Giamarellos-Bourboulis and colleagues 21, we observed that the percentage of NK cells was not different between patients and healthy volunteers throughout the study period. This discrepancy may be explained by the unexpectedly low NK percentage (4.12%) found in the six healthy volunteers, and the old age of the patients (77 years) enrolled in the Greek study. Patients with shock presented with a constantly decreased NK absolute count, without any differences between septic and non-septic shock. Finally, NK cell kinetics was similar between survivors and non-survivors. Taken together, these data suggest a marginal role of NK cells on shock outcome, although the relatively small number of patients enrolled precludes definite conclusion.

To further evaluate the role of Tregs on sepsis outcome, we used a well characterized CLP model of polymicrobial sepsis in mice. In accordance with Scumpia and colleagues 12, we observed a rapid increase in Tregs subset among splenocytes in septic animals. However, despite their increased proportion, antibody-induced depletion of Tregs before the onset of sepsis did not alter survival. Again, these data are in line with those of Scumpia and colleagues 12. The originality of the present study stems from the septic shock model we used: mice were fluid resuscitated and antibiotics given, further approaching the complex physiopathology of human sepsis. Therefore, our data are more able to be extrapolated to human septic shock. In accordance with what we observed in humans, Tregs did not appear to influence cytokine production (TNF-α, IL-1β, IL-6) in septic animals.

In contrast, our animal data seem in discordance with those reported by Heuer and colleagues 11. Several important differences between these two studies need to be underlined. First, Heuer and colleagues performed an adoptive transfer of in vitro prestimulated Tregs and then it is not clear whether the protective role on survival was due to Tregs themselves or to the artificial induction of membrane/soluble mediators. Second, the model of CLP we used here appears to be more stringent and thus, a potential protective role of Tregs may have been masked by the severity of the model. Finally, although we used the classical and well-accepted anti-CD25 antibody approach to deplete Tregs, we must acknowledge that anti-CD25 is not specific for Tregs and leads to the depletion of other important cells (activated T cells). Moreover, anti-CD25 antibody may not have completely eliminated CD25^low Tregs.

Taken together, these data argue against a determinant role of naturally occurring Tregs on inflammatory response and outcome during shock states.

• Percentage of Tregs increases early after the onset of shock whatever its etiology, while their absolute number remains lower than in healthy volunteers.

• Percentage of Tregs is inversely correlated to severity of shock, although the time course of percentage or absolute number of Tregs is similar between survivors and non-survivors.

• There is no correlation between percentage or absolute number of Tregs and plasma cytokine concentrations taken individually or expressed as a Th1/Th2 balance.

• In a mouse model of septic shock, antibody-induced depletion of Tregs before the onset of sepsis does not alter survival.

ELISA: enzyme-linked immunosorbent assay; FACS: fluorescent activated cell sorting; FITC: fluorescein isothiocyanate; IFN: interferon; IL: interleukin; NK: natural killer; PE: phycoerythrin; SAPS: Simplified Acute Physiology Score; SOFA: Sepsis-related Organ Failure Assessment; TGF: transforming growth factor; TNF: tumor necrosis factor; Tregs: regulatory T cells.

The authors declare that they have no competing interests.

FH and SG designed the study, enrolled patients, performed experiments, and wrote the manuscript; FM performed experiments; ACP, DB, BL and PEB enrolled patients and analyzed data. All authors read and approved the final manuscript.