We evaluated patients with sepsis and hospital controls who were part of a previously reported study of endothelial function in sepsis ^[25]. Sepsis patients had suspected or proven infection and the presence of two or more criteria for the systemic inflammatory response syndrome (SIRS) within the last 4 hours ^[1]. Severe sepsis patients had organ dysfunction or shock at the time of enrolment according to the American College of Chest Physicians/Society of Critical Care Medicine criteria ^[1], ^[32]. Sepsis severity was estimated using the Acute Physiology and Chronic Health Evaluation (APACHE) II score from the first 24 hours of admission and daily modified Sequential Organ Failure Assessment (SOFA) score ^[33]. Patients were enrolled within 24 hours of ICU admission or within 36 hours of ward admission. Control subjects were recruited from hospital patients who had not met SIRS criteria within the last 30 days and who had no clinical or laboratory evidence of inflammation or infection. Written informed consent was obtained from all participants or next of kin. All sepsis patients had undergone resuscitation and were haemodynamically stable at the time of study enrolment. The study was approved by the Human Research Ethics Committee of Menzies School of Health Research and the Department of Health and Community Services.

Venous blood was collected in lithium heparin tubes at enrolment, day 2–4, and day 7 until discharge from the hospital or death. Whole blood differential white cell counts were measured by Coulter Counter. Lymphopenia was defined as an absolute lymphocyte count less than 1.2×10³/µL ^[34]. Plasma was separated and stored at −80°C.

Lymphocytes were analysed in more detail in a subset of patients from whom samples were processed within 30 minutes of collection, matched for age and gender. Peripheral blood mononuclear cells were separated using Ficoll-Paque™ Plus (GE Healthcare Biosciences, Uppsala, Sweden) and cryopreserved in fetal calf serum and dimethyl sulfoxide. Cells were thawed and stained with appropriate antibodies and analysed on a FACSCalibur flow cytometer (Becton Dickinson Immunocytometry Systems, MA, USA). Antibodies were sourced from Biolegend, California, USA (CD3, CD16 and CD56) or BD Biosciences Pharmingen, California, USA (CD4 and CD8). Results were analysed using Flow Jo software (Tree Star, Oregon, USA). T cells were defined as CD3+ lymphocytes and natural killer cells were defined as CD3−CD16+CD56+ lymphocytes.

Plasma tryptophan and kynurenine concentrations were measured by High Pressure Liquid Chromatography (HPLC; Shimadzu, Kyoto, Japan) with UV (250 nm) and fluorescence (excitation 250 nm, emission 395 nm) detection, using a method modified from van Wandelen and Cohen ^[35]. The kynurenine to tryptophan (KT) ratio was calculated by dividing the kynurenine concentration (µmol/L) by the tryptophan concentration (µmol/L) and multiplying the quotient by 1000 ^[28], ^[36], ^[37].

Concentrations of plasma IFN-γ, IL6 and IL10 were determined using a cytometric bead array (Human Th1/Th2 Cytokine Kit II, BD Biosciences Pharmingen, CA, USA) and a FACSCalibur flow cytometer (Becton Dickinson Immunocytometry Systems, MA, USA). Results were analysed using FCAP array version 1.0.1 (Soft Flow Hungary for Becton Dickinson Biosciences). The lower limits of detection (LLD) of the assay were 2.5 pg/mL for IFN-γ and 10 pg/mL for IL6 and IL10. Values below the LLD were assigned a value halfway between zero and the LLD for statistical analysis. Cytokines were only measured if plasma had been frozen within 2 hours of collection.

Sepsis patients underwent serial bedside reactive hyperemia peripheral arterial tonometry (RH-PAT) measurements at enrolment, day 2–4, and day 7 ^[25]. Control patients had the same assessment at a single time point. RH-PAT (Itamar Medical, Caesarea, Israel) is a non-invasive operator-independent method of assessing endothelial function. Endothelial function is defined by the ability of blood vessels to vasodilate in response to an ischemic stress, which invasive studies have demonstrated to be dependent on endothelial cell NO production ^[38]. RH-PAT is at least 50% NO-dependent ^[39]. RH-PAT uses finger probes to measure digital pulse wave amplitude detected by a pressure transducer ^[40], and has been validated against the more operator-dependent flow-mediated dilatation method ^[41] and with endothelial function in other vascular beds ^[42].

Predefined groups for analysis were severe sepsis, non-severe sepsis (meaning sepsis without evidence of organ dysfunction or shock at enrolment), and hospital controls. Continuous parametric variables were compared using Student's t-test or ANOVA while continuous non-parametric variables were compared using Mann-Whitney, Kruskal-Wallis or Wilcoxon tests as appropriate. Correlations were examined using Pearson's or Spearman's tests for parametric and non-parametric data respectively. As SOFA score was highly right-skewed and no transformation gave a normal distribution, Kendall's tau coefficient for partial correlation was used for multivariate analysis involving SOFA ^[43]. Linear mixed-effects models were used to examine longitudinal correlations. A 2-sided p-value of <0.05 was considered significant. Analyses were performed using Stata version 10.0 (Stata Corp TX, USA) and Prism version 5.01 (GraphPad Software, CA, USA).

The study included 50 patients with severe sepsis, 30 with non-severe sepsis and 40 hospital controls. The three groups did not differ significantly in age or gender (Table 1). Ninety percent of severe sepsis patients and all non-severe sepsis patients were either orally or enterally fed at the time of enrolment; none were receiving parenteral nutrition.

Plasma tryptophan concentrations were significantly reduced in patients with sepsis (p<0.0001, Figure 1 and Table 2). In all sepsis patients, plasma tryptophan was inversely related to SOFA score (r = −0.45, p<0.0001). There was no difference in the baseline plasma tryptophan concentrations among severe sepsis patients who were orally fed (n = 29), enterally fed (n = 16) or who were nil by mouth (n = 5).

Conversely, plasma kynurenine concentrations were elevated in sepsis patients compared to hospital controls (p<0.0001, Figure 1andTable 2). In all sepsis patients, plasma kynurenine correlated with SOFA score (r = 0.34, p = 0.005). As kynurenine is renally excreted and accumulates in renal failure ^[44], ^[45], kynurenine concentrations were tested for relationships with renal impairment. Kynurenine concentrations were significantly higher in patients requiring continuous renal replacement therapy (CRRT) (median 4.5 µmol [IQR 4–5.3]) than in patients not receiving CRRT (2.8 µmol [2.1–4.4]; p = 0.03). In all sepsis patients, kynurenine concentration correlated with plasma creatinine (r = 0.41, p = 0.0002). Nevertheless, the association between plasma kynurenine concentration and SOFA score remained significant even after controlling for creatinine (ktau = 0.24, p<0.01).

IDO activity was significantly increased in sepsis patients (median KT ratio 141 [IQR 64–235]) compared to controls (36 [28–52]) (p<0.0001) and in severe sepsis compared to non-severe sepsis (p = 0.0006, Table 2). The baseline KT ratio correlated with APACHE II (r_s = 0.51, p<0.0001) and total SOFA scores (r_s = 0.54, p<0.0001) in sepsis patients. The KT ratio positively correlated with the hepatic (r_s = 0.28, p = 0.01), renal (r_s = 0.53, p<0.0001), cardiovascular (r_s = 0.42, p<0.0001) and respiratory (r_s = 0.36, p = 0.0009) components of the SOFA score but not the coagulation component (r_s = 0.13, p = ns).

Of the 80 sepsis patients, 6 died by day 28 of the study. The baseline KT ratio in patients who died (median 270 [IQR 102–431] was not statistically significantly different to those who survived (138 [63–232]; p = 0.2).

In longitudinal analysis of severe sepsis, the KT ratio significantly decreased between day 0 (median 162 [IQR 100–286]) and day 7 (89 [65–139]), p = 0.0006); Figure 1D. Among all sepsis patients, decrease in KT ratio correlated with decrease in SOFA score over time (p<0.0001).

Plasma IFN-γ, IL6 and IL10 were all significantly increased in patients with sepsis (Table 2). Plasma concentrations of IL1, IL2, IL4 and tumour necrosis factor-α were not significantly increased in this cohort and were not analysed further. Both IL6 and IL10 positively correlated with SOFA score (r_s = 0.55, p<0.0001 and r_s = 0.55, p<0.0001 respectively) but there was no association between IFN-γ and SOFA score.

In sepsis patients, the KT ratio correlated with plasma IFN-γ (r_s = 0.44, p = 0.0002), IL6 (r_s = 0.49, p<0.0001) and IL10 (r_s = 0.62, p<0.0001). The associations between KT ratio and IL6 and IL10 remained significant after controlling for SOFA score (ktau = 0.30, p<0.003 and ktau = 0.45, p<0.0001 respectively).

In a univariate mixed effects model, the decrease in KT ratio over time correlated with the decrease in IL6 (p<0.0001) and IL10 (p<0.0001) between day 0 and day 7. In a multivariate model, these relationships remained significant after controlling for change in SOFA score (IL6 p = 0.009; IL10 p = 0.02).

Sepsis patients had increased white blood cell counts (p<0.0001) primarily due to increased circulating neutrophils (p<0.05; Table 2), which proliferate in response to bacterial infections ^[46]. Conversely, sepsis patients had significantly lower total lymphocyte counts compared with hospital controls (p<0.0001, Table 2). In all sepsis patients the baseline KT ratio was weakly associated with absolute lymphocyte count (r_p = 0.26, p = 0.02). In a linear mixed effects model, absolute lymphocyte count increased as the KT ratio decreased over time (p = 0.001). This relationship persisted after controlling for SOFA score (p = 0.008). When all subjects were grouped according to lymphopenia, lymphopenic patients (n = 63) had a median KT ratio of 128 [IQR 63–236], compared with 59 [33–86] in non-lymphopenic patients (n = 57) (p<0.0001).

As IDO activity contributes to T cell apoptosis ^[18], we examined the relationship between KT ratio and lymphocyte subsets. Peripheral blood mononuclear cells were analysed from 23 of the 80 sepsis patients whose blood had been processed within 30 minutes of collection. This subset of patients was representative of the cohort in terms of age, gender distribution, total lymphocyte count and KT ratio. In this subset of patients, the KT ratio negatively correlated with absolute numbers of lymphocytes (r_p = −0.54, p = 0.007), T cells (r_p = −0.53, p = 0.01), CD4+ T cells (r_p = −0.50, p = 0.01), CD8+ T cells (r_p = −0.49, p = 0.02) and natural killer cells (r_p = −0.46, p = 0.03) (Table 2).

In sepsis, the KT ratio at baseline correlated inversely with NO-dependent microvascular reactivity (r_s = −0.45, p = 0.001) even after controlling for disease severity (using SOFA score; p = 0.001). In a multivariate mixed effects model controlling for SOFA score, improvement in KT ratio between day 0 and day 7 correlated with improvement in microvascular reactivity (p = 0.001). In all sepsis patients, there was an inverse association between the baseline KT ratio and mean arterial pressure (r_s = −0.29, p = 0.009) and diastolic blood pressure (r_s = −0.29, p = 0.01) but no association with systolic blood pressure.

IDO activity is elevated in sepsis and associated with disease severity, T cell lymphopenia and microvascular dysfunction. Because excessive IDO activity is associated with both immune and endothelial dysfunction, increased tryptophan catabolism may link these two key aspects of sepsis pathophysiology. Modulation of IDO activity warrants investigation as a therapeutic strategy in sepsis.