Stroke affects 15 million individuals annually world-wide, and the risk of having a stroke more than doubles each decade after the age of 55 1 . Most stroke survivors live with residual impairments that diminish independence and quality of life 2 . For older patients with ischemic stroke, post-stroke cognitive impairment (PSCI) is particularly important and frequent, occurring in approximately one third of all patients 3 and having a significant negative impact on rehabilitation outcomes 4 , quality of life 5 and risk of dementia 6 . A recent meta-analysis has confirmed several risk factors, including previous symptomatic stroke, previous asymptomatic stroke seen on imaging, multiple stroke lesions, aphasia, stroke severity, and stroke location, are associated with PSCI as defined by Mini-Mental State Examination (MMSE) scores less than 24, Diagnostic and Statistical Manual of Mental Disorders IV (DSM-IV) or International Classification of Disease-10 (ICD-10) criteria, within 1 year after stroke 7 . The prevalence of post-stroke MMSE scores less than 24, indicative of significant cognitive impairment, was much higher than the prevalence of dementia diagnosed by standard criteria, which included the DSM-IV or ICD-10 7 . The majority of these and other known risk factors for PSCI including older age, lower level of education, family history of dementia 8 are not readily amenable to treatment. Therefore, there is considerable need to identify pathophysiological mechanisms that may contribute to PSCI.

The systemic inflammatory response to acute ischemic stroke involves increases in several pro-inflammatory cytokines and C-reactive protein (CRP) 9 10 11 , which have also been associated with the development of cognitive deficits and dementia in aging populations 12 . We have recently demonstrated a relationship between PSCI and two inflammatory biomarkers, CRP and interleukin-6 (IL-6) 13 . Pro-inflammatory cytokines can activate the indoleamine-2,3-dioxygenase (IDO) enzyme, leading to the depletion of tryptophan (TRP) and the production of kynurenine, increasing the kynurenine/tryptophan (K/T) ratio in peripheral blood, which acts as a clinical measure of IDO activity 14 . Elevations in the K/T ratio and in the concentrations of several kynurenine metabolites have been observed post-stroke, where they have been associated with mortality 15 . While lower peripheral blood tryptophan concentrations or higher K/T ratios have been associated with dementia and the severity of cognitive symptoms in Alzheimer's disease 16 17 , the association between K/T ratios and PSCI has not been investigated.

Tryptophan metabolites along the kynurenine pathway can produce excitatory and oxidative neurotoxicity, but also protect neurons from inflammatory damage and attenuate excitatory neurotoxicity via NMDA receptor antagonism 18 19 20 21 22 23 . Therefore, it is of interest to determine whether kynurenine production might be associated with clinical cognitive outcomes, and if so, at which concentrations.

The purpose of this study was to test the hypothesis that IDO activation is associated with the presence of cognitive symptoms post-stroke. We measured plasma concentrations of kynurenine and tryptophan in acute ischemic stroke patients and explored the relationship between the K/T ratio and cognitive deficits, as measured by the MMSE, a commonly used cognitive screening instrument 24 .

A total of 41 patients (mean ± SD age 72.3 ± 12.2 years, 53.7% male) were recruited. Tryptophan and kynurenine chromatograms appeared as previously described 33 . The mean K/T ratio was 67.40 ± 42.46 μmol/mmol. Scores on the sMMSE ranged from 13 to 30, with 29.3% of the sample having evidence of significant cognitive impairment (sMMSE ≤ 24). Table 1 reports the results of correlations with sMMSE scores and K/T ratios, as well as between sMMSE scores and demographic and lesion characteristics. In addition to K/T ratio, time since stroke and level of education were associated with sMMSE scores, with a trend for age to be significantly associated (Table 1). No association between lesion volume and sMMSE scores was found.

The bivariate correlation between sMMSE scores and the K/T ratio persisted (β = -.327, P = .037) in a backward stepwise elimination linear regression where age, lesion volume, and stroke severity were removed from the model (adjusted R² = .084). When patients were dichotomized by those in the highest K/T ratio tertile versus others, being in the top tertile emerged as a significant predictor of sMMSE scores (β = -.412, P = .006) when controlling for age (β = -.253, P = .081) with stroke severity (β=-.027, P=0.859) and lesion volume (β=-.066, P=0.659) removed from the final model (adjusted R² = .205, F_1,40 = 6.15, p = .005; Table 2). When exploring gender, time since stroke, level of education, number of cerebrovascular risk factors, history of depression, or CES-D scores as covariates, the association between the K/T ratio top tertile and sMMSE scores remained significant (P < .018). A plot depicting the relationship between the K/T tertiles and sMMSE scores is provided in Figure 1.

In an ROC curve, a K/T ratio of 78.3 μmol/mmol predicted the presence of an sMMSE score ≤ 24 with 67% sensitivity and 86% specificity (Figure 2). The area under the curve was .730, indicating that 73.0% of the subjects will have a K/T ratio that correctly ranks their likelihood of having an sMMSE score ≤ 24.

This study demonstrates an association between elevated K/T ratios and the extent of cognitive impairment among acute ischemic stroke patients. Age, stroke severity and lesion volume did not significantly modify this relationship suggesting that the clinical importance of the K/T ratio may extend beyond their relationships with these established risk factors for PSCI. The association between the K/T ratio and cognition following acute stroke was sufficiently robust to be detected with the MMSE, a practical and clinically meaningful cognitive screening instrument.

Inflammatory activation is a key process in the ischemic cascade that leads to secondary brain damage 39 . We have previously reported that plasma CRP concentrations were able to explain 20% of the variance in MMSE scores among acute stroke patients and that plasma IL-6 concentrations also predicted MMSE scores 13 . Interestingly, high plasma concentrations of the pleiotrophic cytokine, IL-6, have been associated with a greater lesion volume 40 , however lesion volume did not predict sMMSE scores in our population. Pro-inflammatory cytokines induce the expression of IDO which catalyzes the committal and rate-limiting step in the production of kynurenine from tryptophan. In the present study, having a K/T ratio in the top tertile explained 21% of the variance in MMSE scores.

An elevated K/T ratio has been associated with inflammatory and neurodegenerative conditions and with clinically important outcomes, including mortality in the very old 41 , poorer cognitive function in those with Alzheimer's disease 17 , and the severity of depressive symptoms in those receiving interferon therapy 42 or in those with cardiovascular disease 43 . The K/T ratio and poorer MMSE scores may be related to the burden of neurotoxic or neuroactive metabolites such as quinolinic acid (QUIN), an NMDA receptor agonist and excitatory neurotoxin, 3-hydroxyanthranilic acid, an oxidative neurotoxin, or kynurenic acid (KYNA), an antagonist at both NMDA and α7 cholinergic receptors. The production of these neuroactive metabolites has been previously associated with depression 44 45 , cognitive impairment, and deficits in learning, retrieval, and long-term memory 40 46 . For instance, the concentration of KYNA is elevated in the caudate nucleus and putamen of Alzheimer disease patients 47 and it has been suggested that NMDA or α7 cholinergic inhibition by KYNA might impair learning and memory 48 . In a study by Widner et al., Alzheimer's disease patients showed significantly higher peripheral blood K/T ratios as compared to age-matched controls and Alzheimer's disease patients with the lowest MMSE scores had significantly higher K/T ratios than those with the highest MMSE scores 17 . A second study by Gulaj et al. did not find a correlation between K/T ratios and MMSE scores; however they observed an association between the KYNA/KYN ratio and MMSE scores, suggesting a possible protective effect of KYNA relative to the production of kynurenine and other metabolites 16 . This effect may be attributable to opposing effects of KYNA and quinolinic acid at NMDA or α7 nicotinic cholinergic receptors. The association between increased concentrations of kynurenine metabolites, especially KYNA and 3-hydroxyanthranilic acid, and mortality observed post-stroke 15 suggests the potential clinical importance of these metabolites in this population.

Identification of the K/T ratio as a PSCI biomarker may represent an important step in improving stroke outcomes because the kynurenine metabolites might be amenable to pharmacological manipulation 23 . In the central nervous system, kynurenine aminotransferase II, expressed primarily in astrocytes, is responsible for KYNA synthesis while kynurenine 3-monooxygenase (KMO) produces the 3-hydroxykynurenine metabolite that gives rise to QUIN, predominantly from the microglia 49 . In animal models of cerebral ischemia, administration of kynurenine sulfate exerts neuroprotective effects and notably increases cerebral concentrations of KYNA 50 51 . It is now appreciated that cerebral 3-hydroxykynurenine and quinolinic acid concentrations can be manipulated, selectively without altering KYNA concentrations, by inhibition of KMO 49 , an approach which shows neuroprotective effects in animal models of cerebral ischemia 19 . In addition, a newly synthesized compound, ZL006, which acts downstream of NMDA receptor activation, has been demonstrated to ameliorate focal ischemic cerebral damage induced by middle cerebral artery occlusion in mice and rats 52 . The implication of endogenous excitotoxicity in the present study suggests that ZL006 might also be evaluated to improve cognitive outcomes.

In the present study, the K/T ratio was able to identify subjects at risk of significant PSCI (i.e. those with sMMSE scores ≤ 24). While larger studies would be needed to confirm a clinically important cut-off value for the K/T ratio, the ROC presented in the present study suggested a 67% sensitivity and 86% specificity for a K/T ratio cut-off of 78.3 μM/mM. While higher K/T ratios were associated with lower MMSE scores in the entire population, this effect was most significant for subjects with a K/T ratio in the highest tertile. A K/T ratio in this range is comparable to that found in subjects with HIV-1 infection in association with cognitive symptoms and AIDS dementia complex 53 54 55 , but much higher than that found in normal healthy aged controls 15 or subjects with cardiovascular disease without a history of stroke 43 . If replicated in larger populations, these data suggest potential clinical utility of the K/T ratio as a discriminating biomarker.

This study demonstrates a relationship between cognitive impairment and activation of the kynurenine pathway post-stroke. It is, however, limited by its use of a brief cognitive screening instrument and a relatively small sample size. Although stroke severity was assessed and included as a covariate, the timing of administration of the NIHSS relative to the stroke and cognitive assessment varied between assessments. Because NIHSS scores commonly fluctuate post-stroke, this measure may not have been adequate to control for stroke severity. However, in recognition of this, we also evaluated lesion volume and found that it was not significantly related to the severity of PSCI in this population. Additionally, the type of cognitive impairment observed post-stroke varies; often including deficits in measures of global cognition, as well as domain-specific impairments involving executive function, language, visuospatial ability, and memory 56 . Some previously identified risk factors for the development of PSCI, including family history of dementia and individual cardiovascular risk factors 8 , could not be controlled for given the sample size. Although all efforts were made to control for potentially confounding variables, pre-stroke cognitive capacity and white matter disease burden were not assessed. Finally, the cross-sectional design of this study was useful in establishing an initial relationship between PSCI and elevated K/T ratios post-stroke, but subsequent longitudinal studies will be necessary to determine the value of K/T ratio in predicting long-term cognitive outcomes.

In summary, an acute inflammatory response characterized by IDO activation may be relevant to the development of PSCI. Since the neuroactivities of kynurenine metabolites may be amenable to pharmacotherapeutic intervention, the K/T ratio may be a clinically important biomarker. Longitudinal observational and intervention studies will improve our understanding of the role of IDO activity in PSCI.

The authors declare that they have no competing interests with respect to the authorship and/or publication of the article.

This study is based on the original idea of KLL and NH. ABG participated in coordination and ABG and WS performed the statistical analysis and wrote the manuscript. KLL, NH, SEB, GT, and AK made contributions to conception and design and analysis and interpretation of data. RIA provided neuroimaging data. All authors read and approved the final manuscript.