Kynurenic acid signaling expands in human and nonhuman primates and impairs dorsolateral prefrontal cortical cognition that is key to mental illness

Abstract

Cognitive deficits from dorsolateral prefrontal cortex (dlPFC) dysfunction are common in neuroinflammatory disorders, including long-COVID, schizophrenia and Alzheimer’s disease, where impairments are correlated with kynurenine inflammatory signaling. Kynurenine synthesis from tryptophan is increased under conditions of inflammation, then further metabolized to kynurenic acid (KYNA) in brain, where it blocks NMDA and α7-nicotinic receptors (nic-α7Rs). These receptors are essential for neurotransmission in dlPFC, suggesting that KYNA may contribute to higher cognitive deficits in these disorders. The current study employed several methods to examine the expression of KYNA and its synthetic enzyme, KAT II, in primate dlPFC, and to determine its effects on working memory-related dlPFC neuronal firing and cognitive functioning in aging macaques with naturally-occurring neuroinflammation. We found that KYNA, its synthetic enzyme, KAT II, and the gene encoding KAT II (AADAT), have greatly expanded expression in macaque and human dlPFC in both glia and neurons, with AADAT especially prominent in primate neurons compared to rodent PFC. In macaques, like humans, plasma kynurenine/tryptophan ratios increased with age, consistent with age-related increasing inflammation. Local application of KYNA onto dlPFC neurons markedly reduced the delay-related firing needed for working memory via actions at NMDA and nic-α7Rs, while inhibition of KAT II enhanced neuronal firing in aged macaques. Systemic administration of agents that reduce KYNA production similarly improved cognitive performance in aged monkeys. These data show that KYNA inflammatory signaling expands in primate dlPFC, and that inhibition of kynurenine-KYNA production may provide a powerful therapeutic avenue for treating higher cognitive deficits in neuroinflammatory disorders.

Introduction

Kynurenine synthesis from tryptophan is increased under conditions of inflammation [1] (Fig. 1a). Although extensive research has examined kynurenine’s roles in the immune response [2], recent research suggests that it also plays a major role in the cognitive deficits associated with many inflammatory disorders. For example, Post-Acute Sequelae of COVID-19 (PASC, “long-COVID”) is associated with increased kynurenine signaling [3–7], and the cognitive deficits caused by long-COVID correlate strongly with plasma kynurenine levels [3]. A similar relationship is seen in patients with schizophrenia [8], and in aging and Alzheimer’s disease (AD) [9]. These patients all exhibit deficits in the cognitive functioning of the recently evolved dorsolateral prefrontal cortex (dlPFC) which subserves working memory and the executive functions, e.g. [10–17]. However, nothing is known about how kynurenine signaling is expressed in the primate dlPFC, or how it impacts dlPFC neuronal function.

Fig. 1. Kynurenine-KYNA signaling in dlPFC.

a The kynurenine signaling pathway and the pharmacological agents used in the current study. b AADAT (KAT II) expression in mouse frontal cortex and macaque and human dlPFC. Note differences in scale. c KAT II protein expression in macaque layer III dlPFC showing co-expression in a MAP2-expressing pyramidal cell. d ImmunoEM showing KYNA expression (red arrowheads) in astroglial leaflets (green pseudocolor) and in dendritic spines (yellow pseudocolor) near the synapse in macaque dlPFC layer III. Note expression in vesicle-like structures in spines near the PSD (between arrows).

Layer III of the macaque dlPFC contains the recurrent excitatory microcircuits essential to working memory and executive function, allowing neurons to excite each other to maintain firing when memory is needed to keep information “in mind” [18]. The dlPFC contains “Delay cells” that are able to sustain firing across the delay period in a working memory task, e.g. with selective response to a location in space during a visuospatial task [19]. Previous research has shown that the ability to maintain firing across the delay depends on both glutamate stimulation of NMDA glutamate receptors (NMDARs) and cholinergic stimulation of α7-nicotinic receptors (nic-α7Rs), with surprisingly little contribution from AMPA glutamate receptors that usually contribute to glutamate actions [20, 21]. Both NMDAR and nic-α7 R are found within the post-synaptic density (PSD) on layer III dlPFC spines [20, 21], the likely substrate of the recurrent excitatory connections that support delay cell firing. The current study examined whether these circuits are affected by kynurenine signaling.

Kynurenine is actively taken up into brain from blood, and is also synthesized locally in brain, where rodent studies show expression in glia [22]. Kynurenine can be further metabolized to either kynurenic acid (KYNA) by KAT II, encoded by the gene AADAT (Fig. 1a), or to quinolinic acid (QUIN) in a parallel pathway [22]. Longitudinal measures in humans indicate that kynurenine levels increase, and serotonin levels decrease, with advancing age and are associated with increased frailty [23]. There are also reports of age-related increases in KYNA levels in humans and animals [24, 25], including increases in the nonhuman primate cortex [26]. KYNA levels are also increased in brains of those who died from SARS-COV-2 [27], and in the dlPFC of patients with schizophrenia, especially those who exhibit an elevated inflammation e.g. [8], and in AD brain [28]. Thus, it is clinically important to learn how KYNA impacts dlPFC physiology and function.

KYNA and QUIN have opposite effects on NMDAR neurotransmission: QUIN stimulates, while KYNA blocks, NMDARs [29]. KYNA has high affinity for the glycine site on the NMDAR, with much lower affinity for AMPAR and kainate receptors [29]. Especially relevant to dlPFC neurotransmission, KYNA also blocks nic-α7Rs [30]. Based on its NMDAR blocking properties, KYNA has come to be known as the “protective” metabolite, as under extreme conditions of high glutamate release, such as ischemic stroke, or with in vitro models, KYNA blockade of NMDAR can protect neurons from rapid death by apoptosis [29]. However, we have hypothesized that KYNA would be harmful under conditions of normal or reduced glutamate release, especially for dlPFC neurons that require both NMDAR and nic-α7 R neurotransmission.

The current study examined the expression of AADAT in mouse, macaque and human PFC, as well as KAT II and KYNA expression in primate dlPFC, plasma kynurenine levels in macaques across the adult age span, and the effects of KYNA on dlPFC Delay cell firing in young and aged macaques performing a working memory task. Aged macaques are particularly useful for this research, as they naturally develop neuroinflammation, reductions in dlPFC Delay cell firing and cognitive deficits beginning in middle age [31, 32]. Finally, we examined the effects of systemic administration of agents that inhibit KYNA production on working memory performance in aging macaques to facilitate translation to human patients with neuroinflammatory disorders.

Methods

All research was conducted according to USDA and NIH guidelines and approved by the Yale or University of Pittsburgh IACUC (for macaques), the Princeton University IACUC (for mice), and the Harvard Partners Human Research Committee (for humans).

All macaques were housed under standard laboratory conditions with daily veterinary evaluation and daily enrichment including fresh fruits and vegetables and optimal paired housing.

Transcriptomics

We obtained AADAT expression from the following single cell/nucleus RNA sequencing datasets: mouse frontal cortex [33]; dlPFC (BA46) of neurotypical postmortem human donors [34]; and dlPFC (BA46) of macaque monkeys [35]. Preprocessing and clustering assignments were retained as in the original publications.

Anatomical studies

Detailed methods for multiple label immunofluorescence, single label immunocytochemistry and immunoEM, including description of antibodies, can be found in the Supplemental Methods. These studies used dlPFC sections from aged macaques (24-31 y) from the Yale macaque brain bank.

Plasma kynurenine ratios

Plasma levels of kynurenine and tryptophan were assayed in 15 macaques, (n = 7 used in the physiology and behavior experiments), by high performance liquid chromatography using a modified method from Widner et al. [9] (additional details in Supplemental Methods).

Electrophysiology

The electrophysiology experiments were performed with three macaques (Macaca mulatta; Monkey T, age 24, female; Monkey C, age 15, male; Monkey H, age 10, male). Single unit recordings were made from the dlPFC of macaques performing an oculomotor version of a spatial working memory task. Once a Delay cell was identified and characterized, iontophoresis of minute amounts of drug were applied onto the recorded neuron. A brief description of this task and recording paradigm, with images, can be found in the Results (Fig. 2a-c). Drugs used were: Kynurenic acid (Tocris; dissolved at 0.01 M concentration in saline with pH 7-9); PF-04859989 hydrochloride (Sigma-Aldrich; dissolved at 0.01 M concentration in sterile water with pH 3.5-5); D-serine (Tocris; dissolved at 0.01 M concentration in saline with pH 7-9); PHA 543613 hydrochloride (Tocris; dissolved at 0.01 M concentration in sterile water with pH 3.5-5). Note that our previous control experiments showed that saline with the same pH had no specific effects on firing rate or spatial tuning of delay cells. Additional details can be found in the Supplemental Methods.

Fig. 2. KYNA reduces dlPFC Delay cell firing in young macaques.

a-c The paradigm for recording dlPFC Delay cells in macaques performing an oculomotor delayed response task (ODR) with local iontophoretic administration of drug. In the ODR task (panel a), the monkey fixates on a central spot to initiate a trial. A cue briefly (0.5 s) appears in one of eight spatial locations, and the monkey must maintain fixation over the subsequent delay period, holding the spatial location in working memory. The fixation spot then disappears, signaling the monkey to move its eye to the remembered location for juice reward. Single unit recordings are made from the caudal principal sulcal dlPFC, with iontophoretic application of drug onto the recorded neuron (panel b). An example Delay cell is shown with spatially-tuned persistent firing across the delay epoch for its preferred direction (panel c). d An example neuron from a young adult monkey showing that iontophoresis of KYNA markedly reduces Delay cell firing and spatial tuning. e KYNA markedly reduced delay firing and spatial tuning of dlPFC Delay cells in the younger monkey (n = 29).

Drug effects on cognitive performance

The systemic behavioral experiments were performed with 14 Rhesus macaques (9 female, 5 males, age spanning 10-33 years). Rhesus monkeys were pretrained on the delayed response test of visuospatial working memory. Once stable baseline performance was achieved, the monkeys were administered either the KAT-II inhibitor, PF-04859989 (Sigma-Aldrich; 0.003-3.0 mg/kg, sc, 2 h before testing); the IDO inhibitor, INCB024360 (Tocris; 0.1 mg/kg, po 2 h before testing), or the KAT-II inhibitor, N-acetyl cysteine, (Tocris; 1.0-10.0 mg/kg, po, 2 h). All results were compared to vehicle control and monkeys were tested blind to drug conditions. Additional details can be found in the Supplemental Methods.

Results

Expression of AADAT, KAT II and KYNA in PFC

We analyzed single cell RNA sequencing data from mouse frontal cortex, and single nucleus data from the dlPFC of macaque and human to investigate the expression of AADAT, encoding the KAT II enzyme that synthesizes KYNA. AADAT showed marked species differences between mouse mPFC and primate dlPFC. In mouse, there was very low-to-undetectable expression of AADAT with the most expression evident in glia, and limited expression in neurons (Fig. 1b). In contrast, there was much more ( ~ 200x) expression of AADAT in macaque dlPFC, including high expression in astrocytes and oligodendrocytes, as well as in all neuron types (Fig. 1b). Indeed, the cells with the highest levels of AADAT expression were pyramidal cells (layers 3-5) that co-expressed RORB, which is found throughout multiple layers in macaque dlPFC, including CUX2-expressing superficial pyramidal cells e.g. in layer III (Fig. S1). Expression of AADAT was similar in human dlPFC, with relatively high levels in most pyramidal cell and interneuron subgroups, and in glia (Fig. 1b). Thus, there is both a qualitative and quantitative expansion of AADAT expression across PFC evolution.

We assessed KAT II and KYNA protein in macaque layer III dlPFC to confirm the transcriptomic findings. Protein labeling of KAT II in the macaque dlPFC was consistent with the AADAT data, showing KAT II expression in neurons as well as astrocytes in aged macaque layer III dlPFC. Multiple label immunofluorescence (MLIF) showed that KAT II was expressed in pyramidal cells and astrocytes, co-labeled by MAP2 (Fig. 1c, S2) and GFAP (Fig. S2), respectively.

KYNA is also expressed in pyramidal cells and glia in aged macaque dlPFC (Fig. S3). A more in-depth characterization of KYNA localization was performed using immunoEM to visualize its relationship to the glutamate synapse. Nanoscale analysis showed robust expression in astrocytes (Fig. 1di, Fig. S4a-g and Fig. S5b-f), especially in the astrocytic processes immediately adjacent to synapses on spines. KYNA was also seen within a subset of dendrites and spines, including in vesicular-like structures near the PSD (Fig. 1diii, Fig. S4h-j and S5a). A schematized summary of KYNA localization near the glutamate synapse is shown in Figure 1dii.

Plasma levels of kynurenine increase with age in macaques

The kynurenine/tryptophan ratio in plasma showed a significant correlation with age (r = 0.63, n = 15 monkeys, 11-29 years, p < 0.01), increasing from 0.021 ng/ml in an 11-year-old monkey to 0.053 ng/ml in a 29-year-old monkey (Table S1). The average plasma ratio across all monkeys was 0.034, which is remarkably similar to the average plasma ratio of 0.035 across 8089 human subjects [36].

Iontophoresis of KYNA reduces the firing of dlPFC Delay cells needed for working memory

The effects of KYNA on dlPFC neuronal physiology during working memory were examined in three macaques of differing ages: a younger adult macaque (#15 in Table S1, age 10-11 years, male, plasma kynurenine/tryptophan=0.021); a middle-aged macaque (#14 in Table S1, age 15 years, male, plasma kynurenine/tryptophan=0.023); and an aged macaque (#6 in Table S1, age 24-26 years, female, plasma kynurenine/tryptophan=0.044). The monkeys performed an oculomotor visuospatial working memory task (Fig. 2a), while recordings were made from the dlPFC, coupled with iontophoretic administration of drug onto the recorded neuron (Fig. 2b). Delay cells exhibit elevated and sustained neuronal firing across the delay period for their preferred direction, but not other directions (Fig. 2c), and thus are spatially tuned as defined by d’ signal detection analyses.

Iontophoresis of KYNA onto dlPFC Delay cells produced a significant, dose-related reduction in Delay cell firing in all monkeys, with the greatest effects in the younger animal. The striking effects of KYNA are illustrated in a representative Delay cell from the younger macaque in Fig. 2d. The effects of KYNA on 29 Delay cells from this younger animal were highly significant (Fig. 2e; firing rate: F(1,28) = 23.72, p < 0.0001; d’: T_dep = 6.691, p < 0.0001).

KYNA also reduced the firing of Delay cells in the aged macaque, but with less sensitivity. An example of KYNA’s effects on a Delay cell from the aged monkey are shown in Fig. 3a, where application of KYNA at a dose of 15 nA had little effect on delay-related firing (2-ANOVA, Šídák’s multiple comparisons, p = 0.99), while subsequent application of KYNA at 30 nA significantly reduced delay firing for the neuron’s preferred direction only (p = 0.0041), reducing spatial tuning. Consistent results were observed for the entire population of delay cells tested from the aged monkey, where KYNA (30-50 nA) significantly reduced the delay firing of all 19 aged delay cells recorded (Fig. 3b. firing rate: F(1,18) = 42.22, p < 0.0001; d’: T_dep = 7.475, p < 0.0001), while lower doses (10-20 nA) had no effect on Delay cells from the aged monkey (Fig. 3c; T_dep = 0.47, p = 0.384). In contrast, Delay cells in the younger monkey were markedly reduced by very low doses of KYNA (10-20 nA) (Fig. 3c; T_dep = 5.799, p < 0.0001). The lesser sensitivity of neurons in the aged animal is likely due to the greater endogenous expression of KYNA in the aged brain, contributing to reduced neuronal firing.

Fig. 3. KYNA effects on Delay cell firing in the aged macaque.

a An example neuron from the aged monkey showing that iontophoresis of KYNA at 30 nA, but not 15 nA, markedly reduces Delay cell firing. b KYNA (30-50 nA) reduced the delay firing and spatial tuning of Delay cells (n = 19) from the aged monkey. c KYNA at higher doses (30-50 nA) significantly reduced the delay-related firing of dlPFC neurons from either the young or aged monkey, while KYNA at lower doses (10-20 nA) only reduced delay-related firing in the young monkey, not the aged monkey.

The design of the ODR task permitted analysis of KYNA effects on specific epochs that require differing neural operations. The effects of KYNA were most evident in the delay epoch of the working memory task (Fig. S6). Although KYNA significantly decreased neuronal firing across the delay epoch for both the preferred and non-preferred directions, it had much greater effects for the preferred direction. In contrast, there was no direction-specific effect of KYNA on fixation-, cue- or response-related firing (Fig. S6). This pattern is consistent with delay-period firing being especially dependent on NMDAR and nic-α7 R mechanisms [20, 21].

Both NMDAR and nic-α7 R blockade contribute to KYNA’s detrimental actions in primate dlPFC

We further examined the receptor mechanisms underlying KYNA’s detrimental actions in younger adult monkeys. There is longstanding evidence that KYNA blocks the glycine site on NMDAR, which is needed for effective glutamate binding. Thus, KYNA could reduce endogenous NMDAR transmission in dlPFC by blocking this site. D-serine is an agonist at the glycine site; thus, we tested whether KYNA’s detrimental effects could be blocked by co-iontophoresis with d-serine. A single neuron example from the young monkey is shown in Fig. 4a, where d-serine at 20 nA alone and subsequent co-application with KYNA had no clear effect on delay firing (F(2,37) = 0.3975, p = 0.6748) and spatial tuning. However, when d-serine application was terminated, KYNA alone significantly reduced delay firing (F(2,38) = 3.452, p = 0.0419). Similar effects were evident at the population level (Fig. 4b), where co-application of d-serine with KYNA prevented the reducing effects of KYNA (F(2,36) = 0.9167, p = 0.4089); when d-serine was terminated, KYNA alone consistently reduced delay firing (F(1,18) = 18.56, p = 0.0004). These results are consistent with KYNA reducing delay firing through blocking the glycine site on the NMDAR.

Fig. 4. Evidence that both NMDAR and nic-α7 R blockade contribute to KYNA’s detrimental actions in primate dlPFC.

a D-serine is an agonist at the glycine site on the NMDAR. An example neuron from the younger adult monkey showing that co-iontophoresis of d-serine with KYNA prevented the reduction in delay firing that ensued when KYNA was then applied on its own. b The average response of all dlPFC Delay cells in the younger adult monkey treated with d-serine and KYNA, where d-serine prevented the reduction in firing caused by KYNA alone (n = 19 neurons). c The average response of all dlPFC Delay cells treated with PHA and KYNA, where PHA prevented the reduction in firing caused by KYNA alone (n = 17 neurons). d An example neuron from the younger adult monkey showing that co-iontophoresis of the nic-a7R agonist, PHA, with KYNA prevented the reduction in delay firing that ensued when KYNA was then applied on its own.

KYNA has also been shown to block nic-α7 R in some circuits, and as dlPFC Delay cell firing requires nic-α7 R permissive actions for NMDAR neurotransmission, we tested for KYNA interactions at this receptor as well using the nic-α7 R agonist, PHA 543613 (PHA). We applied a low dose (10 nA) of PHA that had little effect on delay firing by itself (control vs PHA: Fig. 4c, population, F(1,16) = 0.8471, p = 0.371; Fig. 4d, single example, F(1,26) = 2.264, p = 0.1445). Co-iontophoresis of PHA with KYNA blocked KYNA’s effects on Delay cell firing and spatial tuning (PHA+kyna vs kyna: Fig. 4c, population, F(1,16) = 25.24, p = 0.0001; Fig. 4d, single example, F(1,25) = 23.94, p < 0.0001), consistent with nic-α7 R blockade also contributing to KYNA’s detrimental actions in primate dlPFC.

Iontophoresis of KAT-II inhibitor enhanced delay firing and spatial tuning of dlPFC Delay cells with greatest effects in the aged monkey

We next examined the effect of the KAT-II inhibitor, PF-04859989 (PF) to see if it would enhance delay-related firing, especially in the aged monkey with naturally-occurring kynurenine inflammation. Indeed, PF had its greatest effects in the aged monkey compared to the young and middle-aged animals (Fig. 5). Single neuron examples following iontophoresis of 30 nA of PF are shown for the aged monkey (Fig. 5a) and the youngest monkey (Fig. 5b). Iontophoresis of PF at 30-40 nA enhanced delay-related firing and spatial tuning, especially in the aged macaque who may have greater levels of endogenous KYNA (Fig. 5c, firing rate: F(1,15) = 14.83, p = 0.001; d’: T_dep = 2.14, p = 0.04). More subtle enhancing effects were seen with the middle-aged (Fig. 5d, firing rate: F(1,7) = 17, p = 0.0044; d’: T_dep = 3.876, p = 0.006) and youngest (Fig. 5e, firing rate: F(1,19) = 10.11, p = 0.0049; d’: T_dep = 2.167, p = 0.043) monkeys, with the enhanced delay firing in the aged monkey significantly greater than that in the young and middle-age monkeys (Fig. 5f; 1-ANOVA, F(2,41) = 11.85, p < 0.0001). This pattern of response supported the hypothesis that endogenous KYNA over-expression in the aged monkey contributes to basal reductions in cell firing. Similar age-related effects of PF on working memory performance were seen with systemic administration.

Fig. 5. The KAT II inhibitor PF-04859989 increases dlPFC Delay cell firing in young and aged macaques with greater efficacity in the aged.

a An example neuron from the aged monkey showing that iontophoresis of PF at 30 nA increased Delay cell firing. b An example of a dlPFC Delay cell from the youngest macaque showing that PF (30 nA) had only subtle effects on delay-related firing. c In recordings of Delay cells from the aged dlPFC, PF (30-40 nA) enhanced delay firing and spatial tuning (n = 16 neurons). d In recordings of Delay cells from the middle-aged dlPFC, PF (30-40 nA) enhanced delay firing and spatial tuning (n = 8 neurons). e In recordings of Delay cells from the young dlPFC, PF (30-40 nA) enhanced delay firing and spatial tuning (n = 20 neurons). f PF produced a significantly stronger enhancement of Delay cell firing in the aged monkey than in the younger animals, consistent with higher plasma kynurenine/tryptophan levels in the aged.

Systemic administration of agents that reduce KYNA production improve working memory performance in aged macaques

The overarching goal of this research is to try to identify pharmacological treatments for neuroinflammatory cognitive disorders. Thus, we examined the effects of systemic administration of agents that reduce the production of KYNA (Fig. 1a) on working memory performance in rhesus monkeys. Rhesus macaques naturally develop working memory deficits with advancing age [32], and we hypothesized that some of these deficits may arise from endogenous KYNA over-expression. Systemic administration of the KAT-II inhibitor, PF, to macaques (10-33 years) improved the performance of the aged monkeys (Fig. 6a-c), with a significant correlation between increased drug efficacy and increased age (Fig.6a, b; r = 0.598, p = 0.039), and the largest effects in the oldest monkeys (Fig. 6c; p < 0.004). These findings are consistent with the greater inflammation [37] and KYNA levels [26] in older animals.

Fig. 6. The effects of systemic administration of the KAT-II inhibitor, PF-04859989 (PF), on working memory performance.

a Systemic injection of PF (0.3 mg/kg, sc, 2 h before testing) on working memory performance in macaques aged 10-33 years (n = 11) significantly correlated with age, with older animals improved. b The effects of the best dose of the KAT-II inhibitor, PF, between 0.003-3.0 mg/kg, (sc, 2 h) on working memory performance in the same animals as in “a”. There was a significant correlation between the degree of improvement and age, with older animals improved. c A best dose of the KAT II inhibitor, PF (0.003-3.0 mg/kg, sc, 2 h), significantly improved working memory performance in aged macaques ( > 21 years, n = 8). d Systemic administration of the IDO inhibitor, INCB024360 (0.1 mg/kg, po 2 h before testing) significantly improved working memory performance in aged monkeys. e N-acetyl cysteine (NAC) inhibits KAT II in addition to its anti-oxidant actions and is approved for human use. NAC (1.0 mg/kg, po, 2 h) significantly improved working memory performance in aged macaques (20-33 years; n = 10); a best dose between 1-10 mg/kg produced optimal performance for each animal.

Additional experiments focused on aged macaques with elevated plasma kynurenine/tryptophan levels. Systemic administration of the IDO inhibitor, INCB 024360 similarly improved working memory performance compared to vehicle control (Fig. 6d; T_dep p = 0.015). Finally, we tested N-acetylcysteine (NAC), as this agent is already approved for human use, and has recently been shown to inhibit KAT II actions and reduce KYNA levels, in addition to its anti-oxidant properties [38]. An acute dose of NAC between 1-10 mg/kg improved working memory performance in aged macaques (Fig.6e; T_dep p < 0.001), e.g. with significant improvement at 1 mg/kg compared to vehicle control (Fig.6e; T_dep p = 0.02).

Discussion

Summary

The current study found a large expansion of KAT II/KYNA signaling in the primate dlPFC, with expression in both neurons and glia, where KYNA produced a marked loss of neuronal firing needed for working memory and higher cognition. The loss of firing arose from KYNA blocking both NMDAR and nic-α7 R, the receptors essential to dlPFC neurotransmission [20, 21]. The immunoEM showed that KYNA is positioned to block these receptors, as it is concentrated in glial leaflets and in dendritic spines near the synaptic cleft, including possible vesicular release from spines (schematized in Fig. 1dii). Inhibition of KYNA synthesis enhanced Delay cell firing in aged macaques, and systemic administration of agents that inhibited KAT II or kynurenine synthesis improved working memory in aged macaques with naturally occurring KYNA expression. These data encourage the development of IDO or KAT II inhibitors for the treatment of inflammatory cognitive disorders such as long-COVID and schizophrenia.

Species differences- expansion of kynurenine KAT II signaling in primates

The current study found striking species differences in the expression of AADAT encoding the synthetic enzyme for KYNA, KAT II. In general, there was very low expression in mouse frontal cortex, with expression focused in glia and some interneurons. In contrast, there was much higher expression in macaques and humans, with the largest levels in a subgroup of pyramidal cells, followed by high levels in astrocytes and oligodendrocytes. These data are consistent with a previous study showing that AADAT expression in astrocytes is increased almost 80-fold in humans compared to mice [39], and that KYNA levels in PFC are 10-20 times greater in humans than rats [40]. The current mouse data are also consistent with protein expression in mPFC of rodent, localized in glia [41, 42], and some interneurons [42]. In contrast, the current study found extensive AADAT expression in both neurons and glia in macaque and human dlPFC, with excitatory neurons having the highest levels.

The large species differences in the capacity to synthesize KYNA are intriguing, given that KYNA blocks NMDAR, and the expression of NMDAR-GluN2B (GRIN2B) also expands in primate dlPFC, increasing across the cortical hierarchy and across primate dlPFC evolution [43–45]. In contrast, rodent mPFC neurons are largely dependent on AMPAR as well as NMDAR [46], and systemic kynurenine produces subtle working memory impairment [47]. Thus, KYNA inflammatory signaling in primates may have expanded in concert with increased NMDAR actions to ensure that dlPFC was silenced under inflammatory conditions when energy is needed elsewhere. The more subtle effects in rodents may help to explain the relative paucity of mechanistic research on how kynurenine inflammatory signaling impacts cognition. The striking species differences in neuronal AADAT expression -mostly interneurons in mice, mostly excitatory cells in human and nonhuman primates- suggests that the overall qualitative effects of KYNA inflammation may also differ between species, with a blockade of NMDAR excitation of GABA interneurons in mice leading to overexcitation of circuits, and a predominant blockade of NMDAR excitation of pyramidal cells in primates leading to underexcitation of circuits e.g. as is seen in schizophrenia [10].

The expansion of KAT II/KYNA signaling in primates may also have great relevance to perinatal inflammatory insults and the development of cortical circuits, as NMDAR are required for the creation of appropriate cortical connections [48]. Thus, expansive KYNA blockade of these receptors while cortical connections are forming may produce wide-ranging deficits in connectivity that could contribute to intellectual disability and other mental disorders, which may be more subtle in rodent models. Indeed, recent analyses suggest that maternal inflammation may contribute to brain abnormalities, e.g. in autism spectrum disorders, through this mechanism [49].

A new view on kynurenic acid as the “protective” metabolite

Traditionally, kynurenine signaling has been examined within the context of the “excitotoxicity” that can occur with very high levels of glutamate release, and where increased kynurenine metabolism to QUIN further drives NMDAR stimulation, calcium entry and overload of mitochondria, and subsequent apoptotic cell death e.g. [29]. Under these conditions, KYNA protects neurons from overexcitation and apoptosis by blocking calcium entry through NMDA [29]. This mechanism is thought to mediate the neuronal death that occurs during a stroke, or other severe, acute injuries to the nervous system [50]. Modeling of these toxic events is usually performed in rodent models, or in cell cultures made from rodent neurons, where healthy neurotransmission has a major AMPAR component.

The current study shows that in the recently evolved dlPFC circuits that rely heavily on NMDAR and nic-α7 R neurotransmission for healthy function, KYNA’s blockade of NMDAR and nic-α7 R is not always protective, but instead, can contribute to loss of firing and impaired cognition. This would be particularly true under conditions of normal or reduced glutamate release, as is likely the case with aging [31], and in long-term inflammatory disorders where pathology occurs more gradually, e.g. with loss of neuronal firing, atrophy, and slow, autophagic degeneration rather than apoptosis. As these conditions are common in human age-related and inflammatory disorders including AD, the current data caution that a revised, more complex view of KYNA actions is needed to understand and treat cognitive deficits in humans.

Relevance to cognitive deficits in long-COVID (PASC)

The cognitive deficits of long-COVID consistently target dlPFC functioning e.g. [16, 17, 51–53], including altered dlPFC activity e.g. [54]. Patients who died from COVID have increased kynurenine signaling [55], including increased KYNA in brain [27]. Fluid biomarker data from patients with long-COVID are just beginning to emerge, but recent findings show that plasma kynurenine levels correlate with symptoms of cognitive deficits [3], as well as depression and anxiety [4, 56], the latter which may be related to decreased PFC top-down control of emotion [57]. Interestingly, KYNA can perpetuate kynurenine signaling by activating IDO metabolism of tryptophan [55], as schematically illustrated in Fig. 1a, and this may sustain symptoms even after the initial cytokine inflammatory response is over. The current data show that the increased plasma kynurenine levels consistently found in patients with long-COVID e.g. [3, 4, 55] could directly impair higher cognitive functioning and top-down regulation of emotion through active kynurenine uptake into the dlPFC and its metabolism to KYNA, reducing dlPFC neuronal firing needed for cognition and top-down control. This interpretation would be consonant with the report by Cysique et al, where cognitive deficits in long-COVID correlated best with levels of plasma kynurenine [3].

Relevance to age-related cognitive dysfunction and AD

Longstanding data from rodents, macaques and humans have shown increased KYNA levels in CNS with advanced age [24–26], consistent with greater inflammation with age [9, 58]. The current data are consonant with these findings, as aged macaques were more responsive to KAT-II inhibitors than younger adult macaques, and showed greater levels of plasma kynurenine. The improvement in both Delay cell firing and working memory performance in aged macaques treated with KAT-II inhibitors suggests that endogenous KYNA contributes to cognitive deficits with age. The improved working memory with PF in the current study is coherent with previous data showing that this compound partially protected against working memory deficits induced by the NMDAR antagonist, ketamine, suggesting competitive receptor interactions [59]. Although the iontophoretic data suggests that improved cognition arises from drug actions in the dlPFC, these compounds may also act at other brain regions and in the peripheral nervous system to enhance working memory performance. NAC may also improve performance by protecting mitochondrial function.

KYNA is also increased in early stages of Alzheimer’s disease [28], where increased kynurenine and its metabolites in plasma correlate with measures of Aβ and neurofilament light chain assays of degeneration [60]. IDO expression is also increased near plaques and tangles in AD brain samples [61], consistent with inflammatory contributions to sporadic AD as well as other neurodegenerative disorders. In this regard, it is of interest that patients who died of severe COVID infection exhibited high levels of phosphorylated tau as well as elevated KYNA in their brains [27].

Relevance to schizophrenia

Schizophrenia is increasingly linked to increased inflammatory signaling e.g. [62] which may contribute to decreased dlPFC dendrites and spines [63]. Recent data suggests that KYNA increases synaptic pruning [64], suggesting that this mechanism may contribute to atrophy as well as cognitive deficits in patients with schizophrenia. Postmortem brain analyses have shown elevated KYNA levels [8, 40, 65], and increased TDO and KAT II message in the dlPFC of patients with schizophrenia, especially in those with high cytokine levels [8], Studies of living patients found elevated kynurenine in the plasma of those with high cytokines which correlated with impaired attentional regulation and reduced dlPFC volume [8]. Thus, KYNA may have a significant contribution to dlPFC deficits in patients with schizophrenia and inflammatory burden.

Taken all together, these findings indicate that KYNA has detrimental actions in primate dlPFC, and suggest that agents that inhibit the production of kynurenine and/or KYNA would be helpful in treating the cognitive deficits of neuroinflammatory disorders, including long-COVID [66]. Currently, there are no selective TDO, IDO or KAT-II inhibitors approved for human use. However, recent data show that NAC, in addition to its anti-oxidant properties, inhibits KAT-II [38], and is being tested for potential protective effects in schizophrenia where it may improve working memory [67, 68]. Open-label studies suggest it may also help longstanding traumatic brain injury [69] or long-COVID [70]. The current data suggest that more targeted compounds may be even more effective, especially given the expansion of the KYNA inflammatory pathway in the primate dlPFC.

Author contributions

MW and AA conceptualized the overall project and wrote the manuscript with input from all co-authors. SY, DD, FK, EL, EW, AM, GA, VG and MW performed experiments, data collection and analysis in consultation with GG, DL, SM, AA. All authors reviewed and approved the final manuscript.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41380-025-03425-y.