OF MICE, RATS AND MEN: REVISITING THE QUINOLINIC ACID HYPOTHESIS OF HUNTINGTON’S DISEASE

Abstract

The neurodegenerative disease Huntington’s Disease (HD) is caused by an expanded polyglutamine (polyQ) tract in the protein huntingtin (htt). Although the gene encoding htt was identified and cloned more than 15 years ago, and in spite of impressive efforts to unravel the mechanism(s) by which mutant htt induces nerve cell death, these studies have so far not led to a good understanding of pathophysiology or an effective therapy. Set against a historical background, we review data supporting the idea that metabolites of the kynurenine pathway (KP) of tryptophan degradation provide a critical link between mutant htt and the pathophysiology of HD. New studies in HD brain and genetic model organisms suggest that the disease may in fact be causally related to early abnormalities in KP metabolism, favoring the formation of two neurotoxic metabolites, 3-hydroxykynurenine and quinolinic acid, over the related neuroprotective agent kynurenic acid. These findings not only link the excitotoxic hypothesis of HD pathology to an impairment of the KP but also define new drug targets and therefore have direct therapeutic implications. Thus, pharmacological normalization of the imbalance in brain KP metabolism may provide clinical benefits, which could be especially effective in the early stages of the disease.

1. Excitotoxins and Huntington’s disease: the beginning

Huntington’s disease (HD), originally termed Huntington’s chorea because of the characteristic involuntary movements shown by affected individuals, is a chronic neurodegenerative disorder, which is inherited in an autosomal dominant fashion. Overt motor symptoms usually develop in mid-life, and patients succumb to the disease after another 15–20 years on average. Long believed to be an exclusive basal ganglia disease because of the substantial and progressive neostriatal shrinkage and the massive dilation of the adjacent lateral ventricles, the neuropathology of HD is now known to involve a large number of brain regions and neuronal populations (Jackson et al., 1995; Rosas et al., 2008; Vonsattel et al., 1985).

For a century following the first description of the disease by George Huntington in 1872 (Huntington, 1872), speculations about the cause of neuronal loss in HD were sporadic and mostly based on clinical observations, occasionally supported by insights from microscopic studies of HD brains (Barbeau et al., 1973). Several of these hypotheses, which focused, for example, on the excessive glucose consumption of patients and on abnormal iron deposits in the basal ganglia, remain attractive to this day and may eventually turn out to be useful for a comprehensive understanding of HD pathophysiology. It was not until the 1970s, however, that a new concept provided a plausible mechanism for the characteristic nerve cell death seen in the disease.

As is often the case in the biomedical sciences, this breakthrough in HD research was made possible by disparate and seemingly unrelated findings. The first clue came from neurochemical post-mortem studies of the HD striatum, which revealed a peculiar sparing of dopaminergic afferent fibers even in the final stages of the disease, when intrinsic striatal neurons, identified by measuring GABA and its biosynthetic enzyme glutamate decarboxylase, were dramatically depleted (Bernheimer et al., 1973; Bird and Iversen, 1974; McGeer et al., 1973). This at the time unprecedented neuropathological feature was reminiscent of the “axon-sparing” lesions caused by glutamate and other excitatory amino acids, which John Olney had described in a series of publications in the late 1960s and early 1970s (Olney et al., 1971; Olney and Sharpe, 1969). In the course of these studies, which were mainly concerned with neuronal damage in the hypothalamus and the retina, Olney made two seminal observations: a) axons of extrinsic origin do not degenerate at the lesion site; and b) the neurotoxic potency of acidic amino acids parallels their neuroexcitatory efficacy. Subsequently, he introduced the operational term “excitotoxicity” to describe neuronal lesions that were characterized by the survival of myelinated axons “en passage” and afferent axon terminals (Olney, 1974). The corresponding excitotoxins ranged from endogenous metabolites such as glutamic, aspartic and cysteine sulfinic acid to more potent exogenous compounds, including the seaweed-derived kainic acid, which was found to be approximately 500 times more effective than glutamate both as a toxin and as an excitant (Olney et al., 1971; Olney et al., 1974). Notably, however, the excitotoxic concept remained merely correlative in nature. No attempts to study the mechanism underlying the characteristic features of excitotoxic lesions were made in those early years when excitatory amino acid receptors were still beyond the reach of neurochemists and molecular biologists.

Interest in excitotoxicity increased substantially in the mid-1970s following the demonstration by Coyle and Schwarcz (Coyle and Schwarcz, 1976), and shortly thereafter by McGeer and McGeer (McGeer and McGeer, 1976), that stereotaxic microinjections of kainic acid cause selective, axon-sparing lesions in the rat striatum. These studies not only introduced kainate as a distinct lesioning tool that was easy to use and applicable to a wide variety of experimental situations (McGeer et al., 1978) but also, based on the duplication of the earlier post-mortem findings, provided a novel animal model of HD (Figure 1). In relatively quick succession, a number of laboratories described further morphological and neurochemical features of the kainate-lesioned striatum and also described motor and other behavioral abnormalities in these animals (Divac et al., 1978; Mason et al., 1978; Sanberg et al., 1979). Jointly, these studies resulted in a remarkably consistent catalog of data, which not only confirmed the similarities between HD and the new animal model (Coyle et al., 1983) but also prompted investigators to study additional human neurodegenerative diseases based on excitotoxic lesions of other brain structures (Bergman et al., 1990; Flicker et al., 1983; Wenk et al., 1984).

Figure 1.

Schematic representation of the “axon-sparing”, excitotoxic nerve cell loss in the HD neostriatum. Degenerated neurons in this simplistic model are indicated by dotted lines.

2. The Swedish connection: rotational behavior and ibotenic acid

2.1. A dopaminergic component of excitotoxicity

The postdoctoral tenure of one of us (R.S.) at the Karolinska Institute in the late 1970s coincided with two significant advances in the excitotoxic model of HD. The first innovation was the development of a rapid, behavioral in vivo test for assessing excitotoxic striatal damage in unilaterally lesioned animals. The underlying idea was an offshoot of the work of Ungerstedt, who had demonstrated a few years earlier that systemic administration of dopaminergic agonists in rats with a unilateral striatal dopamine depletion causes the animals to rotate towards the hemisphere that shows lesser dopamine receptor activity (Ungerstedt, 1971). Since a focal kainate injection kills intrinsic neurons carrying dopamine receptors (Schwarcz et al., 1978), unilaterally lesioned rats have a hemispheric imbalance of dopamine function. Consequently, systemic administration of a dopamine receptor agonist such as apomorphine causes these animals to circle towards the side containing the lesioned striatum (Figure 2). The speed of these rotations is roughly proportional to the lesion size and therefore provides a useful quantitative measure of excitotoxic striatal neurodegeneration (Schwarcz et al., 1979a; cf. also Herrera-Marschitz et al., this volume).

Figure 2.

Rotational behavior in rodents with unilateral striatal excitotoxic lesions. Degenerated neurons are indicated by dotted lines. The arrow indicates the rotational direction following the systemic administration of a dopamine receptor agonist.

2.2. Ibotenate: a new and improved excitotoxin

The second advance from this period was triggered by the realization that the kainic acid model of HD had substantial shortcomings, which were related to the pronounced convulsive properties of the toxin. Thus, intrastriatal kainate injections not only cause epileptiform activity, which is rarely observed in adult-onset HD (Bruyn, 1968), but also result in seizure-induced neurodegeneration in the hippocampus, the olfactory cortex and other limbic structures, which suffer only little or no cell loss in HD (Olney and de Gubareff, 1978a; Schwarcz and Coyle, 1977). Studies at the Karolinska Institute revealed that another exogenous excitotoxin, ibotenic acid, was devoid of these unbecoming properties (Figure 3). Originally isolated from the mushroom Amanita muscaria (Eugster, 1968), ibotenate caused well-circumscribed neuronal loss upon focal injection into various brain regions and was therefore soon recognized as a superior tool for producing excitotoxic lesions in rats (Schwarcz et al., 1979b) and other experimental animals (Ciaramitaro et al., 1997; Wenk, 1993). Ibotenate-induced striatal lesions also created an improved animal model of HD, which became widely used – maybe most famously by Björklund and his collaborators in their first attempts to develop cell transplantation therapies for HD (Isacson et al., 1984).

Figure 3.

Chemical structures of kainic, ibotenic, glutamic and quinolinic acids.

The differences between kainate and ibotenate were not limited to the advantages that the latter provided as a research tool. For example and in contrast to kainate (Biziere and Coyle, 1979; McGeer et al., 1978), ibotenate-induced lesions were not dependent on the integrity of glutamatergic fibers, suggesting that endogenous glutamate was a critical component of kainate – but not ibotenate – neurotoxicity (Köhler et al., 1979). Moreover, while the developing striatum was found to be remarkably resistant to kainate (Campochiaro and Coyle, 1978), no such age-dependent vulnerability was seen with ibotenate (Steiner et al., 1984). These and other distinctions between the two toxins, which included different susceptibilities of neuronal populations throughout the brain (Schwarcz and Köhler, 1983), prompted researchers to explore the nature of excitotoxic neuron death in greater detail. In addition to scientific curiosity, it was also hoped, of course, that a better understanding of the mechanisms underlying diverse excitotoxic phenomena might provide critical clues regarding the pathophysiology of HD and, possibly, other neurodegenerative diseases (Coyle et al., 1977).

3. Glutamate receptors and neuroprotection

3.1. Glutamate receptor subtypes

The development of increasingly selective antagonists in the 1970s allowed the study and categorization of ionotropic (i.e. cation-permeable) excitatory amino acid receptors, which had been previously defined exclusively by electrophysiological means (Curtis et al., 1972; Curtis and Watkins, 1960). By the end of the decade, work by the teams of McLennan and Watkins had identified three distinct receptor subtypes, which were named after their model agonists kainate, quisqualate and N-methyl-D-aspartate (NMDA) (Davies et al., 1979; Hall et al., 1979). This classification was subsequently supplanted by the use of α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) in place of the less specific quisqualate (Honoré et al., 1982) and further refined by the molecular revolution of the 1990s (Seeburg, 1993). Awareness of these three classes of pharmacologically defined glutamate receptors stimulated research into their physiological differences and led to first attempts to describe the anatomical localization of the individual receptors using experimental lesions and the recently developed methods of receptor binding and autoradiography. These studies revealed the existence of both pre- and postsynaptic receptors and, in particular, a heterogeneous distribution of the different receptor subtypes (Biziere and Coyle, 1979; Greenamyre et al., 1984; Monaghan et al., 1983; Schwarcz and Fuxe, 1979; Vincent and McGeer, 1979). This, in turn, led to the idea that the subtyping of glutamate receptors might be used to provide a rational explanation for the qualitative differences in the excitotoxic properties of kainate, which selectively activates its own receptor, and ibotenate, which at the time was believed to be a specific NMDA receptor agonist (Curtis et al., 1972). For example, it was proposed that the prevention of striatal kainate neurotoxicity by prior surgical removal of cortical input was causally related to the presence of functional kainate receptors on these glutamatergic afferents (Biziere and Coyle, 1979; Ferkany et al., 1982). The demonstration that the addition of glutamate can partly restore kainate-induced excitotoxicity in the deafferentated striatum further supported the relevance of these presynaptic receptors and at the same time suggested an adjuvant role of endogenous glutamate in the toxic effects of kainic acid (Biziere and Coyle, 1979). In contrast, NMDA receptors are mainly localized on intrinsic striatal neurons (Samuel et al., 1990), which could account for the fact that ibotenate-induced lesions are not affected by the prior removal of glutamatergic nerve terminals of distant origin (Köhler et al., 1979). These findings not only confirmed the need to elaborate the link between excitatory amino acid-mediated excitation and neurotoxicity in greater depth but also raised the question which, if any, of these experimental toxins provided the best animal model of HD.

3.2. Mechanism(s) of excitotoxic neurodegeneration

It soon became obvious that a better understanding of the cellular and molecular mechanisms underlying selective, axon-sparing neuron death was needed to test the construct validity of the excitotoxic HD model. In the early days, these studies were dominated not only by receptor pharmacologists (see above) but also by microscopists, who set out to define the sequence of degenerative events that are initiated after neural tissue is exposed to an excitotoxic insult. These analyses revealed that acute swelling of dendritic spines and branches constitute the first detectable signs of structural damage (Olney et al., 1979a), suggesting the critical importance of glutamate receptors situated on dendritic elements. Since asymmetric, i.e. excitatory, axo-dendritic synapses are placed in close apposition to presynaptic glutamatergic axon terminals (Sharpe and Tepper, 1998), however, these observations did not resolve the question whether the neurotoxic cascade is triggered by presynaptic receptors stimulating glutamate release or by direct activation of postsynaptic receptors. At the same time, it also became quite clear that a deeper awareness of the molecular events following both normal and excessive glutamate receptor activation was needed to properly evaluate the clinical ramifications of the excitotoxic concept.

In the many years since, these mechanisms have been elaborated in detail using various in vitro preparations and, with increasing frequency, informative in vivo paradigms. The consensus reached by these studies, which have become exceedingly complex in nature and now increasingly focus on a plethora of intracellular processes, is that pathology is initiated by a glutamate receptor-mediated, excessive influx of Ca²⁺ ions (Choi, 1985), which over a period of hours leads to hyperphysiological Ca²⁺ concentrations within the cell. Elevated Ca²⁺ levels then set off a cascade of events – both destructive and protective –, which eventually decide the fate of the affected neuron. These events have been dissected studiously and found to depend to a significant extent on the circumstances and conditions of the excitotoxic insult, and on the energy status and developmental stage of the susceptible neurons. Researchers also identified remarkable and poorly understood species and sex differences, and demonstrated that the mechanism of excitotoxic cell death can change, sometimes significantly, in abnormal (i.e. sick) organisms. Description of these complex events, their relation to apoptotic cell death, contributions of additional membrane proteins, and discussion of differentiating features, are beyond the scope of this article. For further information, the interested reader is referred to authoritative summaries and reviews of the subject (Aarts and Tymianski, 2004; Beal, 1992; Besancon et al., 2008; Bossy-Wetzel et al., 2004; Fagni et al., 1994; Greene and Greenamyre, 1996; Johnston, 2005; Leist et al., 1997; Mattson, 2007; Nicotera and Lipton, 1999).

3.3. An offshoot of the excitotoxic concept: developing the idea of neuroprotection

If excitotoxic mechanisms are indeed causally involved in pathological nerve cell loss, any intervention – pharmacological or otherwise – targeting the chain of events leading to neurodegeneration should prevent or arrest the neuronal demise. Although this conclusion seems trivial in hindsight, its therapeutic implications are substantial and have guided many drug discovery programs and clinical strategies during the past decades. First proposed at a Wenner-Gren symposium in Stockholm in 1978 (Schwarcz et al., 1979c), the idea was experimentally verified by demonstrating that the weak NMDA receptor-preferring antagonist D-α-aminoadipic acid and the more potent D-aminophosphonopentanoic acid (AP5) and D-aminophosphonoheptanoic acid (AP7) prevent the neurotoxic effects of systemically administered N-methyl-DL-aspartate (Olney et al., 1979b; Olney et al., 1981) and of intracerebrally applied ibotenate but not kainate (Schwarcz et al., 1982). Subsequently, Meldrum and co-workers tested the anti-excitotoxic concept in animal models of epilepsy (Croucher et al., 1982) and cerebral ischemia (Simon et al., 1984) and showed remarkable anticonvulsant and neuroprotective efficacy, respectively, of NMDA receptor blockade. These two “Science” publications, together with the demonstration that AP7 prevents hypoglycemia-induced neurodegeneration (also published in “Science”; Wieloch, 1985), convinced both laboratory-based and clinical neuroscientists that the targeting of NMDA receptors provided unprecedented, exciting opportunities for the treatment of neurological disorders (Schwarcz and Meldrum, 1985). During the following years, all major pharmaceutical firms with a presence in central nervous system (CNS) research became engaged in the concept of neuroprotection and, together with medicinal chemists in academia, produced an invaluable armamentarium of glutamatergic compounds for pre-clinical and clinical use. These efforts in chemistry were not restricted to targeting the NMDA receptor. By the end of the 1980s, Danish scientists had succeeded in synthesizing the first selective AMPA (“non-NMDA”) receptor antagonists (Honoré et al., 1988), which, interestingly, also afforded impressive neuroprotection in an animal model of cerebral ischemia (Sheardown et al., 1990). It was later shown that NMDA and non-NMDA receptor interact both functionally and physically (Jonas and Spruston, 1994; O’Brien et al., 1998). Moreover, this interplay is influenced by metabotropic glutamate receptors and additional endogenous factors (Bruno et al., 2001; Caldeira et al., 2007; Salt and Eaton, 1996; Stoll et al., 2007) and is especially complex under pathological conditions (Besancon et al., 2008; Choi, 1993; Pellegrini-Giampietro et al., 1994). Thus, in the case of cerebral ischemia, as well as in HD and a host of other acute and chronic pathologies where anti-excitotoxic interventions are indicated on theoretical grounds, translation into clinically useful strategies is complicated, and predictions of success are difficult (Kalia et al., 2008; Sacco et al., 2001; Villmann and Becker, 2007). The problem is further augmented by the fact that glutamate is the major excitatory neurotransmitter in the mammalian CNS, so that any direct manipulation of its receptors might compromise important physiological functions. These and related considerations prompted investigators to examine excitotoxicity from a different perspective, i.e. to ask questions about upstream events leading to excessive glutamate receptor activation. Initially, these studies were designed to identify endogenous agonists of excitatory amino acid receptors and to evaluate their neurotoxic potential.

4. Endogenous excitotoxins and HD

4.1. Glutamate and other acidic amino acids

Based on Olney’s pioneering findings, glutamate was originally regarded the prime suspect in the search for (an) endogenous excitotoxic agent(s). Present in the mammalian brain at millimolar levels, this simple amino acid could be readily envisioned to accumulate in sufficiently high concentrations to overstimulate its receptors and initiate excitotoxic cell death under adverse conditions. However, since the ambient extracellular concentration of glutamate is held in the low micromolar range by powerful neuronal and glial re-uptake processes (Danbolt, 2001), and in view of the old observation that focal injections of even massive amounts of glutamate cause surprisingly little neuron loss in the normal rat brain (McGeer and McGeer, 1976; Olney and de Gubareff, 1978b), it appears unlikely that relatively brief surges in extracellular glutamate would cause irreparable structural damage to the brain. But it is entirely possible that glutamate, maybe in concert with additional endogenous factors, indeed functions as an excitotoxin in situations where its effectiveness is unduly enhanced. Examples, selected here from the earlier literature, include glutamate receptor supersensitivity (Tsai et al., 1995), impairment of glutamate transport mechanisms (Coyle et al., 1977; Cross et al., 1982; McBean and Roberts, 1985) and abnormal glutamate metabolism (Plaitakis et al., 1982).

The question if endogenous acidic amino acids other than glutamate should also be considered possible endogenous excitotoxins with links to human CNS pathology remains unresolved to this day. Although such a connection, namely a causative role of the obscure excitotoxin cysteine-S-sulfate in sulfite oxidase deficiency, was already postulated in the mid-1970s (Olney et al., 1975), only sporadic efforts have been made so far to examine the possibility that pathophysiology is caused by more abundant excitatory amino acids such as aspartate (Bradford and Nadler, 2004), cysteine sulfinate (Iwata et al., 1982) and homocysteate (Klancnik et al., 1992; Olney et al., 1987). Individually or jointly, and especially under conditions favoring their excitotoxic efficacy, these acidic α-amino acids may yet turn out to play important pathogenic roles in HD and other neurological disorders (May and Gray, 1985).

In order to test the possible involvement of glutamate in HD, a flurry of studies in the early 1980s examined its status and effects in patients using cerebrospinal fluid, blood and skin cells, and post-mortem brain tissue. Analyses included the determination of glutamate levels (Bonilla et al., 1988; Ellison et al., 1987; Kim et al., 1980), glutamate uptake (Comings et al., 1981; Mangano and Schwarcz, 1981), glutamate receptors (Greenamyre et al., 1985), intermediary glutamate metabolism (Mangano and Schwarcz, 1982) and glutamate toxicity (Archer and Mancall, 1983; Stahl et al., 1984). In spite of this rather extensive scrutiny, which is still in progress (Hassel et al., 2008), no convincing evidence of glutamate dysfunction was found, leaving the field in a state of confusion (Perry and Hansen, 1990) and, at the same time, stimulating the quest for alternate links between excitotoxicity and HD pathology.

4.2. Quinolinate

A brief communication reporting NMDA receptor-mediated excitation following the ionophoretic application of the tryptophan metabolite quinolinic acid (pyridine-2,3-dicarboxylic acid; QUIN) to rat cortical neurons (Stone and Perkins, 1981) attracted our attention during our own search for novel endogenous excitotoxins. Together with the earlier demonstration of the powerful convulsant effects of QUIN following intracerebroventricular injection in mice (Lapin, 1978), this finding immediately suggested that the compound might have neurotoxic properties, prompting us to examine the consequences of an intrastriatal injection in rats. Indeed, QUIN proved to be a potent striatal excitotoxin, producing the classic, axon-sparing neurodegeneration seen after a focal injection of kainate or ibotenate (cf. above; Schwarcz et al., 1983). In line with a selective activation of NMDA receptors, QUIN-induced lesions were readily prevented by the co-administration of AP7 (Foster et al., 1983). Moreover, like lesions generated by the NMDA receptor agonist ibotenate, neuron loss caused by QUIN was limited to a well-circumscribed area around the injection site.

It soon became apparent, however, that QUIN’s neurotoxic properties were not identical to those of ibotenate and in fact differed qualitatively from those of any other known excitotoxin. Thus, in spite of its lack of affinity for kainate receptors, QUIN shares a number of features with kainate, including diminished excitotoxic potency in the developing striatum (Foster et al., 1983), prevention of striatal neurotoxicity by cortical deafferentation (Schwarcz et al., 1984), and, in the hippocampus, preferential degeneration of pyramidal cells (Schwarcz et al., 1983). Interestingly, QUIN, like kainate but unlike ibotenate, glutamate and cysteine sulfinate – which serve as precursors of muscimol, GABA and taurine, respectively –, does not yield an inhibitory compound upon decarboxylation. Several of its characteristics may indeed be related to the fact that the pyridine derivative QUIN, again like kainate but in contrast to many other excitotoxins, does not contain a free amino group in α position to a carboxylic acid residue and therefore does not strictly qualify as an amino acid.

The QUIN-lesioned rat striatum bears an unusually close resemblance to the HD neostriatum. This was first recognized by Beal and co-workers (Beal et al., 1986), who described the relative sparing of a small population of medium-sized, aspiny neurons containing the peptides somatostatin and neuropeptide Y in the markedly neuron-depleted striatum. These cells, which also contain NADPH-diaphorase, survive in the HD brain (Ferrante et al., 1987a; Ferrante et al., 1987b) but degenerate in experimental animals following a local injection of kainate, ibotenate or NMDA (Beal et al., 1986). These studies were later complemented by other parallels between the HD neostriatum and the QUIN-lesioned rat striatum, such as the resistance of large cholinergic neurons (Davies and Roberts, 1988; Ferrante and Kowall, 1987) and – less unexpectedly – the presence of hypertrophic astrocytes (Björklund et al., 1986; Vonsattel et al., 1985). Since QUIN also produces a HD-like lesion in the primate striatum (Ferrante et al., 1993), and since the chronically QUIN-lesioned rat striatum maintains the characteristic features of the HD neostriatum (Beal et al., 1991), intrastriatal QUIN injections became the experimental model of choice to test novel HD treatments (Beal et al., 1988) and to elaborate the pathological and functional changes seen in the disease (DiFiglia, 1990). Notably, the QUIN model has remained attractive following the identification of the HD gene (Huntington’s Disease Collaborative Research Group, 1993) and is now increasingly considered in conjunction with genetic factors that determine the etiology and pathophysiology of the disease (Pérez-De La Cruz and Santamaria, 2007; see below).

The most interesting conjecture of these studies from a pathogenic viewpoint was the possibility that QUIN might be the long sought-after endogenous excitotoxin. In human and rat brain, QUIN is present at concentrations in the high nanomolar range (Wolfensberger et al., 1983), and prolonged exposure of rat organotypic cortico-striatal cultures to as little as 100 nM QUIN results in characteristic excitotoxic damage (Whetsell and Schwarcz, 1989). We and others therefore argued that a pathological elevation of QUIN levels may produce excitotoxic neurodegeneration in HD. Disappointingly, this hypothesis was initially not corroborated by QUIN measurements in HD. Thus, QUIN concentrations in patients were found to be unaffected (Reynolds et al., 1988) or even reduced (Heyes et al., 1991) in brain tissue, and remained unchanged in the cerebrospinal fluid (Heyes et al., 1991; Schwarcz et al., 1988a). HD patients were also found to excrete normal amounts of QUIN (Heyes et al., 1985). These results significantly dampened enthusiasm for the QUIN hypothesis of HD, especially since they contrasted with parallel reports of dramatic QUIN increases in the cerebrospinal fluid of AIDS patients (Heyes et al., 1989). Unknown to researchers at the time, however, the hypothesis had not been tested with sufficient rigor and was slated to re-emerge a few years later based on new biological and chemical insights.

5. The kynurenine pathway of tryptophan degradation

5.1. Kynurenines

The kynurenine pathway (KP), named after L-kynurenine, the pivotal metabolite placed at a branching point of the enzymatic cascade (Figure 5), is the major catabolic route of dietary tryptophan in mammals. As one of the downstream metabolites in the KP, QUIN is synthesized by a succession of enzymatic steps, which have long been identified and thoroughly examined in peripheral tissues (Schlossberger et al., 1984). In turn, QUIN is broken down further, eventually forming the common cofactor NAD⁺. This “QUIN branch” of the KP is implicated in a variety of immune functions and may play a particularly important role in innate and adaptive immunosuppression (Belladonna et al., 2007; Gonzalez et al., 2008). In spite of the fact that all individual KP enzymes responsible for the conversion of tryptophan to QUIN and beyond are well characterized, however, their involvement in biological processes is just beginning to be understood (Belladonna et al., 2007; Lopez et al., 2006). This active area of research is currently primarily concerned with indoleamine-2,3-dioxygenase (IDO), one of the two enzymes responsible for the initial oxidative opening of tryptophan’s indole ring (the other one being the more substrate-specific tryptophan-2,3-dioxygenase) (Schröcksnadel et al., 2006; Xu et al., 2008).

Figure 5.

The kynurenine pathway of tryptophan degradation in mammalian cells.

Although the link between KP metabolism, the immune system and non-CNS diseases is increasingly appreciated, most speculations in the literature revolve around the question if the KP plays a role in brain physiology and pathology. This focus is justified considering the fact that the KP contains at least two neuroactive metabolites in addition to QUIN, i.e. the QUIN precursor 3-hydroxykynurenine (3-HK) and, in a sidearm of the pathway, kynurenic acid (KYNA) (Figure 5). The (neuro)biological properties of these compounds, which include the generation of free radicals by 3-HK (Vazquez et al., 2000; cf 7.2) and the inhibition of NMDA and α7 nicotinic acetylcholine receptors by KYNA (Hilmas et al., 2001; Kessler et al., 1989), suggest that they might participate actively in a wide range of brain functions that are unrelated to pathological events (Carpenedo et al., 2001; Leipnitz et al., 2007; Rassoulpour et al., 2005; Valko et al., 2007). Like QUIN, however, both 3-HK and KYNA have until now received more attention for their possible involvement in neurological and psychiatric diseases (Moroni, 1999; Müller and Schwarz, 2007; Németh et al., 2006; Oxenkrug, 2007; Ruddick et al., 2006; Schwarcz and Pellicciari, 2002; Stone and Darlington, 2002). Unless they are relevant to the pathophysiology of HD, these pathological consequences of abnormal brain KP metabolism will not be discussed here.

One of the major challenges facing neuroscientists involved in in vivo studies of brain 3-HK, QUIN and KYNA relates to the presumed interplay between peripheral and central kynurenines (the term used to describe all KP metabolites collectively). Whereas tryptophan, kynurenine and 3-HK readily penetrate the blood-brain barrier, no such transport mechanisms exist for the acidic metabolites QUIN and KYNA (Fukui et al., 1991; Gál and Sherman, 1980). This raises the important question to what extent and by which mechanism(s) fluctuations in the levels of circulating kynurenines affect brain function and, in the case of pathological changes, dysfunction (Kita et al., 2002; Wu et al., 2000). This issue has obvious pharmacological and therapeutic relevance since systemic interventions targeting specific KP enzymes in the periphery may not have the desired effects on neuroactive kynurenines in the brain (Reinhard, 2004). Unfortunately, comparatively little is known about the short- and, especially, long-term effects of such peripheral KP manipulations. Since these operations modulate circulating levels of kynurenine in a predictable fashion, however, some useful inferences can be drawn from studies that examine the fate of blood-derived kynurenine in the brain.

5.2. Kynurenine pathway metabolism in the brain

In spite of the fact that the KP accounts for >95% of dietary tryptophan degradation in the periphery (Schlossberger et al., 1984), intracerebral administration of tryptophan normally has very little effect on the flux through the cerebral KP (Heyes et al., 1993). This implies either that the brain activities of IDO or tryptophan-2,3-dioxygenase are exceedingly low or that exogenously applied tryptophan does not have sufficient access to these oxidative enzymes under physiological conditions. In contrast, cerebral KP metabolism is quite responsive to systemically applied tryptophan (During et al., 1989), suggesting, together with the data from Gál and Sherman (Gál and Sherman, 1980), that kynurenine – rather than tryptophan – initiates the metabolic chain leading to the synthesis of neuroactive KP metabolites within the brain. This was verified experimentally in rats by focal intrastriatal injections of radiolabelled kynurenine, which was first taken up (Speciale and Schwarcz, 1990) and then promptly incorporated into several kynurenines, including 3-HK, QUIN and KYNA (Guidetti et al., 1995). In line with expectations, systemic administration of kynurenine, too, increases the brain levels of KP metabolites (Reinhard and Erickson, 1994; Swartz et al., 1990).

Pioneering studies by Heyes and his collaborators in the 1990s revealed that cerebral KP metabolism is qualitatively impaired in several situation of immune compromise. In a series of elegant experiments, the investigators traced some of the impairments to the induction of IDO (Heyes et al., 1992; Heyes et al., 1993; Saito et al., 1991), though they also reported inflammation-related increases in the activity of the downstream KP enzymes kynurenine-3-monooxygenase (KMO), kynureninase and 3-hydroxyanthranilic acid dioxygenase (3-HAD) (Figure 4; Heyes et al., 1993; Saito et al., 1993). Notably, these enzymatic changes were associated with often dramatic increases in the tissue levels of metabolites of the QUIN branch of the KP including, prominently, QUIN itself. On the other hand, the affected brains showed only comparatively minor or no changes in the levels of KYNA and its biosynthetic enzyme, kynurenine aminotransferase. These findings therefore revealed an interesting dissociation between the two KP branches under conditions of immunological compromise.

Figure 4.

Anti-excitotoxic neuroprotection: effects of an intrahippocampal ibotenate injection (5 μg) in the absence (a) or presence (b) of D-amino-phosphonoheptanoic acid (5 μg) in rats. Arrowheads demarcate the tracks of the injection needles. Bar: 500 μM. (From Schwarcz et al., 1982).

Our laboratory had previously identified the synthetic and degradative enzymes of QUIN, 3-HAD and quinolinic acid phosphoribosyltransferase (QPRT), in the brain (Foster et al., 1985a; Foster et al., 1986). Based on studies in the neuron-depleted striatum, where we found substantial increases in both 3-HAD and QPRT activities, we proposed subsequently that the enzymes had to be localized primarily in glial or other non-neuronal cells (Schwarcz et al., 1989). In light of the results of Heyes’ group, it was likely that the QUIN branch of the KP was mainly localized in microglial cells, which are robustly activated by neuronal damage and degeneration, and in response to immunological challenges (Gehrmann et al., 1995). This was later verified directly, using human microglia in vitro (Heyes et al., 1996). Complementary studies revealed that the KYNA branch of the KP, in contrast, was not present in microglia but largely contained in astrocytes (Guillemin et al., 2001; Kiss et al., 2003; Roberts et al., 1992). This physical segregation of the two pathway branches, which can be traced to a lack of KMO in astrocytes (Guillemin et al., 2001), might have important functional ramifications in CNS physiology and pathology.

5.3. Experimental manipulation of kynurenine pathway metabolism

Selective pharmacological or genetic tools are necessary to explore the intricacies of cerebral KP function and to assess the possible role of kynurenines in brain diseases. Examples of specific inhibitors of mammalian KP enzymes had been reported decades ago (Decker et al., 1963; Kido and Noguchi, 1975), and it was shown, in Drosophila, that genetic deletions of individual KP enzymes resulted in phenotypic changes implying neuronal abnormalities (Savvateeva et al., 2000). Beginning in the 1990s and with increasing frequency, medicinal chemists in academia and industry produced new inhibitors, which specifically targeted several mammalian KP enzymes. Specificity was most often achieved by modifying the chemical structure of the natural substrate of an enzyme, and enzyme inhibition was therefore mostly competitive in nature (for review; Schwarcz and Pellicciari, 2002; Stone and Darlington, 2002). In some cases, however, interesting novel structures were discovered by approaches that were not guided by substrate analogy (Lima et al., 2009; Röver et al., 1997). Moreover, endogenous agents such as steroid hormones and cytokines were found to stimulate the activity of individual KP enzymes and may therefore have distinct effects on the function of neuroactive KP metabolites (Andre et al., 2008; Heyes et al., 1997; O’Connor et al., 2008; Oxenkrug, 2007).

Almost without exception, the usefulness of the new synthetic compounds for the study of brain KP metabolism in vivo is restricted by their difficulty to cross the blood-brain barrier after systemic administration (cf. Németh et al., 2005). Experimental studies are also complicated by the fact that pharmacologically induced changes in the function of peripheral KP enzymes can have secondary effects on brain KP metabolism. Probably the best example so far, and one that is of relevance to HD research (see Section 8.), is the substantial increase in brain KYNA levels that is reliably seen after the peripheral application of a KMO inhibitor (Moroni et al., 1991; Pellicciari et al., 1994; Röver et al., 1997; Speciale et al., 1996). As expected, this effect is associated with an accumulation of the enzyme substrate kynurenine in the circulation and, because of the metabolite’s ready penetration from the blood, in the brain. Without further experimentation, it is therefore impossible to unequivocally trace the rise in brain KYNA levels to either peripherally or centrally derived kynurenine, i.e. to determine whether the initial KMO inhibition had occurred intra- or extracerebrally (or both). To circumvent this complexity and its formidable interpretative challenges, which are even more daunting when peripheral or central kynurenines are implicated in neuropathological events, agents targeting individual KP enzymes have been applied directly into the brain. These studies, which have so far been limited to intracerebral injections of enzyme inhibitors, have met with some success (Hamann et al., 2008; Miranda et al., 1997; Walsh et al., 1994) and are beginning to reveal information about the functional interactions of the two KP branches within the brain (Amori et al., in press). However, newly synthesized pharmacological probes have so far not transformed our understanding of cerebral KP metabolism and function, or of the relationship between central and circulating kynurenines. Knowledge in this regard is therefore still rather rudimentary, and improved chemical tools are clearly needed.

Genetic approaches in mammals have met with some success but their impact has so far also been reduced by the fact that brain enzymes were not targeted specifically. For example, mice with a genomic elimination of kynurenine aminotransferase II (a major enzyme of cerebral KYNA synthesis) have remarkable characteristics, which might be causally related to abnormal functions of kynurenines in the brain (Alkondon et al., 2004). Of particular interest with regard to HD pathology, these animals, which have a diminished ability to form the neuroprotective metabolite KYNA, show enhanced neuronal vulnerability to an intrastriatal QUIN injection. This lesion enlargement is not observed when the striatal KYNA deficit in mutant mice is neutralized by pharmacological means. These results indicate that endogenous KYNA controls the vulnerability of striatal neurons to QUIN (Coyle, 2006; Sapko et al., 2006).

6. The kynurenine pathway in Huntington’s disease

6.1. Studies in humans

Although several studies had failed to reveal increased QUIN accumulation in HD (see 4.2), there were early indications of impaired cerebral KP metabolism in the disease. This included our demonstration of elevated 3-HAD and QPRT activity (Foster et al., 1985b; Schwarcz et al., 1988b) as well as contradictory reports of abnormal KYNA levels (Beal et al., 1992; Beal et al., 1990; Connick et al., 1989) in the HD brain. Most interesting in view of later developments, Reynolds and colleagues reported increased brain tissue levels of 3-HK in HD, which were not, however, accompanied by changes in the activity of 3-HK’s degradative enzyme, kynureninase (Pearson et al., 1995; Pearson and Reynolds, 1992). Notably, no metabolic abnormalities indicative of an overactive QUIN branch of the KP have so far been found in the periphery (Foster and Schwarcz, 1985; Heyes et al., 1985; Stoy et al., 2005).

Triggered by evocative studies in genetic mouse models of HD (see below), we showed more recently that the brain levels of both 3-HK and QUIN are substantially elevated – by three to four-fold – in low-grade (grade 0/1) HD brain (Guidetti et al., 2004; Fig. 1). These changes were seen in the neocortex and in the neostriatum but not in the cerebellum, which suffers only little damage in HD. Compared to age-matched controls, brain 3-HK and QUIN levels were either unchanged or even tended to decrease in grade 2 and advanced grade (grades 3–4) HD brain, and cerebral KYNA levels fluctuated only modestly as the illness progressed. These results suggested a brain region-specific involvement of 3-HK and QUIN especially in the early phases of HD pathophysiology, raising the possibility that timely, targeted interventions aimed at reducing flux through the QUIN branch of the KP might slow or prevent disease progression.

6.2 Studies in genetic mouse models

The discovery that HD is triggered by a polyglutamine expansion in Exon 1 of a protein termed huntingtin (htt) (Huntington’s Disease Collaborative Research Group, 1993) allowed the generation of genetic model organisms that can be used to test potential links between mutant htt and KP dysregulation. We therefore examined the tissue levels of the KP metabolites 3-HK and QUIN longitudinally in four informative brain areas (striatum, cortex, cerebellum and hippocampus) in well-established transgenic (R6/2 and YAC128; Mangiarini et al., 1996; Slow et al., 2003) and knock-in (HdhQ92/HdhQ111; Wheeler et al., 2000) mouse models. These analyses, as well as a more limited, earlier study showing a large increase in brain 3-HK levels in another full-length transgenic mouse model (Guidetti et al., 2000), revealed remarkable biochemical abnormalities in all cases (Guidetti et al., 2006).

In R6/2 mice up to 3 weeks of age, 3-HK and QUIN levels were unchanged in all four brain regions. From 4 through 12 weeks of age, i.e. from the time of earliest phenotypic changes until the animals became moribund, mutant mice had significantly higher 3-HK levels in the cortex, striatum and cerebellum than age-matched wild-type controls, though no changes were observed in the hippocampus. These region-specific elevations in 3-HK were recently shown to be associated with, and may in fact be caused by, increases in KMO and reductions in kynureninase activity (Stachowski et al., 2008). In YAC128 mice, which show slower disease onset and a less dramatic phenotype, 3-HK levels in all four brain regions were indistinguishable from controls at 2 and 5 months but were elevated in cortex, striatum and cerebellum (but again not in the hippocampus) at 8 and 12 months of age. This increase was particularly pronounced in the cortex, where 3-HK levels were increased 3–4 times at 12 months of age. In 8- and 12-month-old YAC128 mice, QUIN levels were also significantly elevated, but this increase was restricted to striatum and cortex, i.e. the regions most affected in HD. Finally, both 3-HK and QUIN levels were also significantly elevated in cortex and striatum of 15–17-month-old Hdh^Q92/Hdh^Q111 knock-in mice, which show the most modest phenotypic changes of the three mutants tested.

Collectively, these studies supported the idea that mutant htt induces cellular alterations, which consistently result in the up-regulation of the neurotoxic branch of the KP in striatum and cortex, i.e. the two brain regions that are most prominently affected in HD. The parallels to the human disease are not likely to be accidental and suggest that genetic HD mouse models can be used for the further elaboration of the role of KP metabolites in pathogenesis.

6.3. Lessons from yeast

The yeast Saccharomyces cerevisiae, which has been used to brew beer and wine for millennia, is commonly used in the laboratory, because numerous intracellular processes are well conserved between yeast and higher organisms, in many cases including humans. In addition, this simple eukaryotic microorganism lends itself to easy genetic manipulation including selective insertion or deletion of genes of interest. Using a simple bioassay (cell toxicity), we recently performed a large-scale screen in yeast that identified mutations in 28 genes that suppress toxicity of a mutant htt fragment (Giorgini et al., 2005). One of the most potent of these mutations was in the gene encoding Bna4, the yeast counterpart of mammalian KMO. We therefore proceeded to measure the formation of 3-HK and QUIN in yeast cells. Compared to cells expressing wild-type htt, and in remarkable agreement with the studies in humans and mice described above, 3-HK levels were ~2.2-fold higher in cells expressing mutant htt. QUIN levels, too, were significantly increased. Moreover, as expected, 3-HK and QUIN were not detected in cells lacking Bna4.

We next assessed the viability of mutant htt-expressing strains in which several additional yeast orthologs of KP enzymes were individually deleted. Elimination of BNA1, the yeast equivalent of 3-HAD (the immediate biosynthetic enzyme of QUIN), partially suppressed mutant htt-mediated toxicity. Biochemical analyses of these mutant cells showed that the deletion of BNA1 had prevented QUIN production as predicted and had caused an accumulation of the upstream metabolite 3-HK. This indicated that the remaining toxicity of mutant mtt in this strain might be related to the presence of excessive concentrations of 3-HK (see 7.2. for further discussion).

Elimination of two other KP genes, ARO9 and NPT1, enhanced mutant htt toxicity in yeast. Deletion of ARO9, the yeast equivalent of kynurenine aminotransferase, should block the irreversible formation of KYNA from kynurenine and therefore increase the levels of kynurenine and its toxic downstream metabolites 3-HK and QUIN. In addition, toxicity might be aggravated because the absence of KYNA deprives the cells of a protective free-radical scavenger (Goda et al., 1999). Deletion of NPT1 blocks the formation of NAD+ from nicotinic acid, forcing all synthesis of NAD+ to occur via the QUIN branch of the KP. Consistent with these explanations, 3-HK levels were increased 90- and 15-fold, respectively, and QUIN was elevated 10-fold in Npt1-deficient cells compared to the parental strain expressing mutant htt. Although 3-HK levels were also higher in Aro9-deficient yeast carrying an empty vector, expression of mutant htt in these cells increased 3-HK levels more than 40-fold.

Of the 28 gene deletions identified in the original suppressor screen in yeast (Giorgini et al., 2005), BNA4 was the only one that affected the KP directly. Subsequent studies revealed, remarkably, that 22 of the remaining 27 suppressors also affected KP metabolism, albeit indirectly. Compared with wild-type cells expressing mutant htt, 61% of the suppressors showed significant reductions in 3-HK alone, 75% in QUIN alone, 82% in 3-HK or QUIN, and 54% in both 3-HK and QUIN (Table 1). These results strongly suggested a central role for the KP in mediating the toxicity of mutant htt in yeast. Additional experiments indicated that the reductions in 3-HK and QUIN were controlled by the activity of histone deacetylases on KP promoters. Finally, histone deacetylase-dependent regulation of the KP was also observed in a mouse model of HD, in which treatment with a neuroprotective histone deacetylase inhibitor blocked KP activation in microglia expressing a mutant htt fragment both in vitro and in vivo (Giorgini et al., 2008). These studies provided further support for the hypothesis that KP impairment might play a pivotal role in HD pathology and, in particular, suggest that regulatory mechanisms might be targeted to normalize brain KP metabolism and provide clinical benefits.

Table 1.

Levels of 3-hydroxykynurenine (3-HK) and quinolinic acid (QUIN) in toxicity suppressor yeast strains expressing mutant huntingtion (103Q). Numbers listed refer to the relative percentage of 3-HK and QUIN levels in deletion strains as compared to the parental strain (BY4741). bna4Δ: deletion of the yeast ortholog of kynurenine 3-monooxygenase.

Strain	3-HK	QUIN
wild-type (BY4741)	100.0	100.0
Vesicular transport, vacuolar protein sorting and import
bfr1Δ	49.0*	51.0*
cyk3Δ	46.8*	53.6*
def1Δ	49.6*	87.6*
mso1Δ	60.2*	69.8*
sna2Δ	21.7*	48.2*
vps53Δ	298.3*	57.9*
Transcription or maintenance of chromatin architecture
mbf1Δ	32.5*	77.5*
nhp6bΔ	87.3	51.6*
paf1Δ	102.9	23.7*
rxt3Δ	98.4	91.3
ume1Δ	13.2*	31.0*
ylr278cΔ	18.0*	42.6*
Known and putative yeast prion genes
def1Δ	49.6*	87.6*
rnq1 Δ	116.0*	63.3*
ybr016wΔ	132.0*	62.8*
yir003wΔ	60.2*	62.2*
ylr278cΔ	18.0*	42.6*
Other cellular processes
arg7Δ	133.0	96.4
bna4 Δ	0*	0*
ecm37Δ	51.6*	67.8*
fmp27Δ	128.2	96.6
hsp104Δ	96.9	58.4*
mgt1Δ	61.4*	73.3*
pho87Δ	16.8*	85.4
rdh54Δ	64.4*	63.6*
ydr287wΔ	53.3*	62.1*
yer185wΔ	26.0*	81.8*
Genes with unknown function
smy2Δ	54.4*	182.2*
ymr082cΔ	171.3*	96.5
ymr244c-aΔ	74.8	79.8

^*P < 0.05 (Student’s t-test). (From Giorgini et al., 2008).

7. Mechanism of quinolinate toxicity

7.1. Quinolinate is not a conventional excitotoxin

The new evidence supporting a pathophysiologically significant role of QUIN in HD rekindled interest in the neurodegenerative properties of the metabolite. As mentioned earlier, the sequelae of an intracerebral QUIN injection are distinct from those of the classic excitotoxins kainate and ibotenate. Surprisingly, in spite of the fact that QUIN is a selective NMDA receptor agonist, its neurotoxic effects also differ substantively from those of NMDA. Thus, unlike QUIN, NMDA is an potent neurotoxin in the cerebellum (Garthwaite et al., 1986), does not distinguish between neuronal populations in the hippocampus (Nadler et al., 1981), is more effective in the immature than in the adult brain (McDonald et al., 1988), and produces nerve cell loss that is significantly less susceptible to blockade by KYNA than QUIN-induced lesions (Foster et al., 1984). The reasons for these remarkable qualitative differences between two supposedly very similar excitotoxins have been tentatively related to the fact that QUIN preferentially activates only a subgroup of NMDA receptors in the forebrain (De Carvalho et al., 1996; Monaghan and Beaton, 1991), whereas NMDA itself, by definition, targets all receptor subtypes. This interpretation would be consistent with the observation that QUIN lesions are always preventable by broad spectrum NMDA receptor antagonists such as AP5 and AP7. It would also be compatible with a role of a subset of NMDA receptors in the pathophysiology of HD (see Fan and Raymond, 2007, for review).

A study by Rios and Santamaria first suggested that the neurotoxic characteristics of QUIN might also be related to its ability to stimulate lipid peroxidation under physiological conditions (Rios and Santamaria, 1991). Although this effect was observed in cell-free brain homogenate, it was blocked by KYNA or specific NMDA receptor antagonists such as MK-801 (Santamaría and Ríos, 1993) and therefore appears to be dependent on receptor activation. Induction of lipid peroxidation, monitored by measuring peroxinitrite production (Pérez-De La Cruz et al., 2005), is iron (II)-dependent (Stípek et al., 1997) and probably related to the formation of specific QUIN-iron complexes. Upon auto-oxidation, these complexes generate reactive hydroxyl radicals through the Fenton reaction (Pláteník et al., 2001; Santamaría et al., 2001). This mechanism may explain the remarkable neurotoxic potency of the weak NMDA receptor agonist QUIN in vivo. In support of this idea, QUIN-induced striatal lesions are significantly attenuated by treatment with agents that enhance peroxynitrite decomposition, bind ferrous iron and/or scavenge reactive oxygen species (ROS) (Müller et al., 2007; Nakao et al., 1996; Santiago-López et al., 2004). It is also tempting to speculate that these protective interventions may be especially effective vis-à-vis QUIN, because they target Fe (II), which is an essential co-factor of 3-HAD and other oxygenases responsible for the biosynthesis of QUIN.

7.2. A cytotoxic role for 3-hydroxykynurenine?

The cytotoxic properties of 3-HK, which is present in the mammalian brain in nanomolar concentration (Pearson and Reynolds, 1991), were first described 20 years ago (Eastman and Guilarte, 1989). The compound readily auto-oxidizes and thereby generates hydrogen peroxide and other ROS, which appear to be singularly responsible for neurotoxicity. Thus, 3-HK has no known direct effect on NMDA receptors, and cell damage in vitro and in vivo can be prevented by co-administration of metal chelators and radical scavengers (Chiarugi et al., 2001; Eastman and Guilarte, 1990; Goldstein et al., 2000; Nakagami et al., 1996; Okuda et al., 1996, 1998). Of particular interest for HD research, exposure to as little as 1 μM 3-HK causes the death of spiny striatal neurons in organotypic cultures, yet is relatively innocuous towards the same small population of medium-sized aspiny neurons that is spared in the HD neostriatum (Okuda et al., 1996; see 4.2). Again in analogy to the HD brain, cerebellar neurons in culture are remarkably resistant to 3-HK-induced oxidative stress (Okuda et al., 1998).

These studies, as well as the increasing appreciation that ROS might play a critical role in HD pathology (Fagni et al., 1994; Wyttenbach et al., 2002), prompted us a few years ago to examine the possibility that 3-HK might exacerbate striatal lesions caused by its downstream metabolite QUIN in rats. Intrastriatal co-application of 3-HK, at a concentration that did not produce neuronal damage on its own, indeed potentiated QUIN-induced neurodegeneration several-fold while maintaining the excitotoxic, i.e. axon-sparing, nature of the lesion (Guidetti and Schwarcz, 1999). Confirming the unique characteristics of QUIN as an excitotoxin, 3-HK was unable to similarly increase the efficacy of intrastriatal NMDA injections. Moreover, we showed that the effect of 3-HK was not due to its in situ conversion to QUIN and demonstrated pharmacologically that the potentiation by 3-HK, which can also be observed in vitro (Chiarugi et al., 2001), was caused by ROS generation. These experiments coincided chronologically with new evidence of enhanced 3-HK levels in vulnerable brain regions of HD patients and HD mouse models (see above) and gave credence to the concept that the two closely related KP metabolites, rather than QUIN alone, should be considered joint pathogens in the disease.

Further support for a role of 3-HK-derived ROS in HD came from additional experiments in yeast. Thus, ROS levels in yeast cells expressing mutant htt were approximately eight-fold higher than in control cells expressing wild-type htt, but this difference was neutralized with the elimination of the KMO ortholog Bna4 in both cells (see 6.3). In a complementary pharmacological study, both the elevated 3-HK and the high ROS levels in mutant htt-expressing cells were reduced by more than 50% by treatment with the selective KMO inhibitor Ro 61–8048 (Röver et al., 1997). This drug, like the deletion of BNA4, also enhanced the viability of yeast cells expressing mutant htt (Giorgini et al., 2005). Thus, either genetic or pharmacological interference with 3-HK synthesis abolishes or significantly attenuates the toxic consequences of mutant htt expression and the excessive formation of both 3-HK and ROS. Together with the finding that deletion of BNA1, the yeast ortholog of 3-HAD, partially suppresses mutant htt-mediated toxicity in yeast (Giorgini et al., 2005; see 6.3), these results indicate a pathophysiologically significant role of ROS, generated by both 3-HK and QUIN, in HD.

7.3. Non-cell-autonomous neurodegeneration

Increased immunostaining for microglia and astrocytes is observed specifically in affected brain regions of HD patients (Myers et al., 1991; Sapp et al., 2001; Simmons et al., 2007; Singhrao et al., 1999; Singhrao et al., 1998; Vonsattel et al., 1985), and prominent glial abnormalities are also seen in HD mouse models (Ma et al., 2003; Minghetti et al., 2007; Reiner et al., 2007; Simmons et al., 2007). Moreover, recent positron emission tomography (PET) studies using a radiotracer specific to activated microglia documented distinct abnormalities in presymptomatic HD gene carriers. Most interestingly, microglial activation in these preclinical HD cases correlated inversely with striatal neuronal dysfunction, as determined by loss of dopamine receptor binding (Tai et al., 2007). These and related studies indicate that glial cells, and especially microglia, might be more than innocent bystanders and mere markers of ongoing or completed neuronal destruction. They may, in fact, actively participate in the pathological process by mobilizing toxic KP metabolites.

Most KP enzymes are constitutively expressed in both microglia and astrocytes (Guillemin et al., 2001; Heyes et al., 1996). However, microglial cells contain far more KMO and 3-HAD than astrocytes (Giorgini et al., 2008; Lehrmann et al., 2001) and are therefore the major source of 3-HK and QUIN, while transamination of kynurenine to KYNA occurs mainly in astrocytes (Guillemin et al., 2001; Kiss et al., 2003). As KP metabolites are deranged in HD, we propose that KMO and other KP enzymes may contribute to neurodegeneration in a non-cell autonomous manner. As illustrated schematically in Figure 6, we speculate that the toxic cascade is likely to be initiated by mutant htt expression in neurons. Perhaps in response to mitochondrial impairment and/or a plethora of other abnormalties, dysfunctional neurons and neuronal activity then create a pro-inflammatory environment leading to the activation of surrounding microglial cells. This, in turn, would lead to an up-regulation of microglial KP metabolism, enhanced microglial release of 3-HK and QUIN and, finally, the demise of already weakened neurons by a combination of excessive ROS production and NMDA receptor activation. As indicated, mutant htt in microglia may also play an active role in these processes (Figure 6).

Figure 6.

Schematic representation of non-cell-autonomous neurodegeneration in HD. (From Guidetti et al., 2006).

8. Kynurenine 3-monooxygenase as a therapeutic target

Inhibition of KMO (E.C. 1.14.13.9) appears to be the best strategy to reduce both 3-HK and QUIN formation in HD. Not only is the enzyme an optimal target for attenuating flux through the entire QUIN branch of the pathway, but reduced KMO activity might also shift KP metabolism towards enhanced KYNA synthesis and thereby boost neuroprotection (cf. Figure 5). Located in the outer mitochondrial membrane (Okamoto and Hayaishi, 1969), KMO is a NADPH-dependent simple flavoprotein, which, as a classic oxygenase, uses molecular oxygen as a co-substrate. Mammalian – including human – KMO has been purified, cloned and functionally expressed in vitro (Alberati-Giani et al., 1997; Breton et al., 2000; Uemura and Hirai, 1998), and its catalytic properties are similar in brain, which contains comparatively low enzyme activity, and peripheral organs (Erickson et al., 1992). Of relevance to the present discourse, the K_m value of 10–20 μM for the substrate kynurenine, determined in vitro, also appears to apply to brain KMO activity in vivo (Guidetti et al., 1995). This implies that the enzyme is not saturated in vivo and will produce increasing amounts of 3-HK when kynurenine levels rise above their ambient tissue concentration (~2 μM; Beal et al., 1990) or when enzyme activity is stimulated by (an) endogenous activator(s) (Connor et al., 2008). One of these possibilities likely accounts for the elevation in 3-HK levels seen in vulnerable regions of the HD brain.

Since KMO knockouts, i.e. genetically engineered animals in which the KMO gene has been inactivated, are just beginning to be investigated (Giorgini et al., in preparation), pharmacological agents that inhibit brain KMO activity are presently the best experimental tools for testing the 3-HK/QUIN hypothesis of HD. After systemic administration, selective KMO inhibitors indeed show impressive neuroprotective efficacy, prevent or ameliorate neurological symptoms, and modulate neurotransmitter release in the brain (Carpenedo et al., 1994; Clark et al., 2005; Cozzi et al., 1999; Erhardt et al., 2001; Pellicciari et al., 1994; Rassoulpour et al., 2005; Wu et al., 2000). As mentioned earlier, however, currently available compounds do not effectively penetrate the blood-brain barrier, so that these effects cannot be unambiguously attributed to the inhibition of the brain enzyme (cf. 5.3). Still, brain KMO is a valid drug target, as evidenced by the fact that enzyme inhibition reduces excitotoxic damage in organotypic hippocampal cultures, and direct intracranial applications of a selective KMO inhibitor attenuate QUIN-induced neurodegeneration in vivo (Carpenedo et al., 2002; Harris et al., 1998; Miranda et al., 1997). However, the synthesis of novel KMO inhibitors with good brain access will clearly be instrumental for studying these neuroprotective phenomena in greater depth.

Proof-of-concept studies will undoubtedly also be aided by KMO knockout animals. Both heterozygous and homozygous KMO mutants will be useful for studying the physiological role(s) of the KP, and for identifying compensatory mechanisms and possible functional isoforms of the enzyme. Experiments with conditional and brain region-specific knockouts should, additionally, clarify the relationship between peripheral and central KP metabolism, the role of the KP in development, and the function of KMO the normal brain. Finally, crosses between HD mice and KMO knockout mice will be especially informative with regard to pathophysiological mechanisms specific to HD.

Jointly, the new genetic and pharmacological tools will make it possible to determine the impact of long-lasting decreases in 3-HK and QUIN levels in the normal and the dysfunctional brain. These studies will not only assess the potential therapeutic value of cerebral KMO inhibition in HD but will also provide information regarding possible chronic treatment hazards. For example, although the two KP branches are acutely segregated in the brain (Amori et al., in press), persistent down-regulation of the QUIN branch may with time increase the bioavailability of KMO’s substrate, kynurenine, for the synthesis of KYNA. The resulting benefit of enhanced neuroprotection would then have to be weighed against the cognitive side effects, which can be expected as a consequence of increased inhibition of NMDA and α7 nicotinic receptors (Hilmas et al., 2001; Kessler et al., 1989). These and similar considerations will help in the design of future clinical trials in which the KP is chronically manipulated in humans.

9. Conclusions and future challenges

The work summarized here suggests that the KP metabolites 3-HK and QUIN, by a combination of excessive NMDA receptor stimulation and ROS production, play a causal role in HD pathology. This hypothesis, which builds on and expands the idea that excitotoxic mechanisms are central to the disease, is supported by studies in model organisms and analyses of brain tissue specimens from HD patients. Since microglial cells produce hyperphysiological amounts of 3-HK and QUIN under pathological conditions, this concept implies an active participation of these neuroimmune cells in the neurodegenerative process. Consequently, the 3-HK/QUIN hypothesis posits that interventions leading to a normalization of microglial KP metabolism will provide therapeutic benefits in HD patients (cf. also Giorgini, 2008).

Several formidable challenges remain. In addition to the development of KMO inhibitors with good brain penetration and possible clinical utility, our laboratories are most interested in the cellular and molecular mechanisms linking mutant htt to impaired cerebral KP metabolism. Our results from studies in animals and humans suggest that KP abnormalities occur early in the disease process (Guidetti et al., 2006; Guidetti et al., 2004), and experiments in yeast indicate that KP dysfunction constitutes a central common mechanism connecting the mutant gene to cell death (Giorgini et al., 2008) (cf. Section 6.). These findings suggest that microglial KP dysfunction could involve currently unknown, yet druggable, regulatory proteins converging on the KP, and also imply that interventions aimed at reducing flux through the QUIN branch of the KP are likely to be most effective in the early stages of HD. With the appropriate experimental tools and paradigms, these hypotheses are testable and should soon reveal interesting novel insights regarding the cell biology and treatment of HD.

Abbreviations

3-HK

3-Hydroxykynurenine

3-HAD

3-Hydroxyanthranilic acid dioxygenase

AMPA

α-Amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid

AP5

D-Amino-phosphonopentanoic acid

AP7

D-Amino-phosphonoheptanoic acid

CNS

Central nervous system

HD

Huntington’s disease

htt

Huntingtin

IDO

Indoleamine-2,3-dioxygenase

KMO

Kynurenine-3-monooxygenase

KP

Kynurenine pathway

KYNA

Kynurenic acid

NMDA

N-methyl-D-aspartic acid

QUIN

Quinolinic acid

QPRT

Quinolinic acid phosphoribosyltransferase

ROS

Reactive oxygen species