Novel aspects of glutamine synthetase in ammonia homeostasis

Elevated blood ammonia (


Introduction
In mammals, ammonia is a biologically significant intermediate in several biochemical processes and not merely a waste product from metabolism of nitrogenous compounds.Hyperammonemia (elevated concentrations of ammonia in blood) may be caused by a number of conditions, such as liver failure, portocaval shunting, colonization of the gut with urease-producing bacteria, and inborn errors of metabolism (Haberle, 2013).In addition, the brain may be exposed to high concentrations of ammonia during bacterial encephalitis and brain abscess formation (Dahlberg et al., 2016).
Like several other biochemicals, e.g.glutamate (for reviews see e.g.: Danbolt, 2001), ammonia is toxic if present at excessive concentrations.The clinical features of hyperammonemia range in severity and include personality changes, confusion, seizures, coma, and death (for reviews see: Cash et al., 2010;Prakash and Mullen, 2010).Some of the symptoms are probably direct consequences of ammonia's interference with cellular processes (Bosoi and Rose, 2009), while others are indirect, likely reflecting tissue edema (Rovira et al., 2008;Alonso et al., 2014;Larangeira et al., 2018).
The toxicity of ammonia was noted more than a century ago when experimental portacaval shunting in dogs caused hyperammonemia and encephalopathy (Hahn et al., 1893).The link between hyperammonemia and encephalopathy has been solidified by a substantial body of evidence from patients (Clemmesen et al., 1999) and animal models of liver disease (for reviews see: Butterworth et al., 1987;Clemmesen et al., 1999;Bernal et al., 2007;Aldridge et al., 2015).Interestingly, other factors contribute to the encephalopathy because the severity of the encephalopathy does not correlate with the blood ammonia concentration (Shawcross et al., 2011), and because there is not a consistent linear relationship between ammonia levels in the blood and the brain (e.g.Cauli et al., 2008;Zwirner et al., 2010;Kanamori et al., 2002;Dahlberg et al., 2016;Shawcross et al., 2011).One possible factor is inflammation (Shawcross et al., 2007; for reviews see: Coltart et al., 2013;Jayakumar et al., 2015;Tranah et al., 2013;Romero-Gomez et al., 2015;Aldridge et al., 2015).Another complicating issue is that the pathogenesis of ammonia-related encephalopathy is not exclusively driven by the liver, but also involves other organs with high ammonia turnover, such as the skeletal muscle and the kidneys (Aldridge et al., 2015).
The scope of the present review is to provide an up to date account of the mechanisms underlying ammonia metabolism, with particular emphasis on the emerging importance of glutamine synthetase (Glul; EC 6.3.1.2) in ammonia homeostasis and brain function.We will not discuss the pathophysiological sequelae of hyperammonemia, as those have been thoroughly reviewed by others (for review see : Visek, 1968;Butterworth et al., 1987;Braissant et al., 2013;Aldridge et al., 2015;Bosoi and Rose, 2009)).

Zonation of ammonia and glutamate metabolism in the liver
The recent availability of genetically modified animals has made it easier to uncover the interaction between ammonia metabolism and glutamate metabolism.For instance, urea cycle mutations (e.g.mice lacking CPS1) are associated with severe increases in blood glutamine and ammonia levels (Khoja et al., 2019).Additionally, mice lacking hepatic glutamate dehydrogenase display altered ammonia homeostasis characterized by increased blood ammonia and reduced conversion to urea (Karaca et al., 2018).Loss-of-function mutation in the glutamine transporter SNAT3 (slc38a3) causes several metabolic perturbations in the liver such as decreased glutamine levels, as well as increased urea levels (Chan et al., 2016).
Anterograde and retrograde perfusion experiments, as well as targeted lesion experiments, have confirmed the metabolic zonation and suggest that the urea cycle pathway is a low affinity, high capacity ammonia clearance system that is preferentially localized to the periportal hepatocytes.In contrast, the glutamine synthetase reaction in the perivenous hepatocytes is believed to represent a high affinity, low capacity ammonia clearance system independent from the urea cycle pathway (Gebhardt and Mecke, 1983;Haussinger, 1983;Gebhardt et al., 1988).

The importance of the urea cycle in ammonia metabolism
In 1932, Krebs and Henseleit used the surviving tissue slice technique of Warburg to show that mammalian liver slices form urea from ammonia and carbon dioxide (Krebs, 1942).Their pioneer study not only refuted the prevailing hypothesis that ammonium cyanate or related compounds were involved in urea formation, but also initiated a new era of biochemistry related to urea biosynthesis by defining the metabolic roles of ornithine, citrulline and arginine.With the efforts from numerous other biochemists, the chemical reactions involving the five key urea cycle enzymes were clarified, i.e. carbamoyl-phosphate synthetase 1 (CPS1), ornithine transcarbamylase (OTC), argininosuccinate synthase (ASS), argininosuccinate lyase (ASL), and arginase 1 (ARG1).As shown in Fig. 2, the initial step in the urea cycle is the synthesis of carbamoyl phosphate from bicarbonate and ammonia via the mitochondrial enzyme carbamoyl phosphate synthetase 1 (CPS1; Krebs, 1942;Cohen and Hayano, 1948;Grisolia and Cohen, 1952).The next step is the formation of citrulline from carbamoyl phosphate and ornithine, catalyzed by ornithine transcarbamylase (OTC) which is the only urea cycle gene found on the X chromosome (KREBS et al., 1955;Marshall and Cohen, 1972).This is followed by the cytosolic components of the cycle, beginning with the enzyme arginosuccinate synthase (ASS), which catalyzes the formation of arginosuccinate from citrulline and aspartate (Saheki et al., 1975;Takada et al., 1979;Rochovansky et al., 1977;Ratner and Pappas, 1949) arginosuccinate lyase (ASL) to arginine and fumarate which enters the citric acid cycle (Ratner et al., 1953;Murakami-Murofushi and Ratner, 1979).Finally, arginine is cleaved by arginase 1 to urea and ornithine (Wu and Morris, 1998).Through extensive enzymological studies, it was also revealed that acetylglutamate, catalyzed by N-acetylglutamate synthase (NAGS; EC 2.3.1.1),acts as allosteric activator of carbamoyl phosphate synthetase (Grisolia and Cohen, 1953;Shigesada and Tatibana, 1971;Hall et al., 1958).Because the urea cycle is carried out in two subcellular compartments (mitochondria and cytosol), mitochondrial transporters had to be important, and in agreement, the two transporters that carry ornithine and aspartate across the mitochondrial membrane were eventually identified.Ornithine reenters the mitochondria via ORNT1 (slc25a15) in exchange for citrulline (Camacho et al., 1999;Kuno et al., 1990), while aspartate is released to cytosol via Citrin (slc25a13; Kobayashi et al., 1999).As expected, congenital or acquired deficiencies in any of the above listed enzymes and transporters can result in various degrees of hyperammonemia and encephalopathy (summarized in Table 1).The history of the science related to the urea cycle and the metabolism of arginine and citrulline has been reviewed in detail elsewhere (Ratner, 1954;Holmes, 1980;Cohen, 1981; Fig. 1.Hepatic zonation of ammonia metabolism and glutamate metabolism.Simplified schematic illustration of ammonia removal via the urea cycle and via the glutamine synthetase pathway in the liver.The urea cycle pathway is a low affinity, high capacity ammonia clearance system that is preferentially localized to the periportal hepatocytes, while the glutamine synthetase reaction in the perivenous hepatocytes represent a high affinity, low capacity system for escaping ammonia.Dysfunctions in either urea cycle or glutamine synthesis induces hyperammonemia and causes encephalopathy (Qvartskhava et al., 2015;Hakvoort et al., 2017).These processes are intermingled with glutamate metabolism via a differential involvement of enzymes and transporter proteins.Amino acids (AA, except glutamate and aspartate) are taken up and catabolized inside periporal hepatocytes (Häussinger et al., 1992).The two well studied examples are abundant plasma amino acids glutamine (Gln) and alanine, which possibly undergo transport inside the cells by slc38a5 (Taylor and Rennie, 1987).Liver glutaminase (GLS2) hydrolyses glutamine to glutamate and ammonia, and aminotransferases (ATs), e.g.alanine aminotransferase, transfer an amino group to alpha-ketogutarate (α-KG) to produce glutamate.The reaction catalyzed by glutamate dehydrogenase (GDH) is reversible.GDH produces ammonia that provides one nitrogen of the urea molecule.Also, it can furnish mitochondrial glutamate, which supports the formation of aspartate for the urea cycle (Nissim et al., 2003;Karaca et al., 2018).Glutamate participates in the urea cycle in several ways.First, it plays a regulatory role in the urea cycle because N-acetylglutamate (NAG) is an obligatory activator of carbamyl phosphate synthetase 1 (CPS1) and is produced from glutamate and acetyl-coenzyme A. Second, glutamate donates its amino group to form aspartate, which then supplies the second nitrogen of the urea molecule.Third, as described above, the GDH reaction glutamate is a donor of NH3.Perivenous hepatocytes, on the other hand, take up glutamate (Glu) and aspartate (Asp) by the EAAT2 (slc1a2) isoform (Stoll et al 1991;Hu et al 2018).However, these cells apparently can produce glutamine in the absence of exogenous glutamate, and release glutamine probably via slc38a3.Besides alpha-ketoglutarate (α-KG) that is produced by Krebs cycle, ornithine is catabolized in ornithine aminotransferase (OAT)-containing hepatocytes to provide a substrate for glutamine synthesis (Kuo et al 1991;O'sullivan et al 1998).The enzymes or transporter proteins that are predominantly present in the periportal zone are indicated in red, while those in perivenous zone are indicated in green, those which are present in both zones are indicated in white and those with little information are indicated in gray.Wu and Morris, 1998).
The current non-surgical treatments of urea cycle disorders include various approaches aimed at lowering ammonia levels (Tuchman et al., 2008;Haberle et al., 2012a) such as ammonia scavenging agents (e.g.sodium phenylacetate and sodium benzoate; Maestri et al., 1991;Enns et al., 2007), a low-protein diet with essential amino acid supplementation (Adam et al., 2013), hemodialysis, and arginine or citrulline supplementation (Brusilow, 1984;Donn and Thoene, 1985;Nagasaka et al., 2006).NAGS deficiency in humans can be effectively treated with N-carbamylglutamate (Caldovic et al., 2004;Heibel et al., 2011), and Citrin deficiency has been successfully treated with a low-carbohydrate diet with arginine supplementation (Saheki and Song, 1993).Even though these therapies have significantly improved the overall morbidity and mortality, residual disease sequelae such as developmental delay and language difficulties often persist (Maestri et al., 1991;Kim et al., 2012).Intriguingly, the sequelae of urea cycle disorders cannot be entirely explained by ammonia toxicity, because patients with ASL and ORNT1 deficiencies may have neurological impairment in the absence of hyperammonemia (Baruteau et al., 2017;Kim et al., 2012).Moreover, some urea cycle enzymes and transporters are not exclusively expressed in the liver.For instance, ASS, CPS1 and ORNT1 show similar patterns of developmental changes in the liver and small intestine (Begum et al., 2002).ARG1 is also found in the brain (Yu et al., 2001) and ASL is present in the brain stem (in locus coeruleus; Lerner et al., 2019), as well as in enterocytes and immune cells (Stettner et al., 2018).Deficiency in these enzymes leads to reduced arginine levels and impaired synthesis of nitric oxide with multi-organ dysfunction as the downstream consequence (Erez et al., 2011).

The importance of liver glutamine synthetase in ammonia metabolism
The urea cycle has long been considered to be the primary pathway of nitrogen disposal.In the same year when Krebs described the biosynthesis of urea from ammonia and carbon dioxide, he also noted that the isolated rat liver removed more ammonia from the nutrient medium than was accounted for by the formation of urea (Krebs and Henseleit, 1932).In 1935 he described the synthesis of glutamine from ammonia and glutamate in the guinea pig kidney, brain, retina and other tissues (Krebs, 1935).As shown in Table 1, the presence of hyperglutaminemia in hyperammonemic patients with urea cycle disorders has been known since the earliest case reports.Speck (1947) and Elliot (1948 ab) independently discovered, respectively, in sheep brain and pigeon liver, an enzyme system which catalyzes the synthesis of glutamine by combining glutamate with ammonia (Elliott, 1948;Elliott and Gale, 1948;Speck, 1947).The enzyme was later termed "glutamine synthetase (GS)" or "glutamate-ammonia ligase (Glul).Meister and co-workers used a highly effective stereochemical approach to understand key aspects of Glul function, such as its substrates, catalytic intermediates, and potential inhibitors (Meister, 1968a(Meister, , 1968b;;Rowe and Meister, 1970;Manning et al., 1969;Wellner and Meister, 1966;Cooper et al., 1976).One such inhibitor was methionine sulfoximine (MSO; CAS15985-39-4) which was produced during chemical bleaching (agenization) of flour with agene (nitrogen trichloride; CAS 10025-85-1) and causes acute and chronic seizures when ingested by or administered to animals (Campbell et al., 1951;Silver and Johnson, 1947).The use of agene as bleaching agent was banned in the USA in 1949.
The significance of the Glul pathway for ammonia removal became evident when 15 N-labeled ammonia was administered to laboratory animals, resulting in more 15 N-labeling of glutamine than of urea (Duda and Handler, 1958;Jahoor et al., 1988).Further, systemic administration of MSO reduces glutamine synthesis in multiple organs and leads to hyperammonemia (Warren and Schenker, 1964;Kant et al., 2014;Blin et al., 2002;Rowe and Meister, 1970;Griffith and Meister, 1978;Cloix et al., 2010;Boissonnet et al., 2012), suggesting an important role of Glul in ammonia metabolism.This notion is further strengthened by the report of three human subjects (not related to each other) with partial congenital Glul deficiency (Haberle et al., 2005;Haberle et al., 2006;Haberle et al., 2012b;reviewed in Spodenkiewicz et al., 2016).The subjects had increased blood ammonia levels, suffered from severe neonatal epileptic encephalopathy, and died young from multi-organ failure.The severe and systemic manifestations of the Glul deficiency are consistent with the widespread distribution and important roles of the enzyme.
Investigations of the physiological consequences of Glul have, until recently, relied heavily on the use of MSO.This approach, however, has limitations.First, systemically administered MSO inhibits Glul in the entire body, and thus it is unclear whether the neurological symptoms from systemic MSO are caused by inhibition of liver Glul, brain Glul, or both.Second, MSO has several known effects other than inhibiting Glul, such as impairing glutathione synthesis by inhibiting ɣ-glutamylcysteine synthetase (Bernard-Helary et al., 2002;Shaw and Bains, 2002;Kam and Nicoll, 2007).
To more accurately assess the roles of Glul in health and disease, several transgenic approaches have been undertaken.The first approach was to completely delete the Glul gene throughout the body from conception.However, such deletion leads to embryonic demise at embryonic day 3 (E3), suggesting that Glul is crucial for life (He et al., 2007).In a series of subsequent experiments that took advantage of the Cre-LoxP technology (Le and Sauer, 2000), several knockout mouse lines with more restricted Glul deletions were independently created by three research groups (He et al., 2010a 2).
Characterization of mice with liver-selective deletion of Glul elegantly revealed the importance of the enzyme for ammonia homeostasis.The enzyme-deficient mice developed hyperammonemia and encephalopathy (Qvartskhava et al., 2015;Chepkova et al., 2017).Further, by giving excess amounts of ammonia to the other liver-selective Glul knockouts (RRID:IMSR_029827), it was discovered that Glul in the liver metabolizes approximately 35% of blood ammonia, the urea cycle metabolizes another 35%, and the rest is degraded elsewhere in the body (Hakvoort et al., 2017).While the data agree with prior estimates that 80-87% of blood ammonia is metabolized by the liver (Aldrete, 1975), the equal contribution of Glul and the urea cycle to the metabolism was unexpected, as most prior studies had suggested that the urea cycle was the most important.
Deletion of Glul from the liver was reported in one study to give pericentral macrovesicular steatosis (Hakvoort et al., 2017).This would be in line with another study showing that hyperammonemia facilitates development of non-alcoholic steatohepatitis (De Chiara et al., 2020).However, when the liver selective deletion of Glul was replicated by us (Fig. 3D) and by Qvartskhava and coworkers (Qvartskhava et al., 2019), steatosis was not observed.We also did not observe postnatal growth retardation in our liver-specific glutamine synthetase knockouts (Fig. 3C) as Hakvoort and colleagues did.Hakvoort and Lamers concluded that the differences in steatosis were unlikely due to the Cre-line (Alfp-Cre) itself.Therefore, they speculated whether the differences could be due to variations in the genetic background of the mice as they had used FVB/N mice while Qvartskhava and we had used C57BL6J mice (Hakvoort and Lamers, 2019).We, however, propose that the difference in steatosis development might be due to the age at which Glul is deleted from the liver: the Alfp-Cre construct initiates deletion of Glul already at embryonic day E9.5 (Kellendonk et al., 2000;Parviz et al., 2003), whereas the other Cre-line [Alb-Cre (B6.Cg-Tg (Alb-Cre) 21Mgn/J, stock no.003574; RRID: IMSR_JAX:003574)] becomes active after birth and results in a gradual deletion which completes at the adult stage (Fig. 3A).
The co-expression of Glul with EAAT2 in perivenous hepatocytes raises the question of how much blood-derived aspartate and glutamate contribute to the synthesis of glutamine.Recent investigations using liver-specific EAAT2 knockouts, suggest that EAAT2 cannot supply enough glutamate to fully exploit the capacity of Glul (Hu et al., 2018).This assumption is based on the facts that the expression of hepatic Glul is orders of magnitude higher than that of EAAT2 (Hu et al., 2018), and that the transport catalyzed by EAAT2 is slow (Otis and Jahr, 1998;Otis and Kavanaugh, 2000;Bergles et al., 2002;Grewer and Rauen, 2005).In line with this idea, mice lacking hepatic EAAT2 do not display locomotor dysfunction (Hu et al., 2017) like mice lacking hepatic Glul (Qvartskhava et al., 2015).The reason why EAAT2 may not be essential for liver function is that hepatocytes can produce glutamate via other pathways such as oxidation of ornithine by ornithine aminotransferase (OAT; O'sullivan et al., 1998) or from α-ketoglutarate (Braeuning et al., 2006) (Fig. 1).

Controversies concerning glutamine as a nitrogen donor for urea synthesis
In addition to ammonia-derived nitrogen, other nitrogen donors for urea synthesis have been proposed.In periportal hepatocytes, glutamate-nitrogen enters the urea cycle predominantly as aspartate (Jahoor et al., 1988).But where does the glutamate-nitrogen come from?Alanine-nitrogen has been suggested because hepatic ammonia uptake is accompanied by uptake of alanine in almost equimolar quantities (see Fig. 1; Yang et al., 2000), and because alanine is a major amino acid extracted by the liver as a substrate for gluconeogenesis (Felig et al., 1970).The in vitro work by Brosnan and coworkers confirms the contribution of alanine-nitrogen to urea and to glutamine synthesis, although the capacity of alanine as a nitrogen source is limited (Brosnan et al., 2004).Considering that periportal hepatocytes can take up glutamine and also express the liver isoform of glutaminase (which catalyzes the conversion of glutamine to glutamate), it follows that the amide-nitrogen of glutamine may represent a source of urea-nitrogen.This concept is experimentally supported by observations that addition of glutamine to liver slices results in considerable enhancement of urea production (e.g.Bach and Smith, 1956;Nissim et al., 1992).However, a major weakness of these studies is a lack of in vivo validation (Kamin and Handler, 1957).By using 15 N-glutamine, Handler concluded that there is negligible formation of urea from 15 N-glutamine in vivo (Duda and Handler, 1958), but this has been contradicted by another in vivo study reporting that glutamine is more important than alanine as a nitrogen donor for urea synthesis (Jahoor et al., 1988).Thus, further investigations are required to fully resolve the role of glutamine in urea synthesis in vivo.

Glutamine synthetase in extrahepatic tissues
Glul is strongly expressed in extrahepatic tissues such as the brain, the kidneys, the skeletal muscle, the adipose tissue, the male reproductive organs, the pancreas, and the gastro-intestinal tract (see Table 3).These extrahepatic tissues are responsible for approximately one-third of the total ammonia detoxification capacity via the glutamine synthetase reaction (Hakvoort et al., 2017).Similar to the liver, the extrahepatic Glul-expressing cells often express glutamate transporters (Tables 3 and 4).Interestingly, the pancreas represents an exception, as glutamate transporters are not expressed in the islets cells and are only present at very low levels in the exocrine pancreas (Zhou et al., 2014).Skeletal muscle and kidneys have a particularly high capacity for ammonia removal (Cruz et al., 2017).Selective deletion of Glul in the murine skeletal muscle suggests that the maximal capacity of ammonia absorption in skeletal muscle is approximately 10% (He et al., 2010b).The data in the figure has only been published in abstract form (Hu et al., 2017).Although this estimate is much lower than what was proposed in the past (50%; Hod et al., 1982;Lockwood et al., 1979), it should be kept in mind that skeletal muscle can upregulate Glul activity and therefore increase ammonia metabolism during liver failure (Hod et al., 1982;Desjardins et al., 1999;Clemmesen et al., 2000;Chatauret et al., 2006).
The kidneys also adapt to the consequences of a failing liver by reducing ammonia release into the systemic circulation by slowing down ammonia production and by increasing ammonium (NH 4
A full discussion of ammonia metabolism in the kidneys and skeletal muscle, however, is beyond the scope of this review, as these topics are covered well elsewhere (e.g.Mutch and Banister, 1983;Graham and MacLean, 1998;van de Poll et al., 2004;Weiner and Verlander, 2013;Weiner et al., 2015;Olde Damink et al., 2002;Adeva et al., 2012).

Glutamine synthetase is essential for brain function
The importance of brain Glul was first uncovered by local administration of MSO to the central nervous system (CNS).Depending on the injection site, MSO caused recurrent seizures, altered responses to graded mechanical stimuli, attenuated nocifensive behavior, or altered EAAT1 is selective for astroglia.Note that there are some reports claiming EAAT1 to be present in neurons, but this has not been supported by later publications.Lehre et al. (1995); Ginsberg et al. (1995); Berger and Hediger (1998); Berger and Hediger (2000); Hanson et al. (2015) EAAT2 EAAT2 protein is strongly expressed in astroglia Danbolt et al. (1992); Levy et al. (1993); Rothstein et al. (1994); Lehre et al. (1995); Berger and Hediger (1998); Schmitt et al. (1997); Berger and Hediger (2000) EAAT2 mRNA is present in astrocytes and in the majority of neurons in multiple CNS regions Torp et al. (1994); Schmitt et al. (1996); Torp et al. (1997); Berger and Hediger (1998); Berger and Hediger (2000); Berger andHediger (2001) Torp et al. (1994); EAAT2 is expressed in brain neurons during development, in neurons in diseased brain tissue and in cultured neurons Northington et al. (1998); Northington et al. (1999); Martin et al. (1997); Mennerick et al. (1998); Plachez et al. (2000) Early electron microscopy suggested that nerve terminals, at least in cortex and striatum, are able to take up glutamate though the responsible carrier(s) was not determined.Beart (1976); McLennan (1976); Gundersen et al. (1993) Some of the EAAT2 protein (about 10%) is found in neurons, and is targeted to the axon-terminals where EAAT2 is the only glutamate transporter.EAAT3 is selective for neurons and present in most if not all neurons, but is targeted to the cell body and dendrites.Kanai and Hediger (1992); Rothstein et al. (1994); Bjoras et al. (1996); Torp et al. (1997)

EAAT1 EAAT3
Rat hearts contain EAAT3 and EAAT1, but at lower levels than in the brain.EAAT2 and EAAT4 were not detected.

Skeletal Muscle
Not reported

EAAT3
EAAT3 is the only EAAT-type of transporters in enterocytes, and is found in the small intestine and colon with the highest levels in the distal ileum.Kanai and Hediger (1992); Hu et al. (2018); Berger and Hediger (2006) Exocrine glands

EAAT2
EAAT2 is present at low levels in salivary glands.Berger and Hediger (2006) Adipose tissue EAAT1 Several EAATs have been reported in adipocytes, but EAAT1 is the only EAAT confirmed by validated in situ hybridization and proteome analysis Berger and Hediger (2006); Adachi et al. (2007)

Pancreas
In-depth proteome analysis and characterization of pancreas-specific EAAT2 knockout mice show that EAAT2 is not expressed in pancreatic islets as some have reported, but may be expressed at low levels in the exocrine pancreas.In fact, the islets do not express any of the EAATs nor any of the VGLUTs  jaw-opening reflex (Table 5) (Eid et al., 2008;Dhaher et al., 2015;Gruenbaum et al., 2015;Chiang et al., 2007;Tsuboi et al., 2011;Mostafeezur et al., 2014).The MSO studies are, however, associated with some uncertainty because MSO not only inhibits glutamine synthetase, but also decreases tissue glutathione (Shaw and Bains, 2002), increases astrocyte glycogen (Bernard-Helary et al., 2002), and excites neurons via a Glul-independent mechanism (Kam and Nicoll, 2007).Depletion of neuronal glutathione is shown to increase cytosolic glutamate and affects excitatory neurotransmission (Sedlak et al., 2019).Several knockout approaches for Glul have recently become available (as explained above), making highly accurate in vivo studies of the enzyme possible.However, deletion of Glul in the entire CNS resulted in early neonatal death (He et al., 2010a).More restricted deletions would thereby be necessary in order to obtain longer survival.Unfortunately, there are not that many astroglial Cre drivers, and that the commonly used ones (e.g.hGFAP-Cre) are active in most CNS astrocytes resulting in too extensive deletion.It was therefore a breakthrough when we (Zhou et al., 2019a) were able to selectively delete Glul in the cerebral cortex (i.e.neocortex and hippocampus) using the Emx1_IRES Cre line.These mice survive well into adulthood and exhibit numerous pathological features such as altered locomotive activity, progressive neurodegeneration, and spontaneous recurrent seizures (Zhou et al., 2019a).In contrast, deletion of glutamine synthetase in the liver (Hakvoort et al., 2017), the kidneys (Lee et al., 2016), the skeletal muscle (He et al., 2010b), the pancreas (Bott et al., 2019) or the adipose tissue (Zhou Y and Danbolt NC, personal observations) barely affected mortality and had a mild phenotype with no evidence of behavioral seizures (Table 2).

Glutamine production and roles of glutamine in neurotransmission
Glutamine is abundantly present in the central nervous system, and the synthesis is largely catalyzed by astroglial glutamine synthetase (Zhou et al., 2019a).The proposed major role is to be a precursor of the neurotransmitters glutamate (excitatory) and ɣ-amino butyric acid (GABA; inhibitory).For more than half a century, compartmentation of glutamate-glutamine metabolism has been central to how we envision glutamine as a precursor for neuronal glutamate (Berl et al., 1962;Berl et al., 1968;Berl et al., 1970;Martinez-Hernandez et al., 1977;van den Berg and Garfinkel, 1971;Bradford et al., 1978;Hamberger et al., 1979).As outlined in Fig. 4, neurotransmitter glutamate is released by exocytosis of synaptic vesicles from nerve terminals (for review : Sudhof, 2014).The surrounding astrocytes rapidly take up the released glutamate via two astroglial glutamate transporters (EAAT1/slc1a3 and EAAT2/slc1a2; Danbolt et al., 1992;Lehre et al., 1995), converts glutamate into glutamine via the Glul pathway (Martinez-Hernandez et al., 1977) and releases glutamine into the extracellular space.Neurons are able to take up glutamine (Schousboe et al., 1979;Dyste et al., 1989) via glutamine transporters.It is possible that several glutamine transporters are involved as there are at least 14 different solute carrier proteins with the ability to transport glutamine (for reviews see: Mackenzie and Erickson, 2004;Bhutia and Ganapathy, 2016;Danbolt et al., 2016a;Hellsten et al., 2017).Inside neurons, glutamine can be converted back to glutamate by phosphate-activated glutaminase (PAG/GLS1; Kvamme et al., 2001).A full discussion of glutamine and glutamate metabolism in the brain have been thoroughly reviewed by others (e.g.Yudkoff, 2017;Hertz and Zielke, 2004;McKenna and Ferreira, 2016;Bak et al., 2018;Hertz and Chen, 2018;Hertz and Rothman, 2016).
The glutamate-glutamine cycle model, however, probably oversimplifies a complex reality, and not all available data fit with this model.(A) Although it is clear that neurons can take up glutamine, it is not known how and to what extent glutamine can enter the nerve terminals as no glutamine transporter protein has so far been shown immunocytochemically at the electron microscopic level to be present in axon-terminals in brain tissue (for review see: Mackenzie and Erickson, 2004;Conti and Melone, 2006;Grewal et al., 2009;Zhou and Danbolt, 2014).Possible candidates comprise SNAT7 (slc38a7) and SNAT8 (slc38a8), but uncertainty remains as the corresponding knockout mice are not yet available as negative controls and quantitative information has yet to be obtained (Hagglund et al., 2011(Hagglund et al., , 2015)).Recently, it has been reported that the glutamine transporter SNAT1 (Slc38a1) participates in the regulation of the GABA levels in GABAergic terminals although the SNAT1 protein itself was not detected at the terminals (Qureshi et al., 2019;Qureshi et al., 2020).There is electrophysiological evidence for glutamine uptake at the Calyx of Held synapse (Billups et al., 2013).However, this synapse consists of a special giant nerve terminal which may not be representative for the majority of synapses in the mammalian CNS.The majority of mammalian synapses have boutons that are very small and we know less about them (von Gersdorff and Borst, 2002).(B) Another unexplored observation is that glutamine administered to the intact brain is mostly metabolized to CO 2 (Zielke et al., 1998), in agreement with an important role for glutamine as a neuronal energy substrate (Bradford et al., 1978;Hamberger et al., 1979;Zielke et al., 2009).(C) Electrophysiological analyses of isolated brain slices suggest that glutamatergic neurotransmission can be sustained in the absence of the cycle (Kam and Nicoll, 2007).(D) The terminals can obtain glutamate independently of glutamine as there is (1) Glutamate (Glu) is released from vesicles in the axon-terminals into the extracellular fluid from where it can (2) bind to glutamate receptors (GLU-R) before it is inactivated by uptake into astrocytes (3) via excitatory amino acid transporters (EAATs).( 4) Inside astrocytes it is converted to glutamine (Gln) via the enzyme glutamine synthetase (Glul).( 5) Gln is next shuttled from astrocytes to neurons via mechanisms that remain to be defined unequivocally.( 6) Once inside the neurons, Gln is converted back to Glu by glutaminase, which is particularly enriched in mitochondria.( 7) Finally, Glu is concentrated in synaptic vesicles via vesicular glutamate transporters (VGLUTs), thus completing the cycle.It should ne noted that (8) axonal terminal expressing EAAT2 will partly short circuit the cycle and that (9) there is an export of glutamine from the brain to the blood.some de novo synthesis (Hassel and Bråthe, 2000), and as they can take up glutamate directly via EAAT2 because a portion of EAAT2 (5-10%) is present in axonal terminals (Furness et al., 2008;Danbolt et al., 2016a;Petr et al., 2015;Zhou et al., 2019b;McNair et al., 2019).(E) As McKenna and Yudkoff pointed out earlier, the model ignores the facts that the cycle is not stoichiometric and that the brain avidly exchanges many metabolites with the blood (McKenna, 2007;Yudkoff, 2017).In line with this, the supply of glutamine to terminals may not always keep up with glutamate release (Waagepetersen et al., 2005;Kam and Nicoll, 2007;Marx et al., 2015), because there is a glutamine loss due to the export from the brain to the blood (Cangiano et al., 1983).This export process is mediated by several transporters including LAT1 (slc7a5), LAT2 (slc7a8) and SNAT3 (slc38a3) which are robustly expressed in brain microvessel endothelial cells (Duelli et al., 2000;Dolgodilina et al., 2016;Dolgodilina et al., 2020).Thus, the relative contributions of the various mechanisms are still debated and a complicating factor is that this may differ between brain regions and in disease.Marx and co-workers has recently tried to put available quantitative data together (Marx et al., 2015).

Glutamine synthetase in the CNS may be expressed in cells other than astrocytes
It is well established that Glul is expressed in astrocytes (Martinez-Hernandez et al., 1977); however, it is still debated whether the enzyme is expressed in significant quantities in other CNS cells such as oligodendrocytes, endothelial cells, microglia and neurons.The lack of a consensus is partly due to how Glul is quantified in vitro and the fact that the expression of the enzyme is regulated by cell to cell interactions and soluble factors that may be very different in vitro versus in vivo (e.g.Barakat-Walter and Droz, 1990;Cahoy et al., 2008).This point was well illustrated by Barakat-Walter and Droz (1990) who found that dorsal root ganglion cells grown in vitro express Glul while those grown in vivo do not.Another factor contributing to the lack of consensus is the different results obtained with immunocytochemistry performed in different laboratories with different antibodies and labeling protocols (Martinez-Hernandez et al., 1977;Norenberg and Martinez-Hernandez, 1979;Cammer, 1990;Tansey et al., 1991;D'Amelio et al., 1990;Yamamoto et al., 1989;Takasaki et al., 2010), suggesting that differences in the specificity of immunohistochemical procedures play a role, as this is extensively discussed in prior publications (e.g.Danbolt et al., 1998;Holmseth et al., 2006;Lorincz and Nusser, 2008;Saper, 2009;Holmseth et al., 2012b;Danbolt et al., 2016b).
It is important to resolve the questions with regards the cellular expression of Glul because if the enzyme is expressed in other cell types, then this will impact the interpretation of a number of studies of the ammonia metabolism.Fortunately, the recent development of knockout models is expected to close key gaps in knowledge and open up for new fields of research that may ultimately translate to novel treatments for ammonia-related disorders.By knocking out Glul in a cell-specific manner, the functional consequences can be studied with a high level of accuracy and the tissue can also serve as a specificity control for immunocytochemical staining of the enzyme.For example, one of the Glul knockout approaches (Emx1Cre/Glul-Flox) targets astrocytes, glutamatergic neurons, and oligodendrocytes in the cerebral cortex, but spares cortical GABAergic neurons and cells in other parts of the brain (Zhou et al., 2019a;Gorski et al., 2002).Another approach selectively deletes Glul in oligodendrocytes (Xin et al., 2019), endothelial cells (Eelen et al., 2018) and macrophages (Palmieri et al., 2017).
Surprisingly, mice lacking oligodendrocyte Glul have about 20% reduction in brain glutamate and glutamine levels and exhibit disturbed glutamatergic neurotransmission, but unaffected myelination (Xin et al., 2019).The finding of normal myelination in vivo is in contrast to an in vitro study where glutamine synthetase deficiency affects differentiation of oligodendrocytes (Saitoh and Araki, 2010).These results emphasize the importance of validating in vitro findings with subsequent in vivo studies.Moreover, the use of Glul knockout tissue as a control for antibody specificity has shown that Glul is expressed in a subpopulation of oligodendrocytes that emerges at P21 and is particularly densely stained in caudal regions of the CNS (Xin et al., 2019).The late onset and regional heterogeneity in Glul expression, may explain earlier inconsistent observations.Finally, selective removal of Glul in endothelial cells impairs retinal vessel spouting during vascular development (Eelen et al., 2018), and deletion of Glul in macrophages in tumor-bearing mice promotes tumor vessel pruning, vascular normalization, accumulation of cytotoxic T-cells, and metastasis inhibition (Palmieri et al., 2017).
The above findings challenge the concept that Glul in the CNS is exclusively expressed in astrocytes and therefore the two-compartment (neuron-glia) model of the glutamate-glutamine metabolic cycle, which is thought to be critical for the synthesis of the major excitatory and inhibitory neurotransmitters glutamate and GABA (e.g.van den Berg and Garfinkel, 1971;Rothman et al., 1999).While knockout approaches can be used to gain a more detailed and accurate understanding of the glutamate-glutamine cycle, they are associated with limitations.For instance, the insertion of a promotor-Cre construct into a genome sometimes disturbs the expression of other genes (Forni et al., 2006;Giusti et al., 2014).It is therefore important to assess whether the Cre promotor interferes with the study.Likewise, when tamoxifen-inducible constructs have been used, possible side-effects of tamoxifen must be carefully controlled for.This is particularly relevant to cancer and immunological research, as tamoxifen is an anti-cancer drug and appears to have immunomodulatory effects which are independent of the estrogen-receptor (Corriden et al., 2015; for review see: Behjati and Frank, 2009).Another reason to be cautious is that gene expression profiles of many cell types, including endothelial cells and macrophages, are significantly influenced by the microenvironment and may therefore differ between locations and functional states (Gordon and Pluddemann, 2017;Garlanda and Dejana, 1997).Thus, the expression patterns of Glul in endothelial cells of the brain and in microglia (a macrophage-like cell residing in the CNS) need further studies.It will be interesting to learn more about the metabolic consequences of deleting glutamine synthetase in endothelial and microglia in vivo.
Selective deletion of hepatic Glul causes a similar syndrome characterized by hyperammonemia with astroglial swelling, no visible neurodegeneration, and a minor reduction in life span (Qvartskhava et al., 2015).The mice display altered behavior characterized by increased locomotion, reduced exploratory activities and delayed habituation to a novel environment as well as impaired fear memory (Qvartskhava et al., 2015;Chepkova et al., 2017).
In contrast, selective deletion of Glul in the whole brain in mice results in a 14-fold decline in cortical glutamine and a 1.6-fold increase in cortical ammonia levels (He et al., 2010a).These mice die about three days after birth unless they are fed by hand.In contrast, when Glul is selectively deleted in the cerebral cortex (Zhou et al., 2019a), there is a 4-fold reduction in cortical glutamine and the majority of the animals survive for several months.There is, however, a progressive gliosis and impaired neurovascular coupling.These animals exhibit overall decreased locomotion with brief episodes of wild running, and many develop spontaneous recurrent seizures and neurodegeneration that Y. Zhou et al. commence around 6 weeks of age.A portion of animals die suddenly and unexpectedly, leading to an overall reduced life-span (Zhou et al., 2019a).
The phenotype of the cortical Glul knockout mice are in good agreement with the observations that patients with partial loss of function mutations in the Glul gene had brain atrophy and severe neonatal epileptic encephalopathy (for review see: Spodenkiewicz et al., 2016).These humans had insufficient Glul activities, not only in the brain, but in the entire body including liver, skeletal muscle, kidneys and skin.They died young from multiple organ failures.Further, the defects were present already at conception, and in this context it should be noted that the deleterious effects of hyperammonemia may be more severe to the developing brain than the mature brain (Braissant et al., 2013).
Patients with mesial temporal lobe epilepsy (MTLE) have a deficiency in glial Glul in the epileptogenic hippocampus (Eid et al., 2004;van der Hel et al., 2005); however, this differs from the above cases as the loss of Glul in MTLE is limited to small regions of the brain and likely occurs later in life when the brain is fully or near-fully developed.
These distinct signatures may suggest different therapeutic approaches.For instance, pharmacological inhibition of Glul was suggested to temporarily relief brain edema by attenuating ammoniainduced astroglial swelling, and by ameliorating some of the reactive astroglial cytoskeletal alterations (Blei et al., 1994;Tanigami et al., 2005).This strategy, however, worsen the situation in awake intact animals (Rangroo Thrane et al., 2013).On the other hand, glutamine supplementation to a child with inherited Glul deficiency partially corrected the glutamine level in the cerebrospinal fluid, improved alertness and EEG characteristics (Haberle et al., 2012b).

Concluding remarks
Glutamine synthetase plays a significant role in ammonia metabolism, and its dysfunction leads to neurological diseases.An improved understanding of the regulation and physiological implications of glutamine synthesis in health and disease will likely uncover novel drug targets for patients who suffer from hyperammonemia.

Fig. 2 .
Fig. 2. The urea synthesis in the periportal hepatocytes.Cartoon depicting the formation of urea with the involvement of intermediate metabolites ornithine, citrulline and arginine.The chemical reactions and the involved enzymes/transporters (highlighted by blue) are detailed in the text.

Fig. 3 .
Fig. 3. Glutamine synthetase (GS) deletion that is controlled by the albumin promotor (Alb-Cre) occurs progressively in the liver during postnatal development, being nearly complete in the young adult stage (about 7 weeks of old).Panel A. Immunocytochemistry analysis of the liver shows the incomplete deletion of the glutamine synthetase protein (red) in the developing liver (P22) and almost complete deletion in mature liver (P55 and P89).Scale bar, 40 μm.Panel B. Immunoblotting analysis of liver-specific GS knockouts (cKO) and wild-type (WT) littermates (58-64 days old) confirm the absence of the glutamine synthetase in the liver and the preservation in the brain.Panel C. The Alb-Cre driven glutamine synthetase knockouts (male, n = 11) have normal postnatal growth as their wildtype littermates do (n = 13) in C57BL6J mice.Panel D. Steatosis is absent in the liver-specific glutamine synthetase knockout fed with regular chows (3 months and 12 months of age; n = 3).The frozen tissue sections are stained for Oil Red O. Livers from high-fat fed mice were used as a positive control (data not shown).Scale bar, 100 μm.

Fig. 4 .
Fig. 4. Overview of the glutamate-glutamine cycle.(1)Glutamate (Glu) is released from vesicles in the axon-terminals into the extracellular fluid from where it can (2) bind to glutamate receptors (GLU-R) before it is inactivated by uptake into astrocytes (3) via excitatory amino acid transporters (EAATs).(4) Inside astrocytes it is converted to glutamine (Gln) via the enzyme glutamine synthetase (Glul).(5) Gln is next shuttled from astrocytes to neurons via mechanisms that remain to be defined unequivocally.(6) Once inside the neurons, Gln is converted back to Glu by glutaminase, which is particularly enriched in mitochondria.(7) Finally, Glu is concentrated in synaptic vesicles via vesicular glutamate transporters (VGLUTs), thus completing the cycle.It should ne noted that (8) axonal terminal expressing EAAT2 will partly short circuit the cycle and that (9) there is an export of glutamine from the brain to the blood.
. Arginosuccinate is cleaved via Y.Zhou et al.

Table 1
Phenotypes of urea cycle disorders in human and mouse.

Table 2
Phenotypes of conditional glutamine synthetase knockout mice.Global (all cells from conception)The blastocysts fail to implant and die at ED3.4.He et al. (2007) Brain/most CNS astrocytes (hGFAP-Cre) The mice die soon after birth.They fail to feed and have slightly increased cortical ammonia levels.He et al. (2010a) Cerebral cortex (glutamatergic neurons and glia; Emx1-IRES-Cre)The mice are viable, but they have reduced life span, develop spontaneous seizures from 6 weeks of age, and have decreased locomotive activities with episodes of wild running.Oligodendrocytes (MOGi-Cre)The mice are viable, but neuronal glutamatergic transmission is disrupted in a myelin-independent way.There are deficits in cocaine-induced locomotor sensitization.

Table 3
Localization of glutamine synthetase in rodents.
Y.Zhou et al.
; Holmseth et al. (2012a) EAAT4 EAAT4 is expressed in cerebellar Purkinje cells with the highest concentrations where Zebrin II is, and is also present in scattered neurons in the fore-and midbrain.It is especially enriched in the parts of the dendritic and spine membranes facing astrocytes.most abundant of the EAATs in the retina and is expressed by the Müller cells.EAAT2 is found in cone photoreceptors and bipolar cells, while EAAT5 has been less studied, possibly expressed in Müller cells.

Table 5
Phenotypes of local administration of MSO.