Distribution of the branched chain aminotransferase proteins in the human brain and their role in glutamate regulation


Address correspondence and reprint requests to Myra E. Conway, Faculty of Health and Life Sciences, University of the West of England, Coldharbour Lane, Bristol, BS16 1QY, UK. E-mail: myra.conway@uwe.ac.uk.


The branched chain aminotransferase enzymes (BCAT) serve as nitrogen donors for the production of 30% of de novo glutamate synthesis in rat brain. Despite the importance of this major metabolite and excitatory neurotransmitter, the distribution of BCAT proteins in the human brain (hBCAT) remains unreported. We have studied this and report, for the first time, that the mitochondrial isoform, hBCATm is largely confined to vascular endothelial cells, whereas the cytosolic hBCATc is restricted to neurons. The majority of hBCATc-labelled neurons were either GABA-ergic or glutamatergic showing both cell body and axonal staining indicating a role for hBCATc in both glutamate production and glutamate release during excitation. Strong staining in hormone secreting cells suggests a further role for the transaminases in hormone regulation potentially similar to that proposed for insulin secretion. Expression of hBCATm in the endothelial cells of the vasculature demonstrates for the first time that glutamate could be metabolized by aminotranferases in these cells. This has important implications given that the dysregulation of glutamate metabolism, leading to glutamate excitotoxicity, is an important contributor to the pathogenesis of several neurodegenerative conditions, where the role of hBCATm in metabolizing excess glutamate may factor more prominently.

Abbreviations used

Alzheimer's disease




branched chain amino acids


branched chain aminotransferase


cornu ammonis area


consortium to establish a registry for Alzheimer's disease






ethylenediaminetetraacetic acid


gamma-amino butyric acid


human cytosolic branched chain aminotransferase


human mitochondrial branched chain aminotransferase




phosphate buffered saline

Studies using in vivo rat brain and ex vivo rat retina models demonstrate that the branched chain aminotransferases (BCATs) [E.C.] are key metabolic proteins, responsible for the production of 30% of de novo brain glutamate (LaNoue et al. 2001; Lieth et al. 2001). They catalyse the reversible transamination of the branched chain amino acids (BCAAs) leucine, isoleucine and valine to their respective α-ketoacids (BCKAs) and glutamate (Scheme 1) (Ichihara and Koyama 1966). These amino acids can cross the blood-brain barrier, where the influx of leucine is considerably higher than that of other amino acids (Oldendorf 1973; Smith et al. 1987). Studies using [15N]leucine have demonstrated that the BCAAs are nitrogen donors for the synthesis of glutamate and glutamine in brain explants and in primary neuronal cultures (Yudkoff et al. 1983, 1996a,b; LaNoue et al. 2001). De novo synthesis of glutamate also involves other enzymes, including pyruvate carboxylase and the TCA cycle enzymes, to produce α-ketoglutarate (LaNoue et al. 2001). The second and rate limiting step in the metabolism of the BCAAs is catalysed by the branched-chain α-keto acid dehydrogenase complex (BCKDC) composed of three subunits (E1–E3) (Harris et al., 1990). A metabolon between hBCATm and the E1 subunit has been reported that facilitates the channelling of branched-chain alpha-keto acids from BCATm to E1 (Islam et al. 2007), the kinetics of which are influenced by glutamate dehydrogenase (Islam et al. 2010).

Scheme 1.

There are two main human isoforms of BCAT, cytosolic BCAT (hBCATc) and mitochondrial BCAT (hBCATm), these enzymes sharing 58% homology in their primary amino acid sequence (Hutson 1988; Hutson et al. 1992; Davoodi et al. 1998). The mitochondrial isoform, hBCATm is expressed in most tissues and plays a significant role in skeletal muscle glutamine and alanine synthesis (Suryawan et al. 1998). These proteins are unique among the aminotransferase family where they are reversibly regulated through a redox sensitive CXXC motif, which is reversibly inhibited through S-nitrosation (Conway et al. 2004, 2008; Coles et al. 2009). The cytosolic isoform, hBCATc, is found predominantly in nervous tissue and accounts for 70% of rat brain BCAT activity (Sweatt et al. 2004a, b; Garcia-Espinosa et al. 2007). Two other variants have also been described that are homologous to hBCATm: a novel alternatively spliced variant found in placental tissue and a splice variant that is reported to act as a co-repressor of thyroid hormone nuclear receptors (Lin et al. 2001; Than et al. 2001).

As a result of the ease with which the BCAAs pass the blood–brain barrier and their role in glutamate metabolism, the subcellular localization of the BCAT proteins, particularly BCATc, in brain has been extensively investigated in rat brain and cell models (Bixel et al. 1997, 2001; Sweatt et al. 2004a, b; Garcia-Espinosa et al. 2007). In rat brain, the expression of BCATc was found to be solely in neurons, its subcellular distribution depending on neuronal type (Sweatt et al. 2004a, b). In glutamatergic neurons, such as granule cells of the cerebellar cortex and dentate gyrus, BCATc was localized to axons and nerve terminals, whereas in GABAergic neurons, such as cerebellar Purkinje cells and hippocampal pyramidal basket cells, BCATc was concentrated in the cell body. BCATc was also found in autonomic nerve fibres of the digestive tract, as well as in axons of the sciatic nerve (Sweatt et al. 2004a, b). Although the mitochondrial isoform was not mapped in these studies, recent work has reported uniform staining of BCATm in the cell bodies of astrocytes throughout rat brain, supporting the hypothesis that BCATm contributes nitrogen to the glutamate-glutamine cycle via astrocytes in rat (Cole et al. 2012). However, in primary cell cultures of rat brain, BCATm was detected in astrocytes and microglia, and BCATc was found not only in neurons but also in developing oligodendrocytes (Bixel et al. 1997, 2001). In the developing rat brain, BCATc expression was up-regulated in neurons and glial cells as they responded to changes in connectivity and growth factors (Garcia-Espinosa et al. 2007).

The cell-specific compartmentalization of BCATc to neurons and BCATm to astrocytes, together with several key metabolic studies, suggested that in astrocytes transamination was in the direction of glutamate and α-keto acid synthesis, the reverse holding true for BCATc transamination in neurons (Chaplin et al. 1976; Brand 1981; Brookes 1993; Yudkoff et al. 1994). These findings led to the development of the BCAA-BCKA shuttle hypothesis, in which the BCAT proteins function as an anaplerotic pathway that interfaces with the glutamate-glutamine cycle to replenish ‘lost’ glutamate (Fig. 2) (Hutson et al. 1998, 2001). This proposed pathway allows for nitrogen transfer for glutamate synthesis in neurons and to buffer neuronal glutamate (Yudkoff 1997). However, in a rat cell culture model, BCATc was also detected at low levels in some astrocyte populations and it has therefore been unclear the extent to which these models are directly applicable to man (Bixel et al. 1997, 2001).

Figure 2.

The glutamate-glutamine cycle and the branched chain amino acid (BCAA)-branched chain α-ketoacid (BCKA) shuttle. Glutamate is released from neuronal cells during excitatory neurotransmission. Excess glutamate not utilized is taken up via specific receptors in astrocytes. In astrocytes, much of the glutamate undergoes amidation to glutamine by glutamine synthetase. Glutamine is then released for uptake by neuronal cells, to replenish the glutamate pool. However, as glutamate can also be lost through oxidation or generation of glutathione/purines in astrocytes the glutamate-glutamine cycle must work with other anaplerotic pathways to regenerate the neuronal pool (McKenna 2007). BCAA metabolism is thought to participate in nitrogen shuttling in the de novo synthesis of glutamate. In brief, it was proposed that mitochondrial branched chain aminotransferase (BCAT) catalyses the transamination of the BCAA in astrocytes generating glutamate, which enters the glutamate-glutamine cycle with BCKA shuttled to neuronal cells for transamination with glutamate back to BCAA, by BCATc which exit the neuron and return to the astrocyte, with α-ketoglutarate entering the TCA cycle (Yudkoff 1997; Hutson et al. 2001).

Our aim in this study was to examine the distribution of the hBCAT proteins in the human brain, to ascertain whether the proposed BCAA-BCKA shuttle might contribute to the regulation of glutamate in the human brain. The regulation of glutamate is of particular importance for the prevention of excitotoxicity, an important contributor to neuronal cell death in brain ischaemia and several neurodegenerative diseases. For the first time, this study maps the hBCAT proteins in the human brain. In contrast to reported findings in the cell models, our findings highlight the presence of hBCATm in endothelial cells throughout the brain vasculature and an absence of detectable labelling in astrocytes. As in the rat, we report that in man hBCATc is restricted to neurons, although widely distributed throughout the brain. We discuss the impact of these findings with respect to the current BCAA-BCKA shuttle and the potential impact on glutamate regulation in the human brain.

Materials and methods


Peroxidase substrate 3,3′-diaminobenzidine (DAB), amino ethyl carbazole (AEC), donkey serum, 4′,6-diamidino-2-phenylindole (DAPI)-containing mounting medium, biotin-labelled secondary antibody raised to IgG and avidin (Vectastain ABC kit) were purchased from Vector labs (Peterborough, UK). Haematoxylin, Clearene and Clearium were purchased from SurgiPath (Peterborough, UK). Rabbit polyclonal antibody to hBCATc and hBCATm were purchased from Insight Biotechnology limited (Wembley, UK). Mouse monoclonal antibody to hBCATm was purchased from Abcam (Cambridge, UK). Mouse antibody to GAD67 was purchased from Millipore (Billerica, MA, USA). Mouse antibodies to HLA and glial fibrillary acidic protein (GFAP) were purchased from Dako (Cambridgeshire, UK). Mouse antibody to microtubule-associated protein 2 (MAP2) was purchased from Zymed (South San Francisco, CA, USA). All other materials were purchased from Fisher Scientific (Loughborough, UK).

Tissue preparation

The study was approved by North Somerset and South Bristol Research Ethics Committee. All brain tissue used in this study was from brains donated with consent to the South West Dementia Brain Bank (SWDBB) in the University of Bristol. The right hemibrain had been fixed in 10% buffered formalin for 3 weeks before tissue blocks were cut and embedded in paraffin wax for detailed neuropathological assessment. For this study, we examined 12 normal brains, from people who were older than 70 years of age who had not had dementia, and which did not show any neuropathological abnormalities with final clinical outcomes that varied between patients (Table 1).

Table 1. Clinical outcome for patients used in this study
PatientAgeGenderPM (delay)Brain weightCause of death
  1. F, Female; M, Male; PM, Post-mortem *1 - Cause of death was determined at autopsy.

Patient 183F24944Ventricular failure
Patient 285M311337Acute myocardial infarction
Patient 387M241364Acute renal failure
Patient 478M481254Prostate cancer
Patient 580M461279Pneumonia
Patient 673M351350Bi ventricular pulmonary disease
Patient 788F281060Chronic obstructive pulmonary disease
Patient 872F241200Age
Patient 989F151135Serious fall
Patient 1080F39750Colon cancer
Patient 1182F351150Left ventricular failure
Patient 1276M231450Cardiac arrest

Serial sections 7 μm in thickness were taken from multiple regions of brain, including the frontal, temporal, parietal and occipital lobes, basal ganglia, thalamus, hypothalamus, midbrain, cerebellum, pons and medulla. The sections were subsequently immunolabelled as described below. The brain samples used in these studies have been examined in other investigations of protein expression, for example, Endothelin-converting enzyme and tumour necrosis factor receptors I and II (Culpan et al. 2007; Palmer et al. 2010).


Sections were placed in a 60°C oven overnight to aid adhesion prior to immunohistochemical staining and subsequently dewaxed in clearene (2 × 5 min) and dehydrated in 100% ethanol (2 × 3 min). Endogenous peroxidase was quenched in 0.09% hydrogen peroxide/methanol solution for 30 min at ∼20°C. The slides were pre-treated with citrate buffer (10 mM sodium citrate, 0.05% Tween 20, pH 6.0). The slides were then washed (2 × 3 min) in phosphate buffered saline (PBS) containing 0.154 M NaCl, 1.86 mM NaH2PO4.2H2O, 7.48 mM Na2HPO4.12H2O, pH 7.1 where non-specific binding sites were blocked with 10% horse serum in PBS for 20 min at ∼20°C. Sections were incubated at ∼20°C overnight (20 h) with primary antibody (1/6000 for hBCATc and hBCATm) in PBS. The sections were washed in PBS (2 × 3 min) then incubated with biotinylated antibody to IgG for 20 min (Vectastain ABC kit). The slides were washed again in PBS (2 × 3 min) and incubated with the avidin–biotin complex in PBS for 20 min (Vectastain ABC kit). Slides were developed with DAB/H2O2 in distilled water (DAB substrate kit) for 10 min prior to immersion in copper sulphate solution (16 mM CuSO4.5H2O, 0.123 M NaCl) for 4 min, and then counter-stained with Harris's haematoxylin (25% Gill haematoxylin). The slides were dehydrated in 100% ethanol (2 × 5 min), cleared in 100% Clearene (2 × 3 min) and mounted in Clearium. Sections were viewed and imaged on a Nikon Eclipse 50i (Nikon UK, Kingston Upon Thames, UK) or Leica DMR microscope (Leica Microsystems, Milton Keynes, UK). For each experimental run a positive hBCAT control was included, together with a secondary antibody control and an antigen/antibody pre-incubation reaction that shows a negative image, confirming and validating the method and antibodies used in these experiments. To avoid duplication of results examples of these control slides are given in Figs 3c, 4b and e.

Figure 3.

Staining of human cytosolic branched chain aminotransferase (hBCATc) in the temporal lobe and cerebellum. (a) The temporal neocortex (inferior temporal gyrus) showing immunopositive neurons. (b) Hippocampal region CA1 showing negative pyramidal cells and positive interneurons. (c) Antigen incubation of serial section of b, at 200X molar excess. (d) Small immunoreactive neurons and a large pyramidal neuron (large arrow) with visible processes. (e) Granular cell layer with positive basket and golgi cells (*) and stellate cells (small arrows). (f) Purkinje cell bodies shown to be weakly immunopositive. (g) Intermittent staining of axons within the white matter suggesting nodal distribution (large arrows). Scale bar: a, 200 μM; e, 100 μM; b, c, d, f and g, 50 μM

Figure 4.

Medulla human cytosolic branched chain aminotransferase (hBCATc) and human mitochondrial branched chain aminotransferase (hBCATm) staining. (a) hBCATc staining of the inferior olivary nucleus. (b) Antigen incubation of serial section of a, at 200X molar excess. (c) Increased magnification of the inferior olivary nucleus showing staining of small neurons (large arrow) and neuropil staining (small arrow) along with immunonegative hylum (*). (d) hBCATm staining of the inferior olivary nucleus. (e) Antigen incubation of serial section of d, at 200X molar excess. (f) Vessel staining (*) within the amiculum of the inferior olivary nucleus. Scale bar: a, b, d and e, 200 μM; c and f, 100 μM.

Scoring protocol

For anatomical reference, labelling was scored in the same region of each section: the collateral sulci of the temporal lobe and the overlying cortex and areas CA1 and CA4 of the hippocampus. The sections were examined under a 20X objective and the degree of labelling designated on an arbitrary semi-quantitative scale: 0 = no labelling, 1 = weak labelling of occasional cells, 2 = labelling of many cells, 3 = intense labelling of most of the relevant cells. Labelling was scored according to cell type: neuronal for hBCATc and vascular for hBCATm. Independent scoring was performed by three immunohistochemists to validate the scoring assigned.


The slides were dewaxed in Clearene and dehydrated in ethanol, pre-treated with EDTA (10 mM Tris base, 1 mM EDTA, 0.05% Tween 20, pH 9.0) and blocked with donkey serum (1/10 dilution). Slides were incubated for 20 h, with primary antibody dilutions of 1/100 for hBCATc and GAD67 respectively. Subsequently, the slides were washed in PBS (2 × 3 min) and incubated in species-specific secondary antibody (Alexafluor 488 green or 555 red) at a concentration of 1/1000, in the dark for 60 min. The slides were washed in distilled water (2 × 5 min) and mounted in aqueous mountant containing DAPI. Images were acquired using Andor IQ software (Cairn Research Ltd, Faversham, UK) and a confocal Nikon Eclipse 80i microscope.


Antibody specificity was tested by western blot analysis using purified over-expressed hBCATc and hBCATm. No cross-reactivity between isoforms was reported at the antibody concentrations used in these experiments and specificity was comparable to a commercially available antibody (Abcam) (Fig. 5a). Furthermore, these proteins were also identified used western blot analysis in tissue homogenates (Fig. 5b). The labelling was completely prevented by prior incubation of the antibody with antigen. For IHC, an identical pattern of labelling to that of the hBCATm antibody from Abcam was reported from antibodies generated from Insight Biotechnology antibody to hBCATm. For further confirmation, antigen absorption was also used for each experimental run (at 200X molar excess) during IHC (Fig. 3c). The figures generated reflect five patient brains and are representative of the staining observed in all 12 patients examined.

Figure 5.

Specificity of the hBCAT antibodies. (a) Western blot analysis of over-expressed human mitochondrial branched chain aminotransferase (hBCATm) and human cytosolic branched chain aminotransferase (hBCATc) (5 and 10 ng respectively) using anti-hBCATc (1/1000) and reprobed with anti-hBCATm (1/1500). (i) (using anti-hBCATc): Lane 1, recombinant hBCATm (10 ng); Lane 2 recombinant hBCATc (10 ng); Lane 3 recombinant hBCATm (5 ng); Lane 4 recombinant hBCATc (5 ng). (ii) (using anti-hBCATm), Lanes 1–4 as for (i). (b) Control brain tissue (temporal cortex) homogenates probed for hBCATc and hBCATm. Western blot was carried out using antibodies raised to hBCATm and hBCATc. A constant 30 ng of hBCATc or hBCATm was used as a positive control (L1) and 20 μg of control patient homogenate was loaded in the other wells (L2 and L3).

Distribution of hBCATc in the human brain

Labelling of hBCATc was confined to neurons, and detected in all regions of all brains examined (Table 2). The cerebral cortex, hippocampal formation, subdivisions of the basal ganglia and diencephalon, the midbrain, cerebellum, pons and medulla all contained hBCATc-positive neurons. The antigen was largely confined to the neuronal soma and proximal dendrites, but there was occasional focal labelling of axons and scanty weak labelling of oligodendrocytes. In the cerebral cortex, most of the positive neurons were small but the antibody labelled scattered larger pyramidal cells and multipolar cells (Fig. 3a). Immunopositive neurons were most numerous in lamina 3 but were also present in lamina 2 and the deeper layers. In the hippocampus, there was variable immunolabelling of pyramidal cells and strong labelling of multipolar interneurons (Fig. 3b and d). Figure 3c shows antigen incubation controls that demonstrated the specificity of the antibody for hBCATc. Within the granule cell layer of the cerebellar cortex the somata of basket, stellate and Golgi neurons were strongly immunopositive (Fig. 3e). The cell bodies of the Purkinje cells were weakly immunopositive, as were the glomeruli (Fig. 3e and f). In some cases, there was distinct focal staining of axons in the white matter, in a pattern suggesting nodal distribution (Fig. 3g).

Table 2. An overview of hBCATc immunoreactivity throughout the human brain
AreaImmunopositive neuronal cellsIntensity of staining
  1. (−), 0 score; (+), 1 score; (++), 2 score; (+++), 3 score; (++++), > 3 score.

Temporal lobe
GABAergic interneurons++++++
Pyramidal neurons+++++
Dentate gyrus (neurons)++
Subiculum subpopulation of neurons+++
Lamina II neurons++++
Cortex and white matter
Cortical pyramidal cells++++
Cortical neurons++++
Axonal staining−/+++
Putamen and basal ganglia
Large neurons++++++
Small neurons++++
Insular cortex (small neurons)+++
Thalamus (neurons)+++
Caudate nucleus (neurons)++
Lateral geniculate nucleus (neurons)++
Widespread neuronal staining+++++++
Periaqueductal grey matter (neurons)++++
Neuropil staining+++
Inferior colliculus (nerve cells)++++
Supraoptic nucleus of the hypothalamus+++++
Paraventricular neurons of the hypothalamus+++++
Oligodendrocyte staining in the white matter+++
Purkinje cells++
Swollen axon terminals++++
Neurons in the dentate nucleus+++
Pontine nuclei (neurons)+++
Tegmental neurons+++
Neuronal processes (nigro striatal processes)+++
Raphe nuclei (neurons)+++
Nucleus basalis of Meynert (cholinergic neurons)+++++
Hypoglossal nucleus (neurons)+++
Dorsal motor nucleus (neurons)+++
Nucleus ambiguus (neurons)+++++
Gracile nucleus++++
Inferior olivary nuclei (neurons)++++
Inferior olivary nuclei neuropil++++

In the putamen, large (aspiny) neurons were strongly labelled, and there was also weaker labelling of smaller neurons and numerous processes within the surrounding neuropil (Fig. 6). The nucleus basalis of Meynert contained large hBCATc-positive neurons. Neurons in the globus pallidus and the thalamus were only weakly labelled. Strong staining of neuronal somata and processes was seen in the hypothalamus (Fig. 7), particularly in the supraoptic and paraventricular nuclei. As the brown pigment in neurons in the midbrain, substantia nigra (SN) and locus coeruleus was difficult to distinguish from the brown peroxidase reaction product, in these regions AEC staining was used in place of DAB. In the SN, hBCATc was detected in the nerve cell bodies and processes (Fig. 8). Labelled neurons were seen throughout the midbrain with strongly labelled nerve cells in the inferior colliculus and relatively weakly labelled nerve cells in the periaqueductal grey matter.

Figure 6.

Distribution of human cytosolic branched chain aminotransferase (hBCATc) in the Basal ganglia. (a) The capsula externa (*) of the basal ganglia showing staining of large neurons (small arrow) and the surrounding processes. (b) Antigen incubation of serial section of a, at 200X molar excess. (c) Increased magnification of a single large neuron (small arrow). Scale bar: a and b, 200 μM; c, 50 μM.

Figure 7.

Distribution of human cytosolic branched chain aminotransferase (hBCATc) in the hypothalamus. (a) The optic tract (*) and the supraoptic nucleus (small arrows) of the hypothalamus. (b) Increased magnification of the neurons of the supraoptic nucleus. (c) Increased magnification of b to show clear neuronal cell body staining (large arrow) and process staining. Scale bar: a, 200 μM; b, 100 μM; c, 50 μM.

Figure 8.

Distribution of human cytosolic branched chain aminotransferase (hBCATc) in the mid brain. (a) Pigment containing neurons (small arrow) of the substantia nigra. (b) Increased magnification of a. Scale bar: a, 200 μM; b, 100 μM.

Within the pons, hBCATc-positive neurons were present in the raphe nuclei, locus coeruleus and basal pontine nuclei (Fig. 9a–g). The labelling of the pontine nuclei was in contrast to the lack of antigen in the corticospinal tract fibres (Fig. 9f). In the medulla there was strong labelling of neurons in the gracile and cuneate nuclei, hypoglossal nucleus and dorsal motor nucleus of the vagus. The medulla also contained many hBCATc-positive neurons and nerve cell processes in the inferior olivary nucleus (Fig. 4a–c).

Figure 9.

Distribution of human cytosolic branched chain aminotransferase (hBCATc) in the Pons. (a) The 4th ventricle of the Pons (*) with the Raphe nuclei (small arrow). (b) The locus ceruleus (large arrow). (c) Increased magnification of the locus ceruleus showing immunopositive neurons (small arrow). (d) Increased magnification of immunopositive neuronal bodies of the locus ceruleus (small arrow). (e) Increased magnification of the raphe nuclei. (f) Immunopositive neurons and processes in the pontine nuclei (large arrow) and immunonegative corticospinal tract fibres (*). (g) Antigen incubation of serial section of f, at 200X molar excess showing almost complete removal of immunoreactivity. Scale: a, b, f and g, 200 μM; e and c, 100 μM; d, 50 μM.

To ascertain the cell specificity of the neuronal labelling, antibodies specific to GABAergic neurons and astrocyes were used in double immunofluorescence studies. This showed that hBCATc co-localized with GAD67, specific for GABAergic neurons, and MAP2, a neuronal marker (Fig. 10a and b). There was no co-localization with the astrocytic antigen GFAP (Fig. 10c).

Figure 10.

Co-localization of human cytosolic branched chain aminotransferase (hBCATc) to cell type. (a) (From left to right). Red immunofluorescence stained for microtubule-associated protein 2 (MAP2), green immunofluorescence stained for hBCATc and finally a merge showing hBCATc is localized in some neurons but not all. (b) (From left to right). Red immunofluorescence stained for GAD67, green immunofluorescence stained for hBCATc and finally a merge showing hBCATc and GAD67 are co-localized in the same neurons. (c) (From left to right). Red immunofluorescence stained for glial fibrillary acidic protein (GFAP), green immunofluorescence stained for hBCATc and finally a merge showing hBCATc and GFAP are not co-localized in the same cells.

Distribution of mitochondrial hBCAT

Immunoreactivity for hBCATm was present throughout the brain and in all brains examined. The most consistent labelling was of vascular endothelial cells in the grey and white matter (Figs 4 and 11). A small population of glial cells in the subpial region in the inferomedial part of the temporal lobe showed coarse granular immunopositivity, but this was not present in all brains (2/12 brains examined). The distinct patterns of immunolabelling for hBCATc and hBCATm are clearly shown in Fig. 4: hBCATm in the walls of blood vessels in amiculum of the inferior olivary nucleus (Fig. 4d and f) and hBCATc in the neurons and neuropil of the nucleus (Fig. 4a and c).

Figure 11.

Human mitochondrial branched chain aminotransferase (hBCATm) staining in the human brain. (a) hBCATm staining of the vasculature surrounding the hippocampus (small arrowhead), the granule cell layer is also shown (*). (b) Increased magnification of (a) showing clear vessel staining (small arrowhead) around the granule cell layer (*). (c) Increased magnification of vessel in (a and b) (small arrowheads). (d) Punctate staining (arrowhead) appearing in the vasculature of the Temporal lobe. (e) Punctate staining (arrowhead) appearing in the vasculature of the parietal lobe. (f) Labelling of pencillar fibres (large arrowheads). (g) Staining appearing in the vasculature of the parietal lobe. Scale bar: a, and f, 200 μm; b, 100 μm; c, d and e, 10 μm; g, 50 μm.

The endothelium of capillaries and larger blood vessels was immunopositive for hBCATm and showed clear, punctate staining in keeping with the mitochondrial location of this enzyme (Fig. 11). Clear staining of blood vessels was noted surrounding the hippocampus (Fig. 11a–c). There was also some labelling of the tunica media (Fig. 4d and f). Where the pencillar fibres were present, weak immunopositive staining was observed in the basal ganglia (Fig. 11e) but elsewhere in all brains the white matter was unlabelled. The striatum showed weak hBCATm labelling of neuronal cell bodies, and there was further weak labelling of neurons in the hypothalamus, periaqueductal grey matter and inferior olive.

In summary, hBCATc was found in all brain regions and was neuron-specific throughout apart from occasional oligodendrocyte staining within the white matter. The intensity of hBCATc labelling varied between individuals; however, there was strongest labelling in putatively GABAergic neurons in the hippocampus, neocortex, putamen, hypothalamus, pons and medulla, with weaker labelling of putatively glutamatergic neurons (11/12 cases examined). hBCATm was predominantly vessel associated, involving the endothelium and tunica media, with no labelling of astrocytes. In one case, some labelling of neurons for hBCATm in the deep cerebral grey matter and brain stem was observed but staining was significantly less intense than reported for hBCATc in neurons and was absent all together from most brains (11/12 cases examined).


Given the key role that BCAT enzymes play in brain glutamate metabolism, it is important to establish the distribution of these isoforms in the human brain relative to rat and cell culture models. Consistent with the rat and cell culture models, hBCATc was found predominantly in neuronal cells, supporting its proposed role in the synthesis of the neurotransmitter glutamate and as a precursor for GABA production. Distribution of hBCATc expression in hormone secreting cells offers the potential for additional roles for hBCATc. Moreover, for the first time we show that unlike the rodent model, which showed expression of hBCATm in astrocytes, there is extensive immunopositive staining for hBCATm in the vasculature of the brain with no evidence of staining within astrocytes, suggesting a different role for hBCATm in glutamate regulation in the human brain. This and other findings are discussed in further detail with respect to published models.

The role of BCAAs in glutamate production in neurons is based on studies that show BCAAs can label a significant proportion of the glutamate pool in neuronal cell culture, which has also been shown in vivo and ex vivo rat retina models (Yudkoff et al. 1983, 1994; Kanamori et al. 1998; LaNoue et al. 2001). Similar to the rat studies by Garcia-Espinosa et al. (2007) the majority of hBCATc-labelled neurons in the human brain were either GABAergic or glutamatergic, with GABAergic neurons shown to be more strongly immunopositive than glutamatergic neurons. However, many neurons that were neither glutamatergic nor GABAergic also labelled for hBCATc. The distribution and the varied intensities of expression in the hippocampus, where staining of the CA3 region was more intense than the CA1 region, mirrored the findings of Castellano et al. (2007) who investigated mRNA expression of hBCATc in post-natal and adult brains of mice. The cell bodies of neuronal cells within the temporal and hippocampus showed intense staining relative to the dendrite regions reflecting the possibility that their primary role would be to contribute to the glutamate metabolic pool used to generate glutamate rather than the glutamate pool used during excitation. Conversely, in the areas of the supraoptic tract intense staining along axons was noted indicating an additional role of hBCATc transamination in glutamate release in this region. Our findings are consistent with the rat and rat metabolic studies suggesting that hBCATc transamination could potentially contribute to the pool of neurotransmitters and the glutamate required for release during excitation.

Magnocellular neuroendocrine cells were shown to express high levels of hBCATc, in the supraoptic nucleus and the paraventricular nucleus of the hypothalamus. In response to osmoreceptors in the brain these cells which are rich in AMPA and NMDA receptors, respond to glutamate stimulation, which act together with intrinsic membrane conductance that control the bursting pulsatile release of oxytocin and vasopressin (reviewed in Pak and Currás-Collazo 2002). Precedence for BCAT metabolites in hormone secretion has been reported, where leucine has been shown to be a potent insulin secretagogue (Xu et al. 2001; Zhou et al. 2010). As intense hBCATc expression in the magnocellular neurons and also in the substantia nigra was reported it will be important to further investigate if these proteins play a role in hormone release in these cells. The presence of hBCATc in the neurons of the nucleus basalis of Meynert and the raphe nuclei raises a similar curiosity about hBCATc in acetylcholine and serotonin release or in intermediate production. Moderate staining of hBCATc in pigmented cells, such as the neurons of the locus ceruleus, that strongly express melanin concentrating hormone (MCH), may be related to the role of BCAAs in melanogenesis, where it has been demonstrated that incubation of melanoma cells with BCAA inhibited the production of melanin (Cha et al. 2012). The mechanism underpinning this inhibition was not determined. Therefore, given the role of leucine in insulin release, their contribution in these other endocrine cells warrants further investigation.

Previous studies using astroglial primary cultures derived from the brains of newborn Wistar rats showed that BCATm was largely located in astrocyte populations, the basis of which formed the hypothetical BCAA-BCKA nitrogen shuttle, where the BCAT enzymes facilitate the de novo synthesis of glutamate (Fig. 2) (Hutson et al. 1998, 2001; LaNoue et al. 2001). In rat, this allows for nitrogen transfer to regenerate glutamate in neurons (Yudkoff et al. 1996a, b; Yudkoff 1997). In contrast, here, no astrocyte staining was detected in the human brain, indicating that hBCAT transamination cannot occur in human astrocytes. Although, the explanation remains unclear, one possibility is that the expression of hBCATm is induced under conditions of cell proliferation, as demonstrated in studies by Perez-Villasenor et al. (Perez-Villasenor et al. 2005), or that the pattern of expression may be altered by placing cells in culture outside the normal milieu. Alternatively, these studies more likely highlight the difference in BCAT expression between species, especially in light of recent studies which have reported positive astrocyte staining for BCATm in rat brain tissue (Cole et al. 2012). Moreover, one must also consider the age of the brains used in this study and appreciate that expression of hBCATm may differ in younger brains of humans.

In these studies, the expression of hBCATm was pronounced throughout the endothelial layer examined in all areas of the human brain in all samples analysed indicating that transamination can occur in these cells, where the direction of metabolism will be directed by substrate concentration and the expression of other key metabolic systems. Cerebral endothelial cells have a high density of mitochondria (Oldendorf et al. 1977), giving greater capacity for energy production than other capillary endothelial cells and explaining the punctuate staining of mitochondrial hBCAT in these cells. The endothelial barrier has membrane-specific transporters, the L-system facilitative transporters (L1), on both the abluminal and luminal side which allow transport of BCAAs, particularly leucine (Oldendorf 1973; Smith et al. 1987; O'Kane and Hawkins 2003; O'Kane et al. 2004; Abbott et al. 2006). Conversely, only minimal glutamate entry is allowed on the luminal side, which is thought to protect the brain from peripheral glutamate plasma changes that could be potentially toxic to brain cells. Although controversy exists regarding the exact EAAT subtype and its expression on the abluminal side (Chaudhry et al. 1995; O'Kane et al. 1999) evidence in support of brain glutamate efflux is gaining impetus (Hosoya et al. 1999; Gottlieb et al. 2003; Uchida et al. 2011), the rate of which is considered to be influenced by blood glutamate (Gottlieb et al. 2003; Zlotnik et al. 2008, 2012; Teichberg et al. 2009; Campos et al. 2012). Currently, the proposed mechanism in animal models by which the endothelial cells remove excess glutamate include direct uptake of glutamate through Na+-dependent transporters, creating a rise in the glutamate gradient in endothelial cells, which has been proposed to subsequently exit into the blood through facilitated transport (O'Kane et al. 1999). Until the study by Helms et al. 2012, the potential for further metabolism of glutamate within endothelial cells was not considered. If glutamate dehydrogenase (GDH) is expressed in endothelial cells glutamate oxidation can occur, dependent on the concentration and redox state. However, our novel finding of the hBCATm transaminase in these cells raises the possibility of an additional mechanism to support glutamate synthesis or metabolism, in the human brain.

For glutamate oxidation to occur in endothelial cells driven by hBCATm, branched chain keto acids are required, although the source remains unclear. Oxidation of glutamate would release α-keto glutatarate that could enter the TCA cycle and the respective BCAA, which could be further oxidized for energy or be taken up by neuronal cells for further metabolism. Metabolism of brain glutamate by hBCATm could serve as an auxillary mechanism to the glutamate-glutamine cycle, the predominant system that regulates brain glutamate between neurons and astrocytes (Danbolt 2001). For this to be possible, levels of glutamate would need to exceed the Km of the glutamate transporter (approximated at 138 μM in rat models), which could only occur, albeit for ≤ 10 ms during excitation (Dzubay and Jahr 1999) or more importantly during periods of excitotoxicity, a feature in the pathogenesis of neurodegenerative conditions such as Alzheimer's disease (Moussawi et al. 2011). Although the kinetics of glutamate efflux may differ in humans, glutamate uptake and metabolism by hBCATm would be of benefit to glutamate homeostasis, in particular when glutamate levels are high, indicating a possible role for this protein as an auxiliary neuroprotection mechanism. Transamination in the opposite direction of BCAA metabolism would require α-ketoglutarate, which would affect TCA cycle activity if it were not replenished. Notably, pyruvate carboxylase could fill this role through carboxylation of pyruvate, but this enzyme has not been mapped to this area and is considered specific to astrocytes. Other potential candidates include malic enzyme or phosphoenolpyruvate carboxylase kinase, but these are thought to generally operate in the direction of decarboxylation. Alternatively, another transamination reaction (e.g. ALT) could operate to transfer nitrogen from glutamate to pyruvate and regenerate α-keto glutarate and alanine. Evidence for nitrogen shuttling has been established in muscle, where ALT transaminases play a pivotal role in mediating the passage of intermediates. Indeed, if GDH is found to be present and BCATm/GDH association occurs, then the end product is ammonia and α-keto glutarate is regenerated (Hutson et al. 2011). Evidently, the metabolic reason for hBCATm expression in the endothelium is open to discussion, clarity of which will only evolve with improved models of human brain metabolism and also mapping of key metabolic proteins to this area.

In summary, we provide the first direct evidence of transaminases in the endothelial layer that can potentially operate as a support network to astrocytes mediating the fine tuning of glutamate homeostatsis. We propose that under normal physiological conditions the BCAAs are taken up by neuronal cells, where hBCATc controlled transamination acts as a pathway to replenish the glutamate pool. These findings may have further implications with respect to our understanding of glutamate toxicity, an important contributor to a range of neurodegenerative and vascular neurological diseases. Further research is needed into the role of BCAT in glutamate regulation and toxicity in the brain.


This study was supported by Bristol Research into Alzheimer's and Care of the Elderly (BRACE) (to M.E.C.). The University of the West of England, Bristol, BS16 1QY, UK. The authors declare that there are no conflicts of interest involved in the study.