Chapter 2. Interaction between peroxiredoxin 5 and dihydrolipoamide branched
2) The effect of hypoxic stress on DBT enzymatic activity
To examine the effect of hypoxia on DBT enzymatic activity, I assayed DBT enzymatic activity in mouse kidneys under normoxic and hypoxic conditions using the method described by Rodriguez-Bayon (20). In briefly, the activity of DBT was determined spectrophotometrically by measuring the rate of NADH production with α-keto isocaproate as substrate. This type of technique has frequently been used to determine the DBT activity in mammalian tissues subjected to different nutritional and hormonal conditions. DBT enzymatic activity in hypoxic and normoxic mouse kidney was 1.20±0.05 and 0.88±0.04 mU/mg, respectively (n=10 for each group; Figure 3). This suggested that hypoxic treatment resulted in a significant increase in DBT activity (about 1.5-fold) versus normoxic conditions (Figure 3, inset bar graph). Taken together, these data indicated that hypoxic treatment induced interaction between Prdx5 and DBT and increased activity of DBT.
3) Confirmation DBT overexpressing construct
To construct the human DBT overexpressing vector, I cloned DBT cDNA into pCMV- myc-tag vector with EcoRI/XhoI restriction enzyme sites using PCR method. Then the ligation product of DBT construct was transformed into XL-10 gold competent cells (Stratagene, California, USA) via heat-shock method. The successfully transformed cells were selected by LB agar supplemented with 100 àg/ml ampicillin. To screen the DBT/pCMV construct transformed clone, the colony PCR was carried out with DBT primers (part 5 of Materials and
38
Methods). The positive clones then were confirmed by restriction enzyme digestion EcoRI/XhoI and sequencing analysis (Figure 4). These results proved that DBT overexpressing vector was successfully constructed.
Before transfected into HEK293 cells, all constructs were confirmed by enzyme digestion with EcoRI/XhoI or EcoRI/BglII (Figure 5 and Figure 4C). Then the Prdx5 and DBT expression of co-transfected were examined by western blot analysis with anti-myc affinity gel (#E6654, Sigma-Aldrich, USA) or Anti-HA agarose (#A2095, Sigma-Aldrich, USA). As shown in Figure 5C, all constructs were successfully co-transfected and overexpressed in HEK293 cells.
4) The co-localization of Prdx5 and DBT under hypoxic stress
To determine subcellular co-localization, co-immunostaining of Prdx5 and DBT was carried in normoxic and hypoxic conditions. HEK293 cells were cotransfected with a myc- tagged DBT and a HA-tagged Prdx5. As shown in Figure 6, DBT (green) was colocalized with Prdx5 proteins (red) in hypoxic HEK293 cells, while the complex was weakly detectable in normoxic cell. Colocalized pixels were quantified as the percentage of overlapping selected red (Prdx5) and green pixels (DBT) and revealed that co-localization of these proteins increased 3.8- fold in hypoxia versus normoxia (inset bar graph, Figure 6). These results were consistent with results collected in mouse kidney model; thus interaction between Prdx5 and DBT occurs not only in mouse model but also in human cellular levels. These data proved that the enhancing linkage between Prdx5 and DBT in hypoxic stress was a common phenomenon, at least in mouse and human.
5) The role of Prdx5 cysteine residues in the Prdx5-DBT interaction
Human Prdx5 has one catalytic cysteine residue, cystein 48, which is highly conserved in all other peroxiredoxins (Prdx1-4 and 6), and this cysteine residue is involved in peroxidase activity (30). Most recently, Choi and collaborators demonstrated that Prdx5 interacted with Jak2 via the catalytic cystein 48 to modulate the Jak2-Stat5 pathway (29). Based on previous observations, I asked whether the catalytic cysteine 48 plays a role in the interaction between Prdx5 and DBT.
To investigate this hypothesis, I transiently co-transfected HEK293 cells with a myc- tagged DBT and a HA-tagged wild-type (WT) or a cysteine mutant Prdx5 (C48S, C73S, C152S, and C48/152S) expression construct. Then, total cell lysate proteins from transfected-HEK293 cells under normoxia and hypoxia were used for immunoprecipitations with anti-myc-conjugated agarose. The cellular expression of DBT and Prdx5 for each transfected construct showed the
39
same background for every Prdx5 mutant (Figure 5C). As shown in Figure 7A, WT Prdx5 bound weakly to DBT under normoxic conditions, but this interaction showed a significant increase under hypoxic stress. In the case of the Prdx5 C152S mutant, the interaction with DBT did not change compared with WT Prdx5. However, Prdx5 C48S mutant and double mutant (C48/152S) showed diminished interactions with DBT under both normoxic and hypoxic conditions. This result is consistent with the previous finding of the cysteine 48 residue’s role in the interaction of Prdx5 and Jak2. Additionally, Prdx5 C73S showed a notable increase in the interaction with DBT in both normoxic and hypoxic conditions. The mechanism by which cysteine 73 mutation can contribute to enhancing of Prdx5-DBT interaction needs further investigation. Using reverse immunoprecipitation, I also found that Prdx5 and DBT interaction was increased in hypoxia.
Although Prdx5 C73S showed increased interaction with DBT, Prdx5 C48S showed decreased interaction with DBT (Figure 7B). Thus, these data indicated that the Prdx5 and DBT interaction during normoxic and hypoxic condition required the catalytic cysteine of Prdx5.
40
Figure 1. Scheme of function of BCKDH complex in BCAAs catabolic pathway. BCKDH complex is rate-limiting enzyme to break down BCAAs (25). In initial step of BCAAs catabolic pathway, BCAAs (leucine (Leu), isoleucine (Ile), valine (Val)) are coverts into their α-keto acid derivatives, such as α-ketoisocaproate (KIC), α-keto-β-methylvalerate (KMV), or α- ketoisovalerate (KIV) by the BCAA aminotransferase isozymes (BCATm and BCATc). In this
A
B
41
transamination reaction, α-ketoglutarate is the α-keto acid acceptor of the BCAA nitrogen group producing glutamate (Glu). Glu then can be converts into glutamine (Gln) or contributes into neurotransmitter substrate cycles as Glu or GABA (gamma amino butyric acid), glucose-alanine cycle (Cahill cycle). The second step is oxidative decarboxylation reaction catalyzing by the BCKDH complex. Both isoleucine and valine can converts to succinate via methylmalonyl-CoA, then entering the tricarboxylic acid (TCA) cycle, whereas leucine and isoleucine metabolize into acetyl-CoA which can enter TCA cycle or contribute into cholesterol synthesis through a cascade of enzymatic reaction (panel A). The BCKDH complex consists of 3 sub-units: E1 decarboxylase, E2 transacylase and E3 dehydrogenase subunits. The activity of BCKDH complex is thought to be regulated by a reversible phosphorylation (inactivation) / dephosphorylation (activation) cycles (2). The reaction also requires some cofactor such as thiamine pyrophosphate (TPP), flavin adenine dinucleotide (FAD), nicotinamide adenine dinucleotide (NAD+), lipoate, coenzyme A (panel B).
42
Figure 2. Coprecipitation of endogenous Prdx5 and DBT in normoxic and hypoxic mouse kidney. Immunoprecipitates identified by DBT from normoxic and hypoxic mouse kidney were probed with a Prdx5 antibody (n=5) (A). Even though there is not significantly different between expression of DBT and Prdx5 in normoxic and hypoxic kidneys, the Prdx5 amount that was co- purified with DBT was higher in hypoxic kidneys than normoxic kidney. Histogram showed numerical data obtaining by densitometry analysis of (A) (B). In brief, Prdx5 that captured by DBT antibody from hypoxic kidney was four-folds higher than normoxic kidney. This result is consistent with my previous study in which interaction between DBT and Prdx5 was found enhanced in hypoxic mouse kidneys. The histogram graph is expressed as mean values ± standard deviation (**p < 0.05). WCE, whole cell extracts of mouse kidney.
B A
43
Figure 3. In vitro assay of DBT enzymatic activity in normoxic and hypoxic mouse kidneys.
Plots show time-dependent changes in absorbance at 340 nm. The histograms in the insets show the slopes (rate) of the changes in the linear portion of the curves. The y-axes in the histograms indicate the specific activity expressed as mU/mg mitochondrial protein. 1 mU is defined as the amount of enzyme needed to reduce 1 nmol of NADP+ per min at 30°C. Assays were carried out with the mitochondrial fraction of homogenized tissue extract, and the reactions were run using sodium α-ketoisocaproate as the substrate. Data are expressed as mean values ± standard deviation (**p< 0.05). Open circles and histogram, normoxia; closed circles and histogram, hypoxia. These data showed that DBT enzymatic activity increased about 1.5 folds in hypoxic kidney (1.20±0.05 versus 0.88±0.04 mU/mg, hypoxic and normoxic kidneys, respectively). This evidence shows the linkage between hypoxia and DBT activity.
44
A
B C
45
hDBT ACATACACACCAAGAGATAAAGGGCCGAAAAACACTGGCAACTCCTGCAGTTCGCCGTCT 539 FL-DBT-F ACATACACACCAAGAGATAAAGGGCCGAAAAACACTGGCAACTCCTGCAGTTCGCCGTCT 524 ************************************************************
hDBT GGCAATGGAAAACAATATTAAGCTGAGTGAAGTTGTTGGCTCAGGAAAAGATGGCAGAAT 599 FL-DBT-F GGCAATGGAAAACAATATTAAGCTGAGTGAAGTTGTTGGCTCAGGAAAAGATGGCAGAAT 584 ************************************************************
hDBT ACTTAAAGAAGATATCCTCAACTATTTGGAAAAGCAGACAGGAGCTATATTGCCTCCTTC 659 FL-DBT-F ACTTAAAGAAGATATCCTCAACTATTTGGAAAAGCAGACAGGAGCTATATTGCCTCCTTC 644 ************************************************************
hDBT ACCCAAAGTTGAAATTATGCCACCTCCACCAAAGCCAAAAGACATGACTGTTCCTATACT 719 FL-DBT-F ACCCAAAGTTGAAATTATGCCACCTCCACCAAAGCCAAAAGACATGACTGTTCCTATACT 704 ************************************************************
hDBT AGTATCAAAACCTCCGGTATTCACAGGCAAAGACAAAACAGAACCCATAAAAGGCTTTCA 779 FL-DBT-F AGTATCAAAACCTCCGGTATTCACAGGCAAAGACAAAACAGAACCCATAAAAGGCTTTCA 764 ************************************************************
hDBT AAAAGCAATGGTCAAGACTATGTCTGCAGCCCTGAAGATACCTCATTTTGGTTATTGTGA 839 FL-DBT-F AAAAGCAATGGTCAAGACTATGTCTGCAGCCCTGAAGATACCTCATTTTGGTTATTGTGA 824 ************************************************************
hDBT TGAGATTGACCTTACTGAACTGGTTAAGCTCCGAGAAGAATTAAAACCCATTGCATTTGC 899 FL-DBT-F TGAGATTGACCTTACTGAACTGGTTAAGCTCCGAGAAGAATTAAAACCCATTGCATTTGC 884 ************************************************************
hDBT TCGTGGAATTAAACTCTCCTTTATGCCTTTCTTCTTAAAGGCTGCTTCCTTGGGATTACT 959 FL-DBT-F TCGTGGAATTAAACTCTCCTTTATGCCTTTCTTCTTAAAGGCTGCTTCCTTGGGATTACT 944 ************************************************************
hDBT TGCTCGTGGAATTAAACTCTCCTTTATGCCTTTCTTCTTAAAGGCTGCTTCCTTGGGATT 956 FL-DBT-R TGCTCGTGGAATTAAACTCTCCTTTATGCCTTTCTTCTTAAAGGCTGCTTCCTTGGGATT 933 ************************************************************
hDBT ACTACAGTTTCCTATCCTTAACGCTTCTGTGGATGAAAACTGCCAGAATATAACATATAA 1016 FL-DBT-R ACTACAGTTTCCTATCCTTAACGCTTCTGTGGATGAAAACTGCCAGAATATAACATATAA 993 ************************************************************
hDBT GGCTTCTCATAACATTGGGATAGCAATGGATACTGAGCAGGGTTTGATTGTCCCTAATGT 1076 FL-DBT-R GGCTTCTCATAACATTGGGATAGCAATGGATACTGAGCAGGGTTTGATTGTCCCTAATGT 1053 ************************************************************
hDBT GAAAAATGTTCAGATCTGCTCTATATTTGACATCGCCACTGAACTGAACCGCCTCCAGAA 1136 FL-DBT-R GAAAAATGTTCAGATCTGCTCTATATTTGACATCGCCACTGAACTGAACCGCCTCCAGAA 1113 ************************************************************
hDBT ATTGGGCTCTGTGGGTCAGCTCAGCACCACTGATCTTACAGGAGGAACATTTACTCTTTC 1196 FL-DBT-R ATTGGGCTCTGTGGGTCAGCTCAGCACCACTGATCTTACAGGAGGAACATTTACTCTTTC 1173 ************************************************************
D
46
Figure 4. Cloning DBT/pCMV construct. The DBT construct was selected with colony PCR (A). Then the positive colony was cultured overnight and collected the plasmid. The DBT construct was reconfirmed by restriction enzyme digestion with EcoRI/NotI (C). In panel C, lane 1 was pCMV-N-myc vector that was cut with EcoRI/NotI as backbone control, lane 2 was DBT/pCMV construct that showed the positive signal in colony PCR and then was cut with EcoRI/NotI. The restriction enzyme digestion generated 2 products in DBT construct: a 3.8 kbps band (pCMV-myc back bone) and a 1.4 Kbps band (full length DBT). After that, DBT construct was confirmed by sequencing analysis with primers used for cloning. The sequence of DBT construct, lower sequence, was aligned with DBT reference sequence, upper sequence, from Gene Bank (Accession number: BC_016675) by ClustalW2 software (D) and the vector map of DBT/pCMV construct also was represented (B). Sequence of primers used in cloning, colony
PCR, nucleotide sequencing analysis: forward: 5’-
GGCCGAATTCGGATGGCTGCAGTCCGTATGC-3’; reverse: 5’-
GGCCGCGGCCGCTCATTTCAGATCTAGTAGCATAAAAGC-3’. M, size marker (D-1035, Bioneer, Korea).
47
A
B
48
Figure 5. Confirmation of WT and mutant Prdx5 constructs. All constructs relating with Prdx5 in this experiment were cloned as previous study (29). The vector maps of all Prdx5 relating constructs were represented (A). To confirm all Prdx5 relating construct, I digested plasmids with restriction enzymes EcoRI and NotI, then the results were observed on 1%
agarose gel (B). The digestion products consist of two bands with 3.8 Kbps (pCMV-N-HA backbone) and 500 bps (Prdx5 WT, C48S, C73S, C152S, C48/152S). The expression of Prdx5 (HA tag) and DBT (myc tag) co-transfected cells also were examined by western blot analysis (C). The results proved that all Prdx5 and DBT construct were successful expressed in HEK cells and there was no different expression level between Prdx5 WT, Prdx5 cysteine mutant when they were co-transfected DBT.
C
49
Figure 6. Co-localization of Prdx5 and DBT in normoxic and hypoxic HEK293 cells.
Confocal images of costaining with antibodies to Prdx5 and DBT. HEK293 cells were co- transfected with Prdx5- and DBT-expressing vectors, and then the transfected cells were exposed to hypoxic stress (1.0±0.2%) for 6 hours. Cells were fixed on polylysine slides and stained with Prdx5 (red) or DBT (green) antibody. These results revealed that co-localization of these proteins increased 3.8-folds in hypoxia versus normoxia and enhancement of hypoxia on DBT-Prdx5 interaction also is observed in cellular level.
50
Figure 7. Comparative interactions of Prdx5 WT or cysteine mutants with DBT in normoxic and hypoxic cells. Coprecipitation of HA-tagged Prdx5 with myc-tagged DBT (A).
Reverse immunoprecipitation of myc-tagged DBT with HA-tagged Prdx5 (B). HEK293 cells were cotransfected with the HA-tagged Prdx5 WT or cysteine mutant (C48S, C73S, C152S, C48/152S) and myc-tagged DBT expression vector, and then the transfected cells were exposed to hypoxic stress (1.0±0.2%) for 6 hours or normoxic condition (20.0±0.2%). WCE, whole cell extract of HEK293 cells. In panel A, WT Prdx5 bound weakly to DBT under normoxic conditions, but this interaction showed a significant increase under hypoxic stress. In the case of the Prdx5 C152S mutant, the interaction with DBT did not change compared with WT Prdx5.
However, Prdx5 C48S mutant and double mutant (C48/152S) showed diminished interactions with DBT under both normoxic and hypoxic conditions. This finding indicates that Prdx5 requires its cysteine residue at position 48 to interact with DBT.
A
B
51 4. Discussion
BCAAs are essential amino acids that serve as substrates for protein synthesis and also function in nutritional signals, regulating carbohydrate metabolism, energy balance and hormone secretion (7-9). Thus, the homeostasis of BCAAs is tightly regulated by the BCKDH complex.
DBT belongs to the transacylase (E2) subunit of BCKDH, the key enzyme in amino acid metabolism. Each E2 subunit consists of three independently functional domains: a lipoyl- bearing domain located in the N-terminal portion, an E1/E3-binding domain, and an inner-core domain at the C-terminal portion, with the three domains tethered by flexible linker regions. The core domains of E2 subunit form a 24-meric scaffold, which is decorated with multiple copies of E1 and E3 attached through the subunit-binding domain. Basic enzymatic activity of DBT is known, along with the protein sequence and structure but the physiological function of DBT in amino acid metabolism under hypoxic stress is still unknown. Although BCAAs are essential amino acids, accumulation of their metabolites can toxic to cells (11-14). Thus, hypoxia-induced BCAA accumulation may promote kidney injury. The regulation of free amino acid level is an important factor in understanding amino acid metabolism in acute hypoxia on tissues. Previous reports have recently clued that BCAAs accumulated in rat plasma or fetal sheep during acute hypoxia (22-23). It can be explained that oxygen deficiency prevents the use of BCAAs as the mitochondrial electron transfer system. Thus, BCAAs accumulate in plasma (22-23).
Additionally, I found that the enzymatic activity of DBT was increased in hypoxia versus normoxia. I suggest that hypoxia plays a role as the factor stimulating BCKDH activity. As the sequence, BCAAs degradation is enhanced and maintained the homeostasis of BCAAs during hypoxic stress. This is an important evidence proves the linkage between hypoxia and BCAAs catabolism. Moreover, I identified that the reinforced interaction between Prdx5 and DBT in mouse kidney was inducible by hypoxic stress. From these results I hypothesis that Prdx5 may be an important molecule that contributes into the regulation of DBT activity through their direct interaction under hypoxic stress.
Recently, Yang and collaborators reported that Prdx5 exerted protective effects in the hypoxic kidney by regulating a variety of individual proteins in a protein network (24). Using shotgun proteomic analysis, it has been shown that knocking down Prdx5 influences the expression of a variety of protein groups associated with oxidative stress, mitochondrial transport, amino acid/nucleic acid metabolism, glycolysis/gluconeogenesis, fatty acid metabolism, and the cytoskeleton (24). Gba2, Pcca, and Pccb can be contributed into energy metabolism which was insufficient under hypoxic stress. Gba2 encodes a glucosylceramidase
52
which can cleaves membrane lipid glucosylceramide into glucose and ceramide, the glucose producing in this process in turn can be converted into pyruvate and contributes into TCA cycle.
On the other hand, Pcca and Pccb are two components of propionyl coA carboxylase complex which converts propionyl coA, metabolite of valine and isoleucine, into D-methylmalonyl-coA, which in turn will be converted L-methylmalonyl-coA and then succinyl coA via methylmalonyl-coA racemase and methylmalonyl-coA mutase, respectively. Succinyl co A is one of the intermediates of TCA cycle. Furthermore, DBT is one of components of BCKDH complex, which plays an important role in BCAAs catabolism. BCAAs are not only the essential amino acid for protein synthesis but also contribute into energy metabolism. As mentioned above, propionyl coA, metabolite of isoleucine and valine, as well as acetyl coA, metabolite of isoleucine and leucine, can contribute into TCA cycle (25). In this study, hypoxia enhances DBT activity, it suggested that BCAAs catabolism also is enhanced and contributed into energy metabolism. Moreover, DBT also is one of putative interactors of Prdx5. Thus, this finding contributes into the function of Prdx5, which appears to be multifunctional but the full spectrum of its cellular functions remains unknown.
Unlike other members of Prdx family, Prdx5 does not contain a cysteine residue corresponding to the second conserved cysteine residue of the 2-Cys subgroup but possesses two additional cysteine residues (C73 and C152) that are lacking in the 1-Cys subgroup; thus, Prdx5 belongs to the atypical 2-Cys family. The atypical 2-Cys Prdx has side chains for all Cys residues (the peroxidatic Cys48, the resolving Cys152, and the additional Cys73) characterized by the presence of a free sulfhydryl group. The peroxidatic residue Cys48 of Prdx5 is a highly conserved Cys residue in all peroxiredoxins (26). This residue is located within the N-terminal part of the α2-helix, twisted at Ala60, in the reduced form, and lies in the positive charged active-site pocket built by surrounding factors such as Arg128, Thr45, and Val40. The resolving residue Cys152, located in the loop between β7 and α6, is associated with the proposed enzymatic mechanism of action of Prdx5. The proposed enzymatic mechanism requires the formation of an intramolecular disulfide bond between these two Cys residues. The additional Cys, Cys73, is found in the C-terminal part of the β4 stand of human Prdx5. However, this residue is not implicated in the enzymatic mechanism, because its mutation does not modify the activity. It is thus surprising to find this Cys73 located close to the peroxidatic Cys48, at the bottom of the active-site pocket (27-28). Although the enzymatic mechanism involving the cysteine residues is well established, the physiological functions under specific stimulation conditions are still unknown. Other investigators have demonstrated that the catalytic cysteine of
53
Prdx5 has an important role in cytoprotection effects via its antioxidant activity. Moreover, a novel function of the peroxidatic Cys48 is reported as a specific regulator of Jak2-Stat5 signaling triggered by lipopolysaccharide (LPS). Notably, it appeared that Prdx5 played a role in the regulation of IL-6 expression through direct binding to Jak2 via its catalytic Cys48 residue, although this did not involve its peroxidase activity (29). Consistent with that, my data show that hypoxia enhanced the direct interaction between Prdx5 and DBT, and this enhancement did require the catalytic cysteine residue, Cys48, of Prdx5 but not the other cysteine residues (Cys73 and Cys152). This finding contributes to the understanding of the role of Cys48 Prdx5, special in interaction with other proteins, such as DBT and Jak2.
54 5. References
1) AEvarsson, A.; Chuang J. L.; Wynn, R. M.; Turley, S.; Chuang, D. T.; Hoi, W. G., Crystal structure of human branched-chain alpha-ketoacid dehydrogenase and the molecular basis of multienzyme complex deficiency in maple syrup urine disease.
Structure, 2000, 8(3): 277-291.
2) Danner, D. J.; Doering, C. B., Human mutations affecting branched chain alpha-ketoacid dehydrogenase. Front Biosci, 1998, 3: d517-d524.
3) Shimomura, Y.; Yamamoto, Y.; Bajotto, G.; Sato, J.; Murakami, T.; Shimomura, N.;
Kobayashi, H.; Mawatari, K., Nutraceutical effects of branched-chain amino acids on skeletal muscle. J Nutr, 2006, 136(2): 529S-532S.
4) Yang, N.; Han, L.; Gu, X.; Ye, J.; Qiu, W.; Zhang, H.; Gong, Z.; Zhang, Y, Analysis of gene mutations in Chinese patients with maple syrup urine disease. Mol Genet Metab, 2012, 106(4): 412-418.
5) Chuang, D. T.; Chuang, J. L.; Wynn, R. M., Lessons from genetic disorders of branched- chain amino acid metabolism. J Nutr, 2006, 136(1 Suppl): 243S-249S.
6) Dursun, A.; Henneke, M.; Ozgul, K.; Gartner, J.; Coskun, T, Tokatli, A.; Kalkanoglu, H.
S.; Dermirkol, M.; Wendel, U.; Ozalp, I., Mapple syrup urine disease: mutation analysis in Turkish patients. J Inherit Metab Dis, 2002, 25(2): 89-97.
7) Nair, K. S.; Short, K. R., Hormonal and signaling role of branched chain amino acids. J Nutr, 2005, 135(6 Suppl): 1547S-1552S.
8) Li, F.; Yin, Y.; Tan, B.; Kong, X; Wu, G., Leucine nutrition in animals and humans:
mTOR signaling and beyond. Amino Acids, 2011, 41(5): 1185-1193.
9) McAllan, L.; Cotter, P. D.; Roche, H. M.; Korpela, R.; Nilaweera, K. N., Impact of leucine on energy balance. J Physiol Biochem, 2013, 69(1): 155-163.
10) Xiao, F.; Yu, J.; Guo, Y.; Deng, J.; Li, K.; Du, Y.; Chen, S.; Zhu, J.; Sheng, H.; Gou, F., Effects of individual branched-chain amino acid deprivation on insulin sensitivity and glucose metabolism in mice. Metabolism, 2014, 63(6): 841-850.
11) Jouvet, P.; Rustin, P.; Taylor, D. L.; Pocock, J. M.; Felderhoff-Mueser, U.; Maarakis, N.
D.; Sarraf, C.; Joashi, U.; Kozma, M.; Greenwood, K.; Edwards A. D.; Mehmet, H., Branched chain amino acids induce apoptosis in neural cells without mitochondrial membrane depolarization or cytochrome c release: implications for neurological impairment associated with maple syrup urine disease. Mol Biol Cell, 2000, 11(5): 1919- 1932.