New insights into human lysine degradation pathways with relevance to pyridoxine‐dependent epilepsy due to antiquitin deficiency

Deficiency of antiquitin (ATQ), an enzyme involved in lysine degradation, is the major cause of vitamin B6‐dependent epilepsy. Accumulation of the potentially neurotoxic α‐aminoadipic semialdehyde (AASA) may contribute to frequently associated developmental delay. AASA is formed by α‐aminoadipic semialdehyde synthase (AASS) via the saccharopine pathway of lysine degradation, or, as has been postulated, by the pipecolic acid (PA) pathway, and then converted to α‐aminoadipic acid by ATQ. The PA pathway has been considered to be the predominant pathway of lysine degradation in mammalian brain; however, this was refuted by recent studies in mouse. Consequently, inhibition of AASS was proposed as a potential new treatment option for ATQ deficiency. It is therefore of utmost importance to determine whether the saccharopine pathway is also predominant in human brain cells. The route of lysine degradation was analyzed by isotopic tracing studies in cultured human astrocytes, ReNcell CX human neuronal progenitor cells and human fibroblasts, and expression of enzymes of the two lysine degradation pathways was determined by Western blot. Lysine degradation was only detected through the saccharopine pathway in all cell types studied. The enrichment of 15N‐glutamate as a side product of AASA formation through AASS furthermore demonstrated activity of the saccharopine pathway. We provide first evidence that the saccharopine pathway is the major route of lysine degradation in cultured human brain cells. These results support inhibition of the saccharopine pathway as a new treatment option for ATQ deficiency.


| INTRODUCTION
Lysine degradation has been considered to have two distinct pathways, through saccharopine formation by ε-deamination, or through pipecolic acid (PA) formation by α-deamination or transamination (Figure 1), although the first enzyme of the PA pathway has not yet been identified. 1 Both proposed pathways lead to formation of Δ 1 -piperideine-6-carboxylate (P6C) and its open form α-aminoadipic semialdehyde (AASA), which is then converted to α-aminoadipic acid (AAA) by the AASA dehydrogenase antiquitin (ATQ). Based on rat and mouse studies, it has long been believed that the PA pathway is predominant in the brain, while in other tissues the saccharopine pathway is more active. [1][2][3][4][5] In ATQ deficiency, the most frequent cause of vitamin B 6 /pyridoxine-dependent epilepsy, AASA and P6C accumulate, leading to inactivation of pyridoxal 5 0 -phosphate (PLP), the active vitamer of vitamin B 6 . This occurs due to a condensation reaction between PLP and accumulated P6C. 6 Despite seizure control and reduction of AASA with highdose vitamin B 6 (pyridoxine) treatment, around 70% of ATQ-deficient patients still suffer from developmental delay. 6 This could be due to AASA, which is highly reactive and possibly neurotoxic and which levels do not normalize. [7][8][9][10][11] In the search for new treatment modalities, the lysine degradation pathway has gained new attention and the activity of the PA pathway has been brought into question by recent isotopic tracing studies. In fibroblasts from ATQ-deficient patients, PA derived only from the saccharopine pathway. 5 PA formation by the saccharopine pathway was demonstrated via conversion of P6C to PA by pyrroline-5-carboxylate reductase (P5CR), 12 in contrast to the previous assumption that PA is exclusively produced by the PA pathway. It was also recently shown that in mouse brain, liver, and kidney, lysine is not degraded to AAA by the PA pathway but only through the saccharopine pathway. 13,14 Development of inhibitors of α-aminoadipic semialdehyde synthase (AASS), the enzyme which converts lysine to saccharopine, was proposed as a possible new treatment strategy in patients with ATQ deficiency. 13,14 As any activity through the PA pathway could result in accumulation of AASA/P6C even if AASS is inhibited, it is important to determine whether the saccharopine pathway is also the major lysine degradation pathway in human brain, as the lysine degradation pathways appear to show species-specific activities. [15][16][17][18][19]33 In this study, we used isotopic tracing studies to examine whether in human brain cells, AAA is formed by the PA pathway as has been deduced from previous studies, or by the saccharopine pathway as was recently demonstrated in mouse brain. It has been shown that ATQ is expressed in human astrocytes but not in neurons 20 ; therefore, we examined the formation of AAA in astrocytes as well as in a human neural progenitor cell (NPC) line derived from the cortical region of human fetal brain (ReNcell CX NPCs), which we demonstrate express AASS and ATQ. In addition, we examined whether activity through the PA pathway might increase when the saccharopine pathway is impaired, by comparing AASS-deficient and ATQ-deficient human fibroblasts to control fibroblasts.

| Cell harvesting and sample preparation for liquid chromatography-mass spectrometry
The volume of cell lysates analyzed was normalized by Hoescht assay as described previously 22 to account for differences in efficiency of cell transfer during harvesting (Supplementary Information). Cells were washed three times with ddH 2 0, then scraped using a rubber scraper into 50% methanol (Sigma-Aldrich). Cell extracts were sonicated for 15 sec then centrifuged at 14,500 g at 4 C for 15 min. The supernatant was evaporated in a Concentrator Plus (Eppendorf) and resuspended in 10 μL of 50% methanol.  (Supplementary Information). For all metabolites, the limit of detection and the limit of quantification were estimated at 0.004 μM (S/N = 3) and 0.013 μM (S/N = 10), respectively. Enrichment of 15 N-metabolites was expressed as molar percent enrichment (MPE), as described previously. 5 The quantification limit of MPE was calculated as <5%.
Qualitative analysis of lysine, PA, AAA, glutamate, and glutamine, as well as their isotopically labeled analogues was performed using an alternative UHPLC-MS/MS approach (Hydrophilic Interaction Liquid Chromatography [HILIC] chromatography utilizing a Thermo UHPLC Ultimate 3000XRS coupled to a QExactive high-resolution mass spectrometer, Supplementary Information).

| Western blot analysis
Western blots were performed following standardized procedures as described in the Supplementary Information. Primary antibodies and dilutions used were ATQ (1:50000; Abcam, Cambridge, UK), PA oxidase (PIPOX) (1:2000; Sigma-Aldrich, Buchs SG, Switzerland), AASS (1:2000; Abcam), and β-Actin (1:10000; Abcam). All procedures followed were in accordance with the ethical standards of the responsible committee on human experimentation (institutional and national) and with the Helsinki Declaration of 1975, as revised in 2000. Informed consent was obtained from the legal guardians of the three patients who provided the fibroblasts for this study.

| Expression of enzymes and activity of the PA and saccharopine pathway in cultured human cells
Western blot analysis showed that the enzymes ATQ, PIPOX, and AASS are expressed in human astrocytes, NPCs, and fibroblasts ( Figure 2A).
In cultivated human astrocytes, NPCs, and fibroblasts, we detected lysine degradation through the saccharopine pathway and not the PA pathway, using isotopic tracing studies with L-[α 15 Figure 2B). 15 N-labeled AAA was detected in astrocytes, NPCs, and fibroblasts only when the media was supplemented with α 15 N-Lys, as shown by 40%-60% MPE of 15 N-labeled AAA ( Figure 2B). The same results were also seen in human neural stem cells derived from human embryonic stem cells (hESCs) (Gibco, ThermoFisher Scientific) with 44% MPE of 15 N-labeled AAA in cells supplemented with α 15 N-Lys ( Figure S2). The 40%-60% MPE of 15 N-labeled AAA indicates that approximately half AAA was unlabelled, which is likely due to isotopic scrambling during the reversible, transaminase-induced conversion of AAA and α-ketoglutaric acid to α-ketoadipic acid and glutamate. In contrast, 15 Nlabeled AAA was below the quantification limit (5% MPE) in cells grown in ε 15 N-Lys, indicating that no PA pathwayderived AAA was detectable.
In astrocytes, low concentrations of AASA/P6C and PA were measured (close to the detection limit), derived exclusively from the saccharopine pathway (80%-100% MPE with α 15 N-Lys, Figure S3). In NPCs, no AASA/P6C or PA was detected in cells incubated with lysine, and in fibroblasts, no AASA/P6C and low levels of PA were measured, only derived from the saccharopine pathway (50% MPE with α 15 N-Lys, Figure S3). In cell culture media from astrocytes, NPCs and fibroblasts, labeled PA was detected, only from cells supplemented with α 15 N-Lys and not ε 15 N-Lys ( Figure S4).
These results were confirmed by qualitative analysis of isotopically labeled PA and AAA in cell lysates using highresolution MS ( Figure S5), and activity of AASS was demonstrated by 15 N-glutamate enrichment studies. Increased 15 N-glutamate enrichment in cells incubated with ε-15 Nlysine ( Figure 3) confirms conversion of saccharopine to AASA and glutamate as this involves transfer of ε-15 N to glutamate (Figure 1). This was detected in all cell types investigated in this study, although absolute levels of enrichment were shown to be rather low. This indicates that glutamate production by the saccharopine pathway contributes relatively little to overall glutamate production. In ATQdeficient fibroblasts, lower 15 N-glutamate enrichment was detected for cells incubated with α 15 N-Lys than in control fibroblasts (Figure 3), indicating that the labeled glutamate results from the transamination of AAA.
Interestingly, 15 N-glutamate enrichment in cells incubated in ε-15 N-lysine containing media was higher in ATQdeficient fibroblasts than control fibroblasts suggesting increased AASS activity in ATQ-deficient compared to control fibroblasts (Figure 3). Glutamine, used as a negative control, did not exhibit enrichment beyond the natural abundance of 0.7% 15 N.

| Activity of the PA pathway in cultured fibroblasts with impaired saccharopine pathway
Furthermore, using isotopic tracing analysis, we examined whether lysine degradation through the PA pathway is increased when the first step of the saccharopine pathway is AASA/P6C and PA were not detectable in AASSdeficient cells, whereas in ATQ-deficient cells, AASA/P6C and PA concentrations were increased compared to control fibroblasts. AAA concentrations were not higher in AASSdeficient fibroblasts than background levels detected in ATQ-deficient fibroblasts (Figure 4). Furthermore, in ATQdeficient fibroblasts, 15 N-AASA/P6C and 15 N-PA were produced only in cells supplemented with α 15 N-Lys therefore derived exclusively from the saccharopine pathway ( Figure 4A and B). Furthermore and as expected, no enrichment of 15 N-glutamate to unlabeled glutamate was observed in AASS-deficient fibroblasts ( Figure S6). These results illustrate that in fibroblasts, degradation of lysine does not occur via the PA pathway even when the saccharopine pathway is impaired.

| DISCUSSION
While it has long been reported that the PA pathway is the predominant route of lysine degradation in the brain, a reappraisal of the importance of the saccharopine pathway in brain is warranted. We examined the route of production of AAA in human astrocytes and in a human NPC line derived from the cortical region of human fetal brain (ReNcell CX NPCs), which express AASS, PIPOX, and ATQ, but not in neurons, as it has been shown that in human brain, neurons do not express ATQ. 20 Conclusions of our study are based on the widely adopted hypothesis that the first step of the PA pathway occurs via lysine α-transamination or deamination.
Data generated within our study revealed that conversion of lysine to AAA is only detectable through the saccharopine pathway in cultured human astrocytes and human NPCs, which is in line with the recent report that formation of AAA in lysine degradation occurs by the saccharopine pathway in mouse brain. 13,14 Our isotopic tracing studies showed conservation of the α-15 N label in all lysine F I G U R E 4 Legend on next coloumn. Glutamate enrichment studies also demonstrated transfer of the ε-15 N to glutamate, indicating conversion of saccharopine to AASA and glutamate as previously shown in mouse cerebellar and cerebral slices. 24 As such we confirmed ε-deamination or transamination of lysine (the saccharopine pathway) as an active route of lysine degradation in human astrocytes and NPCs.
In view of our findings, previous research results deserve special consideration. In 1954, Rothstein and Miller reported that in rat, PA and AAA were derived only from the PA pathway and not through the saccharopine pathway of lysine degradation. 25 This was deduced from analysis of urine after IP injection of α 15 N-or ε 15 N-DL-lysine and was based on the assumption that L-PA derived from L-lysine; however, it was later shown that L-PA is metabolized not only from Llysine but also D-lysine 3,16 by D-amino acid oxidase. 19 Therefore, the formation of AAA by the PA pathway in this study cannot be attributed only to L-lysine degradation and may have resulted from degradation of D-lysine.
It was subsequently reported that the saccharopine pathway of degradation is active in tissues such as liver and kidney, [2][3][4][26][27][28] but that the PA pathway is the predominant pathway in brain. The importance attributed to the PA pathway in brain was deduced from the inability to detect saccharopine in lysine degradation in rat brain in early studies, 3 low expression of AASS in mammalian brain compared to liver, heart, and kidney, 3,4,18,27,29 and the decrease in AASS enzyme activity in the first week of postnatal life. 28 Saccharopine has since been detected in several studies in mouse brain, 13,14,18 and it has been proposed that the difficulty in detecting saccharopine in early studies may have been due to time of sampling, 13,14 as well as the lower efficiency of lysine conversion to saccharopine by saccharopine dehydrogenase than the conversion of saccharopine to AASA by lysine-oxoglutarate reductase, resulting in low concentrations of saccharopine. 30 In mammals, the bifunctional enzyme AASS possesses activities of both these enzymes. 31,32 Furthermore, the detection of PA as a lysine degradation product in brain was considered evidence of activity through the PA pathway of lysine degradation. 3 However, a recent study in ATQ-deficient fibroblasts showed that PA can be produced by the saccharopine pathway, 5 and this was also shown to occur in mouse brain, although more PA was produced by the PA pathway as shown by higher 15 N-PA enrichment in brain from ε-15 N-lysine and α-15 N-lysine treated mice. 13,14,18 Our data in human astrocytes support and extend previous findings in human fibroblasts showing that PA is produced via the saccharopine pathway, by P6C conversion to PA. 5 This indicates species differences in the conversion of L-lysine to PA by the PA pathway, as this was detected in mouse but not in human cells.
We also showed that in fibroblasts with impaired saccharopine pathway activity due to AASS deficiency, there was still no detectable production of PA or AAA by the PA pathway.
Preliminary findings suggest that AASS activity in ATQdeficient fibroblasts could be increased compared to control fibroblasts, as shown by glutamate enrichment studies (Figure 3). This may indicate feedback inhibition by AAA or a downstream metabolite and increased rate of lysine degradation through the saccharopine pathway in ATQ deficiency. Future studies investigating ATQ deficiency in human brain cells would be important to extend these findings to brain, as any increase in AASS activity due to ATQ deficiency would further support the potential benefits of inhibiting AASS (and therefore preventing AASA accumulation) in ATQ deficiency.

| CONCLUSIONS
Our study provides the first evidence that the saccharopine pathway is the main route for lysine degradation to AAA in human astrocytes and a human NPC line derived from the cortical region of human fetal brain (ReNcell CX NPCs), and extends previous data in human fibroblasts. Results of our experiments in addition to previous studies indicate species differences in the activity of steps of the lysine degradation pathway. In mouse, PA is produced by both αand ε-deamination, while in all human cells investigated to date, PA is only produced by ε-deamination. Recognition of the relevance of the saccharopine pathway of lysine degradation in human brain tissue may have been delayed due to differing activity of each step of the PA and saccharopine pathways depending on species, tissue as well as age-dependent effects. 28 While future studies using astrocyte-neuronal cocultures may provide further information on possible shuttling of lysine degradation products between neurons and astrocytes, our findings indicate the importance of the saccharopine pathway in human brain cells, supporting the idea that inhibition of the saccharopine pathway could offer new treatment options for ATQ deficiency.