The retina and retinal pigment epithelium differ in nitrogen metabolism and are metabolically connected

Defects in energy metabolism in either the retina or the immediately adjacent retinal pigment epithelium (RPE) underlie retinal degeneration, but the metabolic dependence between retina and RPE remains unclear. Nitrogen-containing metabolites such as amino acids are essential for energy metabolism. Here, we found that 15N-labeled ammonium is predominantly assimilated into glutamine in both the retina and RPE/choroid ex vivo. [15N]Ammonium tracing in vivo show that, like the brain, the retina can synthesize asparagine from ammonium, but RPE/choroid and the liver cannot. However, unless present at toxic concentrations, ammonium cannot be recycled into glutamate in the retina and RPE/choroid. Tracing with 15N-labeled amino acids show that the retina predominantly uses aspartate transaminase for de novo synthesis of glutamate, glutamine, and aspartate, whereas RPE uses multiple transaminases to utilize and synthesize amino acids. Retina consumes more leucine than RPE, but little leucine is catabolized. The synthesis of serine and glycine is active in RPE but limited in the retina. RPE, but not the retina, uses alanine as mitochondrial substrates through mitochondrial pyruvate carrier. However, when the mitochondrial pyruvate carrier is inhibited, alanine may directly enter the retinal mitochondria but not those of RPE. In conclusion, our results demonstrate that the retina and RPE differ in nitrogen metabolism and highlight that the RPE supports retinal metabolism through active amino acid metabolism.

Vertebrate retinas have an extraordinarily high metabolic rate to maintain ion gradients for phototransduction and to replenish the outer segments, which are shed daily (1)(2)(3). Retinal pigment epithelium (RPE) 4 is a monolayer of pigmented cells interposed between the choroid and the retina. RPE has an active metabolism to transport metabolites and phagocytose outer segment to support neural retina (4 -6). Both retina and RPE are highly vulnerable to metabolic defects. Mutations of metabolic genes or dysfunctional metabolic processes in either retina or RPE could cause inherited retinal degeneration, agerelated macular degeneration, and other retinal diseases (7)(8)(9)(10)(11). The understanding of retina and RPE metabolism is fundamentally important for developing treatment for retinal degenerative disease. However, the metabolic difference between retina and RPE is not well-understood. For example, intervention of energy metabolism by activating mammalian target of rapamycin (mTOR) leads to distinctive phenotypes in retina and RPE: mTOR shows protection from retinopathy when activated in the retina, but it induces retinal degeneration when activated in RPE (12)(13)(14)(15). As metabolically active tissues, retina and RPE robustly consume nutrients to supply carbon, nitrogen, and oxygen, as well as to rapidly produce by-products such as lactate, ammonia, and CO 2 . However, prior studies of retinal metabolism mostly focus on carbon metabolism. How nitrogen is metabolized and recycled in retina and RPE remains elusive.
Cell metabolism constantly produces and consumes nitrogen-containing metabolites including ammonia, amino acids, and nucleotides (16). Human adults produce ϳ1000 mmol of ammonia daily (17), rising from deamination of amino acids, breakdown of purine and pyrimidine, and degradation of urea by the microbiome (16,18). Most ammonia is removed or assimilated by the urea cycle in the liver to synthesize urea and by glutamine synthetase (GS) to form glutamine from glutamate. Both urea and glutamine could be released into plasma and excreted in urine (19,20). Additionally, ammonia could also be utilized for asparagine synthesis by asparagine synthetase and for synthesis of glutamate from ␣-ketoglutarate (␣KG) by glutamate dehydrogenase (GDH), which seems to be important in tumor proliferation (21,22). Under physiological conditions, plasma ammonia concentration is Ͻ50 M in human adults and Ͻ150 M in neonates. With hyperammonemia (pathologically high ammonia), plasma ammonia level can rise to 0.2-1 mM in patients with liver damage and up to 5 mM in patients with inborn errors of the urea cycle (19). Pathological levels of ammonia are toxic to both the brain and retina, which induce structural and functional alterations referred to as hepatic encephalopathy or hepatic retinopathy because of liver insufficiency (20,23). Ammonia is particularly toxic to astrocyte or glial cells, most likely, because they are the only cells expressing GS, which detoxifies ammonia into glutamine in brain or retina (19,24). In honeybee retina, neuronal metabolism stimulates the release of ammonium, which could be transported into glial cells to synthesize glutamine (25,26).
Transaminases (aminotransferases) are essential in amino acid synthesis and mitochondrial intermediate metabolism because they catalyze the reversible transfer of ␣-amino groups from ␣-amino acids to ␣-ketoacids without producing or consuming free ammonium ions. Coupled with the conversion between glutamate/␣KG, aspartate transaminase, alanine transaminase, and branched-chain amino acids (BCAAs) (leucine, isoleucine, and valine) reversibly transport amino group between alanine/pyruvate, aspartate/oxaloacetate (OAA), and BCAAs/branched-chain ketoacids (16). Inhibition of transaminases substantially reduces de novo glutamate synthesis in rat retinas in vitro (27). Pyruvate, ␣KG, and OAA are also key substrates for mitochondrial energy metabolism. We report that inhibition of mitochondrial pyruvate transport massively accumulates aspartate but depletes glutamate and ␣KG, resulting in retinal degeneration (28,29). In liver, alanine could enter mitochondria to bypass the inhibition of mitochondrial pyruvate transport (30 -32). It remains unclear how mitochondrial metabolism influences nitrogen metabolism in the retina.
In this report, we investigate nitrogen metabolism in retina and RPE by using 15 N tracers. We have found that nitrogen metabolism is tissue-specific. [ 15 N]Ammonium is mostly assimilated into glutamine in both retina and RPE. However, retina depends on aspartate transaminase for synthesizing glutamate, aspartate, and other amino acids, whereas RPE employs different transaminases for the synthesis and breakdown of amino acids. Glutamate is the major donor for mitochondrial aspartate synthesis in both retina and RPE. Surprisingly, alanine could not be utilized as mitochondrial substrates in the retina except when mitochondrial pyruvate carrier is blocked.

Impact of ammonium on retinal metabolism
To study how ammonium influences retinal metabolism, we incubated the freshly isolated mouse retina or RPE/choroid with different concentrations of unlabeled NH 4 Cl in the pres-ence of glucose in Krebs-Ringer/HEPES/bicarbonate buffer. We chose 2 h for ex vivo incubation, because our previous studies showed that metabolites almost reach steady state from tracers at 2 h, whereas cultured retina remain healthy under this simple medium (5,(33)(34)(35). After incubation, the steady-state metabolites in nitrogen metabolism, including amino acids, organic acids, and urea (Table S1), were analyzed by GC-MS. In both retina and RPE/choroid, 0.2 mM of NH 4 Cl did not change the level of any metabolites, whereas 1 mM of NH 4 Cl significantly increased glutamine by ϳ2-fold over the control without NH 4 Cl (Fig. 1, A and B). At 5 mM of NH 4 Cl, glutamine was further increased by ϳ4-fold. Both pyruvate and ␣KG were decreased in the retina and RPE/choroid in response to 5 mM of NH 4 Cl (Fig. 1, A and B). Interestingly, aspartate, asparagine, glutamate, and succinate were changed in the retina, whereas only serine was changed in the RPE/choroid at 5 mM of NH 4 Cl. To further study how the nitrogen was utilized, we incubated retina and RPE/choroid with 1 or 5 mM of 15 NH 4 Cl and measured the percentage of 15 N incorporation into nitrogen-containing metabolites by GC-MS (Fig. 1C). The nitrogen-containing metabolites were measured in selective ion monitor mode with their isotopologues listed in Table S2. In both retina and A and B, mouse retina or RPE/choroid (RPE/Cho) was incubated with different concentrations of unlabeled NH 4 Cl for 2 h. The metabolites were analyzed by GC-MS. *, p Ͻ 0.05 versus 0 mM (n ϭ 4, t test). Relative abundance is the ion intensity relative to 0 mM. C, a schematic for incubation of retina and RPE/ choroid with 15 NH 4 Cl. D-G, isolated retina or RPE/choroid was incubated with different concentration of 15 NH 4 Cl for 2 h. The isotopologues were analyzed by GC-MS for percentage of labeling ( 15 N fraction) (D and E) and ion abundance (F and G). Relative abundance is the total ion abundance of the isotopologues relative to 0 mM. *, p Ͻ 0.05 versus 0 mM (t test). Suc, succinate; Pyr, pyruvate.

Nitrogen metabolism in retina and RPE
RPE/choroid, glutamine had the highest 15 N fraction with Ͼ85% of 15 N-labeled by 15 NH 4 Cl (Fig. 1, D and E). In the retina, the 15 N fraction was as high as 98% at 5 mM 15 NH 4 Cl. However, the labeling of 15 N into other metabolites were much lower: Ͻ4% for aspartate at 1 mM of 15 NH 4 Cl and Ͻ12% at 5 mM ( Fig.  1, D and E). Similar to the results of steady-state metabolites, asparagine was only labeled in the retina (Fig. 1D). To test how 15 NH 4 Cl influenced the abundance of the labeled metabolites, we measured the total ion abundance (the sum of ion abundance of isotopologues including both labeled and unlabeled). Similar to unlabeled NH 4 Cl, glutamine was increased in both retina and RPE/choroid, but aspartate and asparagine were only increased at 5 mM of 15 NH 4 Cl (Fig. 1, F and G). These results indicate that ammonium is predominantly utilized in glutamine synthesis, and the recycling of ammonium into glutamate occurs only at toxic level of ammonium.

Tissue-specific nitrogen metabolism in retina, RPE/choroid, liver, and brain in vivo
To study ammonia metabolism in vivo, we intraperitoneally injected 15 NH 4 Cl at 167 mg/kg and harvested tissues and plasma after 5, 15, 30, and 60 min (0 min was injected with PBS) ( Fig. 2A). Similar to ex vivo results, glutamine was labeled the highest among all metabolites in both retina and RPE/choroid (Fig. 2, B and D). The abundance of glutamine was also increased by 1.5-2-fold (Fig. 2, C and E), confirming our finding that the ammonia is predominantly assimilated into glutamine. However, different from ex vivo studies, Ͼ10% urea was labeled with increased total abundance both retina and RPE/choroid in vivo (Fig. 2, B-D).
In the plasma, glutamine, urea, alanine, and oxoproline were the only metabolites that were labeled by 15 NH 4 Cl (Fig. 2F). All  15 NH 4 Cl (167 mg/kg) or PBS, the tissues were extracted to analyze the percentage of labeling ( 15 N fraction) and total ion abundance from 15 NH 4 Cl in the retina (B and C), RPE/choroid (D and E), plasma (F and G), liver (H and I), and brain (J and K). Relative abundance is the total ion abundance relative to the group with PBS injection for 5 min. *, p Ͻ 0.05 versus PBS injection (n ϭ 4 -6, t test). Cho, choroid; Oxop, 5-oxoproline; GABA, gamma-Aminobutyric acid; Asn, asparagine.

Nitrogen metabolism in retina and RPE
these metabolites rose by 1.5-2-fold in abundance at 5 min and then came back to normal except urea, which was increased in all the time points (Fig. 2G). These results suggest that these metabolites are released into blood from tissues that assimilate ammonium.
In the liver, almost 70% of glutamine was labeled at 5 min, but the labeling quickly dropped to ϳ10% within 60 min (Fig. 2H). The abundance of glutamine also quickly rose in 5 min and then came back to normal (Fig. 2I). Urea, glutamate, alanine, aspartate, oxoproline, and GABA were highly labeled at 5 min, but the labeling quickly decreased at 30 min except urea (Fig. 2H). These results suggest that liver quickly responses to the challenge of rising ammonia by assimilating ammonia in multiple pathways. Some assimilated metabolites such as glutamine, urea, alanine, and oxoproline are released into the bloodstream. The brain had similar labeling pattern to the retina except higher labeling of asparagine and other amino acids (Fig. 2J). Different from retina and liver, the abundance of most metabolites peaked at 60 min (Fig. 2K), indicating that the clearance of ammonia in the brain is slow.
Taken together, these results consistent with previous knowledge that liver is the major site for urea cycle. In addition to excreted in urine, the urea in the liver may enter other tissues such as retina, RPE, and brain, because urea was not labeled even at 5 mM of 15 NH 4 Cl in vitro (Fig. 1, D and E).

The fate of [ 15 N]glutamine in retina and RPE/choroid
The transamination of glutamate into ␣KG is key to transferring nitrogen group into ketoacids to synthesize amino acids (Fig. 3A). To study the fate of the nitrogen group in this transamination reaction, we incubated retina and RPE/choroid with [ 15 N](amine)-glutamine that could label glutamate by GS (Fig. 3A). More than 95% glutamine in the media was labeled, but the 15 N fraction in the retina was less than RPE/choroid ( Fig. S1A). Analysis of [ 15 N]glutamine consumption showed that retina had a faster rate of glutamine consumption than RPE/choroid (Fig. 3B). As expected, Ͼ80% of glutamine and its nonenzymatic product 5-oxoproline was labeled with 15 N in the retina and RPE/choroid. However, the pattern of newly incorporated amino acids were significantly different between retina and RPE/choroid (Fig. 3C). Except for GABA, RPE/choroid incorporated more [ 15 N](amine)-glutamine into glutamate, aspartate, alanine, BCAAs, serine, and glycine with corresponding increase of metabolite abundance (Fig. 3, D and E). Apart from oxoproline, glutamate, and aspartate, [ 15 N](amine)-glutamine did not increase the abundance of other metabolites (Fig. 3D).
To study the fate of amide in glutamine, we incubated tissues with 1 mM [ 15 N](amine)-glutamine, which could release free 15 NH 4 ϩ by glutaminase (Fig. 3F). Consistently, retina has a lower 15 N fraction in media but a faster rate of glutamine consumption than RPE/choroid ( Fig. S1B and Fig. 3G). Except for high labeling of glutamine in both tissues and low labeling of asparagine in the retina, glutamate and other metabolites were not labeled by [ 15 N](amine)-glutamine. The increased abundance of glutamate should come from the degradation of glutamine by glutaminase (Fig. 3I). These results indicate that the amine in glutamine is mostly used for amino acid synthesis in both retina and RPE/choroid through transamination, but the amide in glutamine is recycled back to glutamine or asparagine but not glutamate.

The fate of[ 15 N]aspartate in retina and RPE/choroid
To study how the nitrogen-group of aspartate was utilized, we incubated [ 15 N]aspartate in both retina and RPE/choroid. Aspartate should be able to use the reversed reaction of transaminase to produce glutamate and its derived amino acids (Fig.  4A). The medium enrichment of 15 N and aspartate consumption rate were similar between retina and RPE/choroid (Fig. S1C and Fig. 4B). As expected, [ 15 N]aspartate was the mostly labeled metabolites in tissues. Glutamate, glutamine, and 5-oxoproline were also highly enriched with 15 N in the similar rate. However, RPE/choroid had more labeling of alanine, serine, glycine, and BCAAs, whereas retina had higher labeling of ␤-alanine, asparagine, proline, and GABA (Fig. 4C). The abundance of metabolites with the addition [ 15 N]asparate further showed

Nitrogen metabolism in retina and RPE
that aspartate was important for the synthesis of glutamate, glutamine, and asparagine in the retina, whereas aspartate was also an important source for alanine, serine, and glycine in the RPE (Fig. 4, D and E).

The fate of [ 15 N]leucine in retina and RPE/choroid
The catabolism of BCAAs requires branched-chain amino acid transaminases, which transfer amino group to ␣KG to form glutamate (Fig. 5A). We incubated [ 15 N]leucine to trace the fate of nitrogen metabolism in retina and RPE/choroid. The medium 15 N fraction was comparable between retina and RPE/ choroid; however, retina had a significantly higher consumption rate of leucine than RPE/choroid ( Fig. S1D and Fig. 5B). Other than leucine, isoleucine was also highly enriched with 15 N in both retina and RPE/choroid. However, the other BCAA, valine, was labeled at much lower percentage than isoleucine (Fig. 5C). Unexpectedly, [ 15 N]leucine was highly incorporated into many other amino acids such as glutamate, aspartate, glutamine, 5-oxoproline, and alanine in the RPE/choroid, but the labeling of other amino acids in the retina was minimal or none (Fig. 5C). In the retina, only glutamate was slightly increased in abundance by [ 15 N]leucine. However, leucine increased the abundance of multiple other amino acids in the RPE/choroid (Fig. 5, D and E), suggesting that most consumed leucine was not used for amino acid synthesis in the retina. In freshly isolated tissue, retina had ϳ2.5 higher steady-state leucine than RPE/choroid, but glutamate/leucine ratio was substantially higher in the retina than RPE/choroid (32.16 versus 8.37) (Fig.  S2A). The incubation with 1 mM [ 15 N]leucine in this experiment decreased glutamate/leucine ratio in the retina to 1.6 ( Fig.  S2B), indicating that the retina may utilize even less leucine for nonessential amino acid synthesis in vivo. To test how much glutamate comes from leucine compared with other amino acids, we analyzed retina and RPE/choroid incubated with or without 1 mM of leucine, aspartate, alanine, or glutamine. The results showed that in addition to glutamine, aspartate but not leucine or alanine was the primary substrate for glutamate synthesis (Fig. S3).

The fate of [ 15 N]alanine in retina and RPE/choroid
Alanine transaminase reversely catalyzes the conversion of pyruvate and glutamate into alanine and ␣KG. To study the fate of nitrogen group in alanine, we incubated [ 15 N]alanine in retina and RPE/choroid (Fig. 6A). Neither [ 15 N]alanine enrichment in the media nor [ 15 N]alanine consumption was different between retina and RPE/choroid ( Fig. S1E and Fig 6B). Almost all alanine in both retina and RPE/choroid was replaced with [ 15 N]alanine after 2 h of incubation. Surprisingly, [ 15 N]alanine was readily utilized to synthesize other amino acids in the RPE/ choroid but not retina (Fig. 4, B and D). To study why retina could not utilize the nitrogen group in the alanine, we measured the steady-state levels of substrates for alanine transaminase in freshly isolated tissues. Glutamate was ϳ10-fold higher in the retina than RPE/choroid (Fig. S4A). Similarly, the ratio of glutamate to ␣KG and the ratio of glutamate to alanine were much higher in the retina than RPE/choroid (Fig. S4, B and C). However, in the retina incubated with exogenous 1 mM [ 15 N]alanine, the glutamate/alanine ratio decreased to ϳ2.9-fold (Fig. S4D), but [ 15 N]alanine still could not label other amino acids in the retina, indicating that retina could not utilize the nitrogen group in alanine, and the transaminase reaction is confined in the direction of alanine formation.

The nitrogen sources for the increased aspartate by the inhibition of mitochondrial pyruvate metabolism
We reported before that the inhibition of mitochondrial pyruvate carrier (MPC) leads to accumulation of aspartate, pyruvate, and NAD ϩ /NADH but depletion of citrate, ␣KG, and glutamate in the retina (28, 29) (Fig. 7A). We have shown that the carbons of the accumulated aspartate comes from gluta-

Nitrogen metabolism in retina and RPE
mate (28). However, the nitrogen source for aspartate remains undetermined. To test whether the nitrogen was mostly from glutamate, we inhibited retina or RPE/choroid with MPC inhibitor UK5099 in the presence of [ 15 N](amine)-glutamine (Fig.  7A). Similar to our previous report, pyruvate was accumulated and ␣KG was reduced with UK5099, confirming that MPC is inhibited in both retina and RPE/choroid (Fig. 7B). Unlabeled (M0) aspartate was increased by ϳ4.5-fold; however, the 15 Nlabeled (M1) aspartate was dramatically increased by 10-fold in retina and 20-fold in RPE (Fig. 7, C and D), indicating that glutamate is the major nitrogen donor for aspartate synthesis.
Recently, alanine is reported to enter the mitochondria to generate pyruvate by alanine transaminase when MPC is blocked (30,31) (Fig. 7E). To test whether this pathway is active in retina and RPE, we blocked MPC with UK5099 in the presence of [ 15 N]alanine. As expected, inhibition of MPC increased pyruvate but decreased ␣KG in both retina and RPE (Fig. 7F). Interestingly, the abundance of M1 aspartate, which is labeled

Nitrogen metabolism in retina and RPE
from [ 15 N]alanine, was slightly increased in the retina but decreased in the RPE/choroid (Fig. 7, G and H). Similarly, M0 aspartate, which comes from unlabeled glucose and other amino acids, was increased by UK5099 in both retina and RPE/ choroid (Fig. 7, G and H). These results indicate that alanine transamination occurs in the cytosol and enters the mitochondria through MPC in the RPE/choroid. However, a small fraction of alanine may enter retinal mitochondria as a substrate when MPC is inhibited. It is also possible that [ 15 N]alanine can be transaminated into glutamate in the cytosol and incorporated into aspartate through malateaspartate shuttle.

The fate of [ 13 C]alanine in retina and RPE/choroid with or without inhibition of MPC
To confirm whether alanine enters mitochondria to compensate MPC inhibition, we used [3-13 C]alanine to trace its derived metabolites. If [3-13 C]alanine is converted into pyruvate in the cytosol, the inhibition of MPC should decrease the labeling of mitochondrial intermediates such as citrate. However, if [3-13 C]alanine enters mitochondria to bypass MPC, the inhibition of MPC should not affect the labeling of mitochondrial intermediates (Fig. 8A). Both 13 C alanine enrichment and abundance in the incubation media were similar in retina and RPE/choroid (Fig. S5). Consistent with the results of [ 15 N]alanine, RPE/choroid, but not retina, readily utilized [3-13 C]alanine and incorporated them into pyruvate and mitochondrial intermediates (Fig. 7B). Except a small fraction of pyruvate, the labeling of most metabolites in the retina was close to 0. UK5099 accumulated metabolites upstream of citrate but depleted citrate and its derived metabolites, suggesting that MPC is effectively inhibited (Fig. 7, C and D). Consistently, the fraction of labeled aspartate increased in the retina but decreased in the RPE upon inhibition of MPC (Fig. 7, B-D). M1 citrate should be labeled by the first turn of TCA cycle (Fig. 7A). M1 citrate was not labeled in the control retina but significantly increased in both fraction and abundance with inhibition of MPC, supporting our hypothesis that alanine could enter retinal mitochondria (Fig. 7, E and G). On the contrary, M1 citrate was significantly decreased in RPE/choroid after MPC inhibition (Fig. 7, F and G), indicating that alanine is mostly transaminated into pyruvate in the cytosol in RPE/choroid.

Tissue-specific ammonia metabolism
Ammonia is constantly produced by cell metabolism and removed by multiple pathways. This includes the urea cycle to produce urea, GS to convert glutamate into glutamine, asparagine synthetase to convert aspartate into asparagine, and GDH to convert ␣KG into glutamate (16,21). Our results support that ammonia metabolism is tissue-specific as summarized in Fig. 9A: 1) liver uses multiple pathways including urea cycle, GS, and GDH to quickly condense ammonia into urea, glutamine, and other metabolites and export these metabolites into blood; 2) GS is a major pathway in all the tested tissues, especially in retina and RPE; 3) asparagine is synthesized in brain and retina but not liver and RPE; and 4) ammonia could be recycled into glutamate and its derived amino acids through GDH in the liver, but it occurs in the retina only at pathological concentration of ammonia. Consistent with our finding, a recent study in WT and liver-specific GS knockout mice show that urea cycle and hepatic GS remove almost the same amount of ammonia (ϳ35% of whole body ammonia) (36). Why is there little ammonia recycling into glutamate under physiological conditions? The k m of GDH for ammonia is more than 10 mM, and its reaction proceeds predominantly toward the direction of deamination of glutamate into ␣KG in vivo (16,37). This may explain why glutamate and its derived nitrogen-containing metabolites increased transiently at 5 min in the liver arising from drastically increased ammonia from injection. Interestingly, a recent

Nitrogen metabolism in retina and RPE
study show that breast cancer cells could recycle ammonia through GDH to support tumor proliferation (21). Retina has some similar metabolic features to cancer cells such as high metabolic demand and aerobic glycolysis (1,33,35). However, retina does not recycle ammonia through GDH under physiological conditions. The nitrogen metabolism of retina is more similar to brain with excessive reliance on glutamine synthesis to remove ammonia. This could make retina and brain more vulnerable to ammonia toxicity, because glutamate could be drained away from neurotransmission and mitochondrial metabolism, resulting in an energy deficit in hepatic encephalopathy and hepatic retinopathy. Consistently, we found the brain is slow to remove ammonia with nitrogen-containing metabolites peaked at later time point. Moreover, glutamate, ␣KG, and succinate in the retina but not RPE/choroid are severely depleted in response to 5 mM ammonium.

Retina relies on aspartate transaminase in de novo synthesis of glutamate and glutamate-derived amino acids
Glutamate, a key neurotransmitter, also acts as an important substrate for mitochondrial energy metabolism and for synthesis of amino acids and GSH. In the retina, Müller glial cells and photoreceptors synergize to maintain a proper supply of glutamate: GS is only expressed in Müller glial cells, and it condenses glutamate and ammonia into glutamine; photoreceptors metabolize glutamine into glutamate (1). De novo glutamate  4, t test). G, the ion abundance (area under the curve) of M1 citrate in the retina and RPE/choroid with or without UK5099. *, p Ͻ 0.05 versus M1 citrate without UK5099 (n ϭ 4, t test). Cho, choroid; Lac, lactate; Pyr, pyruvate.

Nitrogen metabolism in retina and RPE
synthesis comes from ␣KG by recycling ammonia through GDH or by transferring the amine group through transaminases (21). In the retina, GDH contributes little or nothing to glutamate synthesis (Fig. 1). Our results of labeling 15 N-amino acids (aspartate, alanine, and leucine) show that aspartate is the predominant amino donor for glutamate synthesis in the retina. Leucine contributes a small fraction of glutamate synthesis, but little nitrogen in glutamate is labeled by alanine. Consistently, labeling with H 14 CO 3 Ϫ in isolated rat retina demonstrated that de novo glutamate synthesis depends on transaminases but independent of GDH (27,38). When GS is inhibited, intravitreal injection of an aspartate transaminase inhibitor further reduces glutamate and b-wave in rat retina (39). Alanine transaminase has been reported to involve glutamate synthesis based on the result of a pharmacological inhibitor, L-cycloserine. However, this inhibitor potently inhibits not only alanine transaminase but also aspartate transaminase, branched-chain aminotransferase, GABA transaminase, and pyruvate decarboxylation (27,40,41).
Aspartate transaminase is an important component of the malate-aspartate shuttle (MAS) that transports reducing equivalent from cytosol into mitochondria. Müller glial cells are dependent on exogenous aspartate for MAS because it lacks aspartate-glutamate carrier 1 (AGC1) (33,42). We reported before that aspartate carbons are required for synthesis of glutamate and glutamine in Müller glial cells (33). Our data further support the importance of aspartate in the synthesis of amine group for glutamate and glutamine. Our results and other studies strongly support the conclusion that aspartate transaminase (27,38), but not ammonia, is the major nitrogen provider for glutamate and its derived amino acids including GABA, aspartate, alanine, serine, glycine, and BCAAs.

Nitrogen donors for aspartate in retina versus RPE when MPC is inhibited
Aspartate is mostly synthesized in the mitochondria and exported through AGC1. The deletion of AGC1 causes depletion of aspartate and glutamine in both retina and brain (33,43). Aspartate transaminase is important for synthesis of aspartate. We previously reported that inhibition of aspartate transaminase substantially decreases aspartate in the retina (33). When the conversion of oxaloacetate is blocked because of the inhibition or deletion of MPC, aspartate is drastically accumulated in the retina (28,29). The accumulated aspartate is mostly 15 Nlabeled from glutamine in both retina and RPE, indicating that the amine group in glutamine is the major donor for the accumulated aspartate. These results are consistent with findings that glutamine oxidation is enhanced when MPC is dysfunctional (28,29,44,45). A recent study shows that this compensation could direct glutamine away from GSH synthesis to impair hepatocellular tumorigenesis (46). Apart from glutamine, recent studies in liver show that alanine could compensate dysfunctional MPC through entering mitochondria to produce pyruvate by mitochondrial alanine transaminase (31,47). By using both [ 15 N]alanine and [ 13 C]alanine (Figs. 7 and 8), our results strongly support that a small fraction of alanine could enter mitochondria to donate nitrogen to mitochondrial aspartate in the retina but not RPE. RPE could use cytosolic alanine transaminase and MAS to transfer nitrogen to mitochondrial aspartate. However, when MPC is inhibited, both the accumulated cytosolic pyruvate and the slowdown of MAS caused by increased cytosolic NAD ϩ /NADH repress alanine transaminase to reduce aspartate synthesis from alanine (Figs. 7 and 8). However, it remains unknown why alanine could enter the mitochondria of the retina but not RPE when MPC is blocked.

Nitrogen metabolism in the metabolic communication between retina and RPE
Increasing evidence shows that retina and RPE are metabolically interdependent (6,35,48). Retina generates a massive amount of lactate, which can be exported to RPE to stimulate glucose transport (35). RPE could process phagocytosed outer segment from photoreceptors into ketone bodies or metabolize proline into mitochondrial intermediates to fuel outer retina (12,48). We propose a hypothesized model of the metabolic communication between retina and RPE through nitrogen metabolism based on our data and recent findings in Fig. 9B. Why does alanine transaminase prefer to react in the direction Figure 9. Hypothesized models for nitrogen metabolism. A, ammonia metabolism in blood, liver, brain, retina, and RPE. In the liver, 15 NH 4 Cl is mostly assimilated in urea through urea cycle, glutamine through GS, and glutamate through GDH to synthesize nonessential amino acids. Urea and nonessential amino acids in the liver may be exported into blood to enter other tissues or excreted into urine. The assimilation of ammonium into glutamine is the dominant pathway in the brain, retina, and RPE. However, brain and retina could also utilize ammonium to synthesize asparagine. B, the role of nitrogen metabolism through transamination in the metabolic communication between RPE and retina. In the outer retina, glucose is mostly metabolized into lactate, which could export to RPE to fuel mitochondrial metabolism and preserve glucose (35); pyruvate can convert into alanine coupled with glutamate oxidation, but alanine could not be utilized in the retinal mitochondria except the inhibition of MPC. In the retina, aspartate transaminase is the major pathway in the catabolism and anabolism of aspartate, glutamate, and other glutamate-derived amino acids. By readily utilizing multiple transaminases, RPE could synthesize glutamate and its derived amino acids and may export them to be used in the outer retina (50). Red arrows represent the amino acid communication between retina and RPE. Dashed lines represent hypothesized pathways. Oxop, oxoproline; Pyr, pyruvate; Lac, lactate.

Nitrogen metabolism in retina and RPE
of alanine formation only in the retina? We speculate that this may be caused by the rapid formation of pyruvate from glucose in the retina through the strong aerobic glycolysis (34). Furthermore, retina favors glutamate to fuel its mitochondria (34), and the generation of alanine could efficiently remove the nitrogen in glutamate catabolism (Fig. 8B). However, RPE metabolism relies on mitochondrial utilization of different substrates including proline, lactate, and lipids rather than on glycolysis (13,35), which may make alanine more readily utilized into other amino acids. Additionally, the newly generated amino acids such as glutamate and glutamine may be exported to be used by retina as we reported (4, 50) (Fig. 9B). In this metabolic communication, like lactate, alanine might be recycled between retina and RPE to efficiently support retinal metabolism in a similar way to the Cori cycle and alanine cycle between muscle and liver (51). This interdependent communication may explain why the activation of mTOR could yield opposite outcome in protection in retina versus RPE (12)(13)(14)52).
Our results show that RPE has high flexibility to utilize and synthesize amino acids with different transaminases. Because RPE has easy access to blood nutrients, this flexibility could give RPE advantage to efficiently utilize multiple nutrients from blood and recycle metabolic by-products from photoreceptors to support the high metabolic demand in outer retina. Furthermore, retina primarily depends on aspartate transaminase for amino acid metabolism, and it has very limited ability to synthesize serine and glycine, resulting in its needs for exogenous amino acids. Consistently, we reported that RPE actively synthesizes glutamate, serine, and glycine and exports them toward retina (4,50). A recent study also shows that serine is required for hypoxia-inducible factor(HIF)-mediated protection against retinopathy, but retinal serine depends on exogenous source (53). Additionally, we have found that retina consumes more leucine than RPE, but little leucine is used for catabolism in the retina. To renew its daily shed outer segment in the photoreceptors, retina requires rapid protein synthesis (54,55). It is estimated by using [ 14 C]leucine that 0.1-1.4% of retinal proteins are newly synthesized per hour (54,56). Consistently, a majority of free leucine is used for protein synthesis, and only ϳ1.5% of free leucine is catabolized (57). Our study highlights the importance of RPE in support of retinal metabolism through active nitrogen metabolism.
In conclusion, we have found that nitrogen metabolism is tissue-specific in the retina and RPE. These specialized nitrogen utilization pathways might be biochemically adapted to the metabolic demands in the retina, including a high rate of energy metabolism and protein synthesis. These findings may provide insight in understanding the biochemical mechanisms in RPEinitiated retinal diseases such as age-related macular degeneration and inherited retinal diseases.

Reagents
All the reagents and resources were detailed in the key resources form (Table S4).

Animals
C57 B6/J male mice at 6 -8 weeks were purchased from the Jackson Laboratory. Mouse experiments were performed in accordance with the National Institutes of Health guidelines, and the protocol was approved by the Institutional Animal Care and Use Committee of West Virginia University.
For tracer study in vivo, the mice receive a bolus intraperitoneal injection of 15 NH 4 Cl at 167 mg/kg. Mouse retina, RPE/ choroid, brain (whole), and liver (a small piece from the same lobe) were quickly harvested, weighed, snap-frozen in liquid nitrogen, and stored at Ϫ80°C before use. Blood was drawn from heart into a microtube with 10 l of EDTA (0.5 mM) and centrifuged at 3000 rpm in cold room for 15 min. The plasma in the supernatant was transferred into a fresh tube and stored in Ϫ80°C before use.

Retina and RPE/choroid explant culture
Mouse retina and RPE/choroid were quickly isolated in 200 l of cold Hanks' balanced salt solution as previously described (4,58). The retina and RPE/choroid were transferred into 200 l of preincubated Krebs-Ringer/HEPES/bicarbonate buffer (59) with NH 4 Cl, different tracers, or a MPC inhibitor in the presence of 5 mM glucose, followed by a 2-h incubation at 37°C in a CO 2 incubator. The tissues were then quickly quenched with cold 0.9 mM NaCl, snap-frozen in liquid nitrogen, and stored at Ϫ80°C before use. The incubation media with or without tissue were harvested and stored at Ϫ80°C before use.

Metabolite extraction
Metabolites were extracted in 80% cold methanol as described previously (59). Mouse retinas were homogenized in 120 l of 80% methanol placed on dry ice with a microtube homogenizer. RPE/choroid, brain, and liver were weighed and homogenized with Omni tissue homogenizer on dry ice. RPE/ choroid was homogenized in 200 l of 80% methanol. Every 5 mg of liver or brain was homogenized in 100 l of 80% methanol. The samples were kept on dry ice for 30 min, centrifuged, and filtered with a 0.4-m syringe filter. All RPE/choroid mixture and 50 l of liver or brain mixture were transferred into glass inserts containing 10 l of internal standard (myristic acid d 27 in 1 mg/ml) and dried in a SpeedVac. For plasma and incubation media, 10 l of plasma or medium was mixed with 40 l of cold methanol to precipitate proteins, the mix was centrifuged, and 10 l of mixture was transferred into glass inserts with the internal standard to dry before derivatization.

Metabolite analysis with GC-MS
Dried samples were derivatized by methoxyamine and N-tertbutyldimethylsilyl-N-methyl trifluoro-acetamide and analyzed by the Agilent 7890B/5977B GC-MS system with DB-5MS column (30-m ϫ 0.25-mm ϫ 0.25-m film) as we described in detail before (29,58,59). Mass spectra were collected from 80 to 600 m/z under selective ion monitor mode. Tables S1 and S2 list the detailed parameters including monitored ions for the measured metabolites. The data were analyzed by Agilent MassHunter quantitative analysis software by extracting ion abundance for each monitored ion in Tables S1

Nitrogen metabolism in retina and RPE
and S2. Raw total ion abundance for isotopologues, the purity of the tracers, and derivatization reagents were imported in ISO-COR software (https://isocor.readthedocs.io/en/latest/) 5 to calculate the percentage of labeling by correcting natural abundance and tracer purity (49,59,60). The fragment formula including derivatization reagents was listed in Table S3 for setting up ISOCOR software. The percentage of labeled ions or enrichment was presented as 15 N or 13 C fraction of total. The relative abundance was the specific or total ion abundance normalized by the ion abundance from control samples.

Statistics
The significance of differences between means was determined by unpaired two-tailed t test. p Ͻ 0.05 was considered to be significant using GraphPad Prism 7.