Arginase 1 Insufficiency Precipitates Amyloid-β Deposition and Hastens Behavioral Impairment in a Mouse Model of Amyloidosis

Alzheimer’s disease (AD) includes several hallmarks comprised of amyloid-β (Aβ) deposition, tau neuropathology, inflammation, and memory impairment. Brain metabolism becomes uncoupled due to aging and other AD risk factors, which ultimately lead to impaired protein clearance and aggregation. Increasing evidence indicates a role of arginine metabolism in AD, where arginases are key enzymes in neurons and glia capable of depleting arginine and producing ornithine and polyamines. However, currently, it remains unknown if the reduction of arginase 1 (Arg1) in myeloid cell impacts amyloidosis. Herein, we produced haploinsufficiency of Arg1 by the hemizygous deletion in myeloid cells using Arg1fl/fl and LysMcreTg/+ mice crossed with APP Tg2576 mice. Our data indicated that Arg1 haploinsufficiency promoted Aβ deposition, exacerbated some behavioral impairment, and decreased components of Ragulator-Rag complex involved in mechanistic target of rapamycin complex 1 (mTORC1) signaling and autophagy. Additionally, Arg1 repression and arginine supplementation both impaired microglial phagocytosis in vitro. These data suggest that proper function of Arg1 and arginine metabolism in myeloid cells remains essential to restrict amyloidosis.


INTRODUCTION
Arginine metabolism associates with numerous biological pathways. The impact of arginine levels and arginine signaling warrants further consideration in Alzheimer's disease (AD) because of its high demand for multiple metabolic pathways and recent discoveries of mTORC1 activation associated bona fide and putative arginine sensors that locate at the lysosome [SLC38A9 (1), TM4SF5 (2)], cytoplasm [CASTORs (3)], and plasma membrane [GPRC6A (4)]. In this respect, properly coordinated arginine metabolism remains critical for linear coupling of signaling events linked to nutrients/amino acids sensing and mTORC1 regulation and could be particularly important for immune cells during neurodegeneration. In addition to arginine's role as a substrate for protein synthesis and arginine sensors, at least five enzymes catabolize arginine including arginase 1, 2 (ARG1, ARG2), nitric oxide synthases 1, 2, 3 (NOS1, 2, 3), arginine decarboxylase (ADC), arginine/glycine amidinotransferase (AGAT1), and arginyltransferase (ATE1), suggesting a critical role for numerous biological processes (5,6). Therefore, Arg1 may control the degree to which proteolysis occurs, protein turnover, and thus impact proteinopathies. Dysmetabolism of arginine and polyamines were among the most altered in mild cognitive impairment (MCI) and AD patient blood samples (7)(8)(9), cerebral spinal fluid (CSF) (9,10) and postmortem brain tissues (11)(12)(13)(14). Alterations in arginine metabolism were also observed in mouse models of Ab and tau deposition (15)(16)(17)(18)(19) and in vitro models (20). However, experimental manipulations and animal models concluded different outcomes regarding arginine metabolism in neuropathology (16,17). Other reports indicated that a potential rare arginase 2 allele was linked to an increased risk of AD (21). These reports signify a relationship between arginine metabolism and AD.
Balanced compartmental arginine levels are necessary for normal physiology and likely require cell type specific demands for proper functioning. Several reports link Arg1 to microglia/ monocyte wound repair and phagocytosis (22,23). A previous study by our group showed that Arg1 overexpression decreased neuroinflammation and tau pathology in a tauopathy mouse model (17). Another group reported Arg1 positive microglia were responsible for reducing Ab plague deposition during sustained neuroinflammation in an amyloidosis mouse model (24). Monocytes and other peripheral myeloid cells were found to infiltrate the brain and home to Ab deposits, thus ameliorated amyloidosis pathology (25,26). However, the impact on Ab deposition following the myeloid-specific deletion of Arg1 remains unknown. We determined to assess how the haploinsufficiency of Arg1 in myeloid cells impacts the brain during amyloidosis using conditional LysMcre deletion in APP Tg2576 transgenic mice. We performed immunohistochemical analysis, mouse behavioral assessments, and biochemical analysis in mice of amyloidosis and Arg1 haploinsufficiency. We also performed in vitro microglia phagocytosis assays following Arg1 reduction and arginine supplementation. Our data indicate a critical function for proper arginine and Arg1 levels during amyloidosis.

Behavioral Testing
All mice were aged at 15 months and subsequently exposed to a series of behavioral tasks to evaluate cognition and general activity. All tests were executed in order of increasing stress with appropriate break time between tests to avoid any lingering effects of previous experimental conditions. The experimenter was blinded to the data and experimental groups. The video tracking software ANY-maze (version 4.99, Stoelting Company, Wood Dale, IL) was used.

Open Field
Open field test (Ugo Basile S.R.L., Italy) was performed to measure general ambulatory activity and anxiety-related behaviors. After a 30-min procedure room acclimation period, mice were allowed to investigate the 44 cm 2 maze for 15 min while being recorded by ANY-maze. The locomotive activity was measured for total distance traveled and total time immobile. Anxiety-related behavior was measured by time in the center zone. Increased thigmotaxis indicates anxiety-related behavior and suggests an increased anxiety level.

Y-Maze
The Y-maze test (Ugo Basile S.R.L., Italy) was used to calculate spontaneous alternation to measure spatial working memory and anxiety. Mice were permitted to freely travel the maze (35 cm × 5 cm × 10 cm; L × W × H) for 5 min, while all movements recorded by ANY-maze. Alternation occurs if a mouse entered a different arm during three consecutive entries. The percentage of alternation was calculated (number of alternations divided by the total number of entries minus two). The decreased percentage of alternations suggests impaired working memory.

Radial Arm Water Maze
We performed a radial arm water maze (RAWM) based on the previous protocol with minor changes (28). In brief, six swimming arms are extending from the center. Only the goal arm contains the escaping platform at the end. Each arm is associated with an obvious visual cue in the surrounding environment. On day 1, mice were trained with constant alternating between visible and hidden platforms with the same goal arm for 15 trials. On day 2, mice were exploring the arms with only the hidden platform located at the same goal arm as day 1 for 15 trials. On day 3 (reversal), the platform was hidden at the goal arm opposite to that used on day 1 and day 2 and mice explored the arms for 15 trials. One error is counted if the mouse explores an arm that is different from the goal arm. An approximate average cohort size of 16 per day was designed and mice in the same cohort ran through each trial sequentially. The average and total errors for three continuous trials are calculated into five blocks over three days.

Aversive Conditioning
We performed contextual and cued fear conditioning test to measure the animal ability to learn and recollect links between environmental cues and aversive experiences. For this test, an auditory conditioned stimulus (CS, white noise, 70 dB) was accompanied by an unconditioned aversive stimulus (US, a mild foot shock, 0.5 mA) in a novel environment. On day 1 (Training), we placed mice in the fear conditioning apparatus for 180 s, then a 30 s CS was delivered with the US on the metal floor grid during the final 2 s. The training comprised of two CS, paired with two US, with a 2 min interval in between. On day 2 (24 h post-training), we placed mice in the apparatus, with no cues or stimulus (context fear conditioning), for 60 min while monitored for freezing. Directly after contextual testing, we placed the mice into a novel context and exposed them to CS for 180 s (cued fear conditioning). Freezing behavior related to both contextual and cued stimuli were used to assess learning and fear associated recall.

Detection of Congophillic Deposits
We performed Congo red histology as described previously (30). Briefly, brain sections were mounted and submerged for 20 min in alkaline sodium chloride, then incubated in 0.2% Congo Red solution (Sigma-Aldrich), Congo Red staining kit #HT60-1KT) for 30 min Sections were then rinsed in 95 and 100% ethanol, cleared in xylene for 15 min, and coverslipped with DPX mountant (Electron Microscopy Sciences). Statistical analysis was performed for total, vascular, and parenchymal Ab loads.

Protein Preparation for Biochemical Analysis
We prepared protein samples for ELISA and western blotting from the frozen anterior cortex and posterior cortex as previously described (17). Tissues were weighed and resuspended in RIPA buffer (50 mM Tris pH 7.6, 140 mM NaCl, 1% NP40, 0.5% sodium deoxycholate, 0.1% SDS) at 10% wt/vol, containing a 1% vol/vol protease inhibitor cocktail (Sigma-Aldrich, #P8340), phosphatase inhibitor cocktail 2 (Sigma-Aldrich, #P5726), phosphatase inhibitor cocktail 3 (Sigma-Aldrich, #P0044), and PMSF (Sigma-Aldrich, #10837091001). Tissues were mechanically homogenized and sonicated to obtain the whole cell lysate. An aliquot of the wholecell lysate was centrifuged for 30 min at 40,000 g (4°C). Supernatant detergent soluble (S1) was collected. The resulting pellet became a P1 fraction. The protein concentration of wholecell lysate and S1 fraction was measured by Pierce ™ BCA Protein Assay Kit (Thermo Fisher Scientific, Thermo Scientific ™ , #23225) according to the manufacturer's protocol. The P1 protein fraction was resuspended in 70% formic acid using 20% volume of whole-cell homogenate aliquot, then incubated for 1 h at room temperature. Formic acid was buffered with 1M Tris (pH 7.5) using 25% volume of formic acid. A final pH value at 7.5 was adjusted using NaOH solution (50% w/w) to create the detergent insoluble, formic acid (FA) fraction.

L-Arginine Quantification
Whole-cell lysate samples prepared from the mouse posterior cortex were subjected to liquid chromatography-mass spectrometry/mass spectrometry (LC-MS/MS) using a standard curve of specific amino acid analytes. Quantification of Larginine was performed by Sanford Burnham Prebys (Orlando, FL, USA).

Ab 40 and Ab 42 ELISA
We measured the concentrations of Ab 40 and Ab 42 from the aforementioned detergent soluble S1 and detergent insoluble FA P1 fraction of anterior cortex protein lysate using the Ab 1-40 and 1-42 specific sandwich ELISA kit (Mega Nano Biotech. FL, USA). In brief, we coated each well of a 96 well plate with 50 µl of goat anti-human Ab 1-42 antibody (MegaNano BioTech Inc., AB-001) diluted in 1× PBS at 10 mg/ml and incubated overnight at 4°C. We washed the plate five times and blocked it in 200 ml blocking buffer at 37°C for 1 h. After washing the plate, 50 ml diluted detection antibodies anti-Ab 40 (MegaNano BioTech Inc., Ab40-002) or anti-Ab 42 (MegaNano BioTech Inc., AB42-002) were mixed with either 50 µl diluted peptide standard solution or diluted samples in a preparation plate before loading to the assay plate. Then we incubated the plates at 4°C overnight. The next day, we washed first and then added 100 µl of diluted secondary antibody into each well and incubated for 45 minutes on an orbital shaker at room temperature. The plate was then washed four times before applying the TMB peroxidase substrate (Surmodics, Inc., TMBS-1000), then the plates were incubated at room temperature for 10 min. The reaction was ended by adding 100 ml/well of 0.4 M H 2 SO 4 . Absorbance at 450 nm was measured with a BioTek Synergy H4 microplate reader. The final concentration was calculated based on peptide standards.

Quantitative Real-Time PCR
We extracted total mRNA from cryopreserved hippocampal tissues using AllPrep DNA/RNA/Protein Mini Kit (QIAGEN, #80004) according to the manufacturer's protocol. For qRT-PCR analysis, we probed the mouse Arg1 mRNA from the hippocampal tissues of the aforementioned four mouse groups. The cDNA was generated using the Superscript ™ III First-Strand Synthesis System (Thermo Fisher Scientific, Invitrogen ™ , #18080051). The QuantiTect Primer ® Assay was used for mouse Arg1 gene transcripts (QIAGEN, Mm_Arg1_1_SG, #QT00134288, Lot# 229462537; IDT, Actin, Mm.PT.39a.22214843.g) along with SYBR ® Green Quantitative RT-qPCR Kit (Sigma-Aldrich, #QR0100-1KT). We used a standard curve from a pool of all genotypes to span three logs of dynamic range. Melt curve analysis was performed for primer validation. We used the Opticon 2 ™ Real-Time PCR System (v4.3, Bio-Rad) to detect amplicon.

Primary Macrophage Culture
Bone marrow derived macrophages (BMDMs) were isolated from the femurs and tibias of nTg/Arg1 +/+ mice, nTg/Arg1 fl/fl mice and nTg/Arg1 fl/fl /LysMcre Tg/+ mice at 3-6 months of age. Bones were flushed with a syringe filled with cold washing media (RPMI 1640 supplemented with 10% FBS and 1% penicillin/ streptomycin) to extrude bone marrow into a sterile falcon tube. The bone marrow was then triturated three times using syringes fit with 18 gauge needle and then centrifuged at 1,000 rpm for 5 min at 4°C. After decanting the supernatant, red blood cells were lysed in lysis buffer (0.15 mol/L NH4Cl, 10 mmol/L KHCO3, and 0.1 mmol/L Na2EDTA, pH 7.4) for 3 min. The remaining cells were washed and pelleted in washing media, then resuspended in BMDM differentiation media (RPMI 1640 supplemented with 1% penicillin/streptomycin, 1% HEPES, 0.001% b-mercaptoethanol, 10% FBS, and 20% supernatant from sL929 cells-provides macrophage colony stimulating factor) (32). BMDM cells were plated in T75 flasks at 5-8 × 10 5 cells/ml. Cell culture media was changed on post days of 2, 4, and 6, and then cells were re-plated at the density of 1 × 10 6 cells/ ml on day 7 in differentiation media without sL929 supernatant. The following day, BMDMs were stimulated with LPS (100 ng/ml; Invitrogen) + IFN-gamma (20 ng/ml; eBioscience), IL-4 (20 ng/ml; eBioscience), or LPS (50 ng/ml) + IL-4 (20 ng/ml) in Neuro-2a growth medium or differentiation media without sL929 supernatant as previously described (33). Azithromycin (AZM, Sigma PHR1088, 25 mM) was added to the BMDMs at the time of stimulation. Unstimulated BMDMs were maintained in the appropriate growth medium as controls. 24 h post stimulation, the supernatant of the stimulated macrophages (macrophage conditioned media) was collected and centrifuged to remove the cell debris before being applied to Neuro-2a cells for the measurement of neurotoxicity. BMDM cell lysates were used to assess arginase activity using the QuantiChrom Arginase Assay Kit (Bioassay Systems DARG-200) according to the manufacture's protocol.

Neurotoxicity Assay
Neuro-2a cells were cultured in growth media which contains 45% DMEM and 45% OptiMEM Reduced-Serum Medium (Life Technologies) supplemented with 10% FBS and 1% penicillin/ streptomycin. Experiments were carried out using Neuro-2a cells within 12 passages. Neurotoxicity was assessed after Neuro-2a cells were seeded in 96-well plates at a density of 2 × 10 5 cells/ml for 24 h in Neuro-2a growth media. Then Neuro-2a growth media were replaced by different stimulated macrophage conditioned media for 24 h. The cell viability was measured by using MTT assay (Sigma-Aldrich) according to the manufacturer instructions and as described previously (34).
To determine how Arg1 insufficiency impacts Ab deposition, we measured total Ab, and forms with specific C termini ending at Ab 38, Ab 40 and Ab 42 in the anterior cortex (ACX), hippocampus (HPC), and entorhinal cortex (ECX) of the brain in 16-month-old APP Tg2576 mice with Arg1 sufficiency and insufficiency by immunohistochemistry. Non-transgenic mice had undetectable levels of all Ab species (data not shown). Overall, total Ab was increased in HPC (p = 0.043), ECX (p = 0.034) and trended toward an increase in ACX (p = 0.059) in APP/Arg1 insufficient mice relative to APP/Arg1 sufficient mice ( Figures 1A, B, K). Ab 42 was significantly increased in APP/ Arg1 insufficient mice compared to APP/Arg1 sufficient mice in all three brain regions (ACX, p = 0.038; HPC, p = 0.011; ECX, p = 0.005) ( Figures 1C, D, L). However, we did not detect changes in plaque load measured by Ab 40 ( Figures 1E, F, M) or Ab 38 ( Figures 1G, H, N). Furthermore, we measured compact plaques using the histological stain Congo red and found no changes in the ACX, HPC, and ECX regions between these two groups ( Figures 1I, J, O). Lastly, we measured Ab 42 and Ab 40 by ELISA in both detergent soluble and insoluble (formic acid soluble) fractions. We found increased Ab 42 in the detergent soluble fraction (p = 0.048), but not in the formic acid fraction ( Figure 1P). No changes were detected for Ab 40 in ELISA ( Figure 1Q). Collectively, these data strongly suggest that Arg1 insufficiency during amyloidosis promotes diffuse Ab 42 deposition.

Arg1 Insufficiency During Amyloidosis Hastens the Impairment of Mouse Behavioral Performance
Next, we determined if Arg1 insufficiency impacts behavioral performance in APP and nTg mice. We used open field behavior to measure overall activity by total distance traveled and total immobile time, and anxiety by total time spent in the center zone. The APP mice were more active relative to nTg mice measured by increased total distance traveled (p < 0.0001, Figure  3A) and decreased total time immobile (p = 0.001, Figure 3B). Although we did not find any difference for the time spent in the center zone in mice with APP, we found that mice with Arg1 insufficiency spent less time in the center, indicating increased anxiety (p = 0.039, Figure 3C). We measured mouse general working memory by Y-maze and showed that APP mice demonstrated a decreased percentage of alternation (p < 0.0001), suggesting decreased working memory with amyloidosis ( Figure 3D). Additionally, we found main effects in genotypes of APP (p < 0.0001) and Arg1 insufficiency (p = 0.027), as well as an interaction between genotypes in measuring the number of entries in Y-maze (p = 0.003). Specifically, the APP/Arg1 insufficient mice displayed an increased number of entries compared to APP/Arg1 sufficient mice (p < 0.0001), suggesting APP/Arg1 insufficient mice were more exploratory ( Figure 3E). Furthermore, we tested mice in the radial arm water maze (RAWM), to measure spatial working memory. Mice received two days of training using one goal arm followed by 1 day of reversal training (using the opposite arm as the goal arm). Errors were counted during training until the mouse reached the rescue platform. Repeated measure ANOVA analysis over three days showed that mice with APP performed more errors compared to nTg mice (p < 0.0001, Figure 3F). Examining each day of training separately revealed that APP mice successfully learned procedural aspects of the maze so there were no main genotype effects on Day 1 (p = 0.054) or Day 2 (p = 0.064) of training ( Figure 3G). However, APP mice were less able to learn a new platform location, so there was a main effect of APP genotype on the reversal day of testing (p < 0.0001, Figure   3G). The total number of errors for APP/Arg1 insufficient mice was the highest amongst all groups ( Figure 3G). Lastly, we measured aversive conditioning for fear associated memory. After the initial training trials for all groups, we measured the aversive conditioning by calculating the percentage of freezing time during contextual and cued testing. In the context testing, the APP/Arg1 insufficient mice showed the lowest percentage of freezing time among all groups, and significantly less than nTg/ Arg1 insufficient mice (p = 0.009), which contributed to the overall main effect in APP genotype (p = 0.006) ( Figure 3H). In cued testing, all groups of mice showed a higher percentage of freezing time during the toning phase than the no tone phase, indicating an association with learning behavior ( Figure 3I). During the toning phase, the APP/Arg1 insufficient mice demonstrated the weakest freezing response among all groups and had less total freezing time percentage than nTg/Arg1 insufficient mice (p = 0.009), which accounts for the main effect in APP genotype (p = 0.002) ( Figure 3I). In summary, mice with amyloidosis display impaired behaviors activities, anxiety, and memory, while Arg1 insufficiency exacerbated these effects.

Arg1 Insufficiency Decreases Total Arginase 1 Expression and Increases Arginine in Mouse Brain
To determine if hemizygous deletion of Arg1 in myeloid cells impacts total ARG1 expression and arginine levels in the brain, we measured Arg1 mRNA by qRT-PCR, ARG1 protein by western blotting and arginine by LC-MS/MS. Interestingly, although we did not find changes in Arg1 mRNA between groups, we detected an overall decreased expression of ARG1 protein by the main genotype effect of Arg1 insufficiency (p = 0.032) ( Figures 4A, B, D). We detected a higher level of arginine in the APP/Arg1 insufficient mice relative to nTg/Arg1 insufficient mice (p = 0.003) and the main effect of APP genotype (p = 0.036) ( Figure 4C). We also observed a trend for increased arginine comparing APP/Arg1 insufficient mice to APP/Arg1 sufficient mice (p = 0.095) ( Figure 4C). Overall, hemizygous deletion of Arg1 in myeloid cells modestly decreased total ARG1 expression and increased arginine levels during brain amyloidosis.

Ragulator-Rag Complex Critical for Microglial Phagocytosis Is Increased During Amyloidosis and Blunted in Arg1 Insufficient APP Mice
Several key studies using different models reported that several components of the Ragulator-Rag complex, namely LAMTOR2, LAMTOR3, LAMTOR4, and RAG A were essential regulators for maintaining normal lysosomal activity and microglial phagocytosis; loss of functions inhibit phagocytic digestion, even after engulfment of neuronal debris (38,39). Therefore, we measured the expression of LAMTOR2, LAMTOR3, LAMTOR4, and RAG A by western blot. Densitometry analysis showed that APP/Arg1 sufficient mice had increased expression in LAMTOR4 (p = 0.005), RAG A (p = 0.002) and increased trend of LAMTOR2 (p = 0.095) and LAMTOR3 (p = 0.053) relative to nTg/Arg1 sufficient mice ( Figures 4D-H). However, this effect was blocked in APP/Arg1 insufficient mice compared to APP/Arg1 sufficient mice. Importantly, the APP/ Arg1 insufficient mice had reduced expression in LAMTOR2 (p = 0.031), LAMTOR3 (p = 0.001), LAMTOR4 (p = 0.022) and RAG A (p = 0.012) compared to APP/Arg1 sufficient mice, thus resulted either trending or significant interactions between APP and Arg1 insufficiency genotypes on LAMTOR2 (p = 0.098), LAMTOR3 (p = 0.002), LAMTOR4 (p = 0.008), and RAG A (p = 0.002), respectively ( Figures 4E-H). Despite increased microglia activation measured by IBA1 and CD68, these data strongly suggest that several components of the Ragulator-Rag complex are up-regulated during amyloidosis and reduced with Arg1 insufficiency, indicating a potential deficit in microglial digestion machinery.

Arg1 Repression and Arginine Supplementation Impair Microglia Phagocytosis In Vitro
Two in vitro methods were employed to determine if arginine metabolism impacts microglia phagocytic function measured by pHrodo phagocytosis assay. Fluorescent green E. coli bioparticle conjugates are presented to phagocytic cells and green fluorescence increases as pH decrease from neutral cytoplasm to the acidic lysosome. The phagocytic activity of cells is measured based on the acidification of the bioparticles as they are ingested by phagosomes (pH 6.1-6.5) and digested by phagolysosomes (pH 5.0-5.5) (40). We detected green fluorescence by flow cytometry and calculated the phagocytosis index (%) based on the percentage of pHrodo green fluorescence. First, we employed an immortalized rat microglial HAPI cell line due to high endogenous Arg1 expression. Naïve HAPI cells without siRNAs and bioparticles served as a negative control for green fluorescence detected at an average of 0.2% ( Figures  5A, B, I). HAPI cells treated with siGLO-Green without bioparticles was designed as a positive control for measuring cellular uptake of siRNAs and green fluorescence, which detected at an average of 100% ( Figures 5C, D than cells transfected with siRNA-NT, suggesting that Arg1 repression impaired phagocytosis in HAPI rat microglial cells ( Figures 5G, H, I). Second, we treated primary rat microglia at control (basal arginine concentration at 0.7 mM), 3 mM arginine, and 10 mM arginine in the culture medium followed with pHrodo phagocytosis assay. Cells were gated for total and high pHrodo green fluorescence (GFP Total and GFP High) representing the different extent of phagocytosis activity ( Figures  5J-R). Although the average of GFP Total phagocytosis index was similar among control (88.9%), 3 mM (89.8%), and 10 mM (85%) of arginine, the average of the GFP High phagocytosis index showed larger differences among control (78.4%), 3 mM (77.8%), and 10 mM (30.5%) of arginine ( Figure 5S). Thus, we observed a dramatic 47.9% reduction in GFP High phagocytosis index with 10 mM arginine compared to the control condition ( Figure 5S). Interestingly, 3 mM arginine showed an equivalent GFP High phagocytosis index compared to the control condition ( Figure 5S). A representative histogram of pHrodo green fluorescence clearly showed only the 10 mM arginine treated cells had two small peaks of GFP fluorescence, indicating fewer bioparticles were delivered to the final stage of phagolysosome ( Figure 5T). Collectively, reduced arginine metabolism induced either by Arg1 repression or arginine supplementation impairs microglial phagocytosis in vitro.

DISCUSSION
In the present study, we found that the haploinsufficiency of Arg1 in myeloid cells in a mouse model of amyloidosis increased Ab deposition, activated microglia, and impaired behavioral performance. Mechanistic studies suggest that impaired phagocytosis secondary to reduced Arg1 could contribute to Ab deposition. Overall, we provided evidence that APP/Arg1 insufficiency increased more diffuse Ab deposits measured by total Ab and Ab 42, and also stimulated microglial activation via CD68 and IBA1. The Arg1 insufficiency in APP mice precipitated more behavior impairment evidenced by more anxiety, more exploratory behavior, and decreased fear associated memory. Importantly, the Ragulator-Rag complex was activated during amyloidosis perhaps as a compensatory response to improve microglial phagocytosis and degradation of Ab. However, reduced Ragulator-Rag complex in APP mice with Arg1 insufficiency during amyloidosis indicated that arginine metabolism has an unappreciated role in executing this function at the level of the lysosome. Finally, we showed in vitro evidence that decreased Arg1 expression or increased arginine levels could inhibit microglial phagocytosis. Collectively, these data support the overall concept that Arg1 insufficiency in myeloid cells exacerbates amyloidosis induced neuropathology possibly by activating microglia and impairing phagocytosis. Two isoforms of arginase consisting of cytosolic ARG1 and mitochondrial ARG2 were reported to be increased in the frontal cortex of AD patients (11,21,41). Additionally, a rare allele of ARG2 in males showed increased risk of developing early-onset AD (21). In AD mouse models of amyloidosis, increased Arg1, but not Arg2, was associated with Ab deposition (16,41). Although using DFMO or L-norvaline inhibited arginase 1 and 2 and reduced amyloid deposition in AD mouse models, these compounds both non-specifically inhibit other important targets to reduce amyloid pathology (16,42). Previously we showed that overexpression of Arg1 in the CNS of rTg4510 tau transgenic mice decreased neuroinflammation, reduced tau pathology, and myeloid Arg1 knockout mice signified a distinct microglial phenotype that promoted tau pathology (17). This is consistent with another study that reported Arg1 positive microglia participated in clearing Ab plaques during IL-1b mediated inflammatory response (24). In the current study, by using conditional LysMcre deletion in APP Tg2576 transgenic mice, we are the first to measure the specific effect of reducing Arg1 in myeloid cells in responding to amyloidosis. Our results showed that reduced Arg1 levels in LysM positive cells promoted amyloidosis in CNS. Therefore, we argued that the Arg1 expression in myeloid cells was predicted to be a compensatory response to restrict amyloidosis in AD. The Arg1 promoted the reparative role of macrophages and studies showed neurodegeneration was accelerated in Arg1 global deletion and myeloid Arg1 knockout mice (43). However, although there is a lack of studies in AD mouse models regarding Arg2, Arg2 knockout mice showed neuroprotective effects and reduced neurodegeneration in acute models of retinal injury (44,45). These studies suggest a distinct role of Arg2 in CNS diseases (46). We also know that knocking out Arg1 was lethal in mice due to hyperammonemia caused by arginase deficiency, whereas the Arg2 knockout mice had intact phenotypes and Arg1/Arg2 double knockout mice exhibited the same arginase deficiency phenotype as the Arg1 single knockout mice (47,48). Although myeloid Arg2 was initially reported to be less involved in immune responses, recent findings suggest that arginase isoforms in myeloid cells may play crucial regulatory roles in immune response (49,50). It remains interesting for future studies to investigate the immunological role of myeloid arginase 2 in association with AD.
The lysosome-associated protein CD68 in microglia was often assumed to represent enhanced phagocytosis upon Ab stimulation in AD (51, 52). However, recent work identified lipid droplet accumulating microglia (LDAM) as having phagocytic deficits but show phagosome maturation among the top regulated pathways from LDAM transcriptome. Importantly, LDAM was found to have upregulated endosomal/lysosomal genes including CD68 in BODIPY positive (lipid droplets) microglia (53). Other reports showed that the Ragulator-Rag complex regulates phagocytic flux in zebrafish, and zebrafish lacking functional Ragulator-Rag complexes (i.e. LAMTOR4, RAG A) promoted and expanded lysosomal compartments but were unable to appropriately digest neuronal debris (38). Therefore, this urges certain caution in the appearance of increased lysosomal activity/biogenesis or expanded lysosomal components such as CD68 that may actually indicate improper system functioning or uncoupling of the phagocytosis/digestion process. Interestingly, another report identified three members of the Ragulator complex (LAMTOR2, 3, 4), RAG A, and NRPL2 (GATOR1 complex) as regulators of phagocytosis (39). The authors showed that CRISPR knockout of Ragulator components (LAMTOR2, 3, 4) and RAG A inhibited phagocytosis, whereas knockout of NRPL2 increased phagocytosis (39). In our study, we found increased LAMTOR2, 3, 4, and RAG A in APP/Arg1 sufficient mice, suggesting that Ab deposition induces Ragulator complex machinery possibly for enhanced phagocytosis. However, Arg1 haploinsufficiency in myeloid cells blocked this induction in APP mice. Our in vitro experimental findings also indicated Arg1 suppression via siRNA in microglial cells reduced phagocytosis, providing potential mechanism for the in vivo findings. One result of reduced Arg1 is increased arginine. We also demonstrated that arginine supplementation in primary microglia reduced phagocytosis. Thus, arginine and Arg1 levels impact phagocytic/digestive function. Collectively, this argues that amyloidosis triggers expression in key Ragulator-Rag complex components critical for lysosomal digestion and that proper Arg1 function may be required for this response.
Recent studies with microglia/macrophage markers using single cell RNA-sequencing (scRNA-seq) distinguished brain resident microglia from CNS infiltrated myeloid cells (monocytes/macrophages) (54). The Lyz2 gene increases in active microglial subpopulations particularly during phases of demyelination/remyelination, and associates with Trem2 independent microglia activation (55,56). Since the myeloid marker gene Lyz2 targets both macrophages and microglia, the observed consequence of exacerbated amyloidosis from the current study were possibly attributed to the deficiency of Arg1 in both brain resident microglia and CNS infiltrated macrophages. Furthermore, as one of the earliest groups utilizing peripheral myeloid cells (monocytes/macrophages) to restrict amyloidosis in AD mouse models, we and others discovered that peripheral monocytes/macrophages were recruited to the sites of Ab deposition in the brain to phagocytose Ab plaques and oligomers (25,26,57). Together with resident microglia, infiltrated macrophages promoted degradation of cerebral Ab (58,59). Recent studies showed Arg1 played an essential role in efferocytosis of apoptotic cells in primary microglia/macrophage (60,61). Consistent with the previous report, our current study aligns with the phagocytic process. Therefore, Arg1 in brain myeloid cells may have dual roles in removing Ab plaques by phagocytosis and apoptotic neurons through efferocytosis. Nevertheless, future efforts are warranted to further distinguish the role of Arg1 in brain resident microglia from infiltrated peripheral macrophages in responding to amyloid deposition.
In the past several years, emerging studies reported the antibacterial enzyme lysozyme 2 was also expressed in neurons in addition to myeloid cells, thus complicating the interpretation of results using LysM-Cre mice in studying CNS diseases (62). In one study, researchers found LysM-Cre-driven tdTomato expressed in less than 25% of microglia, but also observed LysM-Cre-driven MeCP2 expression in neurons from several brain regions (63). In another study, researchers found LysM-Cre-tdTomato positive cells were on average account for less than 30% in both microglia and neurons, but could be exclusively expressed in neurons in certain brain regions (64). Most recently, LysM-Cre-tdTomato was confirmed to efficiently target nearly 40% of microglia/macrophages and surprisingly found to also target nearly 10% of neurons (65). Therefore, the expected LysM-Cre specificity in microglia/macrophages and the unexpected LysM-Cre specificity in neurons indicate two limitations in the current study. Indeed, both neurons and microglia are critically involved in AD pathogenesis.
Firstly, all four groups of mice shared the same LysMcre Tg/+ genetic background in which the endogenous Lyz2 gene was disrupted due to the insert of Cre recombinase (66). The lysozyme was previously reported to show protective properties in amyloid pathology by interacting with Ab species to reduce Ab aggregation, and localizing in sites of Ab plaques in AD brains (67)(68)(69). The lysozyme level was also found increased in the CSF of AD patients, presumably produced mainly by brain myeloid cells (67). This assumption was recently confirmed by studying Ab plaqueassociated microglia using scRNA-seq. Converging studies pointed out that Lyz2 was one of the commonly induced microglial genes by disease-associated microglia (DAM) and microglial neurodegenerative phenotype (MGnD), suggesting the microglial expression of Lyz2 played an important role in responding to Ab stimulus in AD mouse models of amyloidosis (55,70,71). These findings argue that Lyz2 gene is mainly expressed in CNS myeloid cells rather than other CNS cells, which could be confirmed in the summarized RNA-seq data website "The Myeloid Landscape 2" (72). Purified CNS cell types showed both human Lyz and mouse ortholog Lyz2 only significantly expressed in CNS myeloid cells rather than other CNS cell types including neurons (GSE73721, GSE52564, GSE75431). It remains unknown if one allele deletion of Lyz2 in our study impacts the function of DAM/MGnD microglial signatures in reacting to amyloidosis. However, the same Lyz2 deficient genotype shared with all mouse groups further lessens this concern.
Secondly, although the exact function of Lyz2 in neurons is not clear and Lyz2 is not dominantly expressed in neurons, we cannot exclude the fact that LysM-Cre recombination reduced Arg1 expression in neurons to a certain extent in our study. Previously we reported that overexpressing Arg1 through pseudotyped AAV9 with neuronal tropism in the hippocampus of rTg4510 tau transgenic mice decreased several neuronal tau pathologies and suggested to increase autophagy through decreased mTORC1 signaling (17). One reason is that overexpression of Arg1 could deplete cellular arginine and eventually decrease arginine sensing mTORC1 activation. In the current study, we inferred that decreased Arg1 in LysM-Cre neurons could accumulate cellular arginine to activate mTORC1 signaling by arginine sensing mechanisms, thus inhibit autophagy activities. Reduced autophagy in neurons had decreased capacity to degrade intracellular Ab aggregates, further exacerbated the amyloid pathology in conjunction with Arg1 deficient myeloid cells. For future studies, myeloid specific mouse models like Cx3cr1-CreERT2 and Tmem119-CreERT2 should be applied to target monocytes/macrophages and resident microglia, respectively (73)(74)(75).
Although arginine and Arg1 comprise of a general arginine metabolism pathway, based on the recent identification of amino acid sensors and signaling components, it remains possible that arginine and Arg1 act through different mechanisms to affect phagocytosis and degradation. For example, differential mechanisms could result from cytoplasmic arginine depletion/ repletion, lysosomal acidification, and GPCR extracellular signaling; however, further studies are required to elucidate whether these emerging pathways converge or remain independent. Key findings reported arginine signaling through the lysosomal transporter SLC38A9-Ragulator-Rag-mTORC1 axis and CASTOR1-Ragulator-Rag-mTORC1 axis, both of which activate downstream mTORC1 signaling and inhibit autophagy (76). In our study, APP/Arg1 insufficient mice showed increased CD68 expression in HPC and ECX compared to APP/Arg1 sufficient mice and nTg/Arg1 insufficient mice, suggesting that amyloidosis and Arg1 deficiency promoted CD68 expression. However, given the relationship between ARG1-arginine and arginine-Ragulator-Rag-mTORC1-axis, it is possible that myeloid-specific demands for Arg1 are required in response to amyloidosis and that uncoupling of phagocytosis/digestion initiates increased Ab deposition. During certain types of inflammation, myeloid/ microglia cells launch iNOS and Arg1 temporally, spatially or even simultaneously, which is thought to provoke either M1 or M2 phenotype to facilitate the clearance of debris and would repair (77). Given this, both enzymes deplete arginine levels, which could essentially promote phagocytosis from both pro and anti-inflammatory states. However, the balance of these enzymes in myeloid/microglial cells in response to amyloidosis may be critical in precipitating the AD phenotype. This is evident in work by Kan et al., which argues that excessive Arg1 also promotes amyloidosis in a NOS2 null background (16). Probably, Arg1 and NOS2 activities, their respective byproducts, and arginine depletion/repletion are temporally critical regarding microglial response to amyloidosis.
Overall, our findings imply that the haploinsufficiency of Arg1 in myeloid cells during amyloidosis exacerbates AD-like neuropathology, neuroinflammation, and behavioral deficits. It remains unclear as to the exact mechanism of how Arg1 or arginine governs phagocytic-digestive function. However, emerging evidence may suggest arginine signaling to mTORC1. Sustained Arg1 would essentially deplete local arginine levels permitting activation of autophagy through mTORC1 inhibition. Therefore, proper Arg1 levels in microglia/myeloid cells become critical for activating phagocytic responses during challenges such as amyloidosis. While failure to mount sustained Arg1 during the Ab challenge could result in local arginine accumulation and phagocytic dysfunction. Importantly, arginine sensors could serve as new therapeutic targets to rebalance failed phagocytic function during amyloidosis or other disorders associated with immune function.

DATA AVAILABILITY STATEMENT
The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

ETHICS STATEMENT
The animal study was reviewed and approved by University of South Florida and University of Kentucky Institutional Animal Care and Use Committee (IACUC). control) showed increased arginase activity post IL-4 treatment with or without LPS (p < 0.0001); however, macrophages of nTg/Arg1 fl/fl /LysMcre Tg/+ mice did not induce arginase activity in both conditions. In addition, in IL-4 treatment with or without LPS co-stimulation, primary macrophages isolated from nTg/Arg1 fl/fl / LysMcre Tg/+ mice showed decreased arginase activity compared to macrophages from either nTg/Arg1 +/+ mice (p < 0.0001) or nTg/Arg1 fl/fl mice (p < 0.0001). n = 3 independent replicates from three independent biological replication experiments. (B) In measuring Neuro-2a viability using macrophage conditioned medium (MCM), neuroprotection drug azithromycin (AZM) co-incubation with LPS/IFN-g increased neuron viability compared to LPS/IFN-g stimulation, a phenomenon only observed using MCM from nTg/Arg1 +/+ mice (p < 0.0001), but not in nTg/Arg1 fl/fl / LysMcre Tg/+ mice. In addition, AZM co-incubation with LPS/IL-4 showed increased trend in neuron viability than LPS/IL-4 stimulation using MCM from nTg/ Arg1 +/+ mice (p = 0.051), but not in nTg/Arg1 fl/fl /LysMcre Tg/+ mice. n = 10 independent replicates from three independent biological replication experiments. ****, p < 0.0001; n.s. (not significant), p > 0.05. Two-way ANOVA followed by multiple comparisons with Dunnett's or Sidak's post-hoc tests. Values represent mean ± SEM.