Acute Hypoxia Alters Extracellular Vesicle Signatures and the Brain Citrullinome of Naked Mole-Rats (Heterocephalus glaber)

Peptidylarginine deiminases (PADs) and extracellular vesicles (EVs) may be indicative biomarkers of physiological and pathological status and adaptive responses, including to diseases and disorders of the central nervous system (CNS) and related to hypoxia. While these markers have been studied in hypoxia-intolerant mammals, in vivo investigations in hypoxia-tolerant species are lacking. Naked mole-rats (NMR) are among the most hypoxia-tolerant mammals and are thus a good model organism for understanding natural and beneficial adaptations to hypoxia. Thus, we aimed to reveal CNS related roles for PADs in hypoxia tolerance and identify whether circulating EV signatures may reveal a fingerprint for adaptive whole-body hypoxia responses in this species. We found that following in vivo acute hypoxia, NMR: (1) plasma-EVs were remodelled, (2) whole proteome EV cargo contained more protein hits (including citrullinated proteins) and a higher number of associated KEGG pathways relating to the total proteome of plasma-EVs Also, (3) brains had a trend for elevation in PAD1, PAD3 and PAD6 protein expression, while PAD2 and PAD4 were reduced, while (4) the brain citrullinome had a considerable increase in deiminated protein hits with hypoxia (1222 vs. 852 hits in normoxia). Our findings indicate that circulating EV signatures are modified and proteomic content is reduced in hypoxic conditions in naked mole-rats, including the circulating EV citrullinome, while the brain citrullinome is elevated and modulated in response to hypoxia. This was further reflected in elevation of some PADs in the brain tissue following acute hypoxia treatment. These findings indicate a possible selective role for PAD-isozymes in hypoxia response and tolerance.


Introduction
Naked mole-rats (Heterocephalus glaber) are a eusocial and subterranean mammalian species with unusual resistance to hypoxia, cancer, and ageing, among other remarkable abilities [1,2]. Indeed, naked mole-rats are one of the most hypoxia-tolerant adult mammals and can tolerate minutes of anoxia, hours at 3% O 2 , and days to weeks at 8% O 2 [3][4][5][6]. As a result, naked mole-rats are emerging as a key model organism for the study of naturally evolved mechanisms of hypoxia tolerance, from whom insight may be gained into novel molecular pathways that are protective against hypoxia-related pathologies, including age-related diseases and neurological disorders [7,8].
Naked mole-rats respond to hypoxia with a suite of metabolic modifications to reduce energy demand and maximize the efficiency of metabolic pathways. For example, naked response to a hypoxia challenge. Recently, we assessed citrullination in plasma and plasma EVs in naked mole-rats at baseline conditions for the first time, linking this modification to a range of metabolic and immune-associated pathways. However, the roles of PADs in hypoxia, and of circulating EV signatures in response to hypoxia, have yet to be explored in this model. Important as well, as roles for PADs in regulating brain responses to hypoxic insult have previously been shown and also linked to neurodegeneration, we sought to investigate the brain citrullinome of naked mole-rats following a hypoxia challenge, as this may reveal a putatively beneficial phenotype for hypoxic injury responses in the CNS.

Extracellular Vesicle Profiles in Naked Mole-Rat Plasma Change in Response to Hypoxia Treatment
Circulating plasma EVs from naked mole-rats treated for 4 h in normoxia or hypoxia were profiled by NTA for assessment of changes in total EV numbers, as well as for EV subgroups based on size. In addition to total EVs (0-1000 nm), EVs were also separately counted in the size ranges of 0-100 nm ("small EVs"); 101-200 nm ("medium-sized EVs") and >200 nm ("large EVs"), based on NTA measurements (Figure 1). A statistically significant change was observed in total EV numbers between the two groups (0-1000 nm), with reduced EV levels in plasma following hypoxia. Changes in release of different EV subsets was also observed, with a significant decrease in small EVs (<100 nm) following hypoxia, significant reduction in medium sized  nm) EVs following hypoxia, while larger EVs (>200 nm) showed also a trend for reduced numbers following hypoxia, albeit not statistically significant (n=10 per group; exact p-values are indicated on the graphs; p <0.05 considered statistically significant; error bars indicate SEM). (A) Number of EVs isolated from naked mole-rat plasma, comparing normoxia and hypoxia conditions. Changes were assessed in release profiles of total EVs (0-1000 nm), small EVs (<100 nm), medium-sized EVs (101-200 nm) and large EVs (201-1000 nm); based on measurement of plasma EVs from 10 animals per group; error bars represent standard error of mean (SEM); t-test, exact pvalues are shown, p < 0.05 considered statistically significant (indicated by *). (B) Representative NTA curves of plasma EVs from naked mole-rats following normoxia or hypoxia treatment, Figure 1. EV profile trends from plasma of naked mole-rats treated for 4 h in normoxia or hypoxia. (A) Number of EVs isolated from naked mole-rat plasma, comparing normoxia and hypoxia conditions. Changes were assessed in release profiles of total EVs (0-1000 nm), small EVs (<100 nm), medium-sized EVs (101-200 nm) and large EVs (201-1000 nm); based on measurement of plasma EVs from 10 animals per group; error bars represent standard error of mean (SEM); t-test, exact p-values are shown, p < 0.05 considered statistically significant (indicated by *). (B) Representative NTA curves of plasma EVs from naked mole-rats following normoxia or hypoxia treatment, respectively; (C) Western blotting analysis of EV markers for naked mole-rat plasma EVs, showing positive for CD63 and Flotillin-1; (D) Transmission electron microscopy (TEM) of plasma-EVs from naked molerats, showing representative images of the differently sized EVs; scale bar indicates 100 nm, black arrows highlight individual EVs.

Proteomic Profiles of Naked Mole-Rat Plasma EVs Change Following Hypoxia
Total proteomic content of plasma EVs from naked mole-rats treated for 4 h in normoxia or hypoxia ( Figure 2A) were compared using LC-MS/MS analysis of total EV protein content (n = 5 per group). Protein hits identified in plasma EVs for both groups are listed in Table 1, which indicates shared and unique hits. respectively; (C) Western blotting analysis of EV markers for naked mole-rat plasma EVs, showing positive for CD63 and Flotillin-1; (D) Transmission electron microscopy (TEM) of plasma-EVs from naked mole-rats, showing representative images of the differently sized EVs; scale bar indicates 100 nm, black arrows highlight individual EVs.

Proteomic Profiles of Naked Mole-Rat Plasma EVs Change Following Hypoxia
Total proteomic content of plasma EVs from naked mole-rats treated for 4 h in normoxia or hypoxia ( Figure 2A) were compared using LC-MS/MS analysis of total EV protein content (n = 5 per group). Protein hits identified in plasma EVs for both groups are listed in Table 1, which indicates shared and unique hits.  Inter-alpha-trypsin inhibitor heavy chain H4 G5AZM7 Alpha-amylase   Figure 2B) and pathway analysis for the EV Figure 3. KEGG pathway analysis for protein network analysis for EV total protein cargo, showing predicted protein networks annotating associated KEGG pathways for total protein of plasma EVs from (A) normoxia-treated mole-rats and (B) hypoxia-treated mole-rats.
The protein cargo of circulating plasma EVs from naked mole-rats from normoxia and hypoxia treatment was then compared for predictive protein networks. Overlapping KEGG pathways for normoxia and hypoxia EV cargo proteomes were complement and coagulation cascades, Staphylococcus aureus infection, cholesterol metabolism, pertussis, vitamin digestion and absorption, systemic lupus erythematosus, African trypanosomiasis, arrhythmogenic right ventricular cardiomyopathy, Chagas disease, and prion disease. KEGG pathways specific to the normoxia EV proteome were proteoglycan in cancer, proximal tubule bicarbonate reclamation, ferroptosis, glycolysis/gluconeogenesis, HIF-1 signalling pathway, thyroid hormone synthesis, phagosome, and carbon metabolism ( Figure 3 and Table 2A). No KEGG pathways were specific to the hypoxia EV proteome only.    The protein cargo of circulating plasma EVs from naked mole-rats from normoxia and hypoxia treatment was then compared for predictive protein networks. Overlapping KEGG pathways for normoxia and hypoxia EV cargo proteomes were complement and coagulation cascades, Staphylococcus aureus infection, cholesterol metabolism, pertussis, vitamin digestion and absorption, systemic lupus erythematosus, African trypanosomiasis, arrhythmogenic right ventricular cardiomyopathy, Chagas disease, and prion disease. KEGG pathways specific to the normoxia EV proteome were proteoglycan in cancer, proximal tubule bicarbonate reclamation, ferroptosis, glycolysis/gluconeogenesis, HIF-1 signalling pathway, thyroid hormone synthesis, phagosome, and carbon metabolism ( Figure 3 and Table 2A). No KEGG pathways were specific to the hypoxia EV proteome only.
The EV-cargo proteomes of normoxia-and hypoxia-treated animals had several overlapping local network clusters (STRING) (Table 2B), including complement and coagulation cascades, serpin, synapse pruning, fibrinolysis, fibrinogen, transferrin, hemopexin, cholesterol metabolism and efflux, regulation of lipoprotein l, triglyceride transport, glycophorin A, intermediate filament protein, and high-density lipoprotein particles. Local network clusters (STRING) specific for the normoxia EV proteome were related to carbon metabolism, apical plasma membrane urothelial plaque, pentose phosphate pathway, glycolytic process, apple domain, antithrombin-iii, kininogen, and mitochondrial glycoprotein. Three STRING pathways were specific for the hypoxia EV proteome and related to keratin type, blood coagulation, fibrin clot formation, alpha2-antiplasmin, intermediate filament protein, and keratinocyte activation (Table 2B).
Common and distinct KEGG and STRING pathways related to total proteomic content of naked mole-rat plasma EVs from normoxia and hypoxia groups, respectively, are further listed and summarised in Table 2A,B.

Citrullinated Protein Profiles of Plasma EVs under Normoxic Versus Hypoxic Conditions in Naked Mole-Rats
F95-enriched proteins identified in naked mole-rat plasma EVs were analysed by SDS-PAGE and silver staining ( Figure 4A). To compare citrullinated proteins from EV cargo of plasma from animals under normoxic and hypoxic conditions, fractions were subjected to LC-MS/MS analysis. A total of 29 protein hits were common between both groups, whereas 21 hits were identified as deiminated only in normoxia and 15 hits were found deiminated in the hypoxia plasma EVs only (n = 5 per group). The protein hits are listed in Table 3, highlighting common and specific hits, and protein interaction networks were created in STRING ( Figure 4B). Pathway analysis in STRING revealed a number of differences between the normoxia and hypoxia groups ( Figure 4C and Table 4), as summarised in the Venn diagram in Figure 4D. The PPI enrichment p-value for both networks was <1.0 × 10 −16 , indicating more interactions than expected for a random set of proteins of similar size, drawn from the genome, indicating that the proteins are at least partially biologically connected as a group. listed in Table 3, highlighting common and specific hits, and protein interaction networks were created in STRING ( Figure 4B). Pathway analysis in STRING revealed a number of differences between the normoxia and hypoxia groups ( Figure 4C and Table 4), as summarised in the Venn diagram in Figure 4D. The PPI enrichment p-value for both networks was <1.0 × 10 −16 , indicating more interactions than expected for a random set of proteins of similar size, drawn from the genome, indicating that the proteins are at least partially biologically connected as a group.   Table 3. F95-enriched proteins identified in plasma EVs from naked mole-rats treated under normoxic and hypoxic conditions, as assessed by LC-MS/MS analysis. Common and specific hits per group are indicated in the table; a tick (V) indicates whether the deiminated protein hit was identified in normoxia or hypoxia plasma EVs, or in both; blue highlighted rows indicate hits identified in the normoxia group only, and pink highlighted rows indicate hits identified in the hypoxia group only.

Protein ID Protein Name Normoxia Hypoxia
Hemoglobin subunit beta V V Table 3. Cont.

G5BUN4
Inter-alpha-trypsin inhibitor heavy chain H4 Desmosome and cell-cell adhesion mediated by cadherin V (3) Furthermore, KEGG analysis of the EV citrullinome highlighted four shared pathways between the groups and one specific pathway (oestrogen signalling) for the normoxia group, as shown in Figure 5.

PAD Isozyme Levels Differ in Naked Mole-Rat Brains from Normoxia and Hypoxia Groups
Using Western blotting analysis of brain protein lysates from normoxia-and hypoxia-treated naked mole-rats (n = 5 per group), protein levels of the five PAD isozymes (PAD1,2,3,4 and 6) were assessed ( Figure 6A-E). Blots were carried out based on crossreactivity with human PAD isozyme-specific antibodies (see Supplementary Figure S1 for phylogenetic analysis of naked mole-rat PAD isozymes in comparison with human PADs, confirming conservation of the corresponding PAD forms between humans and naked mole-rats). Overall, PAD2 showed no significant changes, with a possible trend for reduced levels (albeit high variability was observed between samples, not reaching statistical significance, and one outlier was observed), and PAD4 protein levels were statistically significantly reduced in the hypoxia-challenged brains (p = 0.0031), whereas PAD1, PAD3, and PAD6 levels showed an overall trend for elevation in the hypoxiachallenged brains (albeit not reaching statistical significance and with considerable individual variability observed). Histone H3 citrullination (CitH3) was also assessed by Western blotting as a representative marker for ETosis and epigenetic changes via this post-translational modification (using the anti-histone H3 citrulline R2 + R8 + R17 antibody), indicating no significant changes in the hypoxia brains compared to the controls when normalised against the beta-actin loading control ( Figure 6F). Furthermore, the citH3 antibody showed a number of unspecific bands at higher molecular weight, whereas the approximately 17 kDa band shown in Figure 6F was used for densitometry analysis and was at the expected size reported for citH3.

PAD Isozyme Levels Differ in Naked Mole-Rat Brains from Normoxia and Hypoxia Groups
Using Western blotting analysis of brain protein lysates from normoxia-and hypoxiatreated naked mole-rats (n = 5 per group), protein levels of the five PAD isozymes (PAD1,2,3,4 and 6) were assessed ( Figure 6A-E). Blots were carried out based on cross-reactivity with human PAD isozyme-specific antibodies (see Supplementary Figure S1 for phylogenetic analysis of naked mole-rat PAD isozymes in comparison with human PADs, confirming conservation of the corresponding PAD forms between humans and naked mole-rats). Overall, PAD2 showed no significant changes, with a possible trend for reduced levels (albeit high variability was observed between samples, not reaching statistical significance, and one outlier was observed), and PAD4 protein levels were statistically significantly reduced in the hypoxia-challenged brains (p = 0.0031), whereas PAD1, PAD3, and PAD6 levels showed an overall trend for elevation in the hypoxia-challenged brains (albeit not reaching statistical significance and with considerable individual variability observed). Histone H3 citrullination (CitH3) was also assessed by Western blotting as a representative marker for ETosis and epigenetic changes via this post-translational modification (using the anti-histone H3 citrulline R2 + R8 + R17 antibody), indicating no significant changes in the hypoxia brains compared to the controls when normalised against the beta-actin loading control ( Figure 6F). Furthermore, the citH3 antibody showed a number of unspecific bands at higher molecular weight, whereas the approximately 17 kDa band shown in Figure 6F was used for densitometry analysis and was at the expected size reported for citH3.

The Brain Citrullinome of Naked Mole-Rats Is Increased Following Hypoxic Challenge
F95-enriched proteins isolated from brains of naked mole-rats that had been exposed to hypoxia, relative to the normoxia control group, were analysed by SDS-PAGE and silver staining ( Figure 7A) and by LC-MS/MS for identification of citrullinome brain signatures ( Figure 7B; Supplementary Tables S1 and S2 for full protein lists per group). A total of 852 protein hits were identified in the normoxic brains, whereas 1222 hits were identified in the hypoxic brains. Therein, 34 deimination hits were specific to the normoxic brains and 245 hits were specific to the hypoxic brains (n = 5 brains per group). STRING analysis was used for identification of protein-protein interaction networks and associated pathways ( Figure 7B,C). Local network clusters (STRING) and KEGG pathways identified for all F95-enriched hits in the normoxic brains and hypoxic brains are listed in Tables 5 and 6, respectively; a summary of overlapping and specific protein hits and pathways is presented in Figure 7D. The PPI enrichment p-value for both networks was <1.0 × 10 −16 , indicating more interactions than expected for a random set of proteins of similar size drawn from the genome, indicating that the proteins are at least partially biologically connected as a group. Protein levels were assessed in n = 5 brains per group and normalised against beta-actin protein levels; exact p-values are indicated (t-test; * indicates significance at p < 0.05; circles represent normoxia and squares hypoxia brain samples, respectively) and the error bar represents SD.

The Brain Citrullinome of Naked Mole-Rats Is Increased Following Hypoxic Challenge
F95-enriched proteins isolated from brains of naked mole-rats that had been exposed to hypoxia, relative to the normoxia control group, were analysed by SDS-PAGE and silver staining ( Figure 7A) and by LC-MS/MS for identification of citrullinome brain signatures ( Figure 7B; Supplementary Tables S1 and S2 for full protein lists per group). A total of 852 protein hits were identified in the normoxic brains, whereas 1222 hits were identified in the hypoxic brains. Therein, 34 deimination hits were specific to the normoxic brains and 245 hits were specific to the hypoxic brains (n = 5 brains per group). STRING analysis was used for identification of protein-protein interaction networks and associated pathways ( Figure 7B,C). Local network clusters (STRING) and KEGG pathways identified for all F95-enriched hits in the normoxic brains and hypoxic brains are listed in Tables 5 and 6, respectively; a summary of overlapping and specific protein hits and pathways is presented in Figure 7D. The PPI enrichment p-value for both networks was <1.0 × 10 −16 , indicating more interactions than expected for a random set of proteins of similar size drawn from the genome, indicating that the proteins are at least partially biologically connected as a group. Protein-interaction networks for all deiminated protein candidates identified in naked mole-rat brains following normoxia or hypoxia (brain citrullinome). (C) STRING pathway analysis results for KEGG and GO terms for the full brain citrullinome following normoxia or hypoxia treatment. (D) Venn diagram summarising deimination/citrullination hits (F95) and shared and specific pathways for the citrullinome between normoxic and hypoxic brains (n = 5 brains per group in all experiments). Table 5. Local network cluster STRING networks for all F95-enriched proteins identified in naked mole-rat brains following normoxia and hypoxia treatment. A tick (V) indicates whether the pathway was identified in the respective group; normoxia-specific pathways are highlighted in blue and hypoxia-specific pathways in pink. A full list of the protein hits underlying the network analysis is provided for both groups in Supplementary Tables S1 and S2. (B) Protein-interaction networks for all deiminated protein candidates identified in naked mole-rat brains following normoxia or hypoxia (brain citrullinome). (C) STRING pathway analysis results for KEGG and GO terms for the full brain citrullinome following normoxia or hypoxia treatment.

STRING NETWORKS IN
(D) Venn diagram summarising deimination/citrullination hits (F95) and shared and specific pathways for the citrullinome between normoxic and hypoxic brains (n = 5 brains per group in all experiments). Table 5. Local network cluster STRING networks for all F95-enriched proteins identified in naked mole-rat brains following normoxia and hypoxia treatment. A tick (V) indicates whether the pathway was identified in the respective group; normoxia-specific pathways are highlighted in blue and hypoxia-specific pathways in pink. A full list of the protein hits underlying the network analysis is provided for both groups in Supplementary Tables S1 and S2.  Table 6. KEGG pathways for all F95-enriched proteins in naked mole-rat brains following normoxia or hypoxia treatment. A tick (V) indicates whether the pathway was identified; the observed gene count for each pathway is indicated in brackets. Pathways identified in normoxia brains only are highlighted in blue, and those identified in hypoxia brains only in pink. A full list of the protein hits underlying the network analysis is provided for both groups in Supplementary Tables S1 and S2.

Metabolic pathways
Oxidative phosphorylation V

HYPOXIA (Observed Gene Count)
Histidine metabolism V (4) Cushing syndrome V (11) Growth hormone synthesis, secretion, and action V (9) Purine metabolism V (10) Fatty acid metabolism V (6) Hepatitis B V (11) In addition, a protein interaction network analysis was specifically performed for F95 hits identified only in the normoxia or hypoxia group. Protein networks created for these protein hits identified as deiminated/citrullinated only in the normoxia or hypoxia group therefore excluded all common identified deiminated/citrullinated proteins between the two groups ( Figure 8). Using this analysis approach, no significant functional enrichments were identified for KEGG pathways in the normoxia F95-positive-hits-only network ( Figure 8A; PPI enrichment p-value 0.0326), whereas 50 KEGG pathways were identified for the hypoxia-specific-only F95 positive hits ( Figure 8B; PPI enrichment p-value <1.0 × 10 −16 ); these are listed in Table 7, also indicating the observed gene count for each pathway.   Figure 8B); the observed gene count for each pathway is listed; (*) are unique to this list only.    Figure 8B); the observed gene count for each pathway is listed; (*) are unique to this list only.

Discussion
The current study assessed circulating EV signatures (whole proteome and citrullinated proteins) and changes relating to protein citrullination in the brains of naked mole-rats following hypoxia treatment, compared with normoxia-treated control animals. We identified large-scale changes in EV proteome signatures and the brain citrullinome between normoxia-and hypoxia-treated animals.

EV Proteomic Signatures in Normoxia-and Hypoxia-Treated Naked Mole-Rats
Some shift was observed in plasma-EV profiles released between normoxia and hypoxia treated animals, with an overall decrease of plasma-EVs following hypoxia, which was significant for smaller and medium sized EVs (<100 nm; 101-200 nm), and a trend for reduction in larger EVs (201-1000 nm) was also observed. Increased EV release overall is linked to hypoxia in other models, including human [69,70], while it has been suggested that an increase in larger EVs is linked to inflammatory responses in human disease [71]. Notably, we observed a trend for decrease in numbers of EVs, including larger EVs above 201 nm. The changes in naked mole-rat plasma-EVs with hypoxia may indicate changes in cargo-export and modulated inflammatory responses to hypoxia and/or protective antiinflammatory systemic responses in this species, which shows unusual hypoxic tolerance. This possibility warrants further investigation. Indeed, an interplay between hypoxia and EVs has been reported for many inflammatory conditions, and hypoxia creates an inflammatory environment in other species [72,73]. Hypoxia-induced EVs in other species may also contribute to disease pathogenesis via cargo transfer (reviewed in [69]). However, it must be noted that cargo transfer of proteins, DNA, RNA and non-coding RNA species may vary between the different types of EVs released, and our current study assesses protein content only. Interestingly, we observe a higher protein content (including citrullinated proteins) and a higher number of KEGG pathways relating to the total proteome of EVs from normoxia, versus hypoxia treated animals. Therefore, our findings indicate that in this hypoxia tolerant animal, circulating EV proteomic content is reduced in hypoxia. This could be indicative of systemic responses and redirection of resources in response to hypoxic insult.
To understand differences in pathobiological and physiological pathways associated with the EV proteomes of normoxia versus hypoxia, protein-protein interaction network analysis was carried out. Several KEGG pathways were identified as specific to the total EV proteome of normoxic animals, including phagosome and carbon metabolism, ferroptosis, proteoglycan in cancer, proximal tubule bicarbonate reclamation, glycolysis/gluconeogenesis, thyroid hormone synthesis, and HIF-1 signalling pathway, whereas no KEGG pathways were specific to the hypoxia EV proteome only. Overlapping KEGG pathways for normoxia and hypoxia EV proteome were complement and coagulation cascades, Staphylococcus aureus infection, pertussis, systemic lupus erythematosus, African trypanosomiasis, Chagas disease, prion disease, arrhythmogenic right ventricular cardiomyopathy, cholesterol metabolism, and vitamin digestion and absorption. This indicates that a number of critical immune-related pathways are influenced by EV-mediated proteome transport, and that there is furthermore a higher number of metabolic pathways influenced by EV-transported protein cargo during normoxia compared to hypoxia.
In addition, local network cluster analysis (STRING) indicated that the EV proteomes in normoxia and hypoxia had a number of overlapping STRING pathways, including complement and coagulation cascades, serpin, synapse pruning, fibrinolysis, fibrinogen, transferrin, hemopexin, cholesterol metabolism and efflux, regulation of lipoprotein l, triglyceride transport, glycophorin A, intermediate filament protein, and high-density lipoprotein particles. STRING pathways specific to the normoxia EV proteome were related to carbon metabolism, apical plasma membrane urothelial plaque, pentose phosphate pathway, glycolytic process, apple domain, antithrombin-iii, kininogen, and mitochondrial glycoprotein. Three STRING pathways were specific to the hypoxia EV proteome and related to keratin type, blood coagulation, fibrin clot formation, alpha2-antiplasmin, intermediate filament protein, and keratinocyte activation. As with the KEGG pathway analysis, this analysis indicates that critical pathways for energy and metabolism are influenced by EV protein transport in normoxia, and furthermore that EV protein cargo participates in key immune pathways, as well as metabolic pathways, in normoxia and in response to hypoxia. Furthermore, pathways linked to blood coagulation and fibrin clot formation identified here in the hypoxia EV cargo correlate with previous studies identifying coagulopathies in hypoxic events and infection [74,75].

EV Citrullinome Signatures in Normoxia-and Hypoxia-Treated Naked Mole-Rats
The plasma EV citrullinome was also assessed based on the enrichment of citrullinated/deiminated proteins isolated from the EVs. KEGG pathways for citrullinated EV protein cargo were found to be fewer in the hypoxia compared with the normoxia group, with four common pathways, all linked to immunity (complement and coagulation cascades, African trypanosomiasis, S. aureus infection, and systemic lupus erythematosus); no further pathways were identified for the hypoxia EV citrullinome, and the oestrogen signalling pathway was specific to the normoxia EV citrullinome only. Therefore, some main pathways in immune responses are influenced by deimination in circulatory EV cargo transport both in hypoxia and normoxia conditions. For example, oestrogen signalling has multifaceted functions in health and disease, and beyond endocrinal roles also regulates epigenetic mechanisms [76,77].
STRING pathways specific for the citrullinome of hypoxia EVs were desmosome and cell-cell adhesion mediated by cadherin. This related to the differences observed in target proteins of deimination in the normoxia versus hypoxia EV citrullinome. In hypoxia, several proteins with roles in cytoskeletal function were identified, including tubulin and plectin, which is a key cytoskeleton interlinking molecule with multifaceted roles in mediating intermediate filament network functions in physiological and pathobiological processes [78,79]. EV biogenesis may furthermore be affected by deimination of cytoskeletal proteins [80]. Glyceraldehyde-3-phosphate dehydrogenase, identified in the hypoxia EV citrullinome, is crucial for the glycolysis pathway and is involved in apoptosis and hypoxic responses [81,82]. It has previously been reported as deiminated, including in brain cancer [51], and such a modification may contribute to its multifaceted functions, including the hypoxic response. Apoptosis facilitator Bcl-2-like protein 14 was also identified in the hypoxia EV citrullinome, highlighting roles for deimination in regulating apoptotic processes. Serine/threonine-protein kinases PDIK1L and PCTAIRE-3 were identified in the hypoxia EV-citrullinome. PCTAIRE-3 is mainly expressed in differentiated neurones and has been linked to Alzheimer's disease pathology, both relating to amyloid precursor protein-dependent Alzheimer's pathology via phosphorylation [83] and by regulation of tau phosphorylation [84]. Multidrug resistance-associated protein 7 was identified in the hypoxia EV citrullinome; although deimination needs to be studied in this context, multidrug resistance is linked to hypoxia [85]. G-protein-coupled receptor 19 was identified in the EV citrullinome in hypoxia, and although effects of its deimination have not been reported, it has been linked to neonatal HI-induced brain damage [86]. Fatty acid-binding protein, identified in the EV-citrullinome in hypoxia, is linked to infection and injury in hypoxic challenge [87,88]. Ventricular zone-expressed PH domain-containing protein was identified in the hypoxia EV citrullinome. Although originally identified as having roles in neural cell differentiation, it is a multifaceted adaptor protein with roles in signal transduction in numerous physiological and pathological processes, including via TGF-beta, BMP, and SMAD signalling [89]. VEPH1 has been reported in EV cargo of various cell types, including in hypoxia and normoxia [90], but hitherto not in deiminated form.

Changes in PADs and the Brain Citrullinome
In the hypoxic brains, a trend for elevation in PAD isozymes PAD1, PAD3, and PAD6 was observed (albeit not reaching statistical significance), whereas PAD2 was not significantly changed (but showed some trend for reduction considering outliers) and PAD4 protein levels were significantly reduced. This modulation in PAD protein levels may link to some of the changes observed in the elevated and modified brain citrullinome in response to hypoxia. It must be considered that the PAD isozymes differ somewhat in preference of target proteins, and therefore a change in expression of the different isoforms may contribute to changes to the brain citrullinome. Upregulation of PAD1 and PAD3 has been reported in response to hypoxia in malignant glioma cells [58]. PAD1 has been strongly associated with skin physiology and skin diseases [91] and embryo development [36,92], and novel roles in breast cancer metastasis and epithelial-mesenchymal transition have been identified [93]. However, PAD1 has not received attention in relation to the CNS. PAD3 has previously been linked to CNS regeneration and remodelling, including in hypoxic injury, as well as stem-ness, neural stem cell death, and survival [43,44,51,94,95]. PAD3 has also been linked to different invasion abilities in brain cancer (glioblastoma multiforme) [52]. Interestingly, both PAD1 and PAD3 have recently been reported as upregulated by hypoxia and regulating glycolysis and cancer cell proliferation by citrullinating pyruvate kinase [96]. Glycolysis and pyruvate pathways were deimination-associated in both the normoxia and hypoxia brains. The role for PAD6 is mainly linked to developmental processes, including oocyte formation and embryo pre-implantation [97][98][99], whereas roles for PAD6 in the CNS have not been investigated. PAD2 is considered the most evolutionary conserved PAD isoform, and its elevation has been linked to neurodegenerative diseases [39,40,[46][47][48][49][53][54][55]. Therefore, the observation of little change in this isozyme and possible trend for its somewhat reduced expression in the naked mole-rat brain following hypoxia is of interest. PAD2 is upregulated in vitro in astrocytes in response to hypoxia [56] and in malignant glioma cells in response to hypoxia [58]. The lack of any considerable changes in PAD2 may therefore indicate a protective mechanism in response to hypoxic insult in the naked mole-rat. Besides roles in epigenetic regulation via histone citrullination, a role for PAD4 in ETosis (NETosis and METosis) is part of a double-edged sword in responses to pathological stimuli, including in hypoxia, and may cause damage of self. PAD4 has been shown to be activated in response to hypoxia in cancer [66], linked to hypoxia in autoimmune disease [100], and upregulated in malignant glioma cells in hypoxia challenge [58] as well as in neurodegeneration [55]. Therefore, our current finding that PAD4 was significantly downregulated in the naked mole-rat brain following hypoxia indicates that this may be a protective mechanism and contribute to its unusual hypoxic tolerance.
Changes in PAD regulation may possibly be reflected in the observed changes of the brain citrullinome. The brain citrullinome was isolated from whole-brain-cell lysates using F95 enrichment in conjunction with LC-MS/MS analysis, and we report considerable increases in the number of deiminated protein hits in the hypoxia brains compared to normoxia brains. Whereas 107 KEGG pathways based on these hits were shared between the brain citrullinome of the hypoxia and normoxia brains, 28 KEGG pathways were specific for the total citrullinome of the hypoxia brains and 2 KEGG pathways to the whole citrullinome of the normoxia brains. Furthermore, a total of 50 KEGG pathways were related to protein networks created based on citrulline-specific proteins identified only in the hypoxia brains, whereas no KEGG pathways were found in similar networks created based on citrullinated proteins identified only in the control normoxia brains (this excluded any overlapping targets between the groups from the protein network analysis).
KEGG pathways relating to the hypoxia brain citrullinome related to a range of physiological and pathobiological mechanisms. This included the mRNA surveillance pathway, but also pathways related to hormonal control and metabolism, including oestrogen signalling pathway; GnRH secretion and GnRH signalling pathway; parathyroid hormone synthesis, secretion, and action; growth hormone synthesis, secretion, and action; mineral absorption; renin secretion; bile secretion; amino sugar and nucleotide sugar metabolism; fatty acid degradation and fatty acid metabolism; insulin signalling pathway; alcoholism; Cushing's syndrome; purine metabolism; and histidine metabolism.
KEGG pathways for the hypoxia brain citrullinome relating to hemostasis and immunological/inflammatory pathways included platelet activation, apelin signalling pathway, human T-cell leukemia virus 1 infection, hepatitis B, chemokine signalling pathway, apoptosis, inflammatory mediator regulation of TRP channels, cellular senescence, renal cell carcinoma, and MAPK signalling pathway. Deimination of the apelin signalling pathway in hypoxia may be of considerable interest; deimination has also been identified in pre-motor Parkinson's disease [55]. Although apelin signalling has multifaceted physiological functions, including in the CNS, it is associated with ischaemia as well as neovascularisation events such as retinopathies, tumours, and retinopathies [101], and a range of neurodegenerative diseases [102,103].
Analysis of local network clusters (STRING) identified 19 network clusters shared between the brain citrullinome of the hypoxia and normoxia brains, whereas five STRING network clusters were specific to the total citrullinome of the hypoxia brains (proteasome ribosome and ribosomal protein L23; ubiquinone and zinc-finger domain; mixed, incl. regulation of actin cytoskeleton and actin; septin) and one (GroEL-like equatorial domain superfamily) was specific to the whole citrullinome of the normoxic brains. The deimination of septin-related pathways may be of interest in hypoxic brains, as these play roles in many key cellular processes, regulate protein stability, and also play roles in protecting HIF-1a from degradation [104].

EV Signatures and the Brain Citrullinome in Relation to HIF-1 Regulation
In relation to hypoxia, it is interesting to observe that the HIF-1-signalling pathway was identified in whole EV proteome of normoxia-but not hypoxia-challenged animals. Furthermore, the HIF-1 KEGG pathway came up in the F95-enriched networks of both normoxic and hypoxic brains, although in the hypoxic brains there were 19 underlying genes associated, but only 17 for the normoxia brains. HIF-1 is a master regulator of oxygen homeostasis, contributes to hypoxia adaption of the naked mole-rat [105], and is highly expressed endogenously in naked mole-rats due to mutation in the VHL (Von Hippel-Lindau disease tumor suppressor) domain [106]. Naked mole-rats have a very high expression of HIF mRNA and protein in normoxia relative to mice [105], and interestingly, respond to acute hypoxia of 4 h by decreasing HIF expression in the brain [105]. In other models, HIF-1 signalling has been linked to regulation of EV release during hypoxia [107], including microvesicles (medium/larger EVs) [108] and small EV (exosome) release [109]. Although roles for deimination in regulating the HIF-1 pathway remain to be fully understood, including in the naked mole-rat, the KEGG HIF-1 pathway has previously been linked to deimination in other animal models of hypoxia resistance, including cetaceans [110]. In human cancer cell models, PAD4 is induced by hypoxia in a HIF-dependent manner, performing histone citrullination required for HIF-dependent transcriptional responses to hypoxia as well as for tumour vascularisation [66]. In addition, in malignant glioma cells, PAD1, 2, 3, and 4 were shown to be upregulated in a hypoxia-inducible factor-1-dependent manner at the mRNA level, albeit no verification of protein citrullination is provided in that study [5]. Collectively, a role for PADs in the hypoxic response via HIF-1 is of considerable interest, and findings from the current study point to differences in this interplay in the naked mole-rat, possibly reflecting some of the unusual hypoxic-resistance capacities of these animals.

Histones in EV Signatures and the Brain Citrullinome
Interestingly, in the whole EV proteome, histones H2A, H2B, and H3 were identified as EV cargo in the hypoxia group, whereas only histone H4 was identified as EV cargo in the normoxia group. When assessing deiminated/citrullinated EV cargo, deiminated histone H2A, H2B, and H3 were found in EVs of both groups, and H3.3 as deiminated in the hypoxia EVs only. In the brains, deiminated histones identified by proteomics in both groups included H2A, whereas in the hypoxia group H1.1 and H2B were also identified as deimination candidates, and H3 was identified in the normoxia group only. Using Western blotting, histone H3 deimination was further assessed in the brains of both normoxic and hypoxic brains as an indicator of ETosis, with no significant changes observed. This indicates that H3 deimination may not increase in hypoxia in naked mole-rat brains, contrary to what has been observed in other injury models of CNS and hypoxia, including in a mouse model of hypoxic ischaemic encephalopathy [44], acute ischaemic stroke [60] and spinal cord injury [43], as well as pre-motor Parkinson's disease [55]. This also correlates with PAD4, which is considered the main driver of NETs in inflammation [63,111], being found to be significantly reduced in the hypoxic naked mole-rat brains in this current study, but having been, for example, linked with NETosis in neurological deficits following traumatic brain injury in other models [61]. PAD4 is also linked to hypoxia-induced autophagy [65]. Furthermore, PAD2, which is also linked to histone deimination [112], was observed here to not be affected by hypoxia, with a possible trend for some reduced levels (albeit not statistically significant) in hypoxic naked mole-rat brains. Conversely, PAD2 is upregulated in vitro in astrocytes of other species in response to hypoxia [56]. Roles for deimination of other histones including H1.1 and H2B may warrant further exploration in this context. In addition, whether circulatory histones in EVs observed here play systemic roles in hypoxia responses needs further assessment. The naked mole-rat has a particularly stable epigenome, which may contribute to its unusual physiological responses both to hypoxia as well as cancer resistance and longevity [113].
In summary, our findings indicate that circulating EV signatures are modified and proteomic content is reduced in response to acute hypoxic conditions in the naked mole-rat, including the circulating EV citrullinome, whereas the brain citrullinome is elevated. This is further reflected in changes of PAD isozyme expression in the hypoxic brains, which showed a trend for elevation of some PADs (PAD1, PAD3, and PAD6) and a trend for reduction in other PADs (PAD2 and PAD4) under hypoxic conditions, further indicating a possible selective role for PAD-isozymes and associated isozyme-specific deimination in hypoxia response and tolerance. The current study used acute hypoxia challenge, but longer periods of hypoxia exposure may also be of interest in the naked mole-rat model in future studies. It has to be noted that the current study focussed on EV proteome signatures only (total proteome and the citrullinome) based on the presence/absence of identified proteins in the respective experimental groups, but that EVs carry also a range of other cargo, including lipids, DNA, mRNA, miRNA, and non-coding RNA species. Such EV signatures may further contribute to hypoxia resistance and will remain subject to further investigation.

Animals
Naked mole-rats were group-housed in interconnected multi-cage systems at 30 • C and 21% O 2 in 50% humidity and with a 12L:12D light cycle. Animals were fed fresh tubers, vegetables, fruit, and Pronutro cereal supplement ad libitum. Animals were not fasted prior to experimental trials. All experimental procedures were approved by the University of Ottawa Animal Care Committee (protocol #3444) in accordance with the Animals for Research Act and by the Canadian Council on Animal Care.
Animals (1-2-year-old subordinate males and females weighing 40-60 g) were exposed to either 21% O 2 (normoxia) or 7% O 2 (hypoxia) for 4 h. Each experimental group was comprised of 10 animals. Following treatment, the animals were sacrificed by cervical dislocation followed by rapid decapitation. Blood was collected in heparinised syringes and plasma was extracted by spinning whole blood at 1500 rpm for 15 min. Plasma aliquots were then frozen at −80 • C until analysis. Whole brains were rapidly extracted on ice and similarly frozen in liquid nitrogen, and then stored at −80 • C until analysis.

Extracellular Vesicle Isolation and Characterisation by NTA Analysis, Western Blotting, and Transmission Electron Microscopy
EVs were isolated from individual mole-rat plasma using differential centrifugation as previously described [32]. Per animal, 100 µL of plasma was added to 400 µL DPBS and then centrifuged at 4000× g for 30 min. The supernatant was collected and spun at 100,000× g for 1 h at 4 • C for collection of total EVs. The EV-enriched pellet was resuspended in 500 µL DPBS and centrifuged again at 100,000× g for 1 h at 4 • C. The supernatant was then discarded and the EV pellet was diluted in 100 µL DPBS.
For quantification of EVs by nanoparticle tracking analysis (NTA) in the individual plasma samples, 10 µL of diluted EV pellet was added to 990 µL DPBS and applied to the NS300 Nanosight (Malvern Panalytical Ltd., Malvern, UK) at syringe pump speed 50. Particles were recorded four times for 1 min per sample at camera level 9, and post-analysis was carried out at threshold level 5 with 40-60 particles per window. The four readings were then averaged per sample using the NTA software (version 3, Malvern, UK). EVs were further analysed by Western blotting for two EV surface markers, CD63 (Abcam, 1/1000) and flotillin-1 (Abcam 1/1000; see below for further Western blotting details), and also visualised using transmission electron microscopy (TEM) according to methods previously described in [32]. Together, these approaches meet the minimum requirements for EV characterisation as per the guidelines of the International Society for Extracellular Vesicle Research [114].

Protein Isolation from Brain Tissue and Western Blotting
Proteins were extracted from brain tissue from treated animals (n = 5 per treatment group). Whole brains were homogenised in RIPA+ buffer (Sigma-Aldrich, Gillingham, UK, containing 10% protease inhibitor cocktail, Sigma-Aldrich) in 2 mL Eppendorf tubes on ice using a Mini Handheld Homogeniser (Kimble, DWK Life Sciences, VWR International). The homogenate was then gently pressed through a 23G needle into fresh Eppendorf tubes on ice, followed by gently pipetting up and down to eliminate any tissue clots. For each brain (400 mg tissue), 2 mL of RIPA+ buffer were used. The homogenates were then incubated on a roller for 1.5 h at 4 • C, pipetted up and down at regular intervals, and thereafter spun down at 16,000× g for 30 min at 4 • C for collection of isolated proteins. The extracted proteins were aliquoted and immediately frozen at −80 • C.
For SDS-PAGE and Western blotting, a 100 µL aliquot per sample was diluted with 100 µL 2× reducing Laemmli sample buffer (BioRad; containing 5% β-mercaptoethanol, Sigma-Aldrich) and boiled for 5 min at 100 • C. A 5 µL aliquot per sample was then applied to 4-20% TGX gels (BioRad, Watford, UK). SDS-PAGE was carried out at 165 V for 52 min. Gels were then transferred for Western blotting analysis using semi-dry transfer (1 h at 15V), and even protein transfer was assessed by PonceauS red stain (Sigma-Aldrich). The membranes were blocked in 5% bovine serum albumin (BSA, Sigma-Aldrich) in TBS-T for 1 h at room temperature (RT) and incubated in primary antibodies overnight at 4 • C on a shaking platform.

F95-Enrichment for Deiminated Proteins from Plasma EVs and Brain Tissue
To identify deiminated/citrullinated proteins in plasma EVs and in brain tissue, deiminated proteins were enriched using the F95 pan-citrulline antibody (MABN328, Merck) in conjunction with the Catch-and-Release Immunoprecipitation Kit (Merck). For brain samples, protein extracts from 5 brains were pooled per experimental group. For EVs, EV isolates from 5 animals per group were pooled. Immunoprecipitation was carried out on agarose columns together with the F95 antibody and the affinity ligand overnight at 4 • C on a rotating platform. Proteins were eluted according to the manufacturer's instructions (Merck, Watford, UK) and subjected either to SDS-PAGE and silver staining, or to LC-MS/MS analysis for identification of protein hits.

Silver Staining of Proteins and F95-Enriched Proteins
Total proteins of EVs and F95-enriched protein eluates from plasma EVs and brains were diluted 1:1 in 2× reducing Laemmli sample buffer, boiled for 5 min at 100 • C, and separated on 4-20% TGX gels (BioRad) for 52 min at 165 V. Following electrophoresis, the gels were silver stained using the BioRad Silver Stain Plus Kit according to the manufacturer's instructions.

LC-MS/MS Proteomic Analysis
In-gel digestion was used for LC-MS/MS analysis, carried out by the Cambridge Centre for Proteomics (University of Cambridge, Cambridge, UK). Normoxic and hypoxic samples were subjected to LC-MS/MS analysis to determine (a) whole protein content of the plasma EVs, (b) F95-enriched protein eluates from the plasma EVs, and (c) F95-enriched protein eluates from the brains. The samples were prepared 1:1 in reducing Laemmli sample buffer, boiled and run 0.5 cm into a 10% TGX gel (BioRad), and then cut out as one whole band per sample (whole EV protein, F95-enriched EV proteins, F95-enriched brain proteins for normoxia versus hypoxia groups). Proteomic analysis was carried out by the Cambridge Centre for Proteomics (Cambridge, UK) according to previously described methods [32], and hits were assessed against the naked mole-rat protein database CCP_Heterocephalus_glaber_20190911 (21,449 sequences; 10,466,552 residues). In addition, a common contaminant database was also searched (cRAP 20190401; 125 sequences; 41,129 residues). Protein scores are derived from ion scores as a non-probabilistic basis for ranking protein hits; individual ion scores > 30 indicated identity or extensive homology (p < 0.05).

Protein Interaction Network Analysis
To identify local network clusters (STRING), gene ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways for proteins from EVs (both total protein content and deiminated protein content), as well as deiminated hits from brain tissue, STRING analysis was used (https://string-db.org/, accessed on 10 March 2022). Predicted protein interaction networks were built based on hits identified from the LC-MS/MS analysis using the protein IDs and organism choice Heterocephalus glaber in the STRING software. For protein lists, "multiple proteins" was selected, confidence set at "medium," and network interaction connecting lines were based on known and predicted interactions. Protein networks were annotated for pathway analysis and data were exported as labelled network images for KEGG and/or Excel files for KEGG pathways and STRING network clusters.

Statistical Analysis
To compare datasets from hypoxia-versus normoxia-treated groups, GraphPad Prism version 7 was used. T-tests were used to determine significance between groups for densitometry readings from Western blotting analysis (n = 5 per experimental group) and for the EV NTA count (n = 10 per experimental group). NTA analysis was carried out using the NTA software (version 3, Malvern Panalytical, UK), and is based on 4 reads per sample and presented as an average read from the samples (black line), with the standard deviation represented as a red line. Histograms for NTA analysis are based on NTA results from 10 samples per experimental group (normoxia vs. hypoxia). Statistical significance was regarded as p < 0.05. STRING analysis was carried out with medium confidence in STRING (https://string-db.org/, accessed on 10 March 2022).

Conclusions
This study used the naked mole-rat model to assess CNS-related responses of PADs in hypoxic protection/tolerance and identify whether circulating EV signatures could reveal a fingerprint for whole-body hypoxia-tolerant responses. Furthermore, citrullination-specific signatures in EVs were assessed in animals under normal versus hypoxic conditions. Our findings highlight novel roles for PADs in regulating brain-associated responses to acute hypoxia challenge and modifications in circulatory EV proteome signatures, possibly indicating a shift to re-directing resources systematically in response to acute hypoxic challenge in this hypoxia-tolerant species.