The intra-mitochondrial O-GlcNAcylation system acutely regulates OXPHOS capacity and ROS dynamics in the heart

Protein O-GlcNAcylation is increasingly recognized as an important cellular regulatory 3 mechanism, in multiple organs including the heart. However, the mechanisms leading to O- 4 GlcNAcylation in mitochondria and the consequences on their function remain poorly understood. In this study, we used an in vitro reconstitution assay to characterize the intra-mitochondrial O- 6 GlcNAc system without potential cytoplasmic confounding effects. We compared the O- 7 GlcNAcylome of isolated cardiac mitochondria with that of mitochondria acutely exposed to 8 NButGT, a specific O-GlcNAcylation inducer. Amongst the 409 O-GlcNAcylated mitochondrial 9 proteins identified, 191 displayed increased O-GlcNAcylation in response to NButGT. This was 10 associated with enhanced Complex I (CI) activity, increased maximal respiration in presence of CI 11 substrates, and a striking reduction of mitochondrial ROS release, which could be related to O- 12 GlcNAcylation of subunits within the NADH dehydrogenase module of CI. In conclusion, our 13 work underlines the existence of a dynamic mitochondrial O-GlcNAcylation system capable of 14 rapidly modifying mitochondrial function. in vitro reconstitution assay, which allows to isolate mitochondria from non-mitochondrial O- GlcNAc cycling systems, our study shows that O-GlcNAc cycling enzymes are present in functionally relevant amounts in the mitochondrial compartment and can trigger broad and rapid changes in protein O-GlcNAcylation which are highly reminiscent of the mitochondrial O- GlcNAcylation profile observed in vivo . Our proteomic workflow confirms previously reported O- GlcNAc-modified proteins, and identifies several novel targets related to energy metabolism, and multiple other facets of mitochondrial biology. Importantly, we show that acute hyper-O- GlcNAcylation increases maximal respiratory capacity, and drastically reduces ROS likely through a complex I-mediated mechanism.

GlcNAcylation system. 10 In this study, we therefore took advantage of an in vitro reconstitution assay to characterize 11 the intra-mitochondrial O-GlcNAcylation system in isolated cardiac mitochondria. Our results 12 confirm the presence of a fully functional and dynamic O-GlcNAc cycling system in these 13 organelles. Using comparative O-GlcNAc proteomics (O-GlcNAcylomics), we provide evidence 14 that the local mitochondrial O-GlcNAcylation system can trigger broad and rapid changes in 15 protein O-GlcNAcylation, which are highly reminiscent of the mitochondrial O-GlcNAcylation 16 profile observed in vivo. Importantly, we also reveal that acute hyper-O-GlcNAcylation increases 17 maximal respiratory capacity, and drastically reduces ROS release though a complex-I mediated 18 mechanism, illustrating the capacity of this system to rapidly modify mitochondrial function.

In vitro reconstitution assay allows to target the intra-mitochondrial O-GlcNAcylation
In order to characterize the mitochondrial O-GlcNAcylation system without the potential 4 confounding effects of the O-GlcNAcylation in other cellular compartments, we devised an in vitro 5 reconstitution assay in which isolated cardiac mitochondria were acutely exposed (i.e. 30 min) to 6 the OGT substrate UDP-GlcNAc in presence or absence of the OGA inhibitor NButGT, with the 7 goal of inducing rapid changes in protein O-GlcNAcylation levels. 8 As some controversy exists regarding the expression of the mOGT isoform in murine tissues 9 (Trapannone 2016), and the presence of sufficient OGA in the mitochondrial compartment 10 (Banerjee 2015), the presence of these enzymes was first verified by immunoblotting in whole 11 lysates from crude and Percoll-purified mitochondria (Fig 1A). Specific protein markers were 12 firstly assessed to confirm the purity of the different purified fractions ( Fig 1B). As expected, the 13 mitochondrial marker TOM20 was highly enriched in the crude and Percoll-purified mitochondrial 14 fractions, while histone H3 and alpha tubulin were mostly recovered in the nuclear and cytosolic 15 fractions, respectively. Small amounts of histone H3 and alpha-tubulin remained present in the 16 crude mitochondrial preparation, but were largely removed by the Percoll purification step. As 17 represented in the Fig 1A, immunoblotting of crude mitochondrial fractions with anti-OGT 18 antibody revealed a predominant band at 103 kDa, which corresponds to the expected molecular 19 weight of the mitochondrial isoform mOGT (Hanover 2003). A band was also observed at 116 20 kDa, consistent with the presence of the nucleocytoplasmic isoform (ncOGT). However, this band 21 was absent in Percoll-purified mitochondria, indicating that mOGT is the predominant, if not the 22 sole, isoform present in mitochondria. For OGA, a single band, running at 75 kDa was observed in 23 the mitochondrial fraction, which corresponds to the expected molecular weight of the short OGA isoform (sOGA), also expressed in the nucleus (Comtesse 2001). Conversely, the full length OGA 1 (fOGA) running at 130 kDa was absent from the mitochondrial fraction. (Fig 1A). 14 Immunoprecipitates were resolved by electrophoresis and revealed by fluorescence at the wavelength of 15 TAMRA (570 nm). Following densitometric quantification of band intensity (shown in red and blue), gels 16 were cut in 7 pieces of equal size and further processed for proteomics analysis.

18
To assess whether protein O-GlcNAcylation was increased in our reconstitution assay, crude 19 mitochondrial fractions were lysed after exposure to UDP-GlcNAc in absence or presence of the 20 O-GlcNAc inducer NButGT. Following denaturation, O-GlcNAc-modified proteins were 21 stabilized and labeled with the fluorescence probe TAMRA using Click-iT chemistry (Fig 1C). Following TAMRA-mediated O-GlcNAc specific immunoprecipitation, proteins were separated 1 by gel electrophoresis and visualized by fluorescence. Multiple bands were observed in 2 immunoprecipitates from control mitochondria indicating a baseline level of protein O-3 GlcNAcylation ( Fig 1C). Importantly, staining intensity was consistently increased in NButGT-4 treated mitochondria indicating a broad and rapid rise in protein O-GlcNAcylation levels.

8
To gain knowledge on the repertoire of proteins modified by the mitochondrial O-9 GlcNAcylation system, gels were cut in seven bands of equal size and processed for tandem mass 10 spectrometry (MS/MS) analysis. To maximize stringency, only proteins reliably detected in all 11 experimental replicates from control and NButGT-treated mitochondria were considered. Using 12 this selection criteria, a total of 842 proteins were identified (Fig 2A). Of these, 50% (409) had a 13 known (339) or predicted (70) mitochondrial status in the Mitominer database (Smith 2016), while 14 the remaining were non-mitochondrial (322), or had an unspecified status (111), which can be 15 expected given that crude mitochondrial fractions were used for this analysis. Since these likely 16 contained residual amounts of ncOGT and fOGA outside mitochondria, we sought to determine 17 whether the impact of NButGT on protein O-GlcNAcylation varied according to the localization 18 of these proteins. As shown in Fig 2B,

treatment with NButGT predominantly increased O-19
GlcNAcylation of mitochondrial proteins (i.e. known + predicted mitochondrial status) compared 20 to non-mitochondrial proteins, indicating that the reconstitution assay was effective at targeting the 21 intra-mitochondrial O-GlcNAcylation system. Consistent with this notion, treatment with NButGT 22 significantly increased O-GlcNAcylation of 191 (q<0.05) to 246 (q<0.1) mitochondrial proteins 23 ( Fig 2C), while none of the non-mitochondrial proteins were significantly affected ( Fig S1).
Because mOGT was reported as preferentially associated with the mitochondrial inner membrane 1 (Banerjee 2015), we looked at the sublocalization of mitochondrial proteins. This analysis indicated 2 that a large proportion of the 409 O-GlcNAcylated mitochondrial proteins originated from the 3 matrix (126) and inner-membrane (124), with only a minor proportion coming from the 4 intermembrane space or outer membrane ( Fig 2D). However, the effect of NButGT on O-5 GlcNAcylation level did not differ significantly across submitochondrial compartments (Fig 2E).  Based on these results, the mitochondrial processes targeted by acute mitochondrial O-1 GlcNAcylation were examined. Pathway enrichment analysis and protein network clustering 2 revealed that proteins related to oxidative phosphorylation, tricarboxylic acid (TCA) cycle and 3 pyruvate and fatty acid metabolism were the top enriched pathways (Fig 3A-B). Among the multi-4 proteins complexes of the oxidative phosphorylation machinery, complex I (23 subunits, 60%,), V 5 (8 subunits, 50%), III (6 subunits, 75%), and II (3 subunits, 75%), were predominantly affected 6 with 50-75% of their subunits being significantly more O-GlcNAcylated in response to NButGT 7 compared to only 25% (4 subunits) for complex IV (Fig 4).  2 pathways in NButGT-treated vs control mitochondria performed using g:Profiler. Proteins were input in 3 g:Profiler in order of decreasing q value with a threshold set at q<0.1 (Ordered Query). Maximum size of 4 functional categories was set at 250 to filter out large annotations that provide limited interpretative value.

5
The g:SCS algorithm was used for multiple hypothesis testing corrections using a default alpha threshold 6 of 0.05 for significance. Enrichment is expressed as a rich factor, which represents the ratio of the number 7 of proteins observed for a given pathway term to the total number of proteins for this term. Circle size 8 reflects the number of proteins per pathway, while color indicates the level of significance.

10
Beyond energy metabolism, smaller clusters of proteins related to several other mitochondrial 11 functions displayed increased O-GlcNAcylation levels in response to NButGT (Fig 4). This 12 included proteins related to: i) mitochondrial protein translation, such as proteins associated with

observed following in vivo treatment with OGA inhibitors.
2 To gain insights on the contribution of the intra-mitochondrial O-GlcNAcylation system to 3 protein O-GlcNAcylation in the heart, we sought to compare these results with a methodologically 4 comparable (i.e. identical Click-iT labelling, IP and MS/MS protocol) cardiac O-GlcNAcylomic 5 dataset derived from mice that were subjected to NButGT or vehicle treatment 6 hours prior to 6 sacrifice ( Fig 5A). Of the 409 mitochondrial proteins detected in the in vitro reconstitution assay, 7 85% were also identified as being O-GlcNAcylated in vivo, and among them 122 displayed 8 enhanced O-GlcNAcylation in the two data sets (FC >1.2, Table S1). 9 Comparison was also made with data from a previous study performed by Ma and colleagues 10 in which O-GlcNAc sites on isolated cardiac mitochondria were mapped 12h after in vivo 11 administration of the OGA inhibitor Thiamet G using a BEMAD labelling method (Ma 2015). Of 12 the 88 O-GlcNAc-modified proteins identified by Ma and colleagues, 62 were found to display 13 increased O-GlcNAcylation in our in vitro reconstitution assay, representing a highly significant 14 over-enrichment ( Fig 5B). The majority of shared proteins across the two datasets were 15 components of the oxidative phosphorylation system, TCA cycle, and fatty acid oxidation pathway 16 ( Fig 5C), with a few noticeable proteins related to ROS metabolism (SOD2, PRX3) and 17 permeability transition pore (mPTP)/apoptosis (VDAC1, SLC25A4, ENDOG). Within the 18 OXPHOS system, overlap between the two datasets was observed for subunits located in the 19 NADH dehydrogenase (N) and ubiquinone reductase (Q) modules of complex I (NDUFS1, 20 NDUFAA7, NDUFA9), the F1 sector of complex V (ATP5O, ATP5B, ATP5A1), the hydrophilic 21 head of complex II protruding in the matrix (SDHA, SDHB) and UQCRC2, a matrix facing subunit 22 of complex III (Fig 5D and S2  GlcNAcylation, still conserved after immunoprecipitation. Furthermore, probing with the 6 NDUFS1 antibody revealed a drastic increase in immunoreactivity following exposure to NButGT, 7 without any changes in protein abundance ( Fig 6B). Similarly, O-GlcNAc staining of ATP5A1, 8 UQCRC2, MTCO1, and SDHB (using the antibody mix OXPHOS) was increased following 9 immunoprecipitation with the anti-O-GlcNAc antibody. Similar results were obtained using 10 cardiac lysates from mice injected with NButGT or vehicle 6 hours prior to sacrifice (Fig 6 A-B). 11

Acute stimulation of O-GlcNAcylation enhances maximal electron flux through a
1 complex I-driven mechanism.

2
To determine whether such acute stimulation of O-GlcNAcylation had a functional impact, 3 mitochondria were pre-incubated with NButGT or vehicle during 30 min before monitoring basal, 4 ADP stimulated and CCCP-uncoupled respiration (for the reader's convenience, the different 5 substrates and inhibitors with action sites is represented in Fig 7A). As shown in Fig 7B-D, 6 exposure to NButGT increased maximal ADP-stimulated respiration in presence of substrates 7 feeding complex I (pyruvate -malate). A similar effect was observed when phosphorylation was 8 uncoupled from respiration using CCCP ( Fig 7E-G), which indicated that activation of the electron 9 transport chain (ETC), rather than stimulation of the ATP synthase was responsible for the rise in 10 maximal respiration observed with NButGT. Interestingly, NButGT had no significant effect on 11 ADP-stimulated respiration when mitochondria were energized with complex II substrate 12 (succinate in presence of rotenone), suggesting that the stimulatory effect of NButGT was linked 13 to complex I ( Fig 7H). This was directly confirmed by measuring the activity of respiratory chain 14 complexes in mitochondrial lysates. Following exposure to NButGT, the activity of complex I was 15 increased by ~ 50%, while those of complex II, complex IV and the TCA cycle enzyme citrate 16 synthase were unchanged (

Acute stimulation of O-GlcNAcylation attenuates mitochondrial ROS release
1 To determine whether this had an impact on ROS release, H 2 O 2 production was determined in 2 mitochondria that were pre-incubated with NButGT, Thiamet G or vehicle. H 2 O 2 release was first 3 measured at baseline in presence of pyruvate and malate. As represented in Fig  To test this further, mitochondria were progressively depolarized with the uncoupler CCCP 18 since RET, and therefore ROS release by the IQ site, is exquisitely sensitive to the electrochemical 19 gradient (Goncalves 2015). As expected, abolishing RET through uncoupling caused a drastic 20 reduction of H 2 O 2 release from the IQ site, and restored the 40-50% difference in H 2 O 2 release 21 observed between control and NButGT-or Thiamet G-treated mitochondria in absence of RET (i.e. 22 when pyruvate-malate drive ROS release at the IF site) (Fig 8B-C and S3). d) The electron-transferring flavoprotein ubiquinone reductase (ETFQOR) site within complex 1 III represents another major site of ROS production in the ETC. In isolated mitochondria, this site 2 produces superoxides at high rates only when electron transfer is blocked with complex III 3 inhibitors ( Fig 8A) (Goncalves 2015). For this reason, antimycin-A was next added to fully 4 uncoupled mitochondria in order to assess the impact of NButGT on ROS release from complex 5 III. As expected, addition of antimycin-A caused a drastic rise in H 2 O 2 release reflecting superoxide 6 release from complex III in all groups (Fig 8B-C). In NButGT-treated mitochondria, H 2 O 2 release 7 was reduced by 30% compared to controls suggesting reduced release of ROS from the ETFQOR 8 site (Fig 8B-C). However, this effect was not observed in presence of Thiamet G (Fig S4), which 9 suggests that the main site of action of acute hyper O-GlcNAcylation on ROS production remains 10 complex I. 11 Of note, previous studies reported that increased cellular O-GlcNAcylation protects from 12 mitochondrial permeability transition pore (mPTP) opening (Ngoh 2011, Ma 2015). Mitochondrial 13 swelling assays were therefore performed to determine whether sensitivity to Ca 2+ -induced mPTP 14 opening was altered following acute exposure to NButGT. As shown in Fig 8D,  GlcNAc-modified proteins, and identifies several novel targets related to energy metabolism, and 6 multiple other facets of mitochondrial biology. Importantly, we show that acute hyper-O-7 GlcNAcylation increases maximal respiratory capacity, and drastically reduces ROS likely through 8 a complex I-mediated mechanism. 9 The OGT gene encodes three splice variants whose products vary only in the number of N-10 terminal tetratricopeptide repeat (TPR) domains known to be involved in protein-protein 11 interaction (Bond 2015). The longest splice variant encodes the 116 kDa nucleocytoplasmic 12 ncOGT isoform which is the most abundantly expressed, while the shortest 78 kDa isoform is 13 curiously derived from a longer transcript. In addition, Hanover's group identified a unique start 14 site in the fourth intron of the OGT gene that generates a 103 kDa isoform which was found to be 15 enriched in the mitochondrial fraction of Hela cells (Hanover 2003, Love 2003) and rat heart 16 (Banerjee 2015). However, the existence of this mitochondrial isoform in mammalian tissues has 17 been recently questioned (Trapannone 2016). In this study, endogenous mOGT was reported to be 18 undetectable in several human cell lines and mouse tissues, including heart. Genomic sequence 19 alignments also suggested that the predicted start site for mOGT was likely lacking in most species 20 analyzed except some primates. Based on these data, the authors concluded that the small amounts 21 of ncOGT detected in the crude mitochondrial fraction was likely sufficient for O-GlcNAcylation 22 of mitochondrial proteins. These results are however in contrast with our data and previous studies 23 from Hanover's group. Our data clearly indicate that the 103 kDa mOGT is the main isoform found subunits displaying the highest fold change following in vivo treatment, with Thiamet G were 8 located in the N module and two in particular were common to our dataset, namely NDUFS1 and 9 NDUFA7. It is interesting to note that NDUFS1 harbors an NADH-ubiquinone oxidoreductase 10 domain and Fe-S clusters, in addition to containing a number of phosphorylation (Gowthami 2019) 11 and acetylation sites, which makes it a potentially important candidate for post-translational 12 regulation of complex I activity. As for NDUFA7, recent data indicate that its ablation in the heart 13 increases mitochondrial ROS release, and triggers pathological cardiac hypertrophy (Shi 2020), 14 although it is still unclear whether this is linked to failed assembly of complex I or to altered 15 activity. It should also be noted that phosphorylation of four other subunits within the N module 16 (NDUFS4, NDUFV1, NFUFV3 and NDUFA12), including two that displayed sensitivity to 17 NButGT, were previously shown to regulate complex I activity and ROS release (Papa 2001, Lund 18 2008, Wang 2014). Further studies will clearly be required to fully elucidate the mechanisms by 19 which acute O-GlcNAcylation stimulates complex I. 20 Although in vitro reconstitution assays are well suited for mechanistic studies, they pose 21 obvious limitations as they do not fully mimic the complex conditions encountered in vivo. To 22 tackle this issue, we also compared our in vitro data with a methodologically comparable cardiac 23 Our results reveal a strikingly high degree of overlap (85%) in the O-GlcNAc modified 1 mitochondrial protein identified in the two datasets. Importantly, 122 of these proteins also 2 displayed sensitivity to NButGT. Besides, highlighting the relevance of our reconstitution assay 3 for mechanistic studies, these results highly suggest that the intra-mitochondria O-GlcNAc cycling 4 system is the main mechanism through which mitochondrial proteins become O-GlcNAcylated in 5 vivo. 6

Preparation of isolated cardiac mitochondria
18 Heart mitochondria were prepared as described previously (Marcil 2006). Hearts were rapidly 19 excised and immersed into ice-cold isolation medium (buffer A, in mM: 300 sucrose, 10 Tris-HCl, 20 1 EGTA, pH 7.3) and weighed. Ventricular tissue was minced with scissors in 5 ml of buffer A 21 supplemented with 0.2% fatty acid free bovine serum albumin (BSA) and homogenized using a 22 Polytron tissue tearer (∼ 3 s at a setting of 3). The homogenate was then incubated with the protease 23 volume was completed to 30 ml with Buffer A+ 0.2% BSA and centrifuged at 800×g for 10 min. 1 The pellet was discarded and the supernatant was decanted and centrifuged at 10 000×g for 10 min. 2 The pellet obtained was re-suspended in buffer B (in mM: 300 sucrose, 0.05 EGTA, 10 Tris-HCl, 3 pH 7.3) and centrifuged at 10,000×g for 10 min. After repeating this washing step twice, the final 4 mitochondrial pellet was re-suspended in 0.3 ml of buffer B to a protein concentration of ∼ 20 5 mg/ml. All procedures were carried out at 4°C. Protein determinations were performed using the 6 bicinchonic acid method (Pierce, Rockford, IL, USA), with bovine serum albumin as a standard. 7  Twelve week-old mice (C57BL/6 N, male) from Janvier Labs were treated with NButGT 20 (50 mg/kg) by intraperitoneal injection 6 hours before sacrifice. Mice were anesthetized with an 21 intraperitoneal injection of a mixture of ketamine (150 mg/kg) and xylazine (10 mg/kg) and hearts 22 were washed in PBS before being freeze-clamped in liquid nitrogen and stored at -80°C. 23

O-GlcNAc immunoprecipitation & immunoblotting
1 Lysate supernatants (20 µg of heart homogenate, nuclear fraction, cytosolic fraction and crude 2 mitochondria; 40 µg of purified mitochondria) were loaded on SDS-PAGE gel and transferred onto 3 polyvinylidenedifluoride (PVDF) membrane. After blocking in BSA 5% TBS-Tween 20 0.1%, 4 membranes were then probed with appropriate antibodies to assess total protein level. The 5 appropriate secondary antibody conjugated to HRP and the BM chemiluminescence blotting performed by SDS-PAGE and immunoblotting was realised as mentioned above with NDUFS1 or 20 OXPHOS antibodies.

Mass spectrometry and protein identification
1 Preparation of proteins -Isolated mitochondria pellets or 20 mg of freeze-clamped hearts 2 were homogenized in 200 µL of RIPA lysis buffer (25 mM Tris HCl, 150 mM NaCl, 1% NP-40, 3 1% sodium deoxycholate, 0.1% SDS at pH 7.6) supplemented with a protease/phosphatase 4 inhibitor cocktail (ThermoFisher) and 1 µM of O-GlcNAc cycling enzyme inhibitors (Sigma-5 Aldrich, Saint-Louis, Missouri, United-States). 250 µg of proteins from the lysate were then 6 precipitated using chloroform/methanol (MeOH) precipitation method as follows. 600 µL of 7 MeOH were added to the 200 µL sample, followed by 150 µL of chloroform and 400 µL of 18 8 megaOhm water. Samples were then vortexed briefly and centrifuged for 5 min at 13,000×g. The 9 upper aqueous phase was carefully removed and discarded. Additional 450 µL MeOH were added 10 to pellet the protein after brief vortex and centrifugeation for 5 min at 13,000×g. The supernatant 11 was then removed, and the pellet air dried for 5 min. Finally, proteins were resuspended in 40 µL 12 of 1% SDS in 20 mM HEPES pH 7.9 and heated 5-10 min at 90°C to assure completely 13 resuspension of proteins. 14

Enzymatic labelling and purification of O-GlcNAcylated proteins -O-GlcNAc groups from 15
proteins were labelled with tetramethylrhodamine azide (TAMRA) using the Click-iT® O-GlcNAc 16 enzymatic labelling system kit (C33368) followed by the Click-iT® protein analysis detection kit 17 (C33370) from Invitrogen according to the manufacturer's instructions. SDS was then quenched 18 with NEFTD buffer (100 mM NaCl, 50 mMTris-HCl, 5 mM EDTA, 6% NP-40 at pH 7.4). Before 19 immunoprecipitation of TAMRA labelled proteins, lysate was precleared with washed protein G 20 sepharose beads to avoid non-specific binding of proteins on the beads. Afterwards, supernatant 21 was incubated with pre-washed protein G sepharose beads (10 µL) coupled with anti-TAMRA 22 antibody (10 µg, A6397, Invitrogen) for 1.5 h at 4°C. Following centrifugation (500×g, 1 min), the 23 beads were washed once with NEFTD buffer (100 mM NaCl, 50 mM Tris-HCl pH 7.4, 5 mM EDTA, 6% NP-40) and three times with NEFT buffer (NEFTD without NP-40). The beads were 1 then boiled 5 min in Laemmli buffer (2 mM EDTA, 4% SDS, 20% Glycerol, 0.004% bromophenol 2 blue, 50 mM DTT and 100 mM Tris at pH 6.8) to elute O-GlcNAc proteins. Proteins were then 3 separated on 1 mm on SDS-PAGE gel and stained with Coomassie blue (Sigma-Aldrich). 4 Requests for access to other data should be addressed to senior authors: Yan Burelle 19 (yburell2@uottawa.ca) and Luc Bertrand (luc.bertrand@uclouvain.be). All requests will need to 20 specify how the data will be used and will require approval by co-investigators. 21