Oligodendrocyte–axon metabolic coupling is mediated by extracellular K+ and maintains axonal health

The integrity of myelinated axons relies on homeostatic support from oligodendrocytes (OLs). To determine how OLs detect axonal spiking and how rapid axon–OL metabolic coupling is regulated in the white matter, we studied activity-dependent calcium (Ca2+) and metabolite fluxes in the mouse optic nerve. We show that fast axonal spiking triggers Ca2+ signaling and glycolysis in OLs. OLs detect axonal activity through increases in extracellular potassium (K+) concentrations and activation of Kir4.1 channels, thereby regulating metabolite supply to axons. Both pharmacological inhibition and OL-specific inactivation of Kir4.1 reduce the activity-induced axonal lactate surge. Mice lacking oligodendroglial Kir4.1 exhibit lower resting lactate levels and altered glucose metabolism in axons. These early deficits in axonal energy metabolism are associated with late-onset axonopathy. Our findings reveal that OLs detect fast axonal spiking through K+ signaling, making acute metabolic coupling possible and adjusting the axon–OL metabolic unit to promote axonal health.

Oligodendrocytes (OLs) produce and maintain the myelin sheaths around axons, making fast and economical communication between distant neurons possible.Axonal health is crucial for brain function, and axonal damage is a feature of aging and various neurological disorders 1,2 .Accumulating evidence reveals that, apart from orchestrating axonal signaling speed, OLs have an important role in preserving neural circuits and long-term neuronal integrity [3][4][5][6][7] .In recent years, several studies have indicated that OLs contribute to supporting axonal energy metabolism [8][9][10] .OLs can sustain their functions through aerobic glycolysis alone, given the preservation of white matter integrity in Cox10-mutant mice, in which mitochondrial respiration is specifically perturbed in OLs 8 .One necessary outcome of aerobic glycolysis is the production of lactate, which could serve as an energy substrate for axons 8,9,11,12 .Indeed, OL-specific deletion of monocarboxylate transporter 1 (MCT1) leads to late-onset axonopathy, implying that lactate and/or pyruvate release from OLs has a role in axonal health 9,13 .Glutamatergic signaling has been shown to stimulate the surface expression of glucose transporter 1 (GLUT1) in OLs, suggesting that axonal activity might regulate the metabolic support provided by OLs to axons 10 .Metabolite supply could be facilitated by cytosolic channels within the myelin sheath 14,15 , and a disruption in this myelinic channel system has been associated with axonal damage 16 .Moreover, mice deficient in the myelin proteolipid protein (PLP), a mouse model of spastic paraplegia, develop severe axonal spheroids with age 17,18 , possibly due to deficits in axonal transport 19 , alterations in mitochondrial function 20 and impaired energy homeostasis 21 .Other homeostatic functions carried out by OLs include antioxidant support 22 and K + buffering 23,24 .
Recognizing that OLs detect axonal spiking, we next asked whether heightened axonal activity influences metabolic flux within OLs.Considering that OLs may supply axons with glycolytic products such as pyruvate or lactate, increased axonal spiking could enhance glucose consumption in OLs.To study this, we expressed the glucose sensor FLII12Pglu700µΔ6 (FLIIP) 32 in optic nerve OLs of wild-type mice through adeno-associated virus (AAV) delivery (Fig. 1h,i).CC1 immunolabeling confirmed sensor expression in mature OLs (Fig. 1h).We first assessed the sensor's response to glucose fluctuations.Removing extracellular glucose lowered cytosolic glucose levels in OLs, whereas blocking glycolysis with 1 mM iodoacetate (IA) in artificial cerebrospinal fluid (ACSF) containing 10 mM glucose increased glucose levels (Extended Data Fig. 1a,b), confirming the sensor's functionality for studying glucose metabolism in OLs.To analyze glycolytic flux, we inhibited glucose transporters with 20 µM cytochalasin B (CytoB) and measured the glucose decline rate (Fig. 1i), as previously outlined 33,34 .Remarkably, transient 50-Hz axonal stimulation accelerated oligodendroglial glucose consumption by approximately ninefold compared to the basal rate at 0.1-Hz stimulation (Fig. 1i).Thus, OLs respond to axonal spiking by promptly increasing their glycolytic activity.

OLs detect axonal spiking through [K + ] ext and Kir4.1 channels
Next, we sought to determine the mechanism by which OLs detect axonal spiking.OLs express N-methyl-d-aspartate (NMDA) receptors, which regulate oligodendroglial glucose import 10 and mediate Ca 2+ increase in myelin upon electrical axonal stimulation 35 or chemical ischemia 36 .α-Amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) receptors were also suggested to influence myelinic Ca 2+ dynamics 35 .Therefore, we examined the contribution of glutamatergic signaling.Blocking AMPA receptors with 2,3-dioxo-6-n itro-7-sulfamoyl-benzo[f]quinoxaline (NBQX; 50 µM) caused no change in the OL Ca 2+ response (Extended Data Fig. 2a), indicating that AMPA receptors might not be involved.Additional inhibition of NMDA receptors with 7-chlorokynurenic acid (7-CKA; 100 µM) and d-2-amino-5-phosphonopentanoate (D-AP5; 100 µM), blocking both glycinergic and glutamatergic NMDA receptor binding sites, reduced Despite the existing notion that OLs support axonal energy metabolism, the molecular and cellular events involved in metabolic coupling remain elusive.Whether neuronal activity influences OLs to drive metabolic support is still unclear.Glutamatergic signaling may mediate the long-term adjustment of oligodendroglial glucose uptake capacity 10 , but what controls a rapid and on-demand delivery of metabolites to axons remains unexplored.
Independent of neuronal subtype, a key indicator of axonal activity is transient increases in extracellular K + concentrations ([K + ] ext ), which depolarize the plasma membrane of OLs [25][26][27] .Here, we hypothesized that activity-driven K + signaling triggers rapid metabolic coupling between OLs and axons.We addressed this question through optic nerve electrophysiology and two-photon imaging, a combination previously used to study axonal ATP dynamics 11,28 .We found that high-frequency axonal spiking triggers a Ca 2+ surge and immediately accelerates glucose consumption in OLs.Axonal activity is detected by OLs predominantly through increases in [K + ] ext and activation of Kir4.1 channels.Both high-frequency stimulation and elevated [K + ] ext evoke a lactate increase in axons, which is diminished by pharmacological inhibition of Kir4.1.Moreover, blocking Kir4.1 impairs the recovery of axonal firing from high-frequency stimulation.Using OL-specific Kir4.1-knockout mice (Kir4.1 fl/fl ;MOGiCre, hereafter termed Kir4.1 cKO), we demonstrate that axonal lactate dynamics are controlled by oligodendroglial Kir4.1 and that OLs are the primary cells involved in activity-dependent K + clearance.Furthermore, axonal glucose uptake and consumption are decreased in Kir4.1 cKO mice, revealing that OLs also regulate axonal glucose metabolism.These early deficits in axonal energy metabolism could affect vesicular transport and antioxidant capacity, leading to the late-onset axonal damage detected in Kir4.1 cKO mice.Our findings imply that increased [K + ] ext during fast axonal spiking stimulates axon-OL metabolic coupling and that oligodendroglial K + homeostasis regulates axonal energy metabolism, function and survival.
In summary, our pharmacological results (summarized in Fig. 2g) indicate that OLs detect high-frequency axonal activity through elevated [K + ] ext , leading to depolarization through Kir4.1 channels and Ca 2+ entry chiefly by reverse-mode activation of NCX (Fig. 2h).

Oligodendroglial Kir4.1 regulates axonal lactate dynamics
Given that OLs respond to fast axonal spiking through Kir4.1 channel activation, we speculated that this K + -driven stimulation could regulate metabolite supply to axons (for example, on-demand lactate delivery).To test this, we expressed the lactate sensor Laconic 50 in optic nerve axons using intravitreal AAV delivery (Fig. 3a).Initial tests in wild-type nerves showed that increasing extracellular lactate levels increased axonal lactate levels (Fig. 3a,b), confirming the expression of lactate transporters in axons.Removal of glucose and lactate from the ACSF significantly reduced axonal lactate levels (Fig. 3a,b).Hence, the Laconic sensor is not saturated at baseline, and its dynamic range allows for studying axonal lactate dynamics.Notably, axonal lactate levels increased during high-frequency spiking, more prominently at higher stimulation frequencies (Fig. 3c,d).To see whether elevated [K + ] ext alone could increase axonal lactate independently of spiking, we used TTX to inhibit axonal activity and the associated workload increase.Indeed, [K + ] bath stimulation increased axonal lactate levels, which was mediated by Kir4.1 activity (Extended Data Fig. 5e).Importantly, Kir4.1 inhibition with Ba 2+ specifically diminished the stimulus-evoked OL Ca 2+ response but did not affect the axonal Ca 2+ surge (Fig. 2c and Extended Data Fig. 3d) and thus the axonal workload upon spiking.Yet, Ba 2+ reduced the activity-induced increase in axonal lactate levels by 40% (Fig. 3e,f).This implies that K + -mediated axon-OL signaling facilitates lactate supply to axons during active spiking.
To investigate whether oligodendroglial Kir4.1 specifically controls axonal lactate dynamics, we used Kir4.1 cKO mice 23 (Fig. 3g and Extended Data Fig. 6).First, we inspected optic nerves from ∼3-month-old Kir4.1 cKO mice and littermate controls (Kir4.1 fl/fl ).We found no differences in CAP peak latencies or nerve excitability (Extended Data Fig. 6a-c).Additionally, we observed no overt changes in myelin sheath thickness and the diameter distribution of myelinated axons (Extended Data Fig. 6d-f).At this age, no signs of axonal damage or neuroinflammation were observed (Extended Data Fig. 6d,g,h).The Ba 2+ -mediated impact on delaying CAP peak recovery (Extended Data Fig. 3f) could involve Kir4.1-mediatedK + clearance by OLs and/or astrocytes 43,51 .Notably, Kir4.1 cKO nerves showed significantly slower kinetics of CAP peak recovery after high-frequency stimulation compared to littermate controls (Extended Data Fig. 6i,j), confirming earlier results 23 .Interestingly, wild-type nerves treated with 100 µM Ba 2+ showed the same recovery delays as Kir4.1 cKO nerves (Extended Data Fig. 6i,j), implying that oligodendroglial Kir4.1 is primarily involved in K + buffering in the adult white matter, with little to no contribution of astrocytic Kir4.1 or other Ba 2+ -sensitive Kir channels.The recovery kinetics of CAP peak latency were also affected in Kir4.1 cKO nerves (Extended Data Fig. 7a-d).Poststimulation CAP recovery deficits were also visible at 25 and 10 Hz (Extended Data Fig. 7e-p), demonstrating that oligodendroglial Kir4.1 also governs activity-dependent K + clearance at lower frequencies.
We then examined axonal lactate dynamics and found that both the resting lactate levels (at 0.1-Hz stimulation) and the high-frequency-evoked lactate surge were notably reduced in Kir4.1 cKO nerves compared to controls (Fig. 3h-j).Basal axonal lactate levels were compared after normalizing the Förster resonance energy transfer (FRET) ratios to minimal lactate levels after glucose deprivation (GD; Fig. 3h,i).Lower lactate levels in Kir4.1 cKO axons might stem from decreased OL lactate supply, altered axonal glycolysis or heightened lactate consumption.The latter seems unlikely, as lactate decay during GD was not faster but marginally slower in cKO mice than in controls (Fig. 3h,k).This is further corroborated by a similar CAP decline rate during GD (Fig. 3l).The two genotypes showed the same latency between lactate depletion and CAP decline (Fig. 3m).Notably, the CAP drop was observed only when axonal lactate neared depletion (Fig. 3m).Given the concurrent decline of CAP and ATP levels during GD 11 , this suggests that axonal lactate fuels ATP for action potentials and axons cease firing as lactate supply dwindles, leading to CAP decline (Fig. 3m).This also implies comparable axonal mitochondrial respiration between the genotypes.

Reduced MCT1 and GLUT1 in central nervous system myelin of Kir4.1 cKO mice
Loss of oligodendroglial Kir4.1 may affect axonal integrity with age 24 .Electron microscopy (EM) at age 3 months revealed no signs of impaired axonal integrity in Kir4.1 cKO mice, with normal myelin thickness and axonal diameters (Extended Data Fig. 6d-f).However, by 7-8 months, there was a significant increase in axon/myelin profiles indicative of axonal degeneration, including giant axonal swellings (Extended Data Fig. 8a-c).The age-dependent axonopathy in cKO mice occurred without noticeable myelin abnormalities or thinning (Extended Data Fig. 8d,e).Additionally, increased glial fibrillary acidic protein (GFAP) immunolabeling was observed, consistent with secondary astrocytic reactivity; however, there was no increase in ionized calcium-binding adaptor molecule 1 (IBA1) immunopositivity, suggesting no microgliosis (Extended Data Fig. 8f,g).
The absence of oligodendroglial Kir4.1 in young mice appears to affect axonal energy metabolism before axonal pathology onset.This aligns with the lower axonal lactate levels observed in young cKO nerves, which did not show axonal damage or neuroinflammation.Both pharmacological inhibition and genetic inactivation of Kir4.1 affected the stimulus-evoked axonal lactate surge, but the lower basal lactate level was an additional feature of cKO mice.Besides regulating acute metabolic coupling, oligodendroglial Kir4.1 may also contribute to adjusting the axon-OL metabolic unit.To explore the metabolic changes from Kir4.1 deficiency further, we performed tandem mass tag (TMT)-based quantitative proteomics on optic nerve lysates from 2.5-month-old Kir4.1 cKO mice and littermate controls (Fig. 4a, Extended Data Fig. 9a and Supplementary Data 1).Expectedly, Kir4.1 (Kcnj10) protein levels were lower in the cKO group (Fig. 4b and Extended Data Fig. 9b).Among the top 50 proteins sorted by the false discovery rate (FDR), those associated with vesicular transport and energy metabolism showed reduced levels (Fig. 4b).Gene set enrichment and pathway analyses revealed declines in transmembrane transporter activity, vesicular transport and oxidative phosphorylation pathways (Fig. 4c,d).A lower abundance of MCT1 (Slc16a1; Fig. 4b and Extended Data Fig. 9c) and GLUT1 (Slc2a1; Extended Data Fig. 9d) was observed, but there was no change in the glucose transporter GLUT3 (Slc2a3; Extended Data Fig. 9e).Additional qPCR analysis indicated a slight decrease in Slc2a1 mRNA but not in Slc16a1 mRNA (Extended Data Fig. 8f), implying that changes in transporter abundance might not be solely due to gene expression alterations.Immunoblotting of myelin purified from cKO brains revealed an approximately 50% reduction in both GLUT1 and MCT1 abundance (Fig. 4e).Hence, Kir4.1 loss affected the relative abundance of metabolite transporters in OLs.K + and Kir4.1-mediated signaling could influence gene expression, protein synthesis, surface trafficking and/or turnover of metabolite transporters in OLs.This might result in adaptations aligning the support machinery of OLs to axonal activity levels.

Minor changes in axonal ATP dynamics in Kir4.1 cKO mice
Proteomics analysis indicated alterations in oxidative phosphorylation pathways, leading us to question whether cKO mice also exhibit changes in axonal ATP homeostasis.To investigate this, we expressed Article https://doi.org/10.1038/s41593-023-01558-3 the ATP sensor ATeam1.03(ref.52) in axons through intravitreal AAV delivery 28 .First, we assessed basal axonal ATP levels by normalizing the FRET ratios at 0.1-Hz stimulation in 10 mM glucose against ratios after GD and mitochondrial respiration inhibition (MI) using 5 mM sodium azide (NaN 3 ) to deplete axonal ATP levels (Fig. 5a-c).No genotype difference in basal axonal ATP levels was observed (Fig. 5c).The ATP   decline rate during GD + MI was also comparable between genotypes (Fig. 5b,d), and the ATP recovery rate after NaN 3 washout and 10 mM glucose reperfusion was unaffected (Fig. 5e).Interestingly, cKO nerves showed a marginally quicker onset of ATP recovery compared to controls (Fig. 5b,f); however, after 15-20 min, the ATP levels in control nerves almost completely returned to baseline levels, whereas cKO nerve recovery was incomplete (Fig. 5b,g).The minor deficit in ATP recovery prompted us to examine axonal firing recovery in cKO mice further.The rapid decline in the CAP area after GD + MI, indicating axonal conduction block, was comparable between genotypes, and the onset and recovery kinetics of axonal firing appeared unchanged (Extended Data Fig. 10a).The analysis of the partial CAP (pCAP) area, which reflects the dynamics of the first and second CAP peaks 28 , showed comparable decline rates upon GD + MI, whereas pCAP recovery dynamics differed between genotypes (Extended Data Fig. 10b).Closer inspection of CAP waveforms revealed a striking difference in the recovery kinetics of peak latency (Extended Data Fig. 10c).In controls, the CAP peak amplitude increased gradually as the peak ) and control (n = 6) mice challenged with GD + MI, normalized to the minimum ATP level (R 0 ).Inset, initial ATP recovery dynamics following reperfusion with 10 mM glucose.c, Basal axonal ATP levels were comparable between genotypes (n = 6 mice, P = 0.61, two-sided unpaired t test).d, No difference in the ATP decline rate between genotypes upon GD + MI (n = 6 mice, P = 0.44, two-sided unpaired t test).e, Similar initial ATP recovery rates between genotypes after GD + MI (n = 6 mice, P = 0.51, two-sided unpaired t test).f, Onset of ATP recovery (see dashed arrows in inset in b) differed between control and cKO nerves (n = 6 mice, *P = 0.03, two-sided unpaired t test).g, Lower axonal ATP level recovery in cKO than in controls following GD + MI (n = 6 mice, *P = 0.0135, two-sided unpaired t test).h, Axonal ATP level changes (%) at 50-Hz stimulation relative to baseline.i,j, Similar ATP level decline rates (i) during 50-Hz stimulation (n = 6 mice, P = 0.81, two-sided unpaired t test) and equal initial ATP recovery rates (j) after stimulation (P = 0.54, two-sided unpaired t test).k, Slightly lower axonal ATP level recovery after stimulation in cKO versus controls (n = 6 mice, *P = 0.0498, two-sided unpaired t test).Data are represented as mean ± s.e.m.

Article
https://doi.org/10.1038/s41593-023-01558-3latency decreased; however, this peak latency shift (increase in conduction velocity) was markedly reduced in cKO nerves (Extended Data Fig. 10d), indicating a deficit in adjusting conduction speed following acute energy deprivation.Given that chemical ischemia increases [K + ] ext and affects oligodendroglial K + conductance 53 , K + clearance by OLs upon axon reenergization could be crucial for adjusting conduction speeds.This impaired speed adjustment might contribute to the reduced ATP recovery in cKO axons after chemical ischemia (Fig. 5b,g).
As the recovery of CAP peak latencies after high-frequency stimulation was reduced in cKO mice (Extended Data Fig. 7), we wondered whether this also correlates with axonal ATP recovery.We thus measured ATP dynamics during and after 50-Hz stimulation (Fig. 5h-k).The rate and extent of decreases in axonal ATP levels during high-frequency stimulation were comparable between genotypes (Fig. 5h,i).The initial ATP recovery rate after stimulation was also unchanged (Fig. 5h,j).However, 3-4 min into the recovery phase, cKO axons showed a slightly lower ATP rebound compared to initial baseline values (Fig. 5h,k).Hence, the absence of Kir4.1 and the lower K + clearance rate might increase the energy burden on axons following high-frequency activity or chemical ischemia.

Reduced axonal glucose metabolism in Kir4.1 cKO mice
The lower axonal lactate levels and activity-induced lactate surges in Kir4.1 cKO mice reflect reduced oligodendroglial metabolic support.Recent findings indicate that, besides lactate, OLs may supply axons with glucose 54 .However, the regulation of glucose uptake and glycolysis in myelinated axons and the potential contribution of OLs remain elusive.To study glucose dynamics, we expressed the glucose sensor FLIIP 32 in optic nerve axons through intravitreal AAV delivery (Fig. 6a).The basal axonal glucose levels (at 0.1-Hz stimulation) were comparable between genotypes, with basal levels derived by normalizing FRET ratios in ACSF with 10 mM glucose against the minimum levels after GD (Fig. 6a-c).To assess the sensor's range, we applied IA to inhibit glycolysis, leading to a 250% increase in glucose levels in both genotypes (Fig. 6d).However, the rate of axonal glucose increase was 36% slower in cKO nerves than in controls (Fig. 6d,e), hinting at a lower glucose uptake or hexokinase activity.Measuring glycolytic flux using CytoB, we found a 37% reduction in axonal glucose consumption in cKO nerves (Fig. 6f,g).This implies a similarly decreased glucose uptake in cKO axons to maintain equal basal glucose levels as controls (Fig. 6c), explaining the slower glucose increase with IA treatment (Fig. 6d,e).While testing whether axonal glucose dynamics differed during 50-Hz stimulation, we observed that glucose levels decreased in control axons but remained steady in cKO axons (Fig. 7a,b).This indicates a stronger activation of glycolysis relative to glucose uptake in control axons but being balanced in cKO axons.After stimulation, glucose levels increased above baseline in both groups (Fig. 7a,c), implying continued glucose uptake activation after stimulation ended.The difference in glucose dynamics during high-frequency activity may result from a lower activation of glucose consumption in cKO compared to controls.Indeed, 50-Hz stimulation increased axonal glucose consumption compared to basal (at 0.1-Hz stimulation) glycolytic activity (Fig. 7d,e), but this increase was 40% less in cKO axons (Fig. 7f).Interestingly, both control and cKO nerves exhibited an eightfold increase in glucose consumption rate under 50-Hz stimulation (Fig. 7g).Therefore, although the glycolytic activation machinery remains intact in cKO axons, their overall glucose metabolism is reduced by approximately 40%, both at rest and during activity.Consequently, we conclude that OLs regulate axonal glucose uptake and consumption, a unique metabolic OL-axon interaction that requires Kir4.1 function.

Discussion
Loss of axonal integrity may result from dysfunctions in the axon-OL unit 5,6,15,22,[55][56][57][58] .This study underscores the critical role of OLs in maintaining axonal health.We revealed that OLs respond to fast axonal spiking by initiating Ca 2+ signaling and glycolysis.OLs detect axonal spiking predominantly through [K + ] ext and Kir4.1 channel activation.This axon-OL signaling mechanism facilitates the supply of metabolites to axons.Disruptions in Kir4.1, whether through pharmacological inhibition or OL-specific inactivation, impaired the activity-induced lactate surges in axons.Lack of oligodendroglial Kir4.1 reduced the myelinic abundance of GLUT1 and MCT1, leading to decreased axonal lactate levels and glucose metabolism.These early metabolic deficits in axons are linked to late-onset axonal damage.We provide a working model in which K + signaling through oligodendroglial Kir4.1 governs metabolic coupling in the white matter (Fig. 8), affecting axonal energy metabolism, function and survival.Transient increases in [K + ] ext are a hallmark of axonal activity.OLs, highly permeable to K + , depolarize with elevations in [K + ] ext (ref.25) and axonal spiking 23 , facilitated by Ba 2+ -sensitive Kir channels 26,27 , of which Kir4.1 is predominantly expressed by OLs 23,24,59,60 .Our results suggest that Kir4.1-mediatedK + uptake is pivotal in triggering Ca 2+ influx in OL somas, facilitated by membrane depolarization and reverse-mode NCX activation.This K + -driven signaling regulates metabolic coupling between OLs and axons.Other activity signals, such as glutamate, may induce Ca 2+ activity in microdomains along internodes 61,62 , probably aiding in OL differentiation during development 61,[63][64][65][66][67] and GLUT1 surface expression 10 .In zebrafish, myelin sheaths of new OLs show Ca 2+ signals in response to neuronal activity 68,69 .However, in developing mouse OLs, Ca 2+ transients were reported to occur independent of cortical neuron activity 70 .We found that, in mouse optic nerves, Ca 2+ signaling in mature OLs is elicited chiefly through [K + ] ext surges during high-frequency axonal firing.K + is released at the nodes of Ranvier 71,72 and along juxtaparanodal domains under the myelin sheath 73,74 .Activity-mediated [K + ] ext accumulation around OLs is possibly more pronounced in fully myelinated adult white matter than in sparsely myelinated cortical regions, which merits further investigation.Upon detecting heightened axonal activity, OLs promptly enhance glucose consumption and provide lactate or pyruvate on demand (Fig. 8).This process is likely triggered by [K + ] ext and Kir4.1 activation.Inhibiting Kir4.1, which specifically disrupts axon-OL signaling, reduced the stimulus-and K + -evoked lactate surge in axons.Notably, the pharmacological methods used to evaluate lactate dynamics did not affect axonal activity or workload, which could affect axonal metabolism.Hence, axonal lactate surges involve OL metabolic support, as further evidenced by diminished stimulus-evoked lactate increases in OL-specific Kir4.1 knockouts.[K + ] ext increases can accelerate glycolysis 34,75 and promote lactate release through proton-linked monocarboxylate transporters 15 .Hence, a proton increase in OLs, linked to [K + ] ext (ref.53) and Kir4.1 activity 76 , may facilitate lactate release.Further research is required to clarify a K + -driven lactate or pyruvate release mechanism in OLs.Additionally, the contributing role of astrocytes 34,75,[77][78][79] , specifically in the white matter 12,80 , needs more exploration.Embryonic deletion of Kir4.1 from both astrocytes and OL lineage cells causes severe white matter pathology, including myelin vacuole formation 23,51,81,82 .Fast axonal spiking induces a rapid increase in OL [Ca 2+ ] and glycolysis.OLs primarily detect axonal activity through elevated [K + ] ext and activation of Kir4.1 channels.This K + -mediated signaling facilitates the supply of lactate (or pyruvate) to axons.Apart from regulating acute metabolic coupling, oligodendroglial Kir4.1 adjusts the myelinic levels of MCT1 and GLUT1.In addition to lactate, OLs might supply axons with glucose and/or modulate axonal glucose uptake at the nodes of Ranvier.Oligodendroglial K + homeostasis also influences axonal glycolysis, which is likely critical for preserving axonal integrity through various glucose metabolism-dependent processes, such as antioxidant protection through the pentose phosphate pathway (PPP), biosynthesis of molecules required for structure and function, regulation of the redox state, and vesicular transport.The potential contribution of astrocytes as a source of (glycogen-derived) lactate (or pyruvate) for axons is not depicted in this scheme, pending future studies.Nav, voltage-gated sodium channel; Kv, voltage-gated potassium channel.
Yet, in the adult white matter, the glial syncytium appears less critical for K + homeostasis 84 .The CAP recordings in this study suggest that OLs may independently handle the activity-induced K + clearance in adult optic nerves.Beyond acute metabolic coupling, oligodendroglial Kir4.1 may also adjust the support machinery of OLs, indicated by reduced myelinic GLUT1 and MCT1 levels in Kir4.1 cKO mice.Optic nerve proteome analysis revealed a decrease in proteins essential for intracellular membrane trafficking, including vesicle-associated membrane protein 2 (VAMP2), VAMP3, Ras-related protein Rab-2B (RAB2B) and RAB5A, reported to be expressed by mature OLs [85][86][87][88] , implicating Kir4.1 in regulating metabolite transporter trafficking [89][90][91] .Previous research has linked a 50-80% decrease in myelinic MCT1 or GLUT1 levels to late-onset axonal pathology in mice, typically emerging between 8 and 24 months of age 9,10,13 .In this study, Kir4.1 cKO mice showed signs of axonopathy at 7-8 months, preceded by a 50% reduction in both GLUT1 and MCT1 levels at 2-3 months.Moreover, this is associated with early deficits in axonal energy metabolism.In young Kir4.1 cKO mice, axons exhibited lower basal lactate levels and reduced glucose uptake and consumption.These deficits likely result from impaired OL metabolite supply and axonal glucose metabolism.Despite these metabolic changes, axonal ATP homeostasis appeared unaffected.Notably, glucose metabolism is crucial for various cellular functions beyond ATP production 92,93 , including antioxidant protection through the pentose phosphate pathway 94 and providing glycolytic intermediates for vital structural and functional molecules (Fig. 8).Additionally, axonal glycolysis is key to sustaining fast axonal transport 95 .Hence, impaired axonal glucose metabolism might affect axonal transport and increase vulnerability to oxidative stress, potentially leading to axonal damage with age.The metabolic alterations in the axons of Kir4.1 cKO mice are unlikely due to impaired K + clearance.Extracellular K + accumulation, secondary to defective clearance, should increase axonal workload and stimulate glucose metabolism 96,97 , yet Kir4.1 cKO axons showed decreased glucose metabolism.The primary impact of impaired K + clearance was seen on the speed of CAP peak recovery following high-frequency stimulation, not at basal activity levels.Hence, the reduced lactate levels and glucose metabolism in axons at basal activity are more likely due to deficits in the axon-OL metabolic unit.Furthermore, the unchanged axonal ATP dynamics in cKO mice indicate stable axonal energy demand during spiking.Consequently, the reduced stimulus-evoked lactate and glucose dynamics were not influenced by acute changes in axonal workload.We also found normal mitochondrial respiration in the axons of Kir4.1 cKO mice, with no signs of inflammation or glial activation affecting axonal metabolism at a young age.Notably, this study uncovered that OLs have a crucial role in regulating axonal glucose metabolism, including both uptake and consumption.Experiments in corpus callosum slices have shown that glucose loading in OLs could sustain callosal CAPs in aglycemia conditions, indicating OLs' ability to transfer glucose to axons 54 .The reduced myelinic GLUT1 abundance, along with reduced glucose uptake in cKO axons, suggests that OLs and myelin affect glucose delivery to axons.The influence of OLs on axonal glycolysis needs further investigation.OLs could affect axonal energy homeostasis through exosome signaling 98,99 .Yet, whether [K + ] ext and Kir4.1 functions are involved in exosome release, similar to glutamate signaling 100 , remains to be seen.Considering the role of the Na + pump in regulating neuronal energy metabolism 96,97 , adaptations in axonal Na + pumps might influence axonal glycolysis.Conclusively, this study underscores the intricate metabolic interactions between OLs and axons, inviting further exploration into the regulation of axonal glucose metabolism, which could have implications for axonal degeneration in aging and neurodegenerative disease.

Tamoxifen treatment
PLP-CreERT;RCL-GCaMP6s mice, aged 6-8 weeks, were treated with tamoxifen (Sigma-Aldrich, T5648).Tamoxifen was freshly prepared in corn oil (Sigma-Aldrich, C8267) at a concentration of 10 mg ml −1 for each experimental group.Mice received intraperitoneal injections of 100 mg tamoxifen per kilogram body weight daily for 3 days.Experiments commenced 4 weeks after the initial tamoxifen injection.

AAV vectors
The single-stranded (ss) or self-complementary (sc) AAV vectors used in this study were produced, purified and quantified by the viral vector facility of Neuroscience Center Zurich, as previously described 103 .
Intravitreal AAV injections were performed with undiluted AAV vectors mixed with fluorescein dye (0.01 mg ml −1 in PBS).The green fluorescein dye was added to monitor successful injections into the vitreous.

Optic nerve electrophysiology and two-photon imaging
Acute optic nerve preparations for concurrent electrophysiology and two-photon imaging were conducted as previously described 28 .Following isoflurane anesthesia and decapitation, optic nerves were excised and placed in a modified interface perfusion chamber (Haas Top, Harvard Apparatus), perfused with ACSF at 37 °C using a TC-10 temperature control system (NPI Electronic), and continuously oxygenated with 95% O 2 and 5% CO 2 .Nerve ends were inserted into custom-made suction electrodes filled with ACSF.This setup was integrated with a custom two-photon microscope 106 featuring a Chameleon Ultra II Ti:sapphire laser (Coherent) and a 25× water immersion objective (XLPLN 25×/1.05WMP2, Olympus), operated by ScanImage software (r3.8.1, Janelia Research Campus 107 ).The nerve was allowed to equilibrate in the perfusion chamber beneath the objective for at least 30 min, ensuring stability.A transistor-transistor logic trigger driven by a stimulus generator (STG4002-1.6mA,Multichannel Systems) was used to synchronize the acquisition of both electrophysiology and imaging data.

CAP recordings and analysis
Optic nerves were stimulated using an STG4002-1.6mAstimulus generator (Multichannel Systems) controlled by MC_Stimulus software.We used 50-µs, 0.8-mA square-wave pulses to evoke CAPs, recorded by a USB-ME16-FAI system (Multichannel Systems) with a 4× gain preamplifier.Data were collected at 50 kHz by MC_Rack software and analyzed using a custom MATLAB script available on GitHub (https://github.com/EIN-lab/CAP-analysis). CAP responses, characterized by three peaks indicating different axonal conduction speeds 10,28,108 , decrease peak amplitude and increase latency during high-frequency stimulation 28 .The analysis focused on the second CAP peak, measuring latency https://doi.org/10.1038/s41593-023-01558-3from stimulus onset to peak.The pCAP area was used to integrate the amplitude and latency changes of the first two peaks, reflective of large and medium-sized axons.Stimulation protocols involved an initial 0.4-Hz stimulation for 1 min for baseline values, followed by 30 s or 1 min at 10, 25 or 50 Hz, then 4-5 min recovery at 0.4 Hz.High-frequency CAPs were recorded every second.GD and MI experiments used a 0.1-Hz stimulation protocol.Nerve excitability was assessed by varying stimulus intensity from 0.1 to 1 mA, measuring the AUC of graded CAPs, with the CAP area expressed as a percentage of the maximum CAP area at 1 mA.

Calcium imaging and analysis
GCaMP6s imaging involved exciting the sensor at 940 nm with laser powers between 5 and 10 mW.Fluorescence emissions were captured using a GaAsP photomultiplier tube (Hamamatsu Photonics) and a 520/70-nm band-pass filter (Semrock).Overview images were taken at 512 × 512 resolution with a 3.2-µs pixel dwell time.For Ca 2+ imaging in OLs or axons, the selected field of view, 15-20 µm below the nerve surface, used 7-8× digital zoom, capturing images at 2.96 Hz and 128 × 128 resolution with a 12.8-µs pixel dwell time.Data analysis was conducted using MATLAB (MathWorks, R2015b) with the CHIPS toolbox 109 (https:// github.com/EIN-lab/CHIPS),as previously described 78,84,110 .Regions of interest (ROIs) around OL somas were manually outlined in ImageJ, with ROI masks inputted into the CHIPS pipeline.For axons, whole frames were analyzed.Image sequences underwent motion correction, noise reduction using a two-dimensional spatial Gaussian filter (σxy = 2 µm) and a temporal moving-average filter (width = 1 s), and background noise (defined as the bottom first-percentile pixel value) subtraction.Signal vectors (ΔF/F) from each ROI or frame were computed using ten frames before stimulation onset as the baseline.Ca 2+ responses were quantified by the AUC during a 30-s stimulation period.Pharmacological effects were assessed through paired analysis comparing the responses before and after drug application, averaged over two to three sessions to minimize variability.The numbers of mice and cells are detailed in the figure legends.

Metabolite imaging and analysis
The FRET-based metabolite sensors (Laconic, ATeam1.03,FLIIP) were excited at 870 nm using laser powers of 5-15 mW.Donor and acceptor fluorescence signals were simultaneously collected using two photomultiplier tubes, a 560-nm edge dichroic beam splitter (BrightLine, Semrock), and band-pass filters of 545/55 nm (yellow channel) and 475/50 nm (blue channel).Images were captured at 256 × 256 resolution with a 6.4-µs pixel dwell time, every 2 or 10 s.FRET analysis used a custom MATLAB script, available at https://gitlab.com/einlabzurich/fretanalysis.Image time series were smoothed using a five-image moving average, with the number of images determined after testing various time windows to preserve temporal dynamics.Background removal was achieved by thresholding, with average whole-frame intensities extracted for each channel.Donor-to-acceptor (Laconic) or acceptor-to-donor (ATeam1.03 and FLIIP) ratios were calculated and normalized to baselines or quasi-zero time points.For visualizing OL or axonal structures while maintaining quantitative FRET ratio data, a color scale was applied to ratio images obtained by pixel-by-pixel division of two channels over an average of 20 frames.RGB images were converted to YCbCr color space, where Cb and Cr represent color and Y represents brightness.The y coordinate was replaced with the square root of the sum of donor and acceptor images, and the image was reconverted to RGB.

Immunohistochemistry
Mice were anesthetized with pentobarbital and perfused with 2% paraformaldehyde (PFA) in PBS.Optic nerves were postfixed for 1 h in 4% PFA, embedded in frozen section medium (Thermo Fisher Scientific) and cut into 12-µm longitudinal sections using a Leica CM3050 S cryostat.Sections were placed on SuperFrost Plus slides (Thermo Fisher Scientific).For immunohistochemistry, slides were first treated with 0.3% Triton X-100 in 50 mM Tris buffer (pH 7.4) with 4% normal donkey serum for 1 h at room temperature.Primary antibodies (Table 1) were incubated overnight at 4 °C in the same solution.Secondary antibodies, mixed with 0.05% Tris-Triton and 4% serum, were applied for 1 h at room temperature.DAPI was added for nuclear staining.Confocal images and z stacks (10 µm) were acquired with a Zeiss LSM 700 or Zeiss LSM 800 confocal laser scanning microscope equipped with a 40× objective (Plan-Apochromat, numerical aperture 1.4,Oil DIC (UV) VIS-IR).Image analysis was performed with ImageJ (Fiji version 1.52p).For GFAP and IBA1 analysis, maximum intensity projections were binarized and the fluorescent particle area was determined.For quantifications, two images per section and four sections per animal were analyzed.

EM and analysis
Mice were anesthetized using isoflurane before decapitation and optic nerve extraction.Optic nerves were immediately transferred to a fixative solution (4% PFA, 2.5% glutaraldehyde in phosphate buffer with 0.5% NaCl, pH 7.4) and fixed overnight at 4 °C.Tissue preparation and EM were carried out as previously described 111 .Briefly, following postfixation with 2% OsO 4 (Science Services) in 0.1 M phosphate buffer (pH 7.3) and acetone dehydration, tissue fragments were embedded in EPON (Serva).Ultrathin sections were cut with a Leica UC7 ultramicrotome (Leica) and then stained using UranyLess (Science Services).EM pictures were captured with a Zeiss EM912 electron microscope (Zeiss) equipped with an on-axis 2k charge-coupled device camera (TRS).EM image analysis was performed using ImageJ (Fiji, version 1.52p).Axonal diameters and g ratios (ratio of axon diameter to the diameter including the myelin sheath) were analyzed using five random overview EM pictures (at 8,000× magnification), with 250 axons evaluated per animal.Calculations were based on circular areas equivalent to the measured areas.For axonal pathology assessment, eight to ten EM images per animal were analyzed.The experimenters were blinded to the genotypes.

Proteomics and analysis
Optic nerves from 2.5-month-old Kir4.1 cKO mice and littermate controls were extracted following deep isoflurane anesthesia.For tissue homogenization, we used PreOmics lysis buffer with 0.5-mm zirconium oxide beads (ZROB05, Next Advance) in a Bullet Blender (BBX24, Next Advance), with two 15-s cycles set at speed 10.TMT-based quantitative proteomics analysis was performed at the Functional Genomics Center Zurich.Protein concentrations were measured using the Lunatic UV-Vis spectrophotometer (Unchained Labs).Processing involved the iST kit (PreOmics) and encompassed treatment with 'Lyse' buffer, boiling, digestion and centrifugation.Peptides captured by the iST filter were washed, eluted, dried and labeled with TMT10plex reagent (Thermo Fisher Scientific, 90110) in acetonitrile (Sigma-Aldrich) and 50 mM TEAB (pH 8.5), followed by quenching.Equal TMT channel amounts formed the combined sample, prefractionated using high-pH reverse-phase chromatography on an XBridge Peptide BEH C18 column (Waters).For MS, samples were analyzed on an Orbitrap Fusion Lumos (Thermo Fisher Scientific), with data-dependent acquisition settings optimized for MS spectra and MS/MS recordings in the Orbitrap.Proteomics data were handled using the local laboratory information management system 112 .Proteome Discoverer (version 2.4) processed the raw MS data, with protein identification through the Sequest HT engine against the Mus musculus reference proteome (UniProt, 20190709).Carbamidomethylation and TMT (+229.163Da, peptide N-terminus and K) were set as fixed modifications, whereas methionine oxidation and N-terminal protein acetylation were set as variable modifications.Enzyme specificity was set to trypsin, allowing a minimal peptide length of six amino acids and a maximum of two https://doi.org/10.1038/s41593-023-01558-3missed cleavages.The maximum FDR for peptides was set to 0.01.Protein fold changes were computed based on intensity values.A set of functions implemented in R package prolfqua 113 was used to filter for proteins with two or more peptides.Data normalization used a modified robust z-score transformation, preserving the original data variance.For differential analysis, we fitted the linear model to every protein, computed contrasts, and moderated the variance, t statistics and P values 114 .FDRs were determined from P values using Benjamini-Hochberg adjustment.GSEA was conducted using WebGestalt.org.

Myelin purification and immunoblotting
A myelin-enriched lightweight membrane fraction was purified from the brains of 2.5-month-old Kir4.1 cKO mice and controls, using sucrose density centrifugation and osmotic shocks as previously described 115 .
Brain tissue was extracted after decapitation following deep anesthesia using isoflurane.Protein concentrations in brain lysates and myelin fractions were determined using the DC protein assay kit (Bio-Rad) following the manufacturer's instructions and measured using the Eon high-performance microplate spectrophotometer (BioTek).Immunoblotting was carried out as previously outlined 18 .Myelin fraction samples were diluted in 4× SDS sample buffer (glycerol 40% (w/v), Tris-HCl (pH 6.8) 240 mM, SDS 8% (w/v), bromophenol blue 0.04% (w/v)), with 5% dithiothreitol added as a reducing agent.Before use, samples were heated at 40 °C for 10 min.For protein separation by SDS-PAGE, the Mini-PROTEAN Handcast system (Bio-Rad) was used with self-casted acrylamide gels (10-15%).Samples were loaded at 5-15 µg per well (depending on the protein of interest) next to 5 µl of prestained protein ladder (PageRuler, Thermo Fisher Scientific).Proteins were separated by constant current (200 V) for 45-60 min.Immunoblotting was carried out with a Novex semi-dry blotter (Invitrogen), and proteins were transferred to an activated (100% ethanol, 1 min; followed by two washing steps with water) polyvinylidene fluoride membrane (Roche Diagnostics, cat.no.03010040001) at 20 V for 40 min.After blotting, membranes were blocked in 1× TBS containing 5% nonfat dry milk (Frema) and 0.05% Tween-20 for 1 h at room temperature.Primary antibodies were diluted in 5 ml of blocking buffer and incubated overnight at 4 °C with horizontal rotation.Membranes were washed three times with TBS-T for 5-10 min each and incubated for 1 h with secondary HRP antibodies diluted in a blocking buffer.Membranes were washed three times with TBS-T for 5-10 min.Detection was carried out using enhanced chemiluminescent detection (ECL) according to the manufacturer's instructions (Western Lightning Plus-ECL or SUperSignal West Femto Maximum Sensitivity Substrate, Thermo Fisher Scientific).Immunoblots were scanned using the ECL imager Chemostar (Intas Science Imaging).Normalization was performed with Fast Green total protein stain.For antibody information, see Table 1.

Real-time qPCR
Mice were deeply anesthetized using isoflurane and then decapitated.Next, optic nerves were quickly dissected.The tissues were homogenized in 350 µl RNeasy lysis buffer (RLT, Qiagen) with 0.5-mm zirconium oxide beads (ZROB05, Next Advance) in a Bullet Blender tissue homogenizer (BBX24, Next Advance; two cycles of 15 s on speed 10).

Statistical analyses
Intergroup

Reporting summary
Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Fig. 5 |
Fig. 5 | Minor changes in axonal ATP dynamics in the absence of oligodendroglial Kir4.1.a, AAV-mediated ATP FRET sensor (ATeam1.03)expression in optic nerve axons.Color-coded ratio images from control (top) and cKO (bottom) nerves show ATP levels in ACSF with 10 mM glucose and after GD + MI with 5 mM NaN 3 .hSyn1, human synapsin 1; mseCFP, monomeric superenhanced cyan fluorescent protein.b, Time course of axonal ATP levels in ∼3-month-old cKO (n = 6) and control (n = 6) mice challenged with GD + MI, normalized to the minimum ATP level (R 0 ).Inset, initial ATP recovery dynamics following reperfusion with 10 mM glucose.c, Basal axonal ATP levels were comparable between genotypes (n = 6 mice, P = 0.61, two-sided unpaired t test).d, No difference in the ATP decline rate between genotypes upon GD + MI (n = 6 mice, P = 0.44, two-sided unpaired t test).e, Similar initial ATP recovery rates

Fig. 8 |
Fig. 8 | Activity-mediated model of axon-OL metabolic coupling.The scheme shows a working model in which axon-OL communication and metabolic coupling in the white matter are controlled by K + and Kir4.1-mediated signaling.Fast axonal spiking induces a rapid increase in OL [Ca 2+ ] and glycolysis.OLs primarily detect axonal activity through elevated [K + ] ext and activation of Kir4.1 channels.This K + -mediated signaling facilitates the supply of lactate (or pyruvate) to axons.Apart from regulating acute metabolic coupling, oligodendroglial Kir4.1 adjusts the myelinic levels of MCT1 and GLUT1.In addition to lactate, OLs might supply axons with glucose and/or modulate axonal

Table 1 | Antibody information Antibody Host species, type Method, dilution Source, cat. no.
as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.The images or other third party material in this article are included in the article's Creative Commons license, unless indicated otherwise in a credit line to the material.If material is not included in the article's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.To view a copy of this license, visit http://creativecommons. org/licenses/by/4.0/.MethodsAnimalsAll animal experiments were permitted by the local veterinary authorities in Zurich, in agreement with the guidelines of the Swiss Animal Protection Law, Veterinary Office, Canton Zurich (Animal Welfare Act, December 16, 2005, and Animal Welfare Ordinance,April 23, 2008).PLP-CreERT;RCL-GCaMP6s mice were generated by crossing PLP-CreERT mice (RRID:IMSR_JAX:005975) 101ele with MOGiCre mice101.Kir4.1 fl/fl mice were used as littermate controls.Transgenic mouse lines were maintained on the C57BL/6 background.For experiments in wild types, we used Charles River C57BL/6 mice.Both male and female mice were used for experiments.Mice were kept in an inverted 12-h light/dark cycle at 23 °C and 55% humidity.Food and water were available ad libitum.
comparisons were performed by a two-sided paired t test or Student's t test, as specified in the figure legends.For analyses involving multiple comparisons, data were subjected to either one-way or two-way ANOVA.All statistical analyses were conducted using GraphPad Prism 9 or R (v.3.2.2, R Core Team, 2015).Datasets had to pass the Shapiro-Wilk test for normality before being subjected to Student's t test and ANOVA or undergo nonparametric tests to determine statistical significance.Equal variances were assumed but not tested.Significance levels were defined as follows: *P < 0.05, **P < 0.01, ***P < 0.001.Data are presented as individual values with mean ± s.e.m. or as box-and-whisker plots.In the box-and-whisker plots, the '+' symbol represents the mean; the center line indicates the median; the box extends from the 25th to 75th percentile; and the whiskers extend from the 5th to 95th percentile.P values and sample sizes are stated in the figure legends.Sample sizes were not determined in advance as they were constrained by the availability of age-matched transgenic mouse cohorts.Where applicable, experimenters were blinded to the genotype.