Oligodendrocytes control potassium accumulation in white matter and seizure susceptibility

1Action potentials lead to increases in extracellular K+ that if left unchecked can impair membrane repolarization, induce tonic firing and trigger seizures. The redistribution of extracellular K+ following neuronal activity is mediated through diffusion and uptake into cells (K+ buffering), particularly glial cells which maintain a high resting conductance to K+ and a highly negative resting potential. In gray matter, astrocytes have been shown to participate in K+ buffering, but much less is known about the mechanisms that enable K+ clearance in white matter, where astrocyte access to axons is limited to nodes of Ranvier and the majority of axonal K+ channels are located beneath the myelin sheath. Oligodendrocytes have extended contact with axons and exhibit a high resting conductance to K+, but their contribution to K+ clearance in white matter and the impact of this K+ redistribution on neuronal activity has not been determined.

2K+ entry into glial cells is facilitated, in part, by inwardly rectifying K+ channels (Kir channels) that promote unidirectional movement of K+ across membranes. Kir4.1 is the most abundant K+ channel expressed by astrocytes and oligodendroglia and has been shown to help establish their resting membrane potential and to redistribute extracellular K+ in gray matter following neuronal activity. The importance of Kir4.1 in K+ homeostasis is underscored by the severe neurological phenotype of Kir4.1 knockout mice, which exhibit ataxia, seizures, deafness, widespread myelin pathology, and early death. Many of these features are also observed in human patients with SeSAME/EAST syndrome, a rare genetic disease caused by loss-of-function mutations in Kir4.1. Polymorphisms in the Kir4.1 gene are associated with idiopathic epilepsy and autism spectrum disorder with seizures, futher supporting a close functional link between Kir4.1 dysfunction and neuronal hyperexcitability. Kir4.1 expression levels are reduced in a vast array of CNS pathologies (for review, see ), and restoration of astrocyte Kir4.1 levels in the striatum of Huntington’s disease model mice enhanced neuronal suvival and ameliorated motor deficits, suggesting that impaired K+ buffering may contribute to both genetic and aquired neurological disease.

3Although the pathological sequelae arising from Kir4.1 dysfunction are commonly attributed to astrocytes, oligodendrocyte precursor cells (OPCs) and mature oligodendrocytes have also been shown to express Kir4.1, by immunohistochemistry, reporter gene expression, RNA-Seq and recordings of Kir-mediated currents. OPCs that lack Kir4.1 fail to fully mature into oligodendrocytes in vitro, and white matter tracts in glial-specific Kir4.1 knockout mice exhibit extensive vacuolization, suggesting that this K+ channel performs diverse roles in oligodendroglia. Functional studies indicate that selective depolarization of oligodendrocytes in white matter alters the conduction speed of action potentials, an effect that is abolished by application of a Kir channel blocker, suggesting that oligodendrocyte Kir4.1 channels can shape neuronal activity on a rapid time scale. However, the relative contribution of oligodendroglial Kir4.1 has been difficult to assess in vivo, as pharmacological manipulations and genetic deletion studies, performed in either global knockouts or early progenitor cells that give rise to both astrocytes and oligodendroglia, do not distinguish the contributions of these two cell types. As a result, the role of Kir4.1 channels in regulating oligodendrocyte development and function remains poorly understood.

4Here, we used conditional genetic strategies to define the role of Kir4.1 channels in shaping oligodendrocyte development and maintaining myelin, and the participation of oligodendrocytes in controlling K+ clearance and neuronal activity within white matter. Although selective deletion of Kir4.1 from OPCs raised their membrane potential and membrane resistance, it did not alter their survival, proliferation or their ability to form oligodendrocytes in vivo. Selective deletion of this channel from mature oligodendrocytes was similarly benign, as it had no effect on their survival or ability to form myelin. However, despite the appearance of normal myelin in the absence of oligodendrocyte Kir4.1, these mice exhibited spontaneous seizures, increased chemoconvulsant sensitivity, and activity-dependent deficits in motor behavior. Physiological studies in the corpus callosum and optic nerve revealed that K+ clearance rates in white matter of these mice were markedly slowed and axons were impaired in their ability to sustain repetitive firing. Together, these findings indicate that oligodendrocytes play a crucial role in controlling K+ homeostasis and that disruption of Kir4.1 expression in oligodendrocytes is sufficient to promote neuronal hyperexcitability and seizures.

5Constitutive deletion of Kir4.1 from glia (Gfa2-Cre;Kcnj10fl/fl mice) leads to dramatic changes in the structure of myelin and in vitro studies of oligodendrocytes lacking Kir4.1 suggest that this channel is required for both their maturation and survival. However, the role of Kir4.1 in shaping the development of oligodendroglia in vivo remains largely unexplored. We deleted Kir4.1 from the nervous system by crossing Kcnj10fl/fl mice with Nes-Cre mice, which express Cre recombinase in neural progenitor cells. These nervous-system-specific Kir4.1 conditional knockout mice (termed nKir4.1cKO) exhibited retarded growth, ataxia, tremor, and early mortality, with most mice dying by postnatal day (P) 25 (Figure 1A–B; Figure 1—video 1), similar to the phenotype of Kir4.1 global knockout and glia-specific Kir4.1 knockout mice. Kir4.1 immunoreactivity was no longer observed in the brains of these mice (Figure 1C), demonstrating effective deletion of this channel from the CNS. Immunostaining for myelin basic protein (MBP) revealed dense myelinated tracts in the corpus callosum and striatum, indicating that oligodendrocytes were formed; however, these regions exhibited widespread vacuolization of myelin, with swellings visible along most internodes (Figure 1D–F; Figure 1—figure supplement 1). Despite these striking morphological changes, the density of oligodendrocytes in nKir4.1cKO mice at P24, measured by immunostaining for the oligodendrocyte protein aspartoacylase (ASPA), was comparable to control mice (Nes-Cre;Kcnj10fl/+) (Figure 1G,H), indicating that oligodendrocytes are generated and mature in the absence of Kir4.1.

6Oligodendrocytes are maintained by resident OPCs, which proliferate to replace those that transform into new oligodendrocytes. Because of this robust homeostatic response, oligodendrocyte density can be maintained despite profound loss, which could mask underlying oligodendrocyte death. To assess whether there is accelerated turnover of oligodendrocytes in nKir4.1cKO mice, we examined the proportion of OPCs that were immunoreactive to the cell division marker Ki67 (Figure 1I). This analysis revealed that OPC density was unchanged (Figure 1J) in nKir4.1cKO mice, and their proliferation was reduced, rather than increased (Figure 1K), suggesting that oligodendrocyte turnover is not increased.

7To assess the physiological properties of oligodendrocytes when Kir4.1 is absent from the CNS, we crossed nKir4.1cKO mice with Mobp-EGFP mice, allowing oligodendrocytes to be visualized for targeted recording in acute slices (Figure 1L). EGFP+ oligodendrocytes in the corpus callosum of these mice at P21 were significantly depolarized compared to controls (Nes-Cre;Mobp-EGFP) (control: –77 ± 2 mV, n = 5; cKO: –31 ± 1 mV, n = 4; p=7.6 × 10–7) (Figure 1M), but maintained the membrane resistance and linear current-voltage response characteristic of WT oligodendrocytes (Figure 1M,N). Together these results indicate that deletion of Kir4.1 from the CNS leads to profound changes in the structure of myelin and depolarization of oligodendrocytes but does not impair their maturation or accelerate their degeneration during development.

8Kir4.1 is the most abundant K+ channel expressed by OPCs, and upregulation of these channels, as assessed by the appearance of Ba2+-sensitive Kir currents, correlates with development of a more negative OPC membrane potential and acquisition of a linear I-V curve profile over the course of postnatal development, suggesting that Kir4.1 plays a key role in establishing the membrane properties of these progenitors. To determine if Kir4.1 channels influence the physiological properties of OPCs and their developmental progression in vivo, we selectively deleted these channels in OPCs by crossing Pdgfra-CreER mice with Kcnj10fl/fl mice. A sensitive Cre-dependent EGFP reporter transgene (ROSA26-CAG-EGFP (RCE)) was included to mark cells in which Cre activity was induced and recombination was initiated at P21 with 4-hydroxytamoxifen (4-HT). EGFP+ OPCs in the corpus callosum, hippocampus, and cortex of these conditional knock-out mice (termed pKir4.1cKO) were targeted for whole cell recording in acute slices and compared to OPCs in control mice (Pdgfra-CreER;RCE) (Figure 2A). OPCs were readily distinguished from oligodendrocytes generated in the intervening 2 weeks by their radial morphology (Figure 2B) and presence of a depolarization-induced sodium current (Figure 2C). OPCs in control mice exhibited prominent outward and inward currents in response to current injection, resulting in a near linear current-voltage (I-V) relationship (Figure 2D,F; Figure 2—figure supplement 1A). Application of a low concentration of BaCl2 (100 µM), which inhibits Kir4.1 channels, blocked the hyperpolarization-induced inward current (Figure 2D,F; Figure 2—figure supplement 1A,B). In contrast, the majority of OPCs (49/65, 76%) in pKir4.1cKO mice exhibited greatly reduced inward currents in response to hyperpolarization, and BaCl2 (100 µM) did not alter their I-V relationship, indicating that Kir4.1 was no longer expressed (Figure 2E,G; Figure 2—figure supplement 2B). The remaining 24% of EGFP+ OPCs in pKir4.1cKO mice exhibited a linear I-V profile (Figure 2—figure supplement 2), suggesting that activation of CreER in these cells did not result in excision of both floxed Kir4.1 alleles over this time period.

9OPCs in both gray and white matter that lacked Kir4.1 were ~20 mV more depolarized and exhibited a ~ 10 fold higher membrane resistance (measured at –80 mV with a 10 mV depolarization step) than control OPCs (Figure 2H–I; Figure 2—figure supplement 1C–D). Similar changes were induced in OPCs of control mice when Kir4.1 was inhibited with 100 µM BaCl2 (Figure 2H–I; Figure 2—figure supplement 1C–D), indicating that these changes reflect the direct contribution of Kir4.1 rather than other compensatory changes induced by deletion of these channels. These results indicate that Kir4.1 channels dominate the resting conductance of OPCs and force these progenitors to adopt a more hyperpolarized membrane potential.

10Previous in vitro studies have shown that expression of Kir4.1 is sufficient to hyperpolarize and cause G1/G0 arrest of glioma cells and flux of K+ through voltage-gated K+ channels has been shown regulate OPC proliferation and differentiation in vitro. To determine how the depolarization induced by Kir4.1 deletion influences OPC behavior, we performed in vivo fate tracing using pKir4.1cKO mice bred to R26R-EYFP mice. In the RCE reporter line used for electrophysiological recordings, nearly 100% of OPCs expressed EGFP (Figure 3—figure supplement 1), yet Kir4.1 currents were absent in only 76% of OPCs (Figure 2—figure supplement 2). Therefore, for these studies we used a R26R-EYFP reporter line that exhibits a recombination efficiency (78% of OPCs) comparable to the proportion of OPCs that lack Kir4.1 currents in cKO mice (Figure 3—figure supplement 1). Control (n = 6) and pKir4.1cKO (n = 7) mice received 4-HT at P21 and the density of OPCs in gray and white matter was assessed at P35 by immunostaining for PDGFRα to identify all OPCs, and EYFP to identify OPCs that have expressed Cre recombinase and lack Kir4.1. If OPC proliferation or survival is impaired, a decrease in the relative proportion of EYFP+ OPCs is expected; however, there was no change in the density of OPCs or the proportion of EYFP+ OPCs between control and pKir4.1cKO mice (Figure 3—figure supplement 2).

11To assess proliferation more directly, the proportion of dividing OPCs was determined by administering bromodeoxyuridine (BrdU) twice daily from P28 to P34 (Figure 3A). Although the proportion of BrdU+ OPCs varied by region, from ~45% in hippocampus to ~60% in corpus callosum, no differences were observed between EYFP+ and EYFP– populations in either region (Figure 3B–E; Figure 3—figure supplement 3A–C), indicating comparable rates of proliferation. At this time point, EYFP+ cells consisted of ~75% PDGFRα+ OPCs and ~15% CC1+ oligodendrocytes in the hippocampus, and ~45% OPCs and ~40% oligodendrocytes in the corpus callosum, with no significant differences observed between control and Kir4.1-deleted OPCs (Figure 3F–H; Figure 3—figure supplement 3D–F). Thus, although OPCs lacking Kir4.1 exhibited markedly higher membrane resistance and more depolarized membrane potential, remarkably, these changes did not alter their proliferation or differentiation into oligodendrocytes.

12Kir4.1 continues to be expressed by oligodendroglia after differentiation, and the results above indicate that oligodendrocytes in situ are depolarized when Kir4.1 is removed from the entire CNS (Figure 1), raising the possibility that cell autonomous alterations in their physiological properties induce the malformation of myelin observed in these animals. To test this hypothesis, we crossed Kcnj10fl/fl mice with Mog-iCre mice, which express Cre recombinase specifically in mature oligodendrocytes. Unlike nKir4.1cKO mice, which have severe neurological symptoms and drastically shortened lifespan (Figure 1), these oligodendrocyte-specific Kir4.1 knockout mice (oKir4.1cKO) were grossly normal phenotypically and survived to adulthood.

13To enable analysis of both oligodendrocytes and astrocytes, and thereby verify the specificity of Kir4.1 deletion, oKir4.1cKO mice were crossed to two other lines: the RCE reporter line, to express EGFP in recombined oligodendrocytes, and a Slc1a2-EGFP line, in which astrocytes constitutively express EGFP. EGFP+ oligodendrocytes and astrocytes from the two lines were isolated using fluorescence-activated cell sorting (FACS) from 10-week-old control and oKir4.1cKO mice. Levels of Kir4.1 mRNA were quantified by reverse transcription and quantitative real-time PCR (Figure 4—figure supplement 1). As expected, oligodendrocytes (n = 6 mice) expressed high levels of mRNA for myelin proteins, while astrocytes (n = 5 mice) expressed higher levels of mRNA for glial fibrillary acidic protein (GFAP) (Figure 4—figure supplement 1A–B). Kir4.1 mRNA levels were >10 fold lower in oKir4.1cKO oligodendrocytes (n = 3) than in control oligodendrocytes (n = 3) (RQ = 0.063, 95% CI = 0.035–0.112) (Figure 4—figure supplement 1C), while oKir4.1cKO (n = 3) and control (n = 2) astrocytes showed no difference in Kir4.1 mRNA levels (RQ = 1.52, 95% CI = 0.32–7.10) (Figure 4—figure supplement 1D).

14To assess the role of Kir4.1 in cell-autonomously regulating the membrane properties of oligodendrocytes, EGFP+ cells were targeted for whole cell recording in the corpus callosum of 5 to 6-week-old control;RCE and oKir4.1cKO;RCE mice. In contrast to nKir4.1cKO mice (Figure 1), no significant changes in resting membrane potential (RMP) (Control: –70 ± 1 mV (n = 17) vs. oKir4.1cKO: –68 ± 1 mV (n = 26), p=0.25), membrane resistance (Rm) (Control: 33 ± 3 MΩ vs. oKir4.1cKO: 40 ± 4 MΩ, p=0.22), or I-V response of EGFP+ oligodendrocytes were observed following selective deletion of Kir4.1 (Figure 4A–B). In an independent group of oligodendrocytes recorded in the alveus of the hippocampus, there was similarly no change in Rm after Kir4.1 deletion (Control: 27 ± 2 MΩ, n = 16 vs. oKir4.1cKO: 28 ± 3 MΩ, n = 18; p=0.86) (data not shown). These findings indicate that Kir4.1 does not influence the membrane properties of mature oligodendrocytes, indicating that other K+ conductances dominate at this stage.

15To investigate the role of other barium-sensitive K+ channels in maintaining the membrane conductance of control and oKir4.1cKO oligodendrocytes, membrane resistance was measured before and after application of BaCl2 (100 µM). Application of BaCl2 increased the membrane resistance of control and oKir4.1cKO oligodendrocytes to a similar degree (ΔRmcontrol: 15.2 ± 3.2 MΩ, n = 25 vs. oKir4.1cKO: 19.9 ± 3.5 MΩ, n = 32; p=0.33) (Figure 4—figure supplement 2). These results indicate that the change in membrane resistance produced by BaCl2 is due to a non-cell-autonomous effect, such as inhibition of astrocyte Kir4.1, or that other K+ channels in addition to Kir4.1 are inhibited by BaCl2.

16It has been widely hypothesized that Kir4.1 plays an important role in myelination, as the most striking histological feature of Kir4.1 global knockout and glia-specific knockout mice is severe spongiform vacuolation of the white matter (Figure 1). However, due to the broad expression of this channel, it is unclear whether this pathology is due to defects in oligodendrocytes or a consequence of astrocyte or OPC dysfunction. To determine the role of oligodendrocyte Kir4.1 in myelination, we examined the white matter of oligodendrocyte-specific Kir4.1 knockout mice. To assess whether axons are functionally preserved in the absence of oligodendrocyte Kir4.1, compound action potentials (CAPs) reflecting conduction along myelinated and unmyelinated axons were recorded in the corpus callosum (Figure 4—figure supplement 3A,B). No difference in conduction velocity was observed for myelinated or unmyelinated axon populations between control and oKir4.1cKO mice (Figure 4D).

17In contrast to the striking white matter pathology exhibited by nKir4.1cKO mice, the brain and spinal cord of 10-week-old oKir4.1cKO animals exhibited grossly normal white matter with ample myelin and no apparent vacuolation (Figure 4E–F). Moreover, there was no change in the density of EGFP+ oligodendrocytes (Figure 4G–H), indicating that oligodendrocyte survival was not impaired.

18To examine the structure of myelin at higher resolution, transmission electron microscopy (TEM) was performed on control and oKir4.1cKO mice. Electron micrographs containing cross-sections of corpus callosum from 10-week-old mice demonstrated a complete absence of vacuoles or aberrant myelin in oKir4.1cKO animals (Figure 4I–J). Furthermore, the pattern of myelination, estimated by calculating g-ratios (diameter of axon/diameter of axon + myelin) (Figure 4K) revealed that there was no difference in g-ratio across all axons (Control: 0.743 ± 0.010 (n = 3 mice), oKir4.1cKO: 0.752 ± 0.001 (n = 3 mice), p=0.38) (Figure 4L), or in the relationship between axon diameter and g-ratio (Control: 0.16 ± 0.02 (n = 3 mice), oKir4.1cKO: 0.15 ± 0.01 (n = 3 mice), p=0.81) (Figure 4M). In addition, swollen or degenerating axons were completely absent from corpus callosum of oKir4.1cKO mice (Figure 4I) and immunostaining for SMI32 (non-phosphorylated neurofilament), which is upregulated in degenerating axons, revealed no axonal pathology in the corpus callosum of control or oKir4.1cKO animals (n = 6 for both genotypes) (Figure 4—figure supplement 4A–B). Brain sections were also immunostained for GFAP, as astrocytes become reactive and upregulate GFAP in response to neuronal degeneration. No difference in GFAP expression was observed between control and oKir4.1cKO animals (Figure 4—figure supplement 4C–D). These findings indicate that myelin and myelinated axons are not structurally altered as result of the selective removal of Kir4.1channels from oligodendrocytes.

19Although oKir4.1cKO mice did not exhibit alterations in the structure of myelin or progressive disability, these animals died much earlier than control littermates (Figure 5A). Kir4.1 is a risk factor gene for epilepsy, raising the possibility that Kir4.1 dysfunction in oligodendrocytes is sufficient to increase seizure susceptibility and induce death. Consistent with this hypothesis, oKir4.1cKO mice exhibited occasional seizures and mice that had died were often discovered with limbs extended, suggesting that they had suffered a catastrophic seizure.

20To assess whether seizure susceptibility is enhanced by removal of Kir4.1 from oligodendrocytes, we compared the response of oKir4.1cKO and control mice to the chemoconvulsant pentylenetetrazol (PTZ). A stark difference was observed between groups. Nearly all (11/12) oKir4.1cKO mice exhibited tonic-clonic seizures with full hindlimb extension after being injected with PTZ, while only 1/17 control mice exhibited this level of seizure severity (Figure 5B; Figure 5—video 1). Moreover, 8/12 oKir4.1cKO mice and only 1/17 control mice experienced fatal seizures, and oKir4.1cKO mice had a faster seizure onset (Figure 5C–D), indicating that deletion of this K+ channel from oligodendrocytes enhances neuronal excitability and increases seizure susceptibility.

21Previous studies have shown that Kir4.1 is involved in buffering extracellular K+ in both gray and white matter, as extracellular K+ levels are slower to recover following neuronal stimulation in glia-specific Kir4.1 knockout mice. However, the relative contribution of astrocytes and oligodendrocytes to K+ clearance, particularly in white matter where there is extensive association between axons and oligodendrocytes, has not been assessed. To investigate the role of oligodendrocyte Kir4.1 in K+ buffering, we recorded the response of oligodendrocytes from control and oKir4.1cKO mice to repetitive axonal stimulation (100 Hz for 1 s) in two white matter tracts, the corpus callosum and the alveus of the hippocampus (Figure 6A; Figure 6—figure supplement 1A).

22In accordance with previous studies, oligodendrocytes were progressively depolarized during the stimulus train due to K+ release from active axons and gradually returned to baseline after cessation of the stimulus (Figure 6B–C; Figure 6—figure supplement 1B–C) following the gradual decline in extracellular K+ levels. Oligodendrocyte depolarization occurred more slowly in oKir4.1cKO mice, and the decay of membrane potential following cessation of stimulation was markedly prolonged relative to controls (Figure 6B–D; Figure 6—figure supplement 1B–D) (corpus callosum: 11.6 ± 1.6 s (n = 15) vs. 5.9 ± 0.7 s (n = 15), p=0.0036; alveus: 9.3 ± 0.6 s (n = 16) vs. 6.3 ± 0.4 s (n = 14), p=0.0017), which was also reflected in the area under the curve (Figure 6E; Figure 6—figure supplement 1E). Thus, while oligodendrocytes lacking Kir4.1 still depolarize in response to axonal activity, presumably due to the presence of other pathways for K+ redistribution, their recovery from the stimulus train was slower, suggesting that extracellular K+ clearance is slowed.

23To provide an independent assessment of extracellular K+ dynamics, we recorded the response of astrocytes in the corpus callosum and alveus of oKir4.1cKO mice (crossed to Slc1a2-EGFP mice to allow astrocyte visualization) (Figure 6F; Figure 6—figure supplement 1F). Astrocyte membrane potential follows the concentration of extracellular K+ closely, due to the high permeability of their membranes to K+. Using the same stimulation paradigm, astrocytes were observed to depolarize with a similar time course to oligodendrocytes (Figure 6G–H; Figure 6—figure supplement 1G–H). However, despite the presence of normal Kir4.1 expression by astrocytes, a significant slowing of their decay following the end of the stimulus was observed in oKir4.1cKO mice, comparable to that observed in oligodendrocytes (Figure 6G–J; Figure 6—figure supplement 1G–J). These findings suggest that oligodendrocyte Kir4.1 channels contribute substantially to the removal of axonally released K+ in white matter.

24To assess the functional impact of oligodendrocyte-specific Kir4.1 deletion on axonal activity in white matter, CAPs were recorded from isolated optic nerves from 10-week-old animals using suction electrodes (Figure 7A). Consistent with the behavior of axons in the corpus callosum (Figure 4D), Kir4.1 deletion from oligodendrocytes did not affect the mode (i.e. peak of the CAP) or median baseline conduction velocity of the nerves (n = 5 nerves from three mice for each genotype) (Figure 7B–C).

25It is established that high-frequency stimulation of the optic nerve results in an increase in extracellular K+ levels and a decay of the CAP waveform. In young Kir4.1 KO optic nerves, there is a slower restoration of the extracellular K+ concentration after stimulation, suggesting that Kir4.1 plays an important role in K+ clearance in this tissue. To assess the contribution of oligodendrocyte Kir4.1 to K+ clearance, optic nerves from control and oKir4.1cKO mice were stimulated at high frequency (100 Hz), which caused the CAP waveform to progressively broaden and decrease in amplitude (Figure 7D). After cessation of stimulation, the CAP waveform gradually returned to its previous form (Figure 7E). Although oKir4.1cKO nerves exhibited a similar reduction in peak amplitude during stimulation, the CAP waveform in these nerves took significantly longer to recover after stimulation ended (Figure 7E–F). The CAP integral, representing total activity in the nerve, and the median axon conduction latency also recovered more slowly in oKir4.1cKO mice (Figure 7—figure supplement 1). When the same experiments were repeated using lower frequency stimulation (20 Hz), the recovery of oKir4.1cKO CAPs was also delayed (Figure 7G, Figure 7—figure supplement 1). These findings suggest that oligodendrocyte-specific Kir4.1 deletion alters action potential conduction in both physiological and pathophysiological activity regimes.

26To determine whether removal of Kir4.1 selectively from oligodendrocytes influences normal behavior, we examined the motor function of oKir4.1cKO mice. When placed in an open field chamber, there was no difference in total locomotion or in rearing behavior of oKir4.1cKO mice (Figure 8A). However, when challenged by placement on an accelerating rotarod, oKir4.1cKO mice had significantly shorter latency to fall (Figure 8B). Moreover, when given free access to a running wheel over multiple days, control mice increased their running activity over time, eventually running 10–15 km per 24 hr period (Figure 8C), while oKir4.1cKO mice were unable to achieve the same performance level, and typically did not exceed 5 km per 24 hr. As a result, the cumulative distance run over 7 days was significantly less, as was the average speed during active running periods (Figure 8D–E). These results suggest that, under physiologically relevant behavioral conditions, oligodendrocyte Kir4.1 channels play a vital role in facilitating neuronal activity.

27Oligodendrocytes play a critical role in the mammalian CNS by forming myelin sheaths around axons that enable action potentials to be propagated rapidly with minimal energy expenditure. Although they were once assumed to be merely passive insulators, recent studies indicate that oligodendrocytes also provide metabolic support to axons in the form of lactate and contribute to a larger glial network by coupling to astrocytes through gap junctions. Removal of oligodendrocytes by exposure to the oligotoxin cuprizone, or through genetic ablation in animal models, enhances neuronal excitability that can lead to tremor and death. Moreover, oligodendrocyte loss contributes to neurological disability in diseases such as multiple sclerosis, highlighting the importance of defining the mechanisms that control their form and function. To determine the cell-autonomous role of the inwardly rectifying K+ channel Kir4.1, which has been linked to K+ clearance, neuronal hyperexcitability and myelin disruption, we selectively deleted this abundant channel from OPCs and mature oligodendrocytes in vivo. Despite high expression of this channel by oligodendroglia, genetic removal of Kir4.1 did not alter the developmental trajectory of OPCs or the ability of oligodendrocytes to form and maintain myelin. Nevertheless, mice in which Kir4.1 was selectively deleted from oligodendrocytes exhibited profound motor deficits that emerged with increasing activity, as well as spontaneous seizures that were often fatal. Recordings from both oligodendrocytes and astrocytes revealed that activity-induced K+ clearance in white matter was impaired in these mice, and axons recovered more slowly from repetitive stimulation. Together, these findings indicate that oligodendrocyte Kir4.1 channels are prominent contributors to K+ homeostasis in white matter, and that selective loss of this channel from oligodendrocytes is sufficient to reduce motor performance, enhance neuronal hyperactivity and induce seizures.

28Global removal of Kir4.1 leads to the formation of prominent vacuoles in myelin sheaths, failure of myelin compaction, and ultimately axonal degeneration (Figure 1E–F; Figure 1—figure supplement 1). Several explanations for this dramatic pathology have been proposed to link Kir4.1-deficient oligodendroglia to the abnormal formation and maintenance of myelin. Loss of this channel from OPCs may compromise their viability, impair their ability to generate oligodendrocytes or cause abnormalities in oligodendrocytes that are produced. This hypothesis is supported by the observation that OPCs (‘complex glia’) were rarely encountered in hippocampal recordings from glial-specific Kir4.1 knockout mice, in which Kir4.1 was deleted from both astrocytes and oligodendroglia. However, our studies show that OPC and oligodendrocyte densities were normal in global CNS Kir4.1 knockout (nKir4.1cKO) mice (Figure 1I–K), despite the presence of widespread myelin vacuolization; the proliferation rate of OPCs also was not increased, as would be expected if these progenitors were mobilized to replace dying OPCs and oligodendrocytes. Moreover, when Kir4.1 was specifically deleted from OPCs during the period of rapid oligodendrogenesis (P21), the survival, proliferation, and differentiation of these progenitors was unaffected (Figure 3; Figure 3—figure supplements 2–3), indicating that removal of Kir4.1 specifically from OPCs does not contribute to this white matter pathology.

29Oligodendrocytes in mixed glial cultures from Kcnj10−/− mice are significantly depolarized and fail to mature fully. We found that they were similarly depolarized in nKir4.1cKO mice in situ (Figure 1M), raising the possibility that physiological changes in oligodendrocytes impair their ability to generate or maintain myelin sheaths. However, selective deletion of Kir4.1 from oligodendrocytes did not alter their membrane properties, the structure of myelin (Figure 4), or the baseline conduction velocity of action potentials in white matter (Figure 4D; Figure 7B–C). Notably, recent studies also found that the membrane properties of satellite oligodendrocytes (oligodendrocytes located next to neuronal cell bodies) in the cerebral cortex were unaffected in Plp1-CreER;Kcnj10fl/fl mice. Oligodendrocytes express a variety of K+ leak channels that may compensate for the loss of Kir4.1, including K2P channels and the inward-rectifying channel Kir2.1. However, the pronounced depolarization of oligodendrocytes in Kcnj10−/− and nKir4.1cKO mice (Figure 1M) indicates that this potential compensation is insufficient in the context of widespread Kir4.1 deletion, suggesting that non-cell autonomous effects are crucial contributors to oligodendrocyte depolarization and myelin disruption.

30Astrocytes in Kcnj10−/− mice are also severely depolarized, and their ability to remove extracellular K+ is impaired, consistent with their high expression of this channel. These dramatic changes in membrane potential could decrease their ability to provide metabolic support to surrounding cells, reduce the production of factors that promote oligodendrocyte development and myelination (for review, see ), and disrupt ion and water balance leading to vacuolization. As oligodendrocytes form gap junctions with astrocytes, they may be particularly sensitive to these changes in astrocyte physiology. In support of this hypothesis, deletion of gap junction proteins that link astrocytes and oligodendrocytes produces myelin pathology comparable to what occurs in Kcnj10−/− mice, and heterozygote analysis indicates that there is a close genetic interaction between these connexin genes and Kir4.1 in producing myelin vacuolization. The presence of a glial syncytium may help to maintain normal oligodendrocyte membrane potentials when Kir4.1 is specifically deleted from these cells and may exacerbate oligodendrocyte depolarization when Kir4.1 is deleted globally, as astrocytes have the ability to maintain a highly stable membrane potential through extensive electric and ionic coupling between cells. Our results emphasize that both the structure and function of oligodendrocytes is critically dependent on astrocytes, and that reactive changes in astrocytes associated with reduced Kir4.1 expression may in turn precipitate pathological changes in myelin.

31OPCs are unique among macroglia in that they continue to proliferate throughout life, engaging in a homeostatic response to replace cells lost through differentiation and death. The mechanisms that guide this remarkable behavior are not well understood, but have important consequences for myelin repair, trauma-induced gliosis and brain cancer. Membrane potential fluctuates predictably during the cell cycle in many cell types, and previous studies indicate that K+ channels play an important role in regulating cell cycle progression in OPCs. Pharmacological inhibition of delayed rectifier K+ channels in cultured OPC-like O-2A cells and OPCs in cerebellar slice cultures inhibits their proliferation and differentiation by inducing G1 arrest. Subsequent studies showed that OPCs upregulate Kv1.3 during G1 phase of the cell cycle, and blockade of this channel prevents G1/S transition, while overexpression of certain Kv1 isoforms promotes their proliferation, suggesting that OPC homeostasis is critically regulated by their membrane potential and potassium conductance. Although the effect of Kir4.1 channels on OPC proliferation and lineage progression had not been evaluated, studies from developing astrocytes and glioma cells suggest that this channel is also involved in cell cycle regulation. Expression of Kir4.1 channels during astrocyte development correlates with a negative shift in RMP, cessation of cell proliferation, and increased differentiation, and astrocytes that become proliferative after injury have lower Kir current density than non-proliferating astrocytes near the injury site. In addition, heterologous expression of Kir4.1 in glioma cells induces hyperpolarization and G1 arrest, an effect nullified if cells are treated with BaCl2 to block Kir4.1 or artificially depolarized by high K+. Thus, it is remarkable that removal of Kir4.1 from OPCs in vivo, which led to a profound shift in their RMP and increase in their membrane resistance, had no discernable effect on their proliferation, density or their ability to differentiate into oligodendrocytes (Figure 3; Figure 3—figure supplements 2–3). It is possible that OPCs overcome the depolarization induced by Kir4.1 loss by transiently increasing K+ channel expression during certain phases of the cell cycle or by reducing the activity of other channels normally recruited through depolarization, such as voltage-gated calcium channels. These findings highlight both the extreme behavioral flexibility of OPCs in vivo and the powerful drive to sustain their numbers in the adult CNS. However, our studies also indicate that OPCs are profoundly influenced by their environment, as complete removal of Kir4.1 from the CNS profoundly reduced their proliferation by the third postnatal week (Figure 1K). This phenomenon occurred in the context of widespread depolarization and dysfunction of astrocytes and vacuolization of myelin, suggesting that astrocytes play a crucial role in maintaining a favorable environment for, and perhaps directly facilitating, OPC proliferation and development.

32Although OPC proliferation and differentiation is unchanged following Kir4.1 deletion, it remains to be seen whether other functions of OPCs are altered in these mice. OPCs receive synaptic input from neurons, and the strength of these inputs is likely to be altered by changes in OPC membrane potential and membrane resistance. It has been shown that OPCs are able to regulate glutamatergic neurotransmission by shedding of cleaved NG2 ectodomains, so even subtle alterations in OPC behavior have the potential to affect synaptic communication in the brain.

33Axonal Kv1 channels that control repolarization and limit re-excitation by distal nodes are abundant underneath the myelin sheath, particularly in juxtaparanodal regions that are isolated from nodes of Ranvier by septate/septate-like junctions. Although a single action potential can increase the local extracellular K+ concentration in unmyelinated tissue by as much as 1 mM from a resting level of 3–3.5 mM, and extracellular K+ levels can rise up to 10 mM during pathological conditions such as seizures, the potential for accumulation of K+ underneath myelin sheaths may be even greater due to the small peri-internodal volume and the barriers to diffusion presumed to be created by the paranodal septate junctions (but see;; ). However, the mechanisms that enable clearance of extracellular K+ from these spaces have not been defined. Pharmacological studies of K+ buffering mechanisms suggest that Kir channels and the Na+/K+ ATPase pump have distinct roles in overall K+ clearance. Analysis of K+ dynamics arising from neuronal activity in gray matter suggest that the Na+/K+ ATPase predominates in lower activity regimes, but at higher frequencies and in situations of localized K+ release, Kir4.1 plays a more prominent role in limiting extracellular K+ accumulation.

34Kir4.1 is concentrated within the fine processes of astrocytes that surround synapses, as well as to astrocyte endfeet that contact blood vessels, making it ideally situated to participate in K+ buffering. Indeed, astrocytes appear to dominate this process in gray matter, as deletion of Kir4.1 from oligodendrocytes did not alter local K+ accumulation in the cortex after neuronal activity. However, the access of astrocytes to axons in white matter is limited to nodes of Ranvier, and the processes of these fibrous astrocytes are less ramified and display less Kir4.1 immunoreactivity than protoplasmic astrocytes in gray matter, which may place a greater burden on oligodendrocytes for K+ redistribution. Our recordings in white matter revealed that deletion of Kir4.1 channels from oligodendrocytes slowed both their depolarization in response to repetitive stimulation and their repolarization after the stimulus train ended (Figure 6; Figure 6—figure supplement 1), suggesting that these channels normally limit extracellular K+ transients induced by neuronal activity. These changes were not restricted to the protected space beneath myelin, as the membrane potential of astrocytes, which closely follow the extracellular K+ concentration, exhibited similarly delayed response kinetics in white matter (Figure 6; Figure 6—figure supplement 1). These findings indicate that oligodendrocyte Kir4.1 is an important mediator of white matter extracellular K+ clearance.

35It is noteworthy that these changes in extracellular K+ homeostasis occurred without changes in oligodendrocyte membrane conductance recorded at the cell soma. As discussed above, somatic membrane conductance may be maintained by expression of other K+ channels, such as K2P or other Kir channels, which are present at baseline and may be upregulated in the setting of Kir4.1 knockout, or by the presence of gap junction-mediated connections with astrocytes. If the conductance is dominated by gap junctions, it would explain the dissociation between membrane conductance and K+ buffering, as these intercellular channels do not allow direct flux of K+ to and from the extracellular space. It is also possible that Kir4.1 is the dominant K+ channel within the intermodal membrane and that these regions are electrically isolated from the soma. Loss of Kir4.1 could then slow K+ clearance from the peri-axonal space and impair the ability of myelinated axons to conduct action potentials. Kir4.1 immunoreactivity has been observed at oligodendrocyte cell bodies and proximal processes, but it has not yet been detected within paranodal loops or internode segments of myelin. Future studies using immunogold and freeze-fracture immunolabeling in white matter may help to better define the spatial relationship between oligodendrocyte Kir4.1 channels and axonal Kv channels. In these domains, Kir4.1 may be the primary conduit of K+ uptake or may work synergistically with other modes of K+ clearance that have been proposed, such as the direct movement of K+ from axons into myelin through paired Kv.1:connexin 29 channels. Although removal of Kir4.1 did not alter the resting membrane potential of oligodendrocytes, it is also possible that this manipulation indirectly altered other K+ clearance mechanisms, such as uptake by the Na+/K+ ATPase.

36Mice in which Kir4.1 was removed from oligodendrocytes exhibited rare, but often fatal seizures, and a dramatic reduction in PTZ-induced seizure threshold (Figure 5). There is extensive evidence that Kir4.1 is a key regulator of neuronal excitability, but the involvement of oligodendrocytes in this process has not been demonstrated. Several spontaneously arising rodent models of epilepsy were found to have altered Kir4.1 levels, and a single-nucleotide polymorphism in the Kir4.1 gene (kcnj10) was found to be responsible for differences in seizure susceptibility between C57BL6 and DBA/2 mice. In addition, polymorphisms in KCNJ10 are risk factors for epilepsy in humans, and seizures are a prominent component of the human disorder SeSAME/EAST syndrome, which results from loss-of-function mutations in KCNJ10. It has been assumed that the seizures are primarily caused by astrocyte abnormalities and downstream sequelae. However, our findings indicate that even in the absence of any structural changes to myelin, deletion of Kir4.1 from oligodendrocytes alone is sufficient to dramatically lower seizure threshold and induce spontaneous seizures. These findings suggest that functional changes in oligodendrocytes, which would not be detected through analysis of myelination patterns, g-ratios or myelin protein expression, may contribute to epilepsy susceptibility or pathogenesis.

37Our studies have focused primarily on K+ clearance in white matter, which is important for the spread of activity from an initial seizure focus. The increased severity of seizures after PTZ administration in oKir4.1cKO mice (Figure 5B), as well as the rapidity with which seizures reached maximum severity after onset (see Figure 5—video 1), are consistent with more rapid spread through white matter tracts in these mice. Gray matter oligodendrocytes, particularly those that myelinate near the axon initial segment, that also lack Kir4.1 in these mice may contribute to seizure initiation. However, recent studies indicate that removal of Kir4.1 from oligodendrocytes does not alter the excitability of cortical pyramidal neurons, perhaps due to the local abundance of astrocyte processes. Additional studies in gray matter will help elucidate the mechanisms of seizure initiation in these mice.

38Oligodendrocyte Kir4.1 knockout mice also displayed significant impairments in motor behaviors (Figure 8), despite normal myelination and the absence of obvious neuronal pathology or neurodegeneration (Figure 4—figure supplement 4). These deficits only became apparent during high-intensity activity, consistent with an activity-dependent, rather than a neurodegenerative mechanism. Although we cannot rule out the possibility that these motor impairments result from subtle changes in neuronal health or circuit function that arise secondary to chronically altered K+ homeostasis or occasional seizures, the intense, coordinated neuronal activity required for high-speed running may be particularly dependent on effective K+ homeostasis. Synchronous neuronal activity is also prominent in brain states associated with diverse behaviors such as learning and sleep, raising the possibility that changes in oligodendrocyte Kir4.1 expression due to aging or disease could impair higher order brain function.

39All animal experiments were carried out in a strict compliance with protocols approved by the Animal Care and Use Committee at the Johns Hopkins University School of Medicine.

40The following transgenic mouse lines were used in this study: Nes-Cre mice were crossed with Kcnj10fl/fl mice to generate a CNS-specific knockout mouse (termed nKir4.1cKO), in which Kir4.1 is deleted from the entire central nervous system. Pdgfra-CreER mice were crossed with Kcnj10fl/fl mice to generate an inducible oligodendrocyte progenitor cell-specific knockout mouse (termed pKir4.1cKO), in which Kir4.1 is deleted from oligodendrocyte progenitors upon exposure to 4-hydroxytamoxifen (4-HT). Mog-iCre mice were crossed with Kcnj10fl/fl mice to generate an oligodendrocyte-specific knockout mouse (termed oKir4.1cKO), in which Kir4.1 is deleted from mature oligodendrocytes. In some experiments, these knockout lines were crossed to fluorescent reporter mouse lines, Rosa-CAG-EGFP (RCE) and R26R-EYFP mice, to allow expression of EGFP or EYFP, respectively, in cells that express Cre. For some experiments, knockout mice were crossed to constitutive fluorescent reporter mice, Mobp-EGFP mice, in which EGFP is expressed in mature oligodendrocytes, and Slc1a2-EGFP mice, in which EGFP is expressed in astrocytes.

414-hydroxytamoxifen (4-HT, Sigma-Aldrich, St. Louis, MO) was dissolved in sunflower seed oil (Sigma) and administered to Pdgfra-CreER;Rosa-CAG-EGFP (RCE) and Pdgfra-CreER;R26R-EYFP mice (with or without Kcnj10fl/fl) at P21 by two intraperitoneal (i.p.) injections of 1 mg, ≥8 hr apart. Mice were given BrdU (Sigma) in their drinking water (1 mg/mL, plus 1% sucrose to increase palatability), in addition to i.p. BrdU injections (50 mg/kg of body weight, dissolved in 0.9% saline) twice daily ≥8 hr apart for 7 days (P28-P34). Ki67 immunolabeling rather than BrdU incorporation was used for proliferation analysis nKir4.1cKO mice to avoid additional manipulation due to their fragile state.

42P21, P35, or P35-P42 mice were deeply anesthetized with isoflurane and decapitated; brains were dissected into an ice-cold N-methyl-D-glucamine (NMDG)-based solution containing the following (in mM): 135 NMDG, 1 KCl, 1.2 KH2PO4, 1.5 MgCl2, 0.5 CaCl2, 20 choline bicarbonate, and 13 glucose (pH 7.4, 310 mOsm). Coronal forebrain slices (250 μm thick for whole cell recordings; 450 μm thick for extracellular CAP recordings) were prepared in ice-cold NMDG-based cutting solution using a vibratome (Leica VT1000S, Leica Microsystems, Wetzlar, Germany) equipped with a sapphire blade. After sectioning, slices were transferred to artificial cerebral spinal fluid (ACSF) containing the following (in mM): 119 NaCl, 2.5 KCl, 2.5 CaCl2, 1.3 MgCl2, 1 NaH2PO4, 26.2 NaHCO3, and 11 glucose (292 mOsm), maintained at 37°C for 20 min and at room temperature thereafter. Solutions were bubbled continuously with 95% O2/5% CO2.

43EGFP+ OPCs, oligodendrocytes, and astrocytes were visualized with an upright microscope (Zeiss Axioskop 2 FS plus) equipped with differential interference contrast (DIC) optics and a filter set for GFP (Brightline, GFP-A-Basic-ZHE; Semrock, Rochester, NY). Cells were visualized using a 40x water-immersion objective (Zeiss Achroplan 40x; Carl Zeiss, Oberkochen, Germany) using DIC and GFP fluorescence signals as a guide. For whole cell recordings, the electrode solution consisted of the following (in mM): 120 CH3SO3H (methansulfonic acid, MeS), 10 K-EGTA, 20 HEPES, 1 MgCl2, 2 Na2-ATP, and 0.2 Na-GTP (pH 7.3, 290 mOsm). For extracellular recordings, the electrode was filled with ACSF. Pipette resistance was 3.0–4.5 MΩ for whole cell recordings and 1.5–2.5 MΩ for extracellular recordings. Recordings were made without series resistance compensation. Unless otherwise noted, the holding potential was –80 mV. Whole cell recordings were performed at room temperature, and extracellular CAP recordings were performed at 37°C (in-line heater TC-324B, Warner Instruments, Hamden, CT). Resting membrane potential was measured in current clamp mode within 30 s of establishing the whole cell recording, and membrane resistance were calculated from a 10 mV depolarizing step in voltage clamp mode. To identify OPCs and exclude differentiated cells in Pdgfra-CreER;RCE mice, a 70 mV depolarizing step was applied to detect the presence of a Nav current. The following agents were applied by addition to the superfusing ACSF: BaCl2 (100 μM, Sigma); tetrodotoxin (TTX, 1 μM; Abcam, Cambridge, UK). Drugs were allowed to wash in for ≥10 min before additional recordings were made. The corpus callosum and alveus were stimulated using a bipolar stainless steel electrode (Frederick Haer Co., Bowdoin, ME; tip separation, 150 mm) connected to a constant current isolated stimulator unit (Digitimer, Ltd, Welwyn Garden City, UK) controlled by a Master-8 pulse stimulator (A.M.P.I., Jerusalem, Israel) and pClamp10 software (Molecular Devices, Sunnyvale, CA). For whole cell recordings, stimuli were 200 μA and 50 μs in duration; for extracellular recordings, stimuli varied from 100 to 300 μA. During extracellular corpus callosum CAP recordings, 1 μM TTX was applied at the end of each recording, and the resulting trace was subtracted from the CAP trace to reduce the stimulus artifact and increase clarity of the CAP.

44Responses were recorded using an Axopatch 200B amplifier (Axon Instruments, Union City, CA), filtered at 1 kHz, digitized at 50 kHz (for episodic recordings) or 5 kHz (for gap-free recordings) using a Digidata 1322A digitizer (Axon Instruments), and recorded to disk using pClamp10 software (Molecular Devices). Data were analyzed offline using Clampfit (Molecular Devices) and Origin (OriginLab, Northampton, MA) software. For whole cell recordings, each recorded cell was considered a biological replicate. For extracellular CAP recordings, each brain slice was considered a biological replicate.

45P24, P35, or P70 mice were deeply anesthetized with pentobarbital and transcardially perfused with 4% paraformaldehyde (PFA) in 0.1 M sodium phosphate buffer (pH 7.4). Brains, spinal cords, and optic nerves were isolated and post-fixed in 4% PFA for 15–20 hr at 4°C. Tissue was cryoprotected in a solution containing 30% sucrose and 0.1% sodium azide in phosphate-buffered saline (PBS). After ≥3 days of cryoprotection, brains were embedded in OCT compound (Sakura Finetek USA, Torrance, CA), frozen at −20°C, and sectioned using a cryostat (HM 550; Microm International GmbH, Walldorf, Germany). 35–40 μm thick brain and spinal cord sections were stored free-floating in PBS with 0.1% sodium azide. 14 μm thick optic nerve sections were slide-mounted and stored at −20°C until further processing.

46Sections were permeabilized with 0.3% Triton X-100 in PBS for 10 min at room temperature, then transferred to 0.3% Triton X-100% and 5% normal donkey serum in PBS (blocking buffer) for 1 hr at room temperature. Sections were then incubated in primary antibody diluted in blocking buffer for 16–24 hr at 4°C. See Key Resources Table for a full list of antibodies used. For ASPA immunostaining, tissue sections were incubated in LAB solution (Polysciences, Warminster, PA) for 10 min before the blocking step. For BrdU immunostaining, sections were incubated in 2 N HCl at 37°C for 30 min, followed by neutralization with 0.1 M sodium borate buffer (pH 8.5) before the blocking step. After primary antibody incubation, sections were washed in PBS and incubated for 2–3 hr at room temperature in Alexa Fluor 488-, Cy3-, Alexa Fluor 647- or DyLight 650-conjugated secondary antibodies. Sections were slide-mounted in Aqua-Poly/Mount (Polysciences).

47Images were acquired using an epifluorescence microscope (Zeiss Axio-imager M1) or a LSM 510 Meta confocal microscope (Zeiss). Confocal images represent maximum intensity projections of z-stacks, with step sizes of 0.5–3 μm. Whole brain section images were acquired as tiled arrays using an epifluorescence microscope equipped with a computer-controlled stage (Cell Observer; Zeiss or BZ-X710; Keyence, Japan), and were aligned using the microscope software (Zen software; Zeiss or BZ-X software; Keyence). Images were processed with ImageJ.

48For Ki67+ OPC counting and GFP+ oligodendrocyte counting, two forebrain sections were analyzed for each mouse, and the results averaged. For OPC BrdU and fate tracing studies, two independent areas (one from each hemisphere) were analyzed from two brain sections for each mouse, for a total of four areas, and results were averaged. All cell countings were performed by an experimenter blinded to the animal’s genotype.

49P70 mice were deeply anesthetized with pentobarbital and transcardially perfused with ice-cold Hank's Balanced Salt Solution without Ca2+ or Mg2+ (HBSS; Gibco Laboratories, Gaithersburg, MD). Brains were isolated and single-cell suspensions were generated using the Neural Tissue Dissociation Kit (Miltenyi Biotec, Bergisch Gladbach, Germany), following the manufacturer’s instructions. Myelin debris was removed using a Percoll gradient, and cells were resuspended in neurobasal media (Gibco) with 1% BSA for sorting. EGFP+ cells were isolated using the MoFlo MLS high-speed cell sorter (Beckman Coulter, Brea, CA) at the Johns Hopkins School of Public Health FACS core. RNA stabilization, genomic DNA elimination, and reverse transcription were performed using the FastLane Cell cDNA Kit (Qiagen, Hilden, Germany), following the manufacturer’s instructions. Real time qPCR was performed using the QuantiTect SYBR Green PCR Kit (Qiagen), following the manufacturer’s instructions, and using a StepOnePlus Real-Time PCR System (Applied Biosystems, Foster City, CA). Primers for the following genes were identified from the MGH PrimerBank: Cnp (ID: 6753476a1), Mobp (ID: 6678910a1), Gfap (ID: 196115326 c1), and Kcnj10 (ID: 34328498a1). The following primers for Gapdh were also used: AAGATGGTGATGGGCTTCCCG (forward), TGGCAAAGTGGAGATTGTTGCC (reverse). Each reaction was performed in quintuplicate, the highest and lowest values were excluded from analysis, and the remaining three technical replicates were averaged for each animal. Data were analyzed using ExpressionSuite v1.1 software (Applied Biosystems).

50P70-P100 mice were deeply anesthetized with pentobarbital and transcardially perfused with 10 mL 0.15 M Sorensen’s Phosphate Buffer (SPB) containing heparin (10 IU/mL, Sigma), then with 60 mL 2.5% glutaraldehyde (Polysciences) in 0.15 M SPB (pH 7.4, 37°C then 4°C). Brains were isolated and post-fixed in 2.5% glutaraldehyde for 2–3 hr at 4°C, then transferred to 0.1% PFA in 0.15 M SPB until further processing.

51Fixed tissue was dissected at 4°C, and cut into 100 μm-thick slices using a refrigerated Lancer 1000 Vibratome (Technical Products International, St. Louis, MO). Corpus callosum was further dissected and post-fixed with 1% OsO4 in SPB, rinsed in dH2O, and stained en bloc for 16 hr with 0.5% aqueous unbuffered uranyl acetate (UAc, pH 4.5). Samples stained with aqueous UAc were dehydrated in graded ethanol series. Dehydrated samples were rinsed in 100% acetone, embedded in plastic resin (10% Epon 812, 20% Araldite 502, 70% dodecenyl acetic anhydride) with 1.25–1.5% DMP-30 (dimethyl amino phenol) added as catalyst, and polymerized at 70°C for 24 hr. Silver and pale gold sections (60–100 nm thick) were cut using a Reichert Ultracut E ultramicrotome (Leica), picked up on 200 mesh copper grids, post stained with UAc and lead citrate, air dried, and examined by transmission electron microscopy. Thin sections were examined in a JEM1400 TEM (JEOL, Tokyo, Japan), operated at 100 kV. Digital images were obtained using an 11 MB Orius SC1000 camera (Gatan, Pleasanton, CA). All images were processed using Adobe Photoshop CS5 (Adobe Systems, San Jose, CA), with ‘levels’ used for contrast expansion and ‘brightness/contrast’ used to optimize image contrast.

52TEM images from corpus callosum containing large numbers of myelinated axons in cross-section were selected for g-ratio analysis. g-ratio analysis was performed using ImageJ software. A threshold was applied to binarize the images and custom software (available at https://github.com/valerie-ann-larson/Larson-et-al-eLife-2018; copy archived at https://github.com/elifesciences-publications/Larson-et-al-eLife-2018) was used to identify axons and calculate their cross-sectional area, from which axon diameters were calculated using the formula for the area of a circle, A = πr2. An experimenter blinded to genotype then measured the myelin sheath thickness of each axon, and excluded any improperly detected or obliquely cut axons from analysis. At least three independent images and at least 100 axons were analyzed from each animal. To compare g-ratios between genotypes, average g-ratios and g-ratio vs. axon diameter slopes were calculated for each animal and were treated as single biological replicates.

53Six- to eight-month-old mice were injected intraperitoneally with pentylenetetrazol (PTZ, 40 mg/kg; Sigma), and placed in standard mouse cages without bedding for observation. Up to four mice were placed in each of four cages for simultaneous observation, and their behavior was recorded using a digital video camera (Canon VIXIA HF R400) for 30 min after injection. Seizure activity was scored by a blinded observer using a modified Racine scale with the following scoring levels: Stage 0 = no response, Stage 1 = ear and facial twitching, Stage 2 = convulsive waves axially through the body, Stage 3 = myoclonic jerks and rearing, Stage 4 = turning over into the lateral position, Stage 5 = generalized tonic-clonic seizures with hind-limb extension, Stage 6 = death. For seizure latency calculation, only mice achieving a seizure score of ≥Stage 3 were included. Surviving mice were euthanized at the end of the experiment.

54P70-P75 mice were anaesthetized with isoflurane and sacrificed by cervical dislocation followed by removal of the heart. Optic nerves were then rapidly dissected and incubated at room temperature in oxygenated ACSF for ≥30 min. Nerves were then transferred to a recording chamber superfused with oxygenated ACSF at 37°C (in-line heater TC-324B, Warner Instruments). Using gentle suction, each end of the nerve was drawn into the tip of a flared pipette electrode. The stimulating electrode (containing the retinal end of the nerve) was connected to a constant current isolated stimulator unit (Winston Electronics Co., St. Louis, MO) driven by pClamp9 software (Molecular Devices). CAPs were elicited by a 1 mA, 50 μs current pulse. The recording electrode (containing the chiasmatic end of the nerve) was connected to one input channel of a Multiclamp 700A amplifier (Axon Instruments). A second electrode, placed near the recording electrode but not in contact with the nerve, was connected to the second channel of the amplifier, and the two signals were subtracted on-line by routing through a differential amplifier (Model 440, Brownlee Precision, Santa Clara, CA), significantly reducing the stimulus artifact. Signals were filtered at 1 kHz, digitized at 100 kHz using a Digidata 1322A digitizer (Axon Instruments), and recorded to disk using pClamp9 software (Molecular Devices). Data were analyzed offline using Clampfit (Molecular Devices), Origin (OriginLab), and custom software (available at https://github.com/valerie-ann-larson/Larson-et-al-eLife-2018; copy archived at https://github.com/elifesciences-publications/Larson-et-al-eLife-2018). Data collected from each optic nerve was considered a biological replicate.

55Open field and rotarod tests were performed in the Behavior Core at the Johns Hopkins University School of Medicine. The experimenter was blinded to the genetic background of the animals during testing. Open-field test was performed using a photobeam activity system (San Diego Instruments, San Diego, CA). Mice were placed in the chamber for a single 30-min period, and movement and rearing were automatically recorded as beam breaks. The rotarod test was performed using a Rotamex-5 rotarod (Columbus Instruments, Columbus, OH). Starting speed was 5 rpm, with acceleration of 1 rpm every 5 s. On the first test day, mice were acclimated for 20 min at 5 rpm prior to the first trial. Five trials were performed per mouse on 3 consecutive days. In order to eliminate performance outliers in an unbiased fashion (e.g. single instances of failed starts that would disproportionately affect average time), the best and worst performances on each day were eliminated, and the remaining three trials were averaged for each mouse. Wheel running behavior was recorded using a low-profile wireless running wheel system (Med Associates, Inc., Fairfax, VT) in the home cage for 7 days. Wheel turns were wirelessly recorded in 30 s bins using Wheel Manager software (Med Associates). Data were analyzed post-hoc using custom software (Robert Cudmore, Johns Hopkins University).

56Statistical analysis was performed using Origin (OriginLab, Northampton, MA) and Excel (Microsoft Corporation, Redmond, WA) software. Sample sizes were constrained by availability of cohorts of age-matched transgenic mice and were not determined in advance. Data are expressed as mean ± SEM throughout, except for qPCR RQ data, which are expressed as a mean and 95% confidence interval, and Kaplan-Meier survival curves, in which the dashed lines represent the 95% confidence interval. For maximal seizure score comparison, the non-parametric Mann Whitney test was used, as data were not normally distributed. To compare Kaplan-Meier survival curves, the log-rank test was used. For multiple comparisons, one-way or two-way ANOVA with Bonferroni’s post-test was performed. For optic nerve CAP recordings, rotarod performance, and running wheel performance, two-way ANOVA was performed with simple effects post-test at each time point. All other comparisons were performed using the unpaired student’s t-test, with results considered significant at p<0.05.