Dietary cholesterol does not rescue hypomyelination in PMD patients
Active demyelination with inflammatory responses in patients with PMD  predicts at least minor BBB disturbances. Therefore, in an individual, compassionate use trial, we first asked whether we could translate the dietary treatment approach in two patients with PMD caused by PLP1 duplication. Daily cholesterol supplementation of up to 590 mg/kg was well tolerated and blood lipid and glucose values remained in the normal range (Online Resource Supplemental Fig. 1). No adverse reactions were observed in the PMD patients, in agreement with cholesterol supplementation studies in patients with Smith-Lemli-Opitz disease . However, a therapeutic benefit was not as obvious as in PMD mice. By MRI measurements, hypomyelination remained stable over 2 years of treatment (Online Resource Supplemental Fig. 2) which is consistent with the normal course of PMD disease [59, 66]. Taken together, dietary supplementation with high dose cholesterol was safe in our two PMD patients and long term disease monitoring will decide about the potential therapeutic benefit of this treatment strategy.
In Plp1tg-72 mice, the compromised BBB allowed access of cholesterol from the circulation into the brain . In contrast, in PMD the comparative analysis of CSF and serum did not reveal gross disturbance of BBB integrity (Online Resource Supplemental Fig. 1a), in agreement with the lack of enhancing gadolinium signals in MRI in a PMD patient . In a substrain of Plp1tg-72 mice with preserved BBB integrity (Online Resource Supplemental Fig. 3a-b), the dietary cholesterol supplementation only moderately ameliorated myelin disease (Online Resource Supplemental Fig. 3c-d), most likely reflecting the limited therapeutic success in PMD patients. Of note, PMD patients and both Plp1tg mouse strains show a comparable initial increase in serum cholesterol (Online Resource Supplemental Fig. 1b, Online Resource Supplemental Fig. 3e)  as expected from healthy adult individuals , suggesting that serum cholesterol alone does not suffice to predict treatment efficacy. Since the substrain of Plp1tg-72 mice more adequately models human PMD, we used this line, termed Plp1tgB below, in subsequent experiments.
Ketogenic diet improves pathology in a PMD model with preserved BBB integrity
The consumption of a high-fat/low-carbohydrate ketogenic diet causes the liver to generate ketone bodies. In the brain, ketone bodies such as beta-hydroxybutyrate facilitate sterol synthesis  which is essential for myelin membrane growth . Thus, we asked whether a ketogenic diet that promotes CNS lipid metabolism under conditions of preserved BBB function [22, 32, 50] is also effective in hypomyelinating disease. Plp1tgB and control mice were given a ketogenic diet (KD) or standard a chow diet (SD) between 2 and 12 weeks of age and physiological parameters were monitored weekly (Fig. 1a). Already after 7 days of dietary intervention with KD, blood levels of ketone bodies strongly increased (about fivefold) and serum glucose dropped to about 70% of the level in SD fed mice (Fig. 1b, Online Resource Supplemental Fig. 4a, b). Ketone bodies are imported into the brain by monocarboxylate transporters, mainly MCT1 expressed by endothelial cells . In accordance with ongoing ketosis, mRNA and protein levels of monocarboxylate transporters as well as enzymes of ketone body utilization significantly increased. This included the rate-limiting enzyme of the pathway, 3-oxoacid CoA-transferase 1 (OXCT1, also termed Succinyl-CoA:3-ketoacid coenzyme A transferase 1, SCOT, EC:126.96.36.199) whose abundance was threefold elevated in spinal cord of KD fed Plp1tgB mice, even more obvious by immunolabeling (Fig. 1d, e, Online Resource Supplemental Fig. 4c, d).
We next determined the degree of oligodendroglial defects, the hallmark of PMD pathology, in the corticospinal tract of the spinal cord from 12 weeks old Plp1tgB mice. As expected, the number of Olig2- and CAII-positive oligodendroglia was reduced in SD fed Plp1tgB mice, but feeding the KD diet rescued this defect in mutants (Fig. 2a). Moreover, untreated Plp1tgB mutant mice showed a robust endoplasmic reticulum stress response with about fourfold elevated ATF6 mRNA and protein levels. Both were significantly ameliorated in KD fed mutant mice (Fig. 2b, Online Resource Supplemental Fig. 4e). In agreement with the reduced oligodendroglial pathology, also astrogliosis and microgliosis were ameliorated in KD fed Plp1tgB compared to SD fed mutants (Online Resource Supplemental Fig. 5). Plp1 mRNA overexpression is the primary defect in the classical form of PMD and in our mouse model [11, 51]. Feeding Plp1tgB mutants with KD did not lower Plp1 expression but rather increased Plp1 mRNA levels slightly (Fig. 2c), presumably reflecting enhanced oligodendrocyte survival. Indeed, the severe hypomyelination of untreated Plp1tgB mice was ameliorated by feeding a KD, as evidenced by fewer unmyelinated axons (g-ratio = 1) and more normally myelinated axons (g-ratio < 0.8), resulting in an overall reduction in mean g-ratio, as determined for the corticospinal tract (Fig. 2d, e). We next investigated whether feeding KD also ameliorated the clinical phenotype of PMD mice. When we assessed motor performance by two different behavioral tests, elevated beam test and rotarod test, untreated Plp1tgB mice showed progressive worsening of motor functions. By contrast, KD fed Plp1tgB animals retained their motor fitness at wild type levels (Fig. 2f, g). For comparison, feeding Plp1tgB mice a medium-chain triglyceride diet, which contains less fat than classical KD, elicited only a mild ketosis in our mice. This dietary regimen did not ameliorate motor performance, suggesting a critical role of ketone bodies rather than a direct effect of triglycerides/lipids (Online Resource Supplemental Fig. 6).
Ketogenic diet ameliorates mitochondrial abnormalities in axons of PMD mice
It is unlikely that the moderately improved myelination explains the dramatic improvement in motor functions in KD fed Plp1tgB animals. In PMD patients, motor development is strongly retarded, and progressive axonal loss likely causes the gradual decline of motor functions already achieved . In accordance, the frequent axonal swellings (spheroids) in Plp1tgB mice were strongly reduced when feeding a KD (Fig. 3a). In addition, we observed in Plp1tgB mice that many axons contained enlarged mitochondrial profiles (Fig. 3b), as observed before in other models of PMD [28, 45, 53]. Such morphological alterations can reflect increased activity and/or functional deficits in mitochondria which could both occur in PMD (see below).
By systematic quantification of electron microscopic images, we detected a significant about 20% area increase of mitochondrial profiles in axons from Plp1tgB animals compared to wild type controls. This enlargement was most evident in unmyelinated axons (Fig. 3c, d), and even detectable in young mutants (Online Resource Supplemental Fig. 7). Importantly, KD completely normalized this phenomenon, which was most evident in axons that remained unmyelinated, and correlated well with the reduced volume of mitochondria, modelled from deconvolved confocal image stacks (Fig. 3c–e, Online Resource Supplemental Fig. 8). We also determined mRNA and protein abundance of several mitochondrial markers in spinal cord lysates to correlate mitochondrial morphology to alterations in fusion or fission of mitochondria but neither of the tested markers showed significant differences between the two treatment groups of Plp1tgB animals, likely because this analysis reflect the sum of neuronal and glial mitochondria (Online Resource Supplemental Fig. 9). Notably, feeding a cholesterol supplemented chow to Plp1tg animals (Plp1tgB or Plp1tg-72) did not reverse the increased mitochondrial sizes, even where myelination was improved (Online Resource Supplemental Fig. 3f). This suggests that high-cholesterol and ketogenic diets supported distinct therapeutic pathways.
The latter prompted us to evaluate the impact of KD on axonal integrity in the absence of myelin and analyzed optic nerves of Plp1tgB animals. When untreated, this fiber tract is almost devoid of myelin at the age of 12 weeks but unmyelinated axons appear morphologically healthy and retinal ganglion cells are not degenerated [20, 57]. Here, the KD diet showed only a modest effect on hypomyelination (Fig. 3f), but ameliorated the enlarged axonal mitochondria, similar to our observations in the corticospinal tract (Fig. 3g).
Next, we assessed the functional significance of mitochondrial abnormalities in hypomyelinated axons by sensitive electrophysiological analyses. Specifically, in acute ex-vivo preparations of the optic nerve, we quantified nerve conduction and functional axon integrity, by comparing compound axon potentials (CAP) as described [54, 69]. As expected for hypomyelinated fibers [4, 54, 64], optic nerves of untreated Plp1tgB mice exhibited an abnormally attenuated evoked response with a strong after-hyperpolarization phase caused by elevated activity of potassium channels (Online Resource Supplemental Fig. 10). In KD fed Plp1tgB mutants, in contrast, CAP amplitude and CAP area was significantly increased compared to their SD fed littermates, implying that more axons contributed to evoked impulse propagation (Fig. 3h–j). Nerve conduction velocity of Plp1tgB mice was slightly increased in KD fed animals, likely reflecting the increased density of myelinated axons (Fig. 3f, k).
Glycolytic oligodendrocytes support the axonal energy metabolism by exporting lactate [23, 36]. To bypass this glial support and rather directly assess mitochondrial function in axons, we challenged the axonal energy metabolism during repeated firing. Specifically, we monitored CAPs in the optic nerve in the absence of glucose and in the presence of a suboptimal lactate concentration (3 mM) during 5 Hz stimulation for 75 min , followed by a recovery period in the presence of both lactate and glucose. In wild type optic nerves, this protocol led to a gradual reduction of CAPs to about 25% of baseline values (Fig. 3l), in agreement with recent findings in the corpus callosum . In optic nerves from Plp1tgB mice, however, the resulting CAP decay was even stronger and more rapid than in wild type controls, demonstrating that the PMD pathology affects the ability of axons to utilize lactate to maintain conduction. Feeding a KD to Plp1tgB mice attenuated the CAP decay (of 5 Hz stimulated optic nerves ex vivo), suggesting that optic nerves from KD fed mice are more resilient to this metabolic stress (Fig. 3l, m). Moreover, in this protocol we noted a strong (80%) recovery of the original CAP amplitude in nerves from KD fed mice comparable to wild types, contrasting to only 40% recovery in SD fed mice (Fig. 3l–n). This finding demonstrates that PMD, which primarily perturbs myelination, also indirectly affects axonal energy metabolism. These data also suggest that axonal integrity can be uncoupled from the myelination status by the KD. That is, despite the marginally increased myelination of the optic nerve of Plp1tgB mice when fed the KD, dysmyelinated axons functionally improved, as evidenced by physiological recordings, and this effect was accompanied by morphological normalization of axonal mitochondria.
Ketogenic diet promotes repair in a model of adult remyelination
Because of the therapeutic benefit of KD in the PMD mouse, a model of developmental hypomyelination, we compared in a 4-arm study the therapeutic potential of KD in a model of adult demyelination and remyelination. At the age of 8–10 weeks, wild type mice were demyelinated by feeding cuprizone, contained in a SD, for 4 weeks (acute paradigm) or for 12 weeks (chronic demyelination), followed by cuprizone withdrawal and remyelination with feeding SD or KD (Fig. 4a). Increased blood levels of the ketone body beta-hydroxybutyrate (βHB) and reduced blood glucose levels confirmed ketosis already at 7 days after the diet switch (Online Resource Supplemental Fig. 11a-d).
In agreement with our findings in Plp1tgB mice, mRNA expression levels of monocarboxylate transporters and enzymes of ketolysis were robustly increased in the corpus callosum of KD fed mice in comparison to SD fed controls and untreated wild type mice (Online Resource Supplemental Fig. 11e, f), likely reflecting ongoing ketone body utilization. Cuprizone leads to the death of most mature oligodendrocytes in the corpus callosum [39, 49], and after the withdrawal of cuprizone, myelin repair is reflected in an increasing number of oligodendrocyte lineage cells and gradual remyelination. While the number of Olig2 positive cells was not altered in KD fed mice compared to SD fed controls, the rate of oligodendrocyte precursor cell differentiation to CAII positive oligodendrocytes was strongly increased in KD fed mice (Fig. 4b–d, Online Resource Supplemental Fig. 11-12). In accordance, by feeding KD remyelination was accelerated about twofold at only 7 days after acute demyelination, and at 2 weeks after chronic demyelination (Fig. 4e–g).
In contrast, the degree of astrogliosis and microgliosis and expression of enzymes involved in detoxification of reactive oxygen species appeared unaffected by KD (Online Resource Supplemental Fig. 11 g-k, Online Resource Supplemental Fig. 12). Pro-inflammatory eicosanoids could contribute to pathogenesis in multiple sclerosis and expression of respective biosynthetic enzymes such as ALOX5 (arachidonate 5-lipoxygenase) was also increased in the chronic cuprizone model (Online Resource Supplemental Fig. 11 l) in accordance with previous reports [47, 75]. However, in contrast to a dietary intervention study with KD in MS patients , their expression was not altered by KD in our study. Axonal defects, as assessed by the density of APP positive axonal spheroids, was reduced by about 50% in KD fed mice compared to SD fed controls (Fig. 4h, Online Resource Supplemental Fig. 12). In addition, the cross-sectional size of axonal mitochondria, which were increased in SD fed mice during remyelination, even normalized (Fig. 4g, i). This suggests that the loss of oligodendroglial integrity rather than gliosis correlates with axonal perturbations. Finally, motor abilities, as quantified by an elevated beam test, also improved in KD fed mice (Fig. 4j). Thus, similar to our findings in developmental dysmyelination, KD also promoted functional repair in acute and chronic adult remyelination paradigms.
In this study, we have used two models of white matter disease with distinct etiologies and pathomechanisms, and found that feeding a KD is remarkably efficient in preserving axonal and oligodendroglial integrity. Plp1tgB mutants model the hereditary leukodystrophy PMD, in which overexpression of the myelin protein PLP induces cellular stress in differentiating oligodendroglia followed by dysmyelination and demyelination . In contrast, the pharmacological cuprizone model induces transient demyelination and also assesses remyelination in adult mice [39, 49]. In both disease models, axon damage is likely secondary to the primary oligodendroglial defect [15, 34, 59], but a detailed understanding of the axonal injury is lacking. A combinatory effect of reduced metabolic support by injured oligodendrocytes and the exposure to pro-inflammatory factors, including reactive oxygen and nitrogen species, is most likely [15, 34]. In the PMD mouse model, we found mitochondrial enlargement not only in dysmyelinated axons but also in myelinated axons before overt demyelination, suggesting metabolic problems in axons at an early preclinical stage.
The KD dramatically changes the global metabolism and induces a plethora of alterations in the body including the CNS . While discussed in many neurodegenerative diseases, KD is currently used in the management of epilepsy [3, 16, 70]. The mechanisms of seizure control are likely diverse, involve the gut microbiota  and may act in parallel, including effects on neurotransmission, epigenetic gene regulation, energy metabolism, and anti-inflammatory activity [16, 22, 61]. Our preclinical data reveal that KD provides metabolic support to axons in two distinct myelin pathologies that model key aspects of human myelin disease and mechanisms of recovery. We specifically note that also (chronically) unmyelinated axons appear to profit from KD.
Ketone bodies share some of the molecular targets with certain fatty acids that are contained in or processed from high-fat ketogenic diet regimens [3, 61]. However, in our study, a medium-chain triglyceride diet that only slightly raised blood ketone body levels was therapeutically ineffective, suggesting a critical and direct role of ketone bodies in the therapy of PMD.
We envision the following scenario of causal relations: (1) In PMD, PLP overexpression induces endoplasmic reticulum stress which affects oligodendrocytes and likely disturbs the oligodendroglial trophic support to axons. Chronic damage ultimately leads to loss of Plp1-overexpressing oligodendrocytes and demyelination. In the cuprizone model, pharmacologically induced oligodendrogliopathy causes demyelination. (2) To maintain impulse conduction unmyelinated axons require about five-fold more energy compared to the same tissue volume of myelinated axons . Hence, aberrantly unmyelinated axons suffer from energy deficits because of the insufficient support by mutant oligodendrocytes and increased energy demands. (3) Energy deprived cells typically increase the activity of mitochondria. In addition, demyelination triggers inflammatory responses involving production of reactive oxygen and nitrogen species. Increased mitochondrial activity and oxidative stress are both associated with morphological changes and volume enlargement of mitochondria [38, 48] as observed in axons of both models of myelin disease. Chronic oxidative stress might damage mtDNA and mitochondrial enzymes  that could further trigger compensatory volume increases of mitochondria and aggravate degenerative processes. (4) In contrast to glucose, ketone bodies are directly metabolized by mitochondria fueling into the tricarboxylic cycle and oxidative phosphorylation. Since feeding a KD supports the conduction of optic nerve axons even in the absence of substantial myelination, we propose there is a direct supportive effect of KD on mitochondrial integrity in dysmyelinated axons. Bypassing the need for oligodendroglial trophic support could restore axonal mitochondria function (and morphology) and resolve energy deficits of axons. Relieving oligodendrocytes from the task to provide energetic support to axons might also contribute to the increased survival of mutant oligodendrocytes and to the ameliorated PMD pathology.
In contrast to the developmental PMD model, adult remyelination in the cuprizone model strongly benefitted from the KD, in line with previous findings using a fasting mimicking diet . Here, we suggest that a compromised BBB in cuprizone fed animals allows for the entry of circulating cholesterol into the CNS (further enhanced by the KD that contains 0.1% cholesterol) [5, 6]. While the BBB may remain largely intact in classical human PMD, it is perturbed in other myelin diseases with a stronger inflammatory component. We conclude that there is a clear rationale to consider KD or a derivative as a future therapy for myelin diseases as it combines two important therapeutic targets, cholesterol for the support of remyelinating oligodendrocytes and ketone bodies for the metabolic support of the axonal compartment, and future clinical trials will reveal its feasibility in the management of demyelinating episodes.