2.1. Schwann cell specific knockout of Phb1
Given the involvement of PHB2 in developmental myelination7, we sought to investigate the role PHB1 plays in Schwann cells (SCs) in vivo. To this end, we crossed mice bearing a floxed Phb1 gene12 with mice expressing Cre recombinase under the control of the Mpz promoter13. This allowed us to generate mice in which Phb1 was deleted specifically in SCs (Phb1fl/fl; Mpz-Cre – referred to as Phb1-SCKO throughout) (Fig. 1A). Recombination in sciatic nerves of Phb1-SCKO animals was confirmed by PCR (Fig. 1B) and resulted in significant reduction of Phb1 mRNA (Fig. 1C) and protein (Supplementary Fig. 1A and 1B).
2.2. Deletion of prohibitin 1 in SCs triggers a severe peripheral neuropathy
We then examined the morphology of sciatic nerves of Phb1-SCKO mice at different days of post-natal development: P10, P20, P40, P60, P90 and P120. Ablation of Phb1 does not impairs radial sorting like deletion of Phb27, but leads to delayed myelination with most SCs still at the pro-myelinating stage in P10 animals (Supplementary Fig. 1C and 1D). However, by P20, Phb1-SCKO mice have an equivalent number of myelinated axons when compared to controls (Fig. 1D and E), even if they are slightly hypomyelinated (Supplementary Fig. 1E). Strikingly, this almost complete recovery from the developmental delay is then followed by a rapid and profound demyelination, which leads to a 60% reduction in the number of myelinated fibers of Phb1-SCKO mice between P20 and P60 (Fig. 1D, E). Morphologically, demyelination is suggested by the presence of several large axons devoid of myelin and by SCs containing myelin debris in cytosolic compartments (Fig. 1D-F). Concomitant to the demyelination, we also identified signs of axonal degeneration, such as axonal shrinkage and local accumulation of vesicles and organelles, indicative of transport defects (Fig. 1F). In addition, Phb1-SCKO crossed to a Thy-1 YFP axonal reporter line present with evident axonal swellings and axon fragmentation in tibial nerves at P20 and P40, respectively (Fig. 1G). This peripheral neuropathy also leads to clear functional impairments in Phb1-SCKO mice. Mutant mice show reduced nerve conduction velocity at P20 and P40 and decreased compound muscle action potential (CMAP) amplitude at P40 (Fig. 1H), in addition to reduced motor performance in the rotarod test at P20 (Supplementary Fig. 1F). In line with the typical manifestation of peripheral neuropathy in mice, Phb1-SCKO animals show clenching of hind limbs towards the body when suspended by the tail (Supplementary Fig. 1G). Moreover, these animals show clear gait impairments that progress to hind limb paresis or paralysis (Supplementary Video 1).
2.3. The Peripheral neuropathy caused by Phb1 deletion affects different types of nerve fibers
Previous studies have suggested that different aspects of SC metabolism are important to maintain homeostasis of distinct types of nerve fibers in the PNS. For example, deletion of the mitochondrial transcription factor Tfam in SCs results in a sensory-motor demyelinating neuropathy6. On the other hand, ablation of the metabolic regulator Lkb1 in SCs causes myelination delay later followed by widespread degeneration of Remak bundles, with no demyelination14. Therefore, we first asked if specific types of fibers were more affected in Phb1-SCKO mice. The size distribution of myelinated axons in sciatic nerves, which contain both motor and sensory fibers, was similar between Phb1-SCKO and controls (Supplementary Figs. 2A and 2B), suggesting that demyelination equally affects fibers of all calibers. As previously reported, the distribution of demyelinated axons was smaller than to the distribution of myelinated axons in Phb1-SCKO animals (Supplementary Figs. 2C and 2D), in line with previous findings that demyelination results in reduction of axon caliber due to decreased neurofilament phosphorylation15, which might be a consequence of loss of myelin and trophic support from SCs.
We also directly investigated if the peripheral neuropathy was prominent in both the motor and sensory compartments by comparing the primarily sensory saphenous nerve and the motor branch of the femoral nerve (Supplementary Fig. 3A). Phb1-SCKO animals show a reduction of myelinated fibers and presence of demyelinated fibers in both the femoral motor and the saphenous nerve (Supplementary Figs. 3B and 3C, respectively). Therefore, Phb1 deletion from SCs causes a sensory-motor peripheral neuropathy.
Last, we examined the effect of Phb1 ablation on Remak SCs, which surround non-myelinated small caliber axons. At P20, Phb1-SCKO mice had a typical density of Remak bundles, which also displayed normal organization with well-ensheathed axons in the correct number and size (Supplementary Fig. 4). However, at P40 several Remak SCs of Phb1-SCKO mice seemed to have retracted their processes, resulting in many axons directly abutting each other (Supplementary Fig. 4). This phenomenon was somewhat lessened at P90; but, at this age, Remak bundles of Phb1-SCKO mice were fragmented, which resulted in reduced number of axons per Remak bundle and increased density of Remak bundles (Supplementary Fig. 4B). In addition, at P90, Remak bundles of Phb1-SCKO mice tended to contain abnormally large axons. In combination, these results may reflect an attempt of the Remak SCs to respond to the ongoing demyelination by: 1) transdifferentiating to a SC type that promote nerve repair; 2) ensheathing some small demyelinated axons. Alternatively, Remak SCs of Phb1-SCKO mice may be simply dysfunctional, causing fragmentation of the Remak bundle and axonal swelling.
2.4 Macrophage infiltration, ERK activation and Schwann cell death are unlikely to be causes of the neuropathy in Phb1-SCKO mice
For all the subsequent analyses, we focused on three time points: P20, P40 and P90, which represent beginning, mid and late stages of the peripheral neuropathy, respectively. It is a reasonable assumption that alterations that are already present at P20 are more likely to be causal for the phenotype, while alterations that appear at the P40 or P90 time points are most likely secondary phenomena that originate as a consequence of the pathology.
Given that prohibitins are required for activation of the Raf-MEK-ERK pathway by Ras 16 and that this is an important pathway for myelination17 and demyelination 18, we first asked whether Phb1 deletion in SCs affected ERK1/2 expression or phosphorylation. (Surprisingly, we only found minor changes in p-ERK1/2 and total ERK1/2 at P40 (Supplementary Fig. 5A and 5B).
A common feature of peripheral neuropathies is the presence of macrophages, which infiltrate the nerves to phagocytize myelin and cellular debris. In some occasions, macrophages can be involved with the initiation of peripheral neuropathies and cause demyelination19. Thus, we asked if macrophages were present in the nerves of Phb1-SCKO mice before demyelination. An elevated number of macrophages is detectable in P40 and P90 Phb1-SCKO animals in comparison to controls (Supplementary Fig. 5C), but there are no differences at P20 (Supplementary Fig. 5). The timing of macrophage infiltration (P40) coincides with the elevation of ERK1/2 phosphorylation (Supplementary Fig. 5B) and of Mcp1 upregulation (Supplementary Fig. 5E), a chemokine previously shown to be secreted by SCs and fibroblasts downstream of ERK signaling to promote macrophage recruitment 20. Therefore, it is possible that all these events are linked, but they are more likely a consequence, rather than a cause of demyelination.
Prohibitins are often necessary to maintain cell survival and support cell proliferation 21. In fact, knockdown of either Phb1 or Phb2 in isolated primary rat SCs leads to cell death7. Thus, we investigated whether SC death or proliferation were altered in vivo in nerves of Phb1-SCKO mice. At P20, there were no differences between groups in TUNEL assay and staining for phosphorylated Histone 3 (p-H3; a mitotic marker) (Supplementary Fig. 6A and 6B). However, nerves of Phb1-SCKO animals showed a slight increase in both TUNEL + and p-H3 + cells at P40 and P90 (Supplementary Fig. 6A and 6B). We further confirmed that TUNEL + and proliferating cells were SCs by co-staining with SOX10 (Supplementary Fig. 6C and 6D). Nonetheless, changes in cell survival and proliferation were tardy and modest and thus unlikely to be sufficient and timely to cause the phenotype. In addition, the balance between cell death and cell division seems to be maintained, since nerves of Phb1-SCKO contain normal quantities of SCs at P20 and P40 (Supplementary Fig. 6E).
2.5. Ablation of prohibitin 1 causes mitochondrial damage accumulation
Prohibitins are mainly localized in the mitochondria, where they are essential for its structure and function. Prohibitins are important for many mitochondrial functions, including fusion, cristae morphogenesis 22, mtDNA maintenance 23, stabilization of respiratory complexes 24,25 and prevention of ROS production26. Thus, we expected that mitochondria would be primarily and severely affected in Phb1-SCKO animals. Analyses by electron microscopy revealed aberrant mitochondrial morphology (Fig. 2A), with SCs of Phb1-SCKO mice showing increased mitochondrial perimeter at all evaluated time points (Fig. 2B-D). Mitochondrial swelling is typical of dysfunctional mitochondria and is associated with several pathological conditions 27.
To investigate SC mitochondria in more detail, we crossed Phb1-SCKO animals to PhAM mice 28, a flox-STOP mitochondrial fluorescent reporter. The Mpz-Cre mediated recombination of PhAM exclusively in SCs allows for a selective evaluation of fluorescent SC mitochondria without confounding results from the large number of mitochondria present in axons. We analyzed confocal images of sciatic nerve teased fibers using an automated routine for ImageJ, which allowed quantification of mitochondrial volume in different compartments in the SC. At P20, larger size mitochondria are overrepresented around the endoplasmic reticulum (ER) and in Cajal bands of Phb1-SCKO animals compared to controls, while mitochondria in juxtaparanodes show a shift toward smaller sizes (Fig. 2E, quantified in Supplementary Fig. 7). We also identified a trend toward increased mitochondrial numbers in Cajal bands of Phb1-SCKO mice (Supplementary Fig. 7B). At P40, aberrant SC mitochondrial patterning in Phb1-SCKO mice persists, with altered size distribution of mitochondria in the vicinity of Cajal bands, the nodes/paranodes and by this age present also in Remak SCs, suggesting mitochondrial fragmentation (Supplementary Fig. 9). Strikingly, at P40, many myelinating SCs of Phb1-SCKO mice displayed almost complete absence of PhAM signal in portions of the SC away from the cell body (Fig. 2F, arrows). This was a phenomenon affecting about 20% of all myelinating fibers of Phb1-SCKO animals (Fig. 2G), while it was not detected in P20 Phb1-SCKO mice (data not shown). In addition, other mitochondrial markers, such as HSPD1 and TOM20 were also severely reduced in SCs in which PhAM was undetectable (Supplementary Fig. 8). This likely reflects the amplification of the early mitochondrial dysfunction seen at P20 and consequent elimination of damaged mitochondria. In line with this hypothesis, there is a progressive depletion of mtDNA from sciatic nerves of Phb1-SCKO animals (Fig. 3A). This mosaic pattern of mitochondrial loss could be explained by the progressive accumulation and amplification of mitochondrial derangements in specific SCs.
To investigate if altered mitochondrial dynamics could be contributing to the mitochondrial alterations in Phb1-SCKO mice, we performed live imaging of SCs isolated from Phb1wt/wt; P0-Cre; PhAM (Control) and Phb1fl/fl; P0-Cre; PhAM (Phb1-SCKO) animals. We exposed a focal region of the SCs to the 405 nm confocal laser to promote photoconversion of PhAM from green to red and then monitored cells for 30 min. In control SCs, photoconverted (red) mitochondria dispersed quickly and no peak in the red signal is evident at the end of the experiment (Fig. 3B). On the other hand, red SC mitochondria are still present as a group after 30 min in Phb1-SCKO mice (Fig. 3B and Supplementary Videos 2 and 3). This indicates that deletion of Phb1 in SCs leads to impaired mitochondrial dynamics and/or transport. Impaired mitochondrial dynamics could be a consequence of dysfunctional Opa1, a dynamin-related GTPase involved in fusion of the inner mitochondrial membrane. Deletion of prohibitins has been previously reported to trigger proteolytic cleavage of Opa1, impairing its function 22. In agreement with this hypothesis, Western blots from sciatic nerve lysates show increased proteolytic processing of Opa1 in Phb1-SCKO mice (Fig. 3C).
Next, we asked whether these mitochondrial changes affected mitochondrial physiology and function. We evaluated mitochondrial membrane potential using TMRM in SCs isolated from control and Phb1-SCKO animals. Ablation of Phb1 results in significant reduction in TMRM fluorescence suggesting that mitochondria of Phb1-SCKO mice are depolarized (Figs. 3D and 3E). Since the mitochondrial membrane potential relates to the cell’s capacity to make ATP through oxidative phosphorylation, we also evaluated mitochondrial respiration using the Seahorse Extracellular Flux Analyzer. Since SCs of Phb1-SCKO mice have impaired survival in vitro and a relatively large number of cells were required for this analysis, we instead acutely knocked down Phb1 in primary rat SCs using shRNA. Our analyses revealed normal basal respiration in Phb1-knockdown SCs, but a significant decrease in the spare respiratory capacity compared to cells treated with sh-Control (Figs. 3F-3I). This result indicates that SCs lacking PHB1 may be unable to appropriately respond to changes in metabolic demand.
A common consequence of inefficient respiration in a situation of stress is the production of reactive oxygen species by mitochondria. However, we did not detect any difference in lipoperoxidation or protein oxidation between nerves of Phb1-SCKO mice and controls (Supplementary Fig. 10), suggesting that, at least at the time points examined, oxidative stress may not be present.
Prohibitin 2 has recently been described as a mitophagy receptor at the inner mitochondrial membrane29. Although it is not clear if Prohibitin 1 also participate in mitophagy, we postulated that PHB1-ablated Schwann cells may accumulate damaged mitochondria because they may be unable to perform mitophagy. We thus analyzed the capacity of PHB1-ablated Schwann cells to carry out mitophagy. For these analyses, we made use of a retrovirus system to deliver mt-mKeima, a mitochondrially-targeted pH-sensitive fluorescent protein30. When in the mitochondria (which has a neutral pH), the mt-mKeima excitation peak is at 440 nm. However, when mitochondria are targeted to degradation in the lysosomes (therefore reducing the pH), the excitation peak of mt-mKeima shifts to 586 nm. Thus, a ratio of the mt-mKeima fluorescence at ~ 586 nm over the fluorescence at ~ 440 nm can be used as a proxy for the degree of mitochondrial degradation in the lysosomes. Using this method, we found that silencing of Phb1 by shRNA does not change the ability of SCs to perform mitophagy (Supplementary Fig. 11).
We next tested if there is an association between mitochondrial damage and demyelination in Phb1-SCKO mice. To this end, we analyzed teased fibers from Phb1wt/wt; P0-Cre; PhAM (Control) and Phb1fl/fl; P0-Cre; PhAM (Phb1-SCKO) animals. In Phb1-SCKO animals, fibers with undetectable PhAM were overrepresented among fibers containing myelin ovoids, suggesting an association between mitochondrial damage and demyelination (Supplementary Fig. 12).
Taken together, our results indicate that mitochondria in SCs of Phb1-SCKO mice are severely impaired starting at P20. In addition, we show evidence that mitochondrial dysfunction and demyelination are linked and that the mitochondrial damage is accumulating, possibly because Phb1-SCKO mice have abnormal mitochondrial dynamics.
2.6. Deletion of Phb1 activates a mitochondrial stress response
Although mitochondria have their own DNA, they still depend on genomic DNA to synthetize most of their proteins (only 13 out of more than 1200 mitochondrial proteins are coded by mtDNA) 31. Thus, in order to keep homeostasis, there is a need for bidirectional mitonuclear communication 32. In a similar fashion, when mitochondria are under stress, cells respond with a coordinated attempt to mitigate potential damage. In mammals, this normally involves activation of the integrated stress response (ISR), a general stress pathway working to reduce overall protein synthesis and favor the expression of stress-response genes 33. A previous report by Viader et al. showed that deletion of the mitochondrial transcription factor Tfam in SCs results in activation of the ISR, which they postulated to be a maladaptive mechanism 11. Thus, we asked whether Phb1 deletion in SCs triggers the ISR by assessing the levels of phosphorylated eIF2α, the hallmark of ISR. Indeed, we found that nerves of Phb1-SCKO mice showed early (P20) and continually elevated levels of phosphorylated eIF2α in comparison to littermate controls (Fig. 4A).
We then sought to investigate the downstream events in this mitochondrial stress response. Classically, mitochondrial dysfunction induces the mitochondrial unfolded protein response (UPRmt), which results in upregulation of a set of genes including mitochondrial chaperones (such as Hspd1 and Hspe1, also known as Hsp60 and Hsp10, respectively) and proteases (such as Clpp). Thus, we probed Phb1-SCKO animals for the activation of UPRmt. Surprisingly, we did not detect elevation of any UPRmt effector at protein or RNA level (Figs. 4B-4D). A recent publication implicated ATF4 in the mammalian mitochondrial stress response and characterized the molecular signature of this pathway34. Therefore, we assessed this pathway in nerves of Phb1-SCKO mice. Even though Atf4 expression itself is not altered by deletion of Phb1 in SCs, four out of the five analyzed transcripts regulated by ATF4 are upregulated in nerves of Phb1-SCKO mice (Fig. 4E): Asns (asparagine synthetase); Chac1 (cation transport regulator-like protein 1); Pck2 (phosphoenolpyruvate carboxykinase 2) and Dddit3 (DNA damage-inducible transcript 3; also known as Chop). Importantly, phosphorylation of eIF2α in this context does not seem to be trigged by activation of PERK kinase in the ER, since its phosphorylation is not increased (Supplementary Figs. 13A and 13B). However, Phb1-SCKO mice do show elevated levels of the HSPA chaperone (also known as Bip) (Supplementary Fig. 13C), and upregulated alternative splicing of Xbp1 (Supplementary Fig. 13D), two markers commonly associated with the unfolded protein response in the ER (UPRER). This suggests that mitochondrial dysfunction caused by Phb1-SCKO indirectly leads to ER stress.
The ER is an organelle involved in lipid synthesis, while mitochondria take part in beta-oxidation, the process of breakdown of fatty acids. Therefore, it is possible that the balance of lipid metabolism is altered in Phb1-SCKO mice. If lipid oxidation occurs disproportionally to lipid synthesis, myelin maintenance could be affected due to depletion of important myelin lipids. This mechanism has been previously proposed to underlie the peripheral neuropathy caused by deletion of Tfam in SCs11. Therefore, we evaluated the expression and phosphorylation of Acetyl-CoA carboxylase (ACC). The ACC enzyme is responsible for production of malonyl-CoA, the substrate for biosynthesis of fatty acids and an inhibitor of beta-oxidation. ACC activity is inhibited by phosphorylation at Ser79 by AMPK or at Ser1200 by PKA. Nerves of Phb1-SCKO mice showed increased phosphorylated ACC at P20 (Fig. 5A-D). Moreover, the expression of many genes involved with lipid synthesis is severely reduced (Fig. 5E and 15F) at both P20 and P40, suggesting that reduction of lipid biosynthesis might be a common finding in neuropathies in which SC mitochondria are damaged.
In summary, ablation of Phb1 in SCs leads not only to a mitochondrial stress response involving the ISR, but also to a broader cellular response, affecting the ER and causing reduced expression of enzymes involved in lipid biosynthesis.
2.7. The mitochondrial stress response is beneficial to Phb1-SCKO mice
Given the dramatic implications of the ISR for cells, such as a widespread inhibition of translation, we aimed to test if continuous activation of the ISR was maladaptive in the context of mitochondrial dysfunction in SCs, as suggested by Viader et al.11. To evaluate the role of ISR in Phb1-SCKO mice, we treated animals daily from P20 to P40 with 2.5 mg/kg of ISRIB (for ISR inhibitor). Unphosphorylated eIF2 transfers the methionylated initiator tRNA (Met-tRNA) to the ribosome in a guanosine 5′-triphosphate-dependent manner to start translation 35. Phosphorylation of the alpha subunit of eIF2 leads to competitive inhibition of eIF2B, the guanosine-exchange factor (GEF) for eIF2, halting translation 36. IRSIB is a small molecule known to enhance the GEF activity of eIF2B, thereby alleviating the translation block 37,38 (Fig. 6A).
As expected, ISRIB treatment did not change p-eIF2α levels (Fig. 6B), but significantly reduced the upregulation of ATF4 target genes in Phb1-SCKO mice (Fig. 6C). In tibial nerves, we identified a small reduction in myelin thickness (g-ratio = the ratio between axon and fiber diameters) in larger caliber axons due to Phb1 deletion in SCs, which was accompanied by a reduction in overall axon caliber. However, there was no significant effect of ISRIB on these parameters (Supplementary Fig. 14). Surprisingly, additional morphological analysis of tibial nerves of Phb1-SCKO mice treated with ISRIB, revealed an exacerbation of the demyelination (Fig. 6D). Supporting this conclusion, quantifications showed an increased number of demyelinated axons and of myelin degradation in Phb1-SCKO mice upon ISRIB treatment (Fig. 6E). In line with our morphological findings, Phb1-SCKO animals treated with ISRIB showed a trend toward reduced performance in the rotarod test as compared to Phb1-SCKO mice treated with vehicle (p-values Phb1-SCKO + ISRIB vs Phb1-SCKO + Veh: Day 2 = 0.0936, Day 3 = 0.0979, Day 4 = 0.0723) (Fig. 6F).
Taken together, our results suggest that activation of the ISR not only is not detrimental, but may even be a protective mechanism against demyelination triggered by deletion of Phb1.