Perinatal loss of galactosylceramidase in both oligodendrocytes and microglia is crucial for the pathogenesis of Krabbe disease in mice

Abstract

Lysosomal galactosylceramidase (GALC) is expressed in all brain cells, including oligodendrocytes (OLs), microglia, and astrocytes, although the cell-specific function of GALC is largely unknown. Mutations in GALC cause Krabbe disease (KD), a fatal neurological lysosomal disorder that usually affects infants. To study how Galc ablation in each glial cell type contributes to Krabbe pathogenesis, we used conditional Galc-floxed mice. Here, we found that OL-specific Galc conditional knockout (CKO) in mice results in a phenotype that includes wasting, psychosine accumulation, and neuroinflammation. Microglia- or astrocyte-specific Galc deletion alone in mice did not show specific phenotypes. Interestingly, mice with CKO of Galc from both OLs and microglia have a more severe neuroinflammation with an increase in globoid cell accumulation than OL-specific CKO alone. Moreover, the enhanced phenotype occurred without additional accumulation of psychosine. Further studies revealed that Galc knockout (Galc-KO) microglia cocultured with Galc-KO OLs elicits globoid cell formation and the overexpression of osteopontin and monocyte chemoattractant protein-1, both proteins that are known to recruit immune cells and promote engulfment of debris and damaged cells. We conclude that OLs are the primary cells that initiate KD with an elevated psychosine level and microglia are required for the progression of neuroinflammation in a psychosine-independent manner.

Graphical abstract

Shin and colleagues found that oligodendrocytes initiate Krabbe disease and prompt accumulation of the cytotoxic sphingolipid psychosine, while microglia exacerbate neuroinflammation in response to oligodendrocyte dysfunction by forming pathognomonic globoid cells. Insight into the mechanisms of disease suggest that targeting both cell types is necessary for efficacious therapeutic intervention.

Introduction

Krabbe disease (KD) is a progressive and fatal neurologic lysosomal storage disorder that usually affects infants. KD is caused by mutations in galactosylceramidase (GALC), a lysosomal galactolipid hydrolase.¹ GALC is involved in the normal turnover of myelin by hydrolyzing galactosylceramide (GalCer), a major sphingolipid constituent of myelin that is essential in myelin compaction and homeostasis.² Unlike other lysosomal storage diseases (LSDs), the primary substrate of GALC, GalCer, does not accumulate highly in the tissues of KD patients. Instead, a minor substrate of GALC, galactosylsphingosine (also known as psychosine), accumulates to toxic levels and has canonically been associated with causing extensive demyelination and the bulk of KD pathological findings.³ The majority of psychosine is generated catabolically through the deacylation of GalCer by acid ceramidase.⁴ Hematopoietic stem cell therapy (HSCT) extends the long-term survival with improved quality of life for asymptomatic or early symptomatic KD patients, but it is not a cure, indicating that the mechanism of KD neuropathology is still to be elucidated.

Due to the extensive demyelination in KD, oligodendrocytes (OLs) have been implicated as the primary cells inducing pathogenesis. Interestingly, our previous study showed that GALC protein in the perinatal mouse brain is expressed in neurons (38%–73%), OL-lineage cells (7%–11%), astrocytes (5%–18%), and microglia (4%–7%), demonstrating a ubiquitous expression of GALC and suggesting a cell-autonomous role of GALC in the brain.⁵ However, the cell-specific contributions in disease remain largely unknown. Common models of study such as the twitcher mouse are beneficial to understand KD pathology and explore therapeutic options; nonetheless, the ubiquitous nature of Galc ablation in these models makes deciphering the etiology of KD complex due to their rapid decline. Therefore, we developed the Galc-floxed allele allowing for temporal and cell-specific deletion of Galc, enabling us to critically evaluate the loss of GALC function in neural and glial cells in vivo and identify key mechanisms of KD onset and pathology.^5,6,7 As the first attempt to characterize one of these, we reported a role for GALC in early neuronal development.⁵ Neural GALC expression is essential to maintain and protect neuronal function and influences brainstem development, which is critical for KD pathogenesis. We also showed that neuronal Galc ablation induces neurodegeneration, leading to muscle weakness and ataxia, but does not affect survival,⁶ indicating the necessity for GALC in the maintenance of neuronal homeostasis, which had previously been proposed only through in vitro studies.^8,9,10,11 In addition, we found that HSCT introduces GALC-positive macrophages, which assist with myelin degradation in the peripheral nervous system, rather than cross-correcting GALC to resident GALC-deficient cells.⁷ We showed that the poor cross-correction is because GALC mutant cells are not able to uptake GALC through the mannose 6-phosphate receptor pathway as efficiently as wild-type (WT) cells. These findings prove the value of the Galc-floxed mouse for determining which cells contribute to different disease-related phenotypes in vivo, which was previously masked by the ubiquitous absence of GALC in the twitcher model.

In this study, we investigate the role of GALC in the major glial cells of the mouse brain (OLs, microglia, and astrocytes) using the Galc-floxed mouse to gain insight into their contributions to KD pathology in an unbiased manner. Our study shows that OL-specific Galc CKO results in a phenotype that includes wasting, psychosine accumulation, and neuroinflammation similar to KD pathology, suggesting that GALC deficiency in OLs directly elicits pathogenesis. Although Galc deletion in non-OL glial cells such as astrocytes and microglial alone did not show any specific phenotypes, Galc ablation in both OLs and microglia generated a more exaggerated pathology than OL-specific Galc knockout (Galc-CKO) alone. The exacerbation occurred in the absence of any additional accumulation of psychosine but rather through the induction of chemo-attractants such as osteopontin and monocyte chemoattractant protein-1 (MCP-1), implying that Galc-deficient microglia may inflate neuroinflammation and degeneration in a non-psychosine-mediated mechanism.

Results

Temporal Galc ablation in perinatal OLs induces inflammation in mouse brain

To investigate the role of glial GALC, we crossed the recently developed Galc-floxed mouse model⁵ to cell-specific Cre drivers: Aldh1l1-Cre/ERT2 (JAX#031008) for astrocytes, Cx3CR1-Cre/ERT2 (JAX#021160) for microglia, and PLP1-Cre/ERT (JAX#005975) for OLs.^12,13,14 As the first attempt, we used all Cre/ERT systems to synchronize the activation of Cre recombinase at a specific time for the parallel comparison among different cell types. Although these Cre lines were already well characterized in the literature,^12,13,14 we confirmed the efficiency of each strain. First, to validate the specificity and regional distribution of Cre expression in the PLP1-Cre/ERT line to mature OLs, PLP1-Cre/ERT mice were dissected at P30 and immunostained with the mature OL marker CC1 and Cre in the midbrain. Cre protein driven by PLP1 promoter is efficiently expressed in the majority of OLs (Figure S1A). To further validate the specificity of PLP1-Cre/ERT-mediated recombination to mature OLs, the tdTomato Cre reporter (JAX #007905)¹⁵ and Translating Ribosome Affinity Purification (TRAP) (JAX#024750) mouse models were individually crossed to PLP1-Cre/ERT mice. The TRAP line expresses a GFP-tagged L10a ribosome subunit in a Cre-dependent manner.¹⁶ Tamoxifen was administered to the PLP1-Cre/ERT; tdTomato and PLP1-Cre/ERT; TRAP mice starting at P2 for four consecutive days and tissue collected at P20. Immunostaining for mature OL markers CC1/myelin basic protein (MBP) and OL-lineage marker Olig2 revealed that over 90% of positive cells were co-positive for either tdTomato or TRAP fluorescence, further confirming the specificity of the PLP1-Cre/ERT transgene (Figures S1B and S1C). Additionally, PCR of genomic DNA purified from the whole brain of the PLP1-Cre/ERT; Galc^flox/flox mouse treated with tamoxifen at P2 and dissected at P45 showed that the Galc gene was efficiently deleted in the brain of the mouse model only in the presence of tamoxifen, demonstrating a strict regulation of tamoxifen-induced recombination (Figure S1D). The Cx3CR1-Cre/ERT2 transgene also efficiently underwent recombination in microglia. Tamoxifen was injected into Cx3CR1-Cre/ERT2; tdTomato starting at P2 for four consecutive days. Immunostaining P30 Cx3CR1-Cre/ERT2; tdTomato with Iba1, a pan-microglial marker, revealed greater than 95% recombination efficiency in the microglia of all brain regions tested (Figure S2). Analysis of Aldh1l1-Cre/ERT2; tdTomato was also specific to astrocytes with 60%–80% recombination efficiency in various brain regions. Using cell-specific markers glial fibrillary acidic protein (GFAP) (astrocytes), Iba1 (microglia), Olig2 (OL-lineage cells), and NeuN (neurons), the co-expression with tdTomato was assessed in cryo-sectioned P11 brains wherein colocalization of tdTomato with Olig2, Iba1, and NeuN was negligible compared to GFAP, which saw greater than 50% colocalization in the brainstem, cerebral cortex, corpus collosum, and cerebellum, which had as much as 80% overlap between the markers (Figure S3). The lesser recombination rates observed in Aldh1l1-Cre/ERT; tdTomato mice compared to the other inducible Cre lines may be due to the irregular distribution of GFAP-positive astrocytes.¹⁷ These results suggest that all three Cre/ERT transgenes are efficiently active in each of their respective glial cell targets.

Next, to delete Galc in each glial cell type and investigate their contributions to KD pathogenesis, we crossed each Cre/ERT line with the Galc-floxed mouse (Figure 1A). To maximize the GALC depletion effect, we used the haplodeficient Galc heterozygote: Galc^flox/−. We have previously proved that Galc heterozygotes (Galc^+/−) do not show any difference from Galc^+/+ in respect of their myelin morphology and level of myelin proteins.⁶ Tamoxifen was injected to each inducible knockout mouse along with control Galc^+/− at P2 for four consecutive days and the resultant mice were monitored until 1 year old (Figure 1A). Interestingly, all conditional Galc-KO mouse models showed a trend of decrease bodyweight compared to the Galc^+/− control; PLP1-Cre/ERT; Galc^flox/− were the most drastically affected, followed by Aldh1l1-Cre/ERT2; Galc^flox/− and Cx3CR1-Cre/ERT2; Galc^flox/− (Figure 1B). PLP1-Cre/ERT; Galc^flox/− mice also exhibited a significant increase of GFAP in the brain, which is indicative of reactive astrogliosis (Figure 1C). Neither Aldh1l1-Cre/ERT2; Galc^flox/− nor Cx3CR1-Cre/ERT2; Galc^flox/− elicited any increase of GFAP level (Figures 1C–1E), suggesting that OLs may be the primary glial cell type that induces the pathogenesis of KD. However, PLP1-Cre/ERT-driven Galc-KO did not affect overall survival unlike the global Galc-KO, implying a weak and temporal Galc ablation in OLs may not be sufficient to replicate full KD pathology.

Figure 1.

Temporal Galc ablation in perinatal OLs induces mild inflammation in the brain with a limited pathophysiology

(A) To delete Galc in OLs, astrocytes, and microglia, Galc-floxed mice were crossed to inducible PLP1-Cre/ERT, Aldh1l1-Cre/ERT2, and Cx3CR1-Cre/ERT2 mice, respectively. Then, Cre expression was induced with tamoxifen injections for four consecutive days starting at P2, along with control Galc^+/−. (B) Assessing bodyweights of different lines including tamoxifen-injected Galc^+/− (control) showed that PLP1-Cre/ERT; Galc^flox/− were the most drastically affected followed by Aldh1l1-Cre/ERT2; Galc^flox/−, and Cx3CR1-Cre/ERT2; Galc^flox/−. Only female animals were included. Mann-Whitney-Wilcoxon test was performed; Data are presented as mean values ± SD; ∗p < 0.05,∗∗∗p < 0.001; ns, not significant. (C) Western blot analysis with total brain lysates reveals a dramatic induction of GFAP in the PLP1-Cre/ERT; Galc^flox/− mice model that was not present in Aldh1l1-Cre/ERT2; Galc^flox/−, and Cx3CR1-Cre/ERT2; Galc^flox/− at 6 months old; (D) Quantification of protein band density reveals a significant increase in GFAP levels of PLP1-Cre/ERT; Galc^flox/− mice compared to Galc^+/− (control) n = 3–6 per genotype. ∗∗∗p < 0.001 (one-way ANOVA). (E) Immunohistochemistry against GFAP showed a region-specific increase of astrogliosis in the brainstem, spinal cord, and cerebellum of 6-month-old PLP1-Cre/ERT; Galc^flox/− mice compared to Galc^+/− (control); n = 3 per genotype. Data are presented as mean values ±SEM; ∗p < 0.05 and ∗∗∗p < 0.001 (one-way ANOVA). See also Figures S1–S3.

Constitutive Galc ablation in OLs replicates KD pathology, including neuroinflammation and regional gliosis

We hypothesized that the insufficient phenotype of PLP1-Cre/ERT-driven Galc-KO might be due to limitations of the PLP1-Cre/ERT transgene, while specific, being expressed in only a small subset of OLs, as many ERT-mediated Cre drivers are specific but lack robust and widespread Cre expression. Although there are multiple constitutive Cre mouse lines targeting OLs, none of them is acutely specific to OLs (reviewed in Goebbels and Nave¹⁸). Nonetheless, we decided to use the CNP-Cre mouse as a robust constitutive Cre driver among most OLs.¹⁹ To test the cell specificity of CNP-Cre expression in our hands, we used both the tdTomato and TRAP reporter lines. Both CNP-Cre; tdTomato and CNP-Cre; TRAP mice exhibited specificity for Olig2+ OL-lineage cells over GFAP+, NeuN+, and Iba1+ cells at P11–14 when the brain is actively myelinated (Figures S4A–S4C), suggesting that CNP-Cre is most active in OLs. We also utilized LysM-Cre (JAX#004781) to constitutively delete Galc in microglia in the brain.²⁰ LysM-Cre specificity for microglia assessed by the colocalization of tdTomato with Iba1 in the brains of P10 LysM-Cre; tdTomato mice (Figures S5A and S5B). Also, there was a moderate colocalization of tdTomato in CD206+ perivascular macrophage (PVM) 20.8% ± 5.7% and border-associated macrophage (BAM) 38.0% ± 7.6% seen in P11 LysM-Cre; tdTomato brains (Figures S5C and S5D), indicating a potential contribution of these CNS-resident macrophage populations to the manifestation of pathology.

Next, to confirm cell-specific deletion efficiency of Galc, we crossed the CNP-Cre and LysM-Cre lines with the Galc-flox mouse,⁵ respectively (Figure 2A). Here, we also used the same haplodeficient Galc^flox/− to see the maximal effect of Galc deletion. To validate the reduction of GALC in the CNS of CNP-Cre; Galc^flox/−, we immuno-stained for CC1+ OLs along with GALC at 2 months, a time point when developmental myelination of the CNS is complete.²¹ This analysis showed a significant reduction in the percentage of GALC+ OLs in the CNS of the mutant compared to Galc^+/− control (Figure S6). Immunostaining of Iba1+ microglia with GALC in 2-month-old LysM-Cre; Galc^flox/− also revealed a significant reduction of GALC+ microglia in the mutant compared to Galc^+/− control (Figure S7).

Figure 2.

Ablation of microglial Galc exaggerates neuroinflammation and shortens survival of CNP-Cre-mediated conditional Galc-KO

(A) Galc-floxed mice were crossed to constitutive CNP-Cre and/or LysM-Cre. (B) The CNP-Cre; Galc^flox/− animals survived on average ∼9 months, whereas the concurrent CNP-Cre and LysM-Cre double-conditional Galc-KO mice had an exacerbated phenotype surviving on average ∼5 months. LysM-Cre; Galc^flox/− mice survive more than 1 year, like control Galc^+/−. (C) Body weights of both CNP-Cre; Galc^flox/− and CNP-Cre; LysM-Cre; Galc^flox/− mice were dramatically reduced compared to control Galc^+/−. Female animals only were included. n = 9–13 per genotype. Mann-Whitney-Wilcoxon test was performed;Data are presented as mean values ± SD; ∗∗∗p < 0.001 and ns, not significant. (D) Immunohistochemistry on sagittal brain cryosections revealed a significant increase in the activated microglial markers CD68 and CD163 in the pons, cerebellum, spinal cord, and corpus collosum of 6-month-old CNP-Cre; Galc^flox/− and moribund CNP-Cre; LysM-Cre; Galc^flox/− mice (∼5 months old) when compared to Galc^+/−. Neither genotype elicited an increase in the cerebral cortex. LysM-Cre; Galc^flox/− brain had no increase in any region tested. Nuclei are stained blue with DAPI. Scale bar, 100 μm. (E) Quantitative analysis shows that CD68+ globoid cells are significantly higher in the brainstem, spinal cord, and corpus callosum of CNP-Cre; LysM-Cre; Galc^flox/− mice when compared to CNP-Cre; Galc^flox/−. Data represented as percentage fold change versus Galc^+/+; n = 3 per genotype ∗∗∗p < 001. (F) An increase in the astrogliosis marker GFAP is trending at 2 months and very evident by 6 months in both CNP-Cre; Galc^flox/− and CNP-Cre; LysM-Cre; Galc^flox/− brains. Inflammatory microgliosis marker TLR2 is highly upregulated in both CNP-Cre; Galc^flox/− and CNP-Cre; LysM-Cre; Galc^flox/− brains but not in LysM-Cre; Galc^flox/−. Asterisk (∗) is a non-specific band. Moreover, a decrease in the myelin basic protein (MBP) is seen at 6 months in both CNP-Cre; Galc^flox/− and CNP-Cre; LysM-Cre; Galc^flox/− mice. (G) Analyses of cleaved caspase-3 and TUNEL presence revealed programed cell death present in the brainstem of CNP-Cre; Galc^flox/− and CNP-Cre; LysM-Cre; Galc^flox/− brains at 6 months old. Scale bar, 100 μm. (H) Coimmunostaining of cleaved caspase-3 with cell-specific markers showed that OLs and neurons are the primary cell types dying as result of GALC deficiency. Yellow arrows indicate dying cells. Scale bar, 100 μm. See also Figures S4–S8.

CNP-Cre; Galc^flox/− mice had significantly reduced bodyweight compared to the control Galc^+/− and survived to a maximum age of 10 months. The LysM-Cre;Galc^flox/− mouse model did not demonstrate any defect in bodyweight or survival, suggesting that the loss of GALC in microglia alone is insufficient to induce KD pathogenesis. Interestingly, Galc deletion by both Cre lines at the same time (CNP-Cre; LysM-Cre; Galc^flox/−) had a more severe clinical manifestation resulting in a lower bodyweight and decreased survival (∼4.5 months) compared to the CNP-Cre; Galc^flox/− mice (Figures 2B and 2C). These data indicate that GALC-deficient microglia may exaggerate the phenotype of CNP-Cre-mediated Galc-CKO mice. Neuroinflammation including microgliosis and astrogliosis is a key indicator of KD progression and severity. To assess the extent of gliosis in the CKO models as well as its regional distribution, CD68 and CD163 were assessed via immunohistochemistry in 6-month-old animals. CNP-Cre; LysM-Cre; Galc^flox/− mice were moribund prior to 6 months and were assessed at a humane endpoint around 5 months. Analyses of CD68 and CD163 intensities were highly correlative. The brainstem, cerebellum, spinal cord, and corpus callosum all saw substantial increases in microglial activation in both CNP-Cre; Galc^flox/− and CNP-Cre; LysM-Cre; Galc^flox/− models; however, microglial activation in these regions for LysM-Cre;Galc^flox/− was absent (Figure 2D). Interestingly, CD68+ microglial activation in the brainstem, spinal cord, and corpus callosum in CNP-Cre; LysM-Cre; Galc^flox/− is significantly higher than in CNP-Cre; Galc^flox/− (Figure 2E), indicating that microglial Galc ablation exaggerates the activated globoid cell formation. Inflammatory gliosis markers such as GFAP and Toll-like receptor 2 (TLR2) are highly increased in the end-stage global Galc-KO (Figure S8). The same neuroinflammatory markers are also highly upregulated in the brains of both CNP-Cre; Galc^flox/− and CNP-Cre; LysM-Cre; Galc^flox/− at 6 months of age or when the animals were moribund, as indicated by whole-brain western blot analysis (Figure 2F). Furthermore, western blot analysis of MBP revealed a decrease in myelin protein levels in the 6-month-old brains of both CNP-Cre; Galc^flox/− and CNP-Cre; LysM-Cre; Galc^flox/− (Figure 2F). To assess cell death in vivo, we performed immunostaining of cleaved caspase-3 and utilized the terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL) assay. Signals for cleaved caspase 3 and TUNEL were highly upregulated in the brainstem of both CNP-Cre; Galc^flox/− and CNP-Cre; LysM-Cre; Galc^flox/− models but not in LysM-Cre; Galc^flox/− brains at 6 months old (Figure 2G), indicating induction of classical apoptosis. Furthermore, detailed analysis revealed the activated cleaved caspase-3 signal to be present in either CC1 or NeuN-positive cells in the brainstems of both CNP-Cre; Galc^flox/− and CNP-Cre; LysM-Cre; Galc^flox/− but absent in GFAP- and Iba1-positive cells (Figure 2H), indicating that OLs and neurons could be the primary cells undergoing cell death in the CNP-promoter-driven Galc-CKO brain.

Ablation of Galc in OLs elicits demyelination and neuroaxonal pathology

Demyelination and axonopathy are well-cited hallmarks of KD pathogenesis. Electron microscopic analysis of the optic nerve and cervical spinal cord revealed a significant increase in the percentage of degenerating and demyelinating axons (Figures 3A–3C), with extents comparable to those observed in the induced global Galc-KO mice that our previous study generated (Figure S9A).⁵ There was also an increase in G ratio for both CNP-Cre; Galc^flox/− and CNP-Cre; LysM-Cre; Galc^flox/− models compared to control (Galc^+/−), indicating a thinner myelin sheath. The density of axons was also decreased in the optic nerve of both models compared to the control Galc^+/−, but only a trend of decrease in the spinal cord was observed without reaching statistical significance (Figures 3B and 3C). These findings suggest that CNP-Cre-mediated conditional Galc deletion results in demyelination and axonal degeneration. No significant difference between control (Galc^+/−) and LysM-Cre;Galc^flox/− in any statistic measured was observed, confirming that microglial GALC alone does not directly affect the myelin phenotype of KD. Notably, the CNP-Cre line was generated by replacing the coding region of CNPase gene with Cre;¹⁹ therefore, it is conceivable that the haplo-deficiency of CNPase can contribute to the phenotype of CNP-Cre-mediated Galc-KO. However, the brains of CNP-Cre heterozygote mice did not show any signs of neuroinflammation at 8 months and took until 19 months of age to display a neuroinflammatory phenotype.²² These results suggest that the effects of CNPase haplo-deficiency on 6-month-old Galc-CKO model are likely to be minimal or negligible.

Figure 3.

CNP-Cre; Galc^flox/− brains have morphological signs of demyelination and axonal degeneration

(A) Electron microscopic analyses of the optic nerves and spinal cords in 6-month-old mice shows signs of abnormal myelin sheaths and axonal structures in the brains of both CNP-Cre; Galc^flox/− and CNP-Cre; LysM-Cre; Galc^flox/−. LysM-Cre; Galc^flox/− alone does not have any abnormal morphology. Scale bar, 2 μm. (B) Quantitative analysis of axon density and morphometric properties of the optic nerves indicate a significant decrease in the total number of axons and an increase in the number of degenerating and demyelinating axons in the optic nerves of CNP-Cre; Galc^flox/− and CNP-Cre; LysM-Cre; Galc^flox/− animals compared to control Galc^+/−. G-ratio analysis of optic nerve myelin revealed a significantly thinner myelin sheath in both CNP-Cre; Galc^flox/− and CNP-Cre; LysM-Cre; Galc^flox/− compared to Galc^+/−. Myelin thickness was not changed in the optic nerve of LysM-Cre; Galc^flox/− mice. (C) The spinal cords of CNP-Cre; Galc^flox/− and CNP-Cre; LysM-Cre; Galc^flox/− mice also show a significant increase in both degenerating axons and demyelinating axons, and thinner myelin sheath, compared to control Galc^+/−. Data are represented as mean values ±SEM; ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001; ns, not significant. See also Figure S9A.

Perinatal Galc deletion in microglia exaggerates the OL-specific Galc-KO-driven neuroinflammation

Although CNP-Cre is mostly active in OLs, we cannot exclude other cells’ effects on the pathology of CNP-Cre-mediated Galc ablation. For example, CNP-Cre can be occasionally expressed in motor neurons and radial glia at low levels.^23,24 Our study also showed that 20%–30% of CNP-Cre was active in other brain cell types (Figures S4B and S4C). Therefore, to validate the exacerbated phenotype triggered by microglial Galc deletion upon OL-specific Galc-KO, we utilized the inducible PLP1-Cre/ERT and Cx3CR1-Cre/ERT2 mice and initiated Galc deletion at P2 (Figure 4A). Concurrently induced Galc deletion in both OLs and microglia showed a trend of decrease bodyweight compared to the PLP1-Cre/ERT; Galc^flox/−, but it was not statistically significant (Figure 4B). Also, PLP1-Cre/ERT; Cx3CR1-Cre/ERT2; Galc^flox/− did not affect survival of the mice, unlike the global Galc-KO. Interestingly, the inflammation markers GFAP (astrogliosis) and CD68 (microgliosis) were significantly increased in a region-specific manner of induced PLP1-Cre/ERT; Cx3CR1-Cre/ERT2; Galc^flox/− compared to PLP1-Cre/ERT; Galc^flox/− mice (Figures 4C and 4D). Moreover, the key neuroinflammatory mediator TLR2 was more present in the PLP1-Cre/ERT; Cx3CR1-Cre/ERT2; Galc^flox/− model, confirming that the exacerbated phenotype did indeed result from the knockouts of both microglial and OL Galc. Furthermore, direct comparison of both CD68 and GFAP levels between induced PLP1-Cre/ERT; Cx3CR1-Cre/ERT2; Galc^flox/− and constitutive CNP-Cre; LysM-Cre; Galc^flox/− (Figure 2) mice showed that both proteins are significantly increased in the brainstem of CNP-Cre; LysM-Cre; Galc^flox/− compared to induced PLP1-Cre/ERT; Cx3CR1-Cre/ERT2; Galc^flox/− (Figure S9B), indicating the brainstem of constitutive Galc CKO is more severely affected by GALC deficiency, presumably due to their earlier timing of GALC depletion and a critical role of GALC in the brainstem development.⁵

Figure 4.

Temporal Galc ablation in both perinatal OLs and microglia induces exacerbated gliosis in the brain

(A) To confirm whether temporal Galc ablation in both OLs and microglia have similar exacerbated pathology to CNP-Cre; LysM-Cre; Galc^flox/− over CNP-Cre; Galc^flox/−, Galc-floxed mice were crossed to inducible PLP1-Cre/ERT and Cx3CR1-Cre/ERT2 mice. Galc deletion was induced with tamoxifen injections for four consecutive days starting at P2, along with control Galc^+/−. (B) Body weights of concurrent induced Galc deletion in both OLs and microglia did not show significant change compared to PLP1-Cre/ERT; Galc^flox/−. Only female animals were included. Mann-Whitney-Wilcoxon test was performed. Data are presented as mean values ± SD; ns, not significant. (C) Whole-brain levels of the inflammation markers GFAP (astrogliosis) and TLR2 (microgliosis) show a significant increase in the temporal Galc deletion in both OLs and microglia compared to the PLP1-Cre/ERT; Galc^flox/− mouse model at 6 months old. (D) Immunohistochemistry against GFAP and CD68 shows a region-specific increase of gliosis in the cerebellum and brainstem of 6-month-old PLP1-Cre/ERT; Cx3CR1-Cre/ERT2; Galc^flox/− following deletion in both OLs and microglia compared to PLP1-Cre/ERT; Galc^flox/− mice; n = 3 per genotype. Scale bar, 100 μm. Data are presented as mean values ±SEM; ∗p < 0.05, ∗∗p < 0.01; ns, not significant. See also Figure S9B.

Microglial GALC deficiency increases chemoattractant cytokines but does not affect psychosine level

Accumulation of the cytotoxic lipid psychosine has long been considered the major driver of Krabbe pathology according to the psychosine hypothesis.³ Psychosine, which accumulates in KD because of the loss of GALC, can damage membranes of myelinating glia and neurons in the nervous systems^8,9,10,11 and is thus a major culprit for demyelination and neurodegeneration in the disease.³ Therefore, to examine whether psychosine accumulation in the conditional Galc-KO models is the root of pathology, cervical spinal cord tissue was analyzed for its psychosine content via high-performance liquid chromatography tandem mass spectroscopy (LC-MS/MS) and compared to WT (Galc^+/+), heterozygotes (Galc^+/−), and global-knockout (Galc^−/−) mice. LC-MS/MS analysis revealed an elevation of psychosine concentrations in all OL-specific knockout models (PLP1-Cre/ERT; Galc^flox/−; tamoxifen treated starting at P2, CNP-Cre; Galc^flox/− and CNP-Cre; LysM-Cre; Galc^flox/−) to within the range of Galc^−/− mice at ∼400 pmol/mg of protein each (cf. purple, brown, and indigo bars in Figure 5A). The data show that a substantial proportion of the cytotoxin that accumulates with GALC deficiency occurs with the loss of the enzyme from OLs. The additional loss of GALC from microglia did not contribute further to psychosine accumulation (cf. brown bar and indigo bar). Indeed, the amounts of psychosine in samples from mice with microglial Galc CKO were not different from those in controls (Galc^+/+ and Galc^+/−). These data suggest that there is a psychosine-independent mechanism involving microglia that contributes to KD pathogenesis. No other model inducible (tamoxifen treated at P2) or constitutive CKO mice had elevated psychosine levels comparable to those seen in the global-knockout animals (Figure 5A). These data indicate that OLs are responsible for the bulk of psychosine production; however, this is not a correlative indicator of disease severity considering the relatively mild phenotype observed in PLP1-Cre/ERT; Galc^flox/− mice (Figure 1).

Figure 5.

Microglial GALC deficiency increases recruiting signals with reduced anti-inflammatory cytokine levels but does not affect psychosine accumulation

(A) Psychosine levels were analyzed in the cervical spinal cord of 6-month-old mice for each respective genotype in the study, except Galc^−/−, which were moribund at P37–38. Psychosine was significantly elevated in PLP1-Cre/ERT; Galc^flox/− (tamoxifen treated at P2–5), CNP-Cre; Galc^flox/−, and CNP-Cre; LysM-Cre; Galc^flox/− animals compared to control Galc^+/−. (B) RT-qPCRs of cytokines in the brainstem of 6-month-old mice gave insight into the neuroinflammatory phenotypes of all conditional knockouts generated. Noteworthily, the pro-inflammatory TNF-α and chemoattractant MCP-1 are more expressed in the brainstem of CNP-Cre; LysM-Cre; Galc^flox/− mouse than CNP-Cre; Galc^flox/−, whereas anti-inflammatory protein TGF-β1 is reduced in CNP-Cre; LysM-Cre; Galc^flox/−. MMP-3 and IL-6 are similarly elevated in both CNP-Cre; Galc^flox/− and CNP-Cre; LysM-Cre; Galc^flox/− compared to all other models tested. Data are presented as mean values ±SEM; ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001; ns, not significant.

CNP-Cre; LysM-Cre; Galc^flox/− mice have a much more severe pathology with increased CD68+ globoid cells (Figure 2) but lack an increase in psychosine levels compared to CNP-Cre; Galc^flox/− (Figure 5A), indicating the presence of unknown factors contributing to the difference in pathology. To determine whether there are any differentially expressed cytokines between the models, we performed RT-qPCR analysis of cytokine gene expression in the brainstem of 6-month mice. The brainstem is the initial region of KD pathology manifestation⁵ and had a dramatic increase in globoid cell formation in CNP-Cre; LysM-Cre; Galc^flox/− mouse compared with the CNP-Cre; Galc^flox/− (Figure 2). This analysis revealed a corresponding trend in the expression of colloquial inflammatory genes between the Galc^−/−, CNP-Cre; Galc^flox/−, and CNP-Cre; LysM-Cre; Galc^flox/− mouse models (Figure 5B). Interestingly, transcripts for the pro-inflammatory tumor necrosis factor (TNF)-α and the chemoattractant MCP-1 were more produced in the brainstem of the CNP-Cre; LysM-Cre; Galc^flox/− mouse than CNP-Cre; Galc^flox/−, whereas transcripts for the anti-inflammatory protein transforming growth factor (TGF)-β1 was lower in CNP-Cre; LysM-Cre; Galc^flox/−. Similarly, MCP-1 was also significantly upregulated in the PLP1-Cre/ERT; Cx3CR1-Cre/ERT2; Galc^flox/− brain when compared with PLP1-Cre/ERT; Galc^flox/− mice. These results suggest that the concomitant microglial GALC deficiency boosts recruiting signaling with reduced anti-inflammatory capacity. Other well-documented inducers or effectors of neuroinflammation in KD, such as matrix metalloproteinase (MMP)-3, interleukin (IL)6, and arginiase-1, were similarly elevated in both the CNP-Cre; Galc^flox/− and CNP-Cre; LysM-Cre; Galc^flox/− or PLP1-Cre/ERT; Galc^flox/− and PLP1-Cre/ERT; Cx3CR1-Cre/ERT2; Galc^flox/−, but their levels were much higher in CNP-Cre; LysM-Cre; Galc^flox/− than PLP1-Cre/ERT; Cx3CR1-Cre/ERT2, emphasizing their correlation with disease severity. GFAP level was also much higher in the brain of CNP-Cre; Galc^flox/− than PLP1-Cre/ERT; Galc^flox/− (Figure S9C). To further examine the inflammatory mediators involved in the phenotype difference in an unbiased manner, we used the Proteome Profiler Mouse Cytokine Array analysis, which can detect 111 different mouse cytokines simultaneously. Interestingly, secreted phosphoprotein (SPP1), so called osteopontin (OPN), is significantly upregulated in CNP-Cre; LysM-Cre; Galc^flox/− brains compared to CNP-Cre; Galc^flox/− (Figures 6A and 6B). Since OPN is known to be secreted into the extracellular matrix from activated microglia to recruit immune cells²⁵ and is required for microglia to engulf other cells,²⁶ we speculate that microglia with GALC deficiency secrete factors including OPN and form globoid cells that contribute to the progression of KD. OPN is also significantly upregulated in cultured GALC-deficient microglia compared to WT microglia (Figure 6C).

Figure 6.

Osteopontin is significantly upregulated in the brain of CNP-Cre; LysM-Cre; Galc^flox/− mouse

(A) Proteome Profiler Mouse Cytokine Array, which can detect 111 different mouse cytokines simultaneously, shows that SPP1, so called osteopontin (OPN), is more upregulated in the CNP-Cre; LysM-Cre; Galc^flox/− brains than CNP-Cre; Galc^flox/− at 6 months old. (B) Quantification of protein dot density revealed a significant increase in OPN levels of CNP-Cre; LysM-Cre; Galc^flox/− compared to CNP-Cre; Galc^flox/−. Data are presented as mean values ±SD; ∗∗∗p < 0.001. (C) Cultured Galc^−/− microglia express more OPN than Galc^+/+ control. Scale bar, 20 μm. Data are presented as mean values ±SEM; ∗∗∗p < 0.001.

Globoid cells are formed from GALC-deficient microglia cocultured with degenerating OLs

Next, we cocultured microglia with mature OLs (3 days of in vitro differentiation) from Galc-KO and WT mice and found dramatic differences in the morphological change between control and GALC-deficient microglia. Notably, the Galc-KO microglia became multinucleated globoid cells in the presence of Galc-KO OLs (Figure 7A). Immunocytochemistry with Iba1 (microglia) and MBP (mature OLs) along with DAPI shows multinucleated globoid cells that contain OLs and microglial markers. An orthogonal magnified view of the globoid cells shows multi-nucleation with multiple nuclei enveloped within a distinct boundary of microglia and MBP-positive myelin debris (Figure S10A). The globoid cell phenotype was unique to the double-knockout condition and not observed in other combinations of OLs and microglia with or without GALC deficiencies (Figure 7B). The quantification of total microglia numbers for each condition was not significantly different (Figure S10B), indicating GALC deficiency does not affect microglial survival in vitro. Interestingly, OPN was significantly increased in cultures of Galc-KO microglia cocultured with Galc-KO OLs (Figure 7C), supporting the result that Galc-KO microglia are the source of OPN production in the brain of CNP-Cre; LysM-Cre; Galc^flox/− mice (Figure 6). The gene expression levels of the chemoattractant MCP-1 and pro-inflammatory IL-6 are also highly upregulated in the coculture of both GALC-deficient microglia and OLs, but other cytokines were not (Figure 7D), indicating that globoid cells may contribute to the production of MCP-1 and IL-6 in KD. There were also increasing trends in the levels of TNF-α, MMP-3, and TGF-β1 cytokines in the double-knockout coculture conditions, but they were not significant compared to controls.

Figure 7.

Galc-KO microglia transformed into globoid cells overproduce MCP-1 and IL-6

(A) Analysis of cultured microglia with mature OLs (3 days in vitro differentiation) from Galc-KO and WT mice shows that the Galc-KO microglia became multinucleated globoid cells in the presence of Galc-KO OLs. Immunocytochemistry with Iba1 (microglia) and MBP (mature OL) along with DAPI shows multinucleated globoid cells that contain OLs and microglial markers. Scale bar, 50 μm. (B) Quantification of multinucleated globoid cell formation reveals that more than 50% of Galc-KO microglia with Galc-KO OLs become globoid cells. However, the globoid cell phenotype was not observed with other combinations of OLs and microglia with/without GALC deficiencies. (C) OPN level was significantly increased from Galc-KO microglia in the presence of Galc-KO OLs. (D) The chemoattractant MCP-1 and pro-inflammatory IL-6 were highly upregulated from Galc-KO microglia in the presence of Galc-KO OLs, but other cytokines were not. Data are presented as mean values ±SEM; ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001; ns, not significant. See also Figure S10.

Discussion

Considering the ubiquitous nature of GALC expression in the CNS, the current study looked to expand the understanding of glial cells’ contributions to KD pathology in an autonomous or combined manner, which remains largely unstudied. Using both tamoxifen-inducible and constitutively active Cre drivers, we were able to efficiently induce recombination of the floxed Galc gene in all major glia. Herein, we show that OLs are the primary driver of KD pathogenesis; however, OL-specific Galc deletion alone is insufficient to recapitulate the full disease state. While the induced PLP1-Cre/ERT-mediated Galc-KO model is adequate to incite KD pathology, the constitutive CNP-Cre-mediated Galc-KO model more fully recapitulates disease severity indicated by induction of severe neuroinflammation, demyelination, and shorter lifespan (summarized in Table S1). Since CNP-Cre can be active in other brain cell types, including motor neurons and radial glia,^23,24 it is conceivable that the severe pathology of CNP-Cre-mediated Galc-KO mice implies that the contribution of other non-OL cell types is critical for the progression of KD. This notion is further exemplified by the exacerbated disease state of both CNP-Cre; LysM-Cre; Galc^flox/− and PLP1-Cre/ERT; Cx3CR1-Cre/ERT2; Galc^flox/− when compared to CNP-Cre; Galc^flox/− and PLP1-Cre/ERT; Galc^flox/−, respectively (Figures 2 and 4). Moreover, the discrepancy in disease severity between the CNP-Cre; Galc^flox/−, CNP-Cre; LysM-Cre; Galc^flox/− and PLP1-Cre/ERT; Galc^flox/− mice is particularly of interest considering the similar accumulation of psychosine between the models (Figure 5A). These data provide insight into the long standing psychosine hypothesis, which states that accumulation of the cytotoxic lipid psychosine is the etiological driver of Krabbe pathology.^3,27 Herein, we demonstrate that this conclusion may not be as cut and dried as previously described considering the more severe disease state seen in the CNP-Cre; LysM-Cre; Galc^flox/− compared to the CNP-Cre; Galc^flox/− model and both in comparison to PLP1-Cre/ERT; Galc^flox/− mice. These findings, in corroboration with our recent study using the pan-neuronal and constitutive Syn1-Cre, which demonstrated that GALC-deficient neurons are sufficient to induce neurodegeneration and neuroinflammation, show that non-OL cell types are key contributors to the overall phenotype observed in the global Galc-KO.⁶ Although cell-autonomous Galc-KO in either microglia or astrocytes alone was insufficient to elicit a pathological phenotype, additional microglial Galc deletion elicited the worst severity of illness upon CNP-Cre-mediated Galc-KO. Through the use of complementary inducible and constitutive Cre systems, we are able to infer that OLs are the primary drivers of KD pathogenesis and the contribution of other Galc-KO cell types further exacerbates disease severity. Moreover, we demonstrate that Galc-KO microglia are key to eliciting an enhanced severe disease state seemingly in response to OL dysfunction and demyelination further exacerbating the KD phenotype.

In KD, accumulation of the toxic lipid psychosine is due to the deacylation of GalCer by acid ceramidase and causes extensive demyelination.^3,4 Therefore, the extent of psychosine accumulation and CNS demyelination and axonal degeneration are key indicators of KD pathology. Direct comparison of the electron microscopic analyses of demyelinated axons between CNP-Cre; LysM-Cre; Galc^flox/− and the induced global Galc-KO from our previous study⁵ revealed that, despite the milder phenotype, the densities of both demyelinating and degenerating axons in the optic nerves and spinal cords of CNP-Cre; Galc^flox/− mice were comparable to the levels seen in the induced global Galc-KO model (Figure S9A). These data further emphasize that fully developed myelin pathology is insufficient to recapitulate a full KD phenotype. Further comparison of regional gliosis on brain sections revealed that both CD68 and GFAP were elevated in the brainstem of CNP-Cre; LysM-Cre; Galc^flox/− to the levels observed in the global Galc-KO, but much less activated in cerebral cortex (Figure S9B). CD68 was also significantly lower in the cerebellum of CNP-Cre; LysM-Cre; Galc^flox/− compared to Galc-KO, whereas GFAP was not different. Levels of both proteins were reduced in the brain of PLP1-Cre/ERT; Cx3CR1-Cre/ERT2; Galc^flox/− compared to Galc-KO. These results suggest that, while the brainstem is a crucial component of KD pathology, the full severity of KD pathogenesis requires input from all regions of the CNS. Moreover, the relatively limited recombination efficiency of LysM-Cre and absence of a constitutive astrocyte model leave open questions about the extent to which inflammatory gliosis is responsible for KD pathophysiology. These data may also indicate that the regional difference of gliosis for each CKO model may be dependent on Cre activity.

The enhanced disease state observed in the CNP-Cre; Galc^flox/− mouse may be attributed to the increase in pro-inflammatory transcripts. Increases in the global levels of TNFα, MCP-1, and MMP-3 were unique to the CNP-Cre; Galc^flox/− and CNP-Cre; LysM-Cre; Galc^flox/− models (Figure 5B) and may not only facilitate the recruitment and activation of gliosis shown by immunostaining (Figure 2D) but also impair future oligodendroglia differentiation.²⁸ Especially, TNFα and MCP-1 are uniquely upregulated in both inducible and constitutive OLs and microglia double-CKO mice compared to each corresponding OL-specific CKO, suggesting that microglial GALC depletion further exaggerates the induction of both cytokines upon OL CKO. These findings further highlight the contributions of non-oligodendroglia cell types to KD pathology and their implications in disease progression. Future insight into the molecular mechanism inducing rampant gliosis that appears to be psychosine independent as seen in the CNP-Cre; Galc^flox/− model will likely contribute significantly to our understanding of KD pathogenesis. Furthermore, the similar accumulation of psychosine between the constitutive CNP-Cre; Galc^flox/− and inducible PLP1-Cre/ERT; Galc^flox/− models gives insight into the time course of psychosine accumulation. Considering the induction scheme utilized with tamoxifen administration to the PLP1-Cre/ERT; Galc^flox/− mouse model beginning at P2 and proceeding for four consecutive days, we can assume that a majority of psychosine accumulation occurs in the time period following postnatal Galc deletion. This agrees with previous findings indicating that a vast increase in psychosine accumulation is seen during the second week of life, a time when developmental myelination is intensifying.²⁹

Multinucleated globoid cells are a prominent feature of KD³⁰ and found prior to the onset of overt demyelination^31,32 and damage-associated inflammation^33,34 in KD fetal tissues, indicating their primary role in KD pathogenesis. However, the impact of these cells on the disease is largely unknown. Despite efforts to determine the cell type or types that give rise to globoid cells and the mechanism by which they form,^35,36 there still remains inconclusive evidence to support whether globoid cells in the CNS are produced by the activated microglia or derived from the perivascular macrophages. Furthermore, how the globoid cells affect KD pathology is still a topic of debate.^36,37,38,39 In this study, we found that CD68+ globoid cells are more prevalent in the brainstem, spinal cord, and corpus callosum in CNP-Cre; LysM-Cre; Galc^flox/− compared to CNP-Cre; Galc^flox/− mice, indicating that GALC-deficient microglia may exaggerate the activated globoid cell formation. A significant increase of MCP-1 in the brainstem of CNP-Cre; LysM-Cre; Galc^flox/− from this analysis indicates that GALC-deficient microglia may attract more immune cells as MCP-1 is a pro-inflammatory chemokine that is involved in recruitment and activation of microglia. Elevated expression of MCP-1 in microglia has been implicated in multiple neurodegenerative disorders, such as Alzheimer’s disease (AD), multiple sclerosis, and traumatic brain injury.^{40,41,42,43,44} Interestingly, MCP-1 is known to be essential for osteoclasts to form “multinucleated” morphology.⁴⁵ MCP-1 mediates the remodeling of the extracellular matrix and cell-cell fusion,⁴⁶ suggesting a similar mechanism involving MCP-1 may contribute to the formation of multinucleated globoid cells in KD. Also, a reduced production of TGF-β1 in CNP-Cre; LysM-Cre; Galc^flox/− suggests an inefficient anti-inflammatory reaction of Galc-KO microglia. Proteome Profiler Mouse Cytokine Array analysis revealed an increase of OPN in the brain of CNP-Cre; LysM-Cre; Galc^flox/− (Figure 6). Further analysis of microglia cocultured with OLs in vitro revealed that Galc-KO microglia generate a globoid cell morphology and increase production of OPN and MCP-1 in response to Galc-KO OLs (Figure 7), indicating that globoid cells are the major producer of these cytokines in KD. In particular, as a secretory protein from activated microglia, OPN has been linked to pathological effects such as immune cell recruitment²⁵ and engulfment²⁶ in multiple sclerosis⁴⁷ and AD,²⁶ suggesting a similar mechanism of action by the globoid cells in KD. Additionally, an upregulation of the pro-inflammatory cytokine IL-6 in these cultures gives insight to the mechanism by which globoid cells promote a neuroinflammatory phenotype. It is likely that globoid cells may attract more microglia to the damaged site by production of OPN and MCP-1, thereby further facilitating globoid cell formation in a feedforward loop and thus exacerbating inflammation in an IL-6-dependent manner.

The Cre-LoxP system is a cornerstone for in vivo genetic manipulation and provides valuable insights into disease mechanisms; nonetheless, several limitations of the systems used in this study restrict a comprehensive analysis of all glial cells within the CNS and should be addressed. Specifically, the haplo-deficiency associated with CNP-Cre and Cx3CR1-Cre/ERT2 knockin alleles has been suggested to provoke phenotypes of their own, potentially complicating interpretation. Additionally, the LysM-Cre allele elicited less than 50% recombination in Iba1+ microglia while simultaneously provoking recombination in BAM (38.0% ± 7.6%) and PVM (20.8% ± 5.7%) (Figure S5). Further, we had previously assessed LysM-Cre activity in peripheral macrophage and found it to elicit greater than 80% recombination in splenic macrophage and 90% in sciatic macrophage populations.⁷ It is therefore key to acknowledge the limitations of LysM-Cre and the potential influence of peripheral cell populations and CNS-resident macrophage to the phenotype observed, especially considering that the current standard of care for KD is hematopoietic stem cell transplant. One notable caveat to the study relates to recent findings indicating that PVM is the primary cell type responsible for producing OPN in AD, wherein the secreted protein promotes microglial synaptic engulfment and drives pathology.^26,48 This prompts speculation to whether the exacerbated phenotype of CNP-Cre; LysM-Cre; Galc^flox/− can be attributed to OPN derived from recombined CNS-resident macrophages. Further analysis will need to be done to examine whether a more robust deletion of Galc in microglia further recapitulates disease severity or, conversely, could ablation of Galc in peripheral hematopoietic cells and/or CNS macrophage populations be responsible for exacerbating pathology by an OPN-dependent mechanism and, therefore, more specific microglial ablation may actually be less deleterious? These questions require further investigation utilizing more specific and robust Cre drivers to delineate the precise role of immune cells in driving KD pathology. Lastly, although the phenotype of Aldh1l1-Cre/ERT2; Galc^flox/− with gliosis markers such as GFAP, CD68, and CD163 were carefully evaluated, we were unable to utilize a constitutive Cre line specific to astrocytes since it was not available at the time of study. Extensive fibrillary astrogliosis is a prominent feature of KD, but the contribution of GALC-depleted astrocytes into Krabbe pathogenesis is largely unknown. It was previously reported that MMP-3 mediates psychosine-induced globoid cell formation in twitcher (Galc-KO) mice and that MMP-3 mostly comes from astrocytes.^35,36 Our study also shows an increase of MMP-3 unique to both CNP-Cre; Galc^flox/− and CNP-Cre; LysM-Cre; Galc^flox/− mice (Figure 5B), but it was not specific to Galc-KO microglia (Figure 7D), indicating the production of MMP-3 from other cells such as astrocytes. A recent study has shown that microglial debris is cleared largely by astrocytes due to the contact inhibition of neighboring microglia in vivo.⁴⁹ Astrocyte-mediated microglial corpse clearance is facilitated by C4 opsonization and is degraded via non-canonical autophagy. Therefore, additional loss of GALC-positive astrocytes may then render the cell unable to degrade dying microglia and thus further aggravate KD pathology. RT-qPCR analysis of astrocyte activation status⁵⁰ between CNP-Cre; Galc^flox/− and CNP-Cre; LysM-Cre; Galc^flox/− mice revealed no difference in the transcripts abundance associated with pro-inflammatory (H2D1, H2T23, GBP2, and SGRN) or anti-inflammatory (STAT3, AQP-4, and STAT6) astrocyte polarization status (Figure S11). Prior studies have demonstrated that astrocytes are competent phagocytes of myelin in vivo; moreover, the uptake of myelin debris induces the nuclear translocation of nuclear factor κB (NF-κB) in purified astrocytes, eliciting their activation and proliferation.^51,52 These findings suggest that GALC-deficient OLs are sufficient to provoke astrocyte activation, which was not further exaggerated by the simultaneous ablation of Galc in microglia.

Although we showed that the cross-correction of GALC still occurs even at a lower level in KD cells in vitro in the previous study,⁷ it was not enough to hinder any appearance of Galc-CKO phenotype in this study. That may indicate that endogenous GALC was unable to sufficiently cross-correct GALC-deficient cells. Indeed, a significant reduction in the levels of GALC in the OLs of CNP-Cre; Galc^flox/− and in the microglia of LysM-Cre; Galc^flox/− to the level of Galc-KO by immunostaining for CC1+ OLs (Figure S6) and Iba1+ microglia (Figure S7) in the CNS along with GALC, respectively, suggests that GALC is efficiently removed in each targeting cell type and cross-correction from WT cells to GALC-null cells may be limited at least for OLs and microglia. Whether this phenomenon extends to other LSDs should be further studied, and this may explain why certain corrective therapies have had only modest clinical improvements despite overall increased enzymatic activity.

In conclusion, our study found that OL-specific Galc deletion in mice results in a phenotype that includes wasting, psychosine accumulation, and neuroinflammation, suggesting a primary role of OLs in KD. We also showed that OL and microglia double-Cre-driven Galc-CKO mice triggers a more severe phenotype than OL-specific Galc CKO, demonstrating that GALC-deficient microglia are critical in driving disease progression. However, their phenotype was much milder than the full Galc-KO, although the level of psychosine was not significantly different from that of the full Galc-KO, indicating that psychosine may not be sufficient to drive the entirety of disease burden in KD. Therefore, the influence of another cell type or type(s) that express GALC and psychosine-independent mechanisms are likely vital elements in the development of KD and thus crucial factors to consider when designing therapies. Our studies using the Galc-floxed allele have laid the groundwork for the in vivo analysis of cell-specific contributions to KD and propose specific molecular mediators of pathogenesis. Further studies will aim to elaborate on the mechanisms by which OLs- or microglia-specific GALC depletion triggers cellular and lysosomal pathogenesis in KD and whether similar processes occur in other LSDs.

Materials and methods

Animals

Experiments were conducted according to the protocols approved by the Institutional Animal Care and Use Committee of University at Buffalo (SUNY) and Roswell Park Cancer Institute (protocol approval nos. UB1188M, UB1254M, and PROTO202000063). Mice were housed under specific pathogen-free conditions at 70°F, 50% room humidity, 12-h light/12-h dark cycle, and received ad libitum access to water and food. All animals were maintained on the congenic background of C57BL/6N. Breeder C57BL/6N mice were purchased from Charles River (Wilmington, MA). Galc-flox mouse line was generated as described in Weinstock et al.⁵ PLP1-Cre/ERT (JAX#005975), ALDH1L1-Cre/ERT2 (JAX#023748), Cx3CR1-Cre/ERT2 (JAX#025524), LysM-Cre (JAX#004781), TRAP or EGFP-L10a (JAX#024750), and tdTomato (JAX#007905) mice were purchased from The Jackson Laboratory (Bar Harbor, ME). CNP-Cre mouse was a gift from Dr. Klaus-Armin Nave.¹⁹ To maximize the GALC depletion effect, the haplodeficient Galc heterozygote Galc^flox/− was used, with Galc^+/− control. Galc^+/− did not show any difference to Galc^+/+ in respect of the myelin morphology and level of myelin proteins.⁶ In addition, the expression of myelin proteins such as myelin-associated glycoprotein (MAG) and MBP was not different between both haplodeficient Galc^+/− and Galc^flox/−, suggesting the unrecombined floxed allele itself does not affect myelin phenotype.⁶ For induced Galc-CKO models, perinatal Cre/ER^T; Galc-floxed mice were injected intraperitoneally for four consecutive days (total four times) with 25 μg/g body weight starting at P2 along with Galc^+/− control. This was the maximum achievable dosage in our hands while avoiding tamoxifen-induced gastric toxicity.⁵³

Survival and body weight measurements

Animals had their bodyweight recorded bi-weekly using a cleaned plastic beaker and scale. Measurements were taken only on females to remove any gender bias and were recorded up to 1 year or time of death. This allowed survival tracking of cohorts of animals from the same litter or closely born litters.

Tissue and immunohistochemistry

Mice at defined ages were anesthetized, sacrificed, and then perfused with ice-cold phosphate-buffered saline (PBS) followed by 4% paraformaldehyde (PFA). If the tissue was for protein or RNA analysis, only PBS was perfused, and the tissue was snap frozen in liquid nitrogen and stored at −80°C for later analysis. Brains and spinal cords were dissected, post-fixed in 4% PFA for 48 h, dehydrated in 30% sucrose at 4°C, embedded in optimal cutting temperature (OCT) freezing medium (Lecia), and processed as cryosections with a thickness of 15 μm. For immunohistochemistry, cryosections were rehydrated in PBS, then permeabilized and blocked in 0.1% Triton X-100, 3% bovine serum albumin, and 2% normal goat serum for 1 h at room temperature. For GALC immunostaining, the blocking buffer also contained 0.05% saponin. The primary antibodies were then incubated overnight at 4°C. After washing (three times for 10 min) with PBS, sections were incubated with fluorophore-conjugated secondary immunoglobulin (Ig) Gs (Jackson Laboratories). After washing with PBS, coverslips were mounted with Vectashield (Vector Laboratories) mounting medium and DAPI. Primary antibodies used were GALC,⁵⁴ Iba1(Wako), CD68 (Bio-Rad), CD163 (Bio-Rad), GFAP (Sigma-Aldrich), MBP (EMD Millipore), cleaved caspase-3 (Cell Signaling Technology), CC1 (Calbiochem), Cre (Novus), NeuN (EMD Millipore), CD206 (Proteintech), and Laminin (R&D Systems). For colorimetric cleaved caspase-3 staining, SignalStain Boost IHC Detection Reagent (Cell Signaling Technology) was used with hematoxylin counterstain. Images were acquired and all analysis was performed blinded to genotypes. Detection of apoptotic cells with the DeadEnd Fluorometric TUNEL System (Promega) was used according to the manufacturer’s directions. Quantification of CD68, CD163, and GFAP fluorescence intensity was measured using FIJI (NIH). Briefly, the respective antibodies were imaged with identical parameters including gain and laser intensity, then background fluorescence was subtracted using a control wherein the same staining protocol was followed with exception of primary antibody addition. Finally, the respective genotype intensities were normalized to the WT control values before statistical analyses; the values represented in the graph are percentage change compared to the WT. For cellular quantification of GALC, maximum intensity images were segmented with ilastik (v1.4b3).⁵⁵ The segmented images were processed with FIJI (NIH) to quantify the intensity of the GALC staining per cell.

Microglia purification and culture

Primary microglia are purified as described in Saura et al.⁵⁶ Briefly, whole brain from P5 mouse was minced to liquefy and cultured in Dulbecco’s modified Eagle’s medium supplemented with 10% fetal bovine serum (Gibco), 1% sodium pyruvate, 1% penicillin-streptomycin, 2× Glutamax (Invitrogen), and 1% amphotericin B until reaching confluency. Once the culture reached confluency, microglia were collected from the mixed culture using sequential trypsinization with 0.25% and 0.05% Trypsin.

RT-qPCRs

RNA was isolated using TRIzol reagent following the manufacturer’s instructions (Thermo Fisher Scientific). RNA was then analyzed for purity on a NanoDrop 2000 spectrophotometer (Thermo Fisher Scientific). cDNA was synthesized from 500 ng of RNA using the Superscript III kit (Thermo Fisher Scientific) according to the manufacturer’s instructions. qPCR was performed using the SYBR Green PCR Master Mix (Applied Biosystems) with β-actin as a reference. All samples were analyzed with at least three technical replicates and the relative expression was calculated using the ΔΔC_t method as described in Livak and Schmittgen.⁵⁷ The sequences and efficiencies of all PCR primers are provided in Table S2.

Morphological analysis

Mice were anesthetized with 250 mg/kg body weight Avertin (Sigma-Aldrich) and then perfused with PBS and 2.5% gluta-aldehyde in phosphate buffer and stored at 4°C until processing. After being post-fixed, the spinal cord and optic nerve were dissected, incubated in 1% osmium tetroxide, dehydrated using sequential incubation in ethanol of increasing concentration, and embedded in Epon resin using propylene oxide as a transition solvent. Semithin sections were cut with 1-μm thickness and stained with 2% toluidine blue. Ultrathin sections of thickness 80–85 nm were stained with uranyl acetate and lead citrate to be examined by a Tecnai electron microscope. For G-ratio analyses of myelin sheath, ultrathin electron microscopy (EM) images were captured at ×2,900 and manually measured using FIJI (NIH).

Western blot analyses

After homogenizing whole brains in pre-chilled radio-immuno-precipitation assay (RIPA) lysis buffer containing protease inhibitors (Roche) and PMSF?, total protein extracts were separated by SDS-PAGE, transferred to polyvinylidene fluoride (PVDF) membrane (EMD Millipore), and blocked with 5% skim milk or BSA in TBS-Tween20. Primary antibodies used were β-tubulin (Novus Biologicals), GFAP (Sigma-Aldrich), TLR2 (Abcam), CNPase (Cell Signaling Technology), MAG (Zymed), PLP (hybridoma AA3), and MBP (EMD Millipore). Specific protein bands were quantified utilizing FIJI (NIH) and Image Studio (LI-COR Biosciences), and the values (in pixels) obtained were normalized on those of the corresponding β-tubulin bands. Normalized values were then expressed as the percentage of values obtained from matched bands of control tissues.

Proteome cytokine analysis

Whole-brain tissue samples were homogenized in ice-cold PBS with protease inhibitors (Roche), PMSF, and 1% Triton X-100. The homogenates were frozen at −80°C, thawed, and centrifuged at 10,000 × g for 5 min at 4°C and the supernatant was collected. The supernatant was analyzed using the Proteome Profiler Mouse XL Cytokine Array #ARY028 (R&D Systems) according to the manufacture’s protocol. Briefly, the nitrocellulose membranes of the array kit contain capture antibodies spotted in duplicate against 111 cytokines. Each membrane was incubated with 300 μg of total brain lysate overnight at 4°C. Following extensive washing, the membrane was incubated with the biotinylated detection antibody cocktail. The membranes were then incubated with Pierce ECL western blotting substrate (Thermo Fisher Scientific) and visualized on a C-DiGit Blot Scanner (Li-COR Bioscience) and specific protein dots were quantified via Image Studio and the values (in pixels) were normalized to the reference spots values. Normalized values where then expressed as percentage of values from matched CNP-Cre; Galc^flox/− animals.

Measurement of psychosine

Psychosine level in tissues were analyzed as described in Weinstock et al.⁷ Briefly, cervical spinal cords were homogenized in PBS. A fraction of PBS homogenate was refrozen and shipped for analysis by Dr. Michael Gelb’s lab at the University of Washington. The other fraction of PBS homogenate was mixed with 10× RIPA lysis buffer (Cell Signaling Pathway) to make 1× RIPA buffer. Samples were then sonicated and analyzed for protein quantification. For psychosine analysis, 250 μL of 1 nM d₅-psychosine (Avanti Polar Lipids) in methanol was added to 5 μL of tissue/PBS homogenate. Psychosine was extracted at 37°C for 2 h with orbital shaking and subsequent centrifugation. The supernatant was loaded onto a methanol-preconditioned Oasis MCX column (1 mL, 30 mg, Waters Corp., #186000252). After sample loading, the cartridge was washed with 1 mL of water with 2% formic acid, 1 mL of methanol, and then 1 mL of 80:20 methanol:water (v:v) with 5% NH₄OH. The column was washed with 0.8 mL of methanol with 5% NH₄OH, which was collected and solvent evaporated using a SpeedVac vacuum concentrator. The residue was reconstituted with 100 μL of mobile phase B prior to ultra-performance liquid chromatography (UPLC)-MS/MS analysis. An ACQUITY UPLC I-Class system from Waters was used for the separation of glucosyl- and galactosyl-sphingosine (psychosine). The UPLC system was coupled to a Xevo TQ-S (Waters) tandem mass spectrometer, which was operated in the multiple reaction monitoring (MRM) mode.

Statistical analyses

Data collection and analysis were performed blind to the conditions of the experiments. Two-tailed unpaired Student t test and one-way ANOVA with Bonferroni correction were used for the differences among multiple groups according to the number of samples, except G-ratio analysis, in which Welch’s t test was used. Mann-Whitney-Wilcoxon test was used for body-weight comparison. Values of p < 0.05 were considered to represent a significant difference. Statistical tests were run in Prism version 8.0 (GraphPad). Data are presented as mean ± SEM unless otherwise specified.

Data and code availability

The data that support the findings of this study are available from the corresponding author upon reasonable request.