Growth Differentiation Factor 15 Elevation in the Central Nervous System Is Associated With Failure to Thrive in Alexander Disease

ABSTRACT

Objective

Alexander disease (AxD) is a severe neurodegenerative disorder caused by gain‐of‐function mutations in the gene for GFAP, which lead to protein aggregation and a primary astrocytopathy. Symptoms vary, but failure to thrive (FTT) and frequent emesis are common and cause significant morbidity. Here we investigate GDF15, a member of the TGFβ superfamily, which regulates energy balance and appetite, as a potential mediator of FTT in AxD.

Methods

In this study, we use the Gfap ^+/R237H rat model (R237H), in which pups fail to gain weight after weaning and become frail and impaired as they mature, to assess muscle atrophy, energy expenditure, and feeding behavior in AxD. We measure GDF15 in brain and cerebrospinal fluid (CSF), assess activation of its receptor GFRAL in area postrema neurons, and use GFAP suppression to correlate FTT phenotypes with GDF15 expression. Finally, we measure GDF15 in patients with AxD.

Results

R237H rats show reduced lean and fat mass and muscle atrophy despite reduced energy expenditure, and at an early age exhibit pica and anorexia. GDF15 is expressed by R237H rat astrocytes and is elevated in brainstem and CSF, but not in plasma. Neurons expressing GFRAL, a mediator of GDF15‐induced appetite suppression, are activated in the area postrema. Suppression of GFAP using antisense oligonucleotides normalizes weight gain and GDF15 levels in brainstem and CSF. In human AxD, GDF15 is elevated in CSF, but not in blood.

Interpretation

GDF15 is associated with FTT in AxD and provides both a target and useful biomarker for the development of future therapeutics.

1. Introduction

Alexander disease (AxD) is a leukodystrophy caused by dominant gain‐of‐function mutations in the gene for glial fibrillary acidic protein (GFAP), the major intermediate filament of astrocytes in the central nervous system (CNS). AxD‐associated mutations cause GFAP aggregation and astrogliosis with a marked increase in stress and neuroinflammatory markers, including GFAP itself [1, 2], and studies in mouse models suggest this feed‐forward mechanism exacerbates mutant protein toxicity and contributes to disease severity [3, 4]. Disease symptoms vary with age of onset and lesion location and can involve decreased appetite, intractable vomiting, and poor weight gain leading to frank failure to thrive (weight gain below the 2nd percentile for age) with significant morbidity [5, 6]. Intractable vomiting in AxD is frequently associated with lesions detectable on MRI in the area postrema (AP) [6], a circumventricular organ in the brainstem known to regulate emesis. Recent studies have shown that growth and differentiation factor 15 (GDF15), a distant member of the TFGβ family of cytokines, is a mediator of the stress response that regulates appetite, nausea, and energy balance [7, 8, 9], independent of physiological appetite regulation, via cholecystokinin neurons harboring the GDF15 receptor GFRAL (GDNF family receptor α‐like) and its transmembrane coreceptor RET in the AP and adjacent nucleus tractus solitarius (NTS) [10, 11].

We have developed a Gfap ^+/R237H rat model of AxD with a mutation equivalent to the severe R239H variant in the human GFAP sequence (referred to hereafter as R237H). These rats appear normal at birth but suffer a precipitous decline after weaning, with little to no weight gain, increasing motor impairment, abnormal gait, and white matter deficits [12]. Reactive astrocytes with GFAP pathology are prevalent throughout the R237H rat CNS, including the brainstem, and elevated GDF15 transcript expression has been demonstrated in both rat and mouse models of AxD [13, 14, 15] as well as cellular models of astrogliosis [16]. In this report, we show that the R237H rat exhibits a cachexia‐like phenotype with severe anorexia, nausea, sarcopenia, reduced lean mass, and decreased energy expenditure. We demonstrate elevation of GDF15 in the CNS but not in the periphery in both the rat model and humans with Alexander disease. In addition, we show that failure‐to‐thrive (FTT) phenotypes, including elevated GDF15, are prevented by GFAP suppression in the rat model. These results give us a better understanding of mechanisms involved in central regulation of appetite and energy balance in AxD and potentially other neurodegenerative diseases with astrogliosis and appetite loss.

2. Materials and Methods

2.1. Rats

The Gfap ^+R237H rat model of AxD was generated as previously described [12]. For this study, animals were housed under microisolation with a 12‐h light cycle in an AAALAC accredited facility. Rodent chow (Teklad Irradiated Global 19% Protein Extruded Rodent Diet) and water were provided ad libitum. Animals were maintained as heterozygotes in a Sprague–Dawley genetic background (Charles River CD IGS rat). All animal studies were conducted in accordance with the United States Public Health Service's Policy on Humane Care and Use of Laboratory Animals and were approved by the Institutional Animal Care and Use Committees under the College of Letters and Sciences and Vice Chancellor Office for Research at the University of Wisconsin–Madison or the William S. Middleton Memorial Veterans Hospital. For all experiments, staff were kept blinded to genotype whenever possible. The numbers of animals assigned to each experiment were based on previous experience with the model.

2.2. Body Composition Analysis and Metabolic Phenotyping

For dual‐energy x‐ray absorptiometry (DEXA) scans to generate lean and fat map images, animals were either anesthetized with isoflurane (8 weeks of age) or euthanized by CO₂ asphyxiation (3 and 4 weeks of age) immediately before imaging with an UltraFocus system (Faxitron) within the UW‐Madison Carbone Cancer Center (CCC) Small Animal Imaging and Radiotherapy Facility.

A Body Composition Analyzer (EchoMRI‐700; Echo Medical System, Houston, Texas) was used to quantify lean and fat mass in unanesthetized rats, and the Comprehensive Laboratory Animal Monitoring System (Oxymax‐CLAMS‐HC, Columbus Instruments, Columbus, Ohio) was used to perform indirect calorimetry analysis within the UW‐Madison Comprehensive Diabetes Center Metabolic Phenotyping and Surgery Core (MPSC). Rats were transported to the MPSC immediately following weaning for body composition measures and 24‐h acclimation to the CLAMS chambers. Measures of energy expenditure, food intake, and activity were collected over the next 24 h from the beginning of the 12‐h light cycle. ANCOVA analysis for energy expenditure was provided by the NIDDK Mouse Metabolic Phenotyping Centers (MMPC, www.mmpc.org) using their Energy Expenditure Analysis page (http://www.mmpc.org/shared/regression.aspx).

2.3. Kaolin Consumption as a Measure of Pica

To analyze pica behavior, rats were housed individually in cages with raised wire floors and given measured amounts of their normal rodent chow (Teklad 2019) and kaolin diet (Research Diets K50001) with equal access to both from a wire hopper. Cage liners were placed under the wire floors to absorb urine and drinking water and keep chow crumbs dry for weight measures, in addition to food remaining in the hopper, to determine consumption over a given period. For initial experiments, kaolin and chow consumption were monitored for 7 days (P25–P32) and the percentage of kaolin consumed reported for the initial 72‐h period (P25–28). For measures at P56 and P112, kaolin and chow consumption were monitored for 72 h only. Animals were monitored daily for food and water levels.

2.4. Protein Quantification by ELISA

For rat tissue collections, animals were euthanized by asphyxiation with CO₂ and rapidly dissected to remove the brain. The brainstem was isolated on ice, frozen on dry ice, and stored at −80°C prior to further analysis.

GDF15 quantification was performed using Mouse/Rat (MGD150) or Human GDF‐15 (DGD150) Quantikine ELISA Kits (R&D Systems, Bio‐Techne) for CSF and plasma, and the Mouse GDF‐15 DuoSet ELISA (DY6385, R&D Systems, Bio‐Techne) was used to analyze rat brainstem lysates. CSF and plasma were analyzed according to the manufacturer's protocol. For tissue analysis, brainstem was homogenized using a Geno/Grinder bead mill (SPEX SamplePrep) in 100 mg/mL phosphate buffered saline (PBS) with 1% Triton‐X‐100, 1 mM PefaBloc SC (Sigma‐Aldrich), and Complete Protease Inhibitor Cocktail (Roche). Homogenates were frozen at −80°C before centrifugation at 10,000 g for 5 min at 4°C and collecting supernatants for protein quantification and analysis. Total protein was quantified for each sample using the bicinchoninic acid (BCA) assay (Pierce, Thermo Fisher Scientific) and diluted to ~2.5 mg/mL in PBS with 1% BSA for analysis. GDF15 values were normalized to the amount of protein used to assay individual tissue samples or volume for biofluids. GFAP protein was quantified by ELISA as previously described [17].

2.5. Histology and Image Analysis

For initial characterization of the R237H rat body phenotype, animals were euthanized with CO₂ at 8 weeks of age, and various tissues including gastrocnemius, heart, and interscapular brown fat were fixed in methacarn before embedding in paraffin for sectioning and staining with hematoxylin and eosin (H&E) by the UWCCC Experimental Animal Pathology Lab. For subsequent analysis of skeletal muscle myocytes, gastrocnemius was removed (including the soleus and plantaris) from R237H rats and wild type littermates (male and female, n = 5) at 8 weeks of age and fixed in 4% paraformaldehyde. A transverse section of the muscle was trimmed from the widest portion near the center before embedding in paraffin for H&E staining and image analysis. To compare myocyte size in wild type and R237H rats, brightfield images were acquired from H&E sections with a SPOT camera (Diagnostic Instruments) on a Nikon Microphot. Area measures were taken from myocyte cross sections from the lateral and medial gastrocnemius using Fiji (ImageJ). For randomized selection, a grid was placed on the image and cells at the cross lines were traced to determine area measures. A total of 20 cells (10 medial, 10 lateral) were measured, and values averaged for each animal.

2.6. Immunolabeling and Image Analysis

For GDF15/GFAP and GFRAL/cFos immunolabeling, animals were euthanized with CO₂ at 8 weeks of age, the brainstem isolated rapidly and fixed in 4% paraformaldehyde overnight, followed by cryoprotection in 10% and 30% sucrose. The caudal portion of the brainstem was sectioned at 40 μm with a sliding microtome and sections stored in cryoprotectant (0.1 M Na‐phosphate buffer pH 7.4 with 25% glycerol, 25% polyethylene glycol) at −20°C. For labeling, floating sections were washed in PBS, blocked and permeabilized in 5% normal donkey serum with 0.3% TX‐100, and incubated with primary antibodies diluted in PBS with 1% BSA for 72 h at 4°C. Sections were washed in PBS with 0.05% Triton‐X‐100 before adding diluted secondary antibodies and incubating for 24 h at 4°C. Finally, sections were incubated with 1 μg/mL DAPI for 10 min, washed, and mounted with ProLong Gold (Invitrogen, Thermo Fisher Scientific). Primary antibodies included sheep anti‐GDF15 (5 μg/mL, R&D Systems AF6385), sheep anti‐GFRAL (5 μg/mL, R&D Systems AF5728), rabbit anti‐cFos (1:1500, Cell Signaling 2250), rabbit anti‐S100β (1:500, Sigma HPAD15768), and mouse anti‐GFAP (1:1000, GA5, Sigma G6171). Secondaries included Alexafluor 488 conjugated donkey anti‐sheep IgG (1:500, Invitrogen A11015), Alexafluor 647 donkey anti‐mouse (1:500, Invitrogen A31571), and CF‐568 conjugated donkey anti‐mouse IgG (1:500, Sigma SAB4600075) and donkey anti‐rabbit IgG (1:500, Sigma SAB4600076). Images were acquired using a Nikon A1R‐HD or Leica Stellaris confocal microscope system within the Waisman Center Cellular and Molecular Neuroscience Core and are presented as maximum intensity projections (8 μm z‐stack with a 1 μm interval, 20× objective; 6.5 μm z‐stack with a 0.5 μm interval, 63× objective).

For quantification of GDF15 expressing cells, 11 μm z‐stacks were collected at a 0.69 μm interval (20× objective) to analyze the NTS in the region between the AP and central canal. For quantification of GFRAL neurons, 2 sets of 11 μm z‐stacks were collected at a 0.69 μm interval (20× objective) to image both the AP and adjacent NTS from the left and right sides of the midline. All images for specific comparisons were taken with equivalent laser power and intensity. ImageJ (Fiji) was used to identify and count labeled cells within each z‐stack per indicated region. To identify GDF15 positive cells, equivalent threshold settings were used across images to subtract background labeling.

For POMC and cFos immunolabeling, animals were anesthetized with isoflurane before transcardial perfusion with saline followed by 4% paraformaldehyde. Brains were removed, fixed overnight, cryoprotected, and sectioned. Sections from the middle of the median eminence were labeled as described above, with the exception that normal goat serum was used to block non‐specific binding. Primary antibodies used were rabbit anti‐POMC (1:500, Phoenix Pharmaceuticals H‐029‐30) and mouse anti‐cFos (1:1000, Abcam ab208942). Secondary antibodies used were Alexafluor‐568 conjugated goat anti‐rabbit IgG (1:500, Invitrogen A11036) and Alexafluor‐488 conjugated goat anti‐mouse IgG (1:500, Invitrogen A11029). Images were acquired as above and are presented as either single optical slices or maximum intensity projections (10.5 μm z‐stack with a 0.875 μm interval).

2.7. Quantitative PCR

Gastrocnemius was collected for molecular analysis from the same animals used to measure myocyte size. Approximately 50 mg of muscle was collected on ice and frozen at −80°C before further processing. A rotor‐stator homogenizer was used to extract RNA from tissue in 1 mL of Trizol following the manufacturer's protocol (Invitrogen, Thermo Fisher Scientific). Purified RNA was quantified, and 1 μg was used as a template for cDNA synthesis using Maxima H Minus Reverse Transcriptase (Thermo Fisher Scientific) in a 20 μL reaction including 50 ng random hexamers and 200 ng oligo(dT)₁₅ primers. Resulting cDNA was analyzed by quantitative PCR (320 nL per 20 μL reaction) with Fast Advanced Master Mix and the appropriate TaqMan probe (ABI, Thermo Fisher Scientific) for amplification and analysis on a ViiA 7 Real‐Time PCR system (ABI, Thermo Fisher Scientific). Probes used to assess muscle atrophy included Trim63 (Rn00590197_m1), Fbxo32 (Rn00591730_m1), and Mstn1 (Rn00569683_m1). Ppia (Rn00690933_m1) was used for normalization (ABI, Thermo Fisher Scientific).

2.8. Gfap‐ASO Treatment

Animals were treated with Gfap‐targeting antisense oligonucleotides (ASO) as previously described [12]. For these experiments, rats received intracerebroventricular (ICV) injections at P21, just after weaning. Animals within sex and genotype groups were alternately assigned to treatment groups based on the order of their identification number given between P5 and P7, prior to any discernible phenotypes. Treatment groups included 300 μg ASO and PBS‐vehicle as a control, which were delivered as 30 μL bolus ICV injections [12].

2.9. Human Subjects

CSF and plasma samples were acquired for our previous analysis of GFAP as a potential biomarker in AxD [18]. Informed consent was obtained for biomarker analysis in the original study according to protocols approved by the institutional review boards (IRB) involved [18]. Additional details for patient and control samples are given in Tables S1–S4.

3. Results

3.1. Cachexia‐Like Phenotype With Sarcopenia in R237H Rats

We have previously shown that R237H rats fail to thrive shortly after weaning, with pups becoming frail and gaunt during a period of normal maturational growth and development [12]. Dual‐energy x‐ray absorptiometry (DEXA) scans to assess fat and lean body maps further demonstrate the contrast between mutant and wild‐type animals as early as 4 weeks of age, and at 8 weeks show an emaciated profile for the R237H rat (Figure 1A). Quantification with EchoMRI shows significant decreases in both lean and fat mass at 3 weeks of age and dramatic differences at 8 weeks (Figure 1B). Histological analyses of skeletal and cardiac muscle show smaller myocytes in R237H rats, and although white fat is virtually absent, brown fat is apparent but devoid of lipid storage (Figure 1C). Measures of cross‐sectional area of myocytes from gastrocnemius further demonstrate the difference in cell size between genotypes (Figure 1D), and transcriptional markers of atrophy and wasting, including Trim63 (MURF1), Fbxo32 (Atrogin‐1), and Mstn (myostatin), are elevated in the R237H rat skeletal muscle (Figure 1E).

FIGURE 1.

Cachexia‐like phenotype with sarcopenia in R237H rats. (A) Fat and lean maps from DEXA scans at P21, P28, and P56 in male wild type (WT) and R237H littermates. Note wild type rat at P56 was larger than the field of view. (B) Quantification of body, lean, and fat mass by EchoMRI in male and female wild type and R237H rats at 3‐weeks (top panel) and 8‐weeks (bottom panel) of age (n = 10) Data were analyzed with a two‐way ANOVA and Šídák's multiple comparisons test. (C) Representative H&E stain of gastrocnemius, heart, and brown fat in wild type and R237H rats at 8 weeks of age. (D) Quantification of myocyte cross‐sectional area in gastrocnemius at 8 weeks of age. (E) Quantification of Trim63, Fbxo32, and Mstn by qPCR as markers of muscle atrophy and wasting. Values are normalized to the wild type average and represented as a fold increase. D and E include males and females, n = 5 per sex and genotype group, and a t‐test was used for comparisons. **p < 0.01, ***p < 0.001, ****p < 0.0001.

3.2. Reduced Energy Expenditure in R237H Rats

To identify potential mechanisms responsible for the body condition phenotype, we performed metabolic phenotyping using indirect calorimetry at 3 weeks, an age at which body weight differences are minimal. Food consumption was reduced in both male and female rats, although activity was only modestly reduced (Figure 2A,B). The respiratory exchange ratio was also decreased, suggesting a shift from carbohydrate to fat or protein catabolism, and energy expenditure was similarly reduced when normalized to body weight (Figure 2C,D). Given the potential effect of reduced body weight on energy expenditure, we analyzed energy expenditure over time without normalizing to body weight (Figure 2E) and performed an analysis of covariance (ANCOVA) with lean mass as a covariate (Figure 2F). Both analyses gave similar results, showing reduced energy expenditure in R237H rats of both sexes.

FIGURE 2.

Reduced energy expenditure in R237H rats at 3 weeks of age. (A–D) Comparisons of rodent chow consumption (A), ambulatory activity (B), respiratory exchange ratio (C), and energy expenditure normalized to body weight (D) in male and female rats at P22–23 over a 24‐h period including 12 h of light and 12 h of dark (n = 10). Data were analyzed with a 2‐way ANOVA and Šídák's multiple comparisons test, *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. (E, F) Energy expenditure per hour (E, p values represent a 2‐way repeated measures ANOVA) and analysis of covariance (ANCOVA) for energy expenditure with lean mass as a covariate (F, p values indicate differences between group EE values adjusted for lean mass). All analyses are for the same cohort of animals consisting of five different litters.

3.3. Pica and Anorexic Behavior Indicate Nausea at an Early Age

Given that energy expenditure is reduced in the R237H rat, we wanted to further investigate food consumption as a cause for the lack of weight gain. Differences in chow consumption were apparent at postnatal day (P) 23 in the indirect calorimetry experiments (Figure 2A), and daily measures starting at P25 show significant differences through P32 (Figure 3A). R237H rats fail to increase their daily consumption of chow during what should be a dramatic growth period and display little to no increase in body weight, with significant differences apparent by P28 (Figure 3B). Given the reduced appetite, lack of body fat, and sarcopenia—phenotypes often associated with cancer cachexia—we sought to investigate whether R237H rats exhibited signs of nausea. As a surrogate measure of nausea, since rodents do not vomit, we measured consumption of kaolin, an inert white clay used to assess pica behavior in response to nausea‐inducing reagents such as chemotherapy drugs [19]. At P28, R237H rats showed increased kaolin consumption, suggesting that nausea may be causing or contributing to their anorexic behavior (Figure 3C). Pica persisted at 8 weeks of age but was diminished at 16 weeks when R236H rats show weight gain and improved body condition.

FIGURE 3.

Pica and anorexic behavior indicate nausea at an early age. (A, B) Daily food consumption (A, grams of chow) and body weights (B) are indicated for male and female R237H and wild type (WT) littermates from P25 to P32. Total chow consumed between P25‐28 (72 h) is indicated as an average per 24 h in (A). Data were analyzed by a two‐way repeated measures ANOVA with Šídák's multiple comparisons test in A and B. (C) Pica behavior is shown as the percentage of kaolin consumed compared with the total amount of normal diet chow at P28 (n = 8 males, 8 females per genotype), P56 (n = 8 males, 9 females per genotype) and P112 (n = 6 males, 6 females per genotype). Data were analyzed with a two‐way ANOVA and Šídák's multiple comparisons test, ***p < 0.001, ****p < 0.0001.

3.4. GDF15 Expression in Brainstem and CSF and GFRAL‐Neuron Activation in the AP/NTS

Recent studies have identified GDF15 as an important contributor in anorexia and cachexia. The receptor for GDF15, the GFRAL/RET heterodimer, is expressed by cholecystokinin neurons located specifically in the AP and NTS of the brainstem. Reactive astrocytes express GDF15 [16], and transcript analysis demonstrates that it is elevated in rodent models of AxD [13, 14, 15]. To determine whether GDF15 is elevated in R237H rats and whether it could impact signaling in neurons expressing GFRAL, we analyzed GDF15 protein expression in the brainstem and found that it was increased by 3 weeks of age in R237H rats compared to wild‐type littermates and remained increased at 8 and 16 weeks of age (Figure 4A). The choroid plexus is also a significant source of GDF15 in cerebrospinal fluid (CSF) in response to stress [20], and the AP is located in the caudal floor of the fourth ventricle and in direct contact with CSF. Measures of GDF15 protein in CSF also showed significant increases at 3, 8, and 16 weeks of age in R237H rats compared to wild‐type (Figure 4B).

FIGURE 4.

GDF15 expression in R237H rat brainstem and CSF. (A‐B) Quantification of GDF15 protein in brainstem at P21 (n = 10 males, 10 females per genotype), P56 (n = 10 males, 10 females per genotype), and P112 (n = 8 males, 8 females per genotype) and CSF at P21 (n = 3 male and 3 female WT; n = 3 male and 6 female R237H), P56 (n = 10 male and 12 female WT; n = 10 male and 9 female R237H) and P112 (n = 8 male and 9 female WT; n = 8 male and 8 female R237H) by ELISA. Data were analyzed with a two‐way repeated measures ANOVA with Šídák's multiple comparisons test, ****p < 0.0001. (C, D) Immunolabeling GFAP and GDF15 in the AP/NTS (bregma −13.7 mm) in WT and R237H rats, with low (C) and high (D) magnification images for the two genotypes. Outlined regions of the NTS in (C) are shown in (D); the AP showed no GDF15 labeling. Scale bars in the first image of (C) and (D) apply to all images within each panel. (E, F) Maximum intensity projections of a GDF15 positive cells co‐labeled with GFAP and S100β, representing astrocytes expressing GDF15 (E), and quantification of GDF15 positive cells and those labeled with GFAP/S100β (F, females, n = 6). Data were analyzed with a t‐test, ***p < 0.001.

To determine whether GDF15 is expressed by reactive astrocytes in the rat model, we analyzed the caudal brainstem in the region of the AP (bregma −13.7 mm) by fluorescence immunolabeling for GDF15 and GFAP. Immunofluorescence analysis showed a patchy distribution of GDF15‐labeled astrocytes in the NTS (Figure 4C,D) and other regions of the brainstem in R237H rats with minimal labeling in the AP. Quantification of cells expressing GDF15, GFAP, or S100β (Figure 4E) confirmed that the majority of GDF15‐positive cells in the R237H rat were astrocytes (Figure 4F).

To determine whether neurons expressing the GFRAL receptor for GDF15 were activated, we analyzed the AP/NTS region for cFos expression. Immunolabeling showed high levels of GFRAL expression in R237H rats, with many neurons exhibiting nuclear cFos localization in both the AP and NTS (Figure 5A), and both the number of GFRAL neurons and the percentage of activated neurons were increased (Figure 5B,C). cFos labeling was also apparent outside of GFRAL‐expressing neurons, suggesting activation of other cell types in R237H rats, while wild‐type rats showed very little cFos expression in this region.

FIGURE 5.

GFRAL‐neuron activation in the AP/NTS of R237H rats. (A‐C) Immunolabeling GFRAL and cFos (A) in the AP/NTS (bregma −13.7 mm) of WT and R237H rats, and quantification of GFRAL expressing neurons and the number and percentage of cFos labeled GFRAL neurons in both the AP (B) and NTS (C) in R237H rats (n = 3 males and 3 females per genotype). Data were analyzed with a t‐test, **p < 0.01, ***p < 0.001.

To assess whether the hypothalamus may also contribute to appetite suppression, we analyzed cFos expression in POMC neurons in the arcuate nucleus, which regulate appetite by promoting satiety, and found no evidence of POMC neuron activation in R237H rats (Figure S1). Similar to the brainstem, activation of cells in other regions of the R237H rat hypothalamus was apparent, but not in POMC neurons.

3.5. GFAP Suppression Prevents GDF15 Elevation and Failure to Thrive Phenotypes

We have shown that GFAP suppression with antisense oligonucleotides (ASO) can both prevent and rescue astrocyte pathology and AxD phenotypes in the rat model [12]. To test whether GFAP suppression also reduces GDF15 expression, anorexia, and pica, we treated R237H rats and littermate controls with Gfap‐targeting ASO by ICV injection at P21. As in our previous studies, R237H rat body weights were normalized by GFAP suppression (Figure 6A,B; 8 weeks of age). We measured chow and kaolin consumption at 6 weeks of age, 3 weeks after treatment, and found ASO‐treated R237H rats consumed the same amount of chow and kaolin as their wild‐type littermates (Figure 6C,D). Quantification of GDF15 in brainstem and CSF showed the expected increase in R237H rats receiving vehicle and a reduction to wild‐type levels in those receiving Gfap‐ASO (Figure 6E). It was also notable that GDF15 measures in plasma showed no changes between genotype or treatment groups (Figure 6F).

FIGURE 6.

GFAP suppression prevents GDF15 elevation and failure to thrive phenotypes. (A) Body weights at 8 weeks of age in R237H rats after GFAP suppression by Gfap‐ASO injection at P21 compared to wild type (WT) littermates and rats treated with vehicle (PBS) as a control. (B) Quantification of GFAP by ELISA in brainstem from rats analyzed in A. (C) Rodent chow consumption and (D) pica behavior, as indicated by kaolin consumption, after treatment with ASO compared to vehicle. (E) GDF15 quantification in brainstem, CSF, and plasma in rats treated with ASO compared to vehicle. N = 4 for all groups (sex, genotype, treatment) with the exception of n = 5 male R237H rats treated with PBS. A two‐way ANOVA with uncorrected Fisher's LSD tests was used for comparisons between genotype and treatment groups, *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

3.6. GDF15 In Human Alexander Disease

Given the results in the rat model, we wanted to determine whether GDF15 expression might be relevant in human Alexander disease. CSF and plasma samples collected from patients with AxD and control subjects for our previous analysis of GFAP as a biomarker [18] were assayed in this study for GDF15 by ELISA (Tables S1–S4). GDF15 was significantly elevated in CSF from individuals with early (0–9 years of age) and juvenile (10–18 years) onset of AxD compared to the control group (0–19 years, Figure 7A). GDF15 levels in CSF from two adult onset cases were the lowest among the patients tested, although they were higher than all but one control CSF sample. In agreement with the rat model, plasma GDF15 levels were not increased in AxD patients, regardless of age of onset, and are shown collectively in Figure 7B. It should be noted that the age of onset is different from the age at the time of collection of CSF and plasma samples from patients with AxD (Tables S1 and S3). Linear regression analysis comparing GDF15 levels in patient CSF with age of onset or age at collection did not show a correlation with the limited sample size. Comparisons of CSF levels of GDF15 and GFAP from our previous study [18] demonstrate a correlation between the two measures (Figure 7C), suggesting GDF15 may also be useful as a potential biomarker.

FIGURE 7.

GDF15 in human Alexander disease. (A, B) Quantification of GDF15 in human CSF (A) and plasma (B) from patients with AxD versus control subjects. For CSF, AxD patient samples were classified as early, juvenile, or adult onset cases and an ANOVA with Kruskal‐Wallis test and Dunn's multiple comparisons was used for analysis (**p < 0.01, ***p < 0.001). An unpaired two‐tailed t‐test with Welch's correction was used to compare plasma samples. (C) Linear regression analysis comparing CSF levels of GDF15 and GFAP from our previous study [18]. p = 0.011 indicates the slope is significantly non‐zero.

4. Discussion

The rat model of AxD meets normal developmental milestones during the first postnatal weeks but fails to gain weight after weaning. Disease pathogenesis in rodent models of AxD coincides with the increase of GFAP expression during astrocyte maturation. In this early postnatal period, astrocytes respond to mutant GFAP by activating stress and innate immune response pathways, producing an array of cytokines and small chemokines, including growth and differentiation factor 15. GDF15 is known to regulate energy balance, and elevated levels are associated with mitochondrial disease, senescence, frailty, cancer cachexia, heart failure, and all‐cause mortality. Given the presence of both reactive and senescent astrocytes in the CNS [21] and the declining body condition and cachexia‐like phenotype of the R237H rat, we sought to determine whether GDF15 is elevated as part of the astrocyte stress response to mutant GFAP and potentially a driver of failure to thrive phenotypes in AxD.

In this report, we show that GDF15 is expressed in a subset of reactive astrocytes, and increased protein levels are found in brain and CSF. We also show that nausea and reduced caloric intake correlate with poor body condition and failure to thrive in the rat model. Furthermore, GDF15 is elevated in CSF from patients with the disease. In other disorders with marked GDF15 elevation, GDF15 is elevated systemically in tissues and blood. In contrast, in both the rat model of AxD and the human disease, we do not find elevation of plasma GDF15, suggesting that anorexia and FTT phenotypes could be centrally driven. Increased expression of markers of atrophy in skeletal muscle indicates wasting, and combined with the lack of white fat, suggests that animals are catabolizing protein. Although we cannot distinguish fat and protein metabolism, the reduction in RER also supports this conclusion. The degree to which starvation contributes to the motor phenotypes and even myelin deficits in the model is not clear but would clearly exacerbate underlying neurological dysfunction.

Neurons harboring the GDF15 receptor GFRAL/RET complex in the AP and NTS mediate appetite suppression, nausea, and vomiting in response to circulating hormones and chemical stimuli. Failure to thrive and frequent vomiting are both common features of AxD, especially with early or juvenile onset, and brainstem lesions in the AP have been linked to recurrent emesis in patients with the disease [22, 23, 24]. Here, we show cFos expression in GFRAL neurons in the AP/NTS of the R237H rat, suggesting activation by either paracrine signaling from local astrocytes or from GDF15 in CSF. Although tanycyte‐like cells expressing tight junctions have been reported to form a barrier between the AP and NTS [25, 26], other reports suggest macromolecules in CSF can penetrate beyond the AP potentially via the perivascular space [27, 28], and GFRAL neurons are activated in both regions, regardless of the source.

Astrocytes of the brainstem and hypothalamus play a critical role in central regulation of appetite and energy balance [29, 30, 31, 32], and R237H rats demonstrate widespread GFAP accumulation with astrogliosis throughout the CNS. Although the hypothalamus is a region of interest, we did not observe cFos activation in anorexigenic POMC neurons in the arcuate nucleus, and astrocyte reactivity in this region is generally associated with neuroinflammation in diabetes, increased appetite, obesity, and insulin resistance [33, 34, 35]. In this report, we focus on the brainstem based on the findings of elevated GDF15, pica behavior, and cFos activation in GFRAL neurons of the rat AP/NTS. GFRAL neurons in the AP project to the NTS, and also to the parabrachial nucleus (PBN) of the pons, where they target calcitonin gene‐related peptide (CGRP)‐expressing neurons [36, 37], known to relay alarm signals to the central amygdala [38] and mediate the aversive and anorexic effects of GDF15 [36, 39]. The severe appetite suppression observed in R237H rats is consistent with chronic pathophysiological GDF15 signaling and GFRAL neuron activation. GDF15 and GFRAL neutralizing therapies are under development, with clinical trials for cancer cachexia already completed [40], and may be beneficial in controlling nausea, intractable vomiting, and failure to thrive in AxD and other disorders with area postrema syndrome such as neuromyelitis optica [41].

GDF15 is also considered a marker of mitochondrial dysfunction and the integrated stress response (ISR) [42, 43, 44, 45], which could be manipulated for neurological benefits beyond appetite and failure to thrive, and ISR regulating compounds are currently in clinical trials for amyotrophic lateral sclerosis (ALS; ABBV‐CLS‐7262 in NCT04948645; IFB‐088 in NCT05508074). In addition, it is noteworthy that mutations in eIF2B cause vanishing white matter disease, a leukodystrophy where the ISR signature and GDF15 expression are predominantly found in astrocytes [45], and astrocyte dysfunction precedes disease onset and white matter loss [46]. A recent study also demonstrated elevation of GDF15 in brain and blood in the hSOD1^G93A mouse model of ALS, as well as in patients with the disease, further suggesting a connection with metabolic dysfunction and weight loss in ALS [47]. GDF15 was expressed by multiple CNS cell types including astrocytes and microglia, and silencing Gfral in hSOD1^G93A mice improved body condition and motor function and delayed paralysis and mortality. Gfral expression was elevated in the hSOD1^G93A model during early and symptomatic stages of disease progression, which could reflect an increase in GFRAL neuron number similar to our observations in the R237H rat.

Although progress has been made with GFAP antisense therapeutics for AxD [12], GFAP suppression alone may not completely reverse disease pathology or symptoms, and alternative approaches will complement ongoing efforts. Furthermore, current biomarker explorations in AxD are centered around GFAP, which is elevated in CSF and plasma of AxD subjects but is not a prognostic biomarker or a surrogate endpoint for disease outcomes. In addition to being a potential new target for therapy, GDF15 represents a new and alternative biomarker for monitoring disease status, progression, and response to therapeutics, all of which could complement efforts directed at GFAP suppression (clinical trial NCT04849741). A caveat of the current analysis is the limited number of patient samples analyzed and a lack of balance in age‐matched controls (Tables S1–S4). Future studies will include stratification by failure to thrive phenotypes, including weight loss, frequent vomiting, and lesion locations on MRI.

Failure to thrive in AxD is typically treated by nutritional supplementation or percutaneous endoscopic gastrostomy (PEG) feeding tube [6], but a better understanding of the underlying mechanisms and centrally driven appetite regulation may lead to improved therapeutic strategies for AxD and other neurodegenerative diseases with astrogliosis.

Author Contributions

T.L.H., A.M., D.A.H, and D.W.L. contributed to the conception and design of the study. T.L.H., M.M.S., C.L., and F.Z. contributed to the acquisition and analysis of data and preparing the figures. All authors contributed to drafting the text.

Conflicts of Interest

D.W.L. has received funding from, and is a scientific advisory board member of, Aeovian Pharmaceuticals, which seeks to develop novel, selective mTOR inhibitors for the treatment of various diseases.

Supporting information

Figure S1: Lack of POMC neuron activation in the arcuate nucleus. (A) Immunolabeling with POMC and cFos in the hypothalamus near the 3rd ventricle (3 V) of wild type (WT) and R237H rats. The square in the merged image is magnified in (B) and is representative of the arcuate nucleus. (C) POMC and cFos labeling in the dorsomedial nucleus demonstrates a positive cFos signal in a R237H rat. N = 3 female rats per genotype at 8 weeks of age.

Table S1: CSF samples from AxD patients.

Table S2: Control CSF samples.

Table S3: Plasma samples from AxD patients.

Table S4: Control plasma samples.

Funding Statement

This work was funded by National Institute of Diabetes and Digestive and Kidney Diseases grants DK076169, DK115255, and DK125859; Eunice Kennedy Shriver National Institute of Child Health and Human Development grants HD076892 and HD105353; Wisconsin Partnership Program at the UW School of Medicine and Public Health grant ID 6770‐2024; National Institute on Aging grants AG056771, AG062715, AG081482, AG084156, AG085898, AG088813, AG092586, and AG094153; National Institute of Neurological Disorders and Stroke grants NS110719 and NS136328; National Cancer Institute grant CA014520; U.S. Department of Veterans Affairs grants I01‐BX004031 and IS1‐BX005524; University of Wisconsin Carbone Cancer Center ; University of Wisconsin‐Madison ; Department of Surgery .

Data Availability Statement

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