Glucose hypometabolism prompts RAN translation and exacerbates C9orf72-related ALS/FTD phenotypes

Abstract

The most prevalent genetic cause of both amyotrophic lateral sclerosis and frontotemporal dementia is a (GGGGCC)_n nucleotide repeat expansion (NRE) occurring in the first intron of the C9orf72 gene (C9). Brain glucose hypometabolism is consistently observed in C9-NRE carriers, even at pre-symptomatic stages, but its role in disease pathogenesis is unknown. Here, we show alterations in glucose metabolic pathways and ATP levels in the brains of asymptomatic C9-BAC mice. We find that, through activation of the GCN2 kinase, glucose hypometabolism drives the production of dipeptide repeat proteins (DPRs), impairs the survival of C9 patient-derived neurons, and triggers motor dysfunction in C9-BAC mice. We also show that one of the arginine-rich DPRs (PR) could directly contribute to glucose metabolism and metabolic stress by inhibiting glucose uptake in neurons. Our findings provide a potential mechanistic link between energy imbalances and C9-ALS/FTD pathogenesis and suggest a feedforward loop model with potential opportunities for therapeutic intervention.

Keywords: ALS; C9orf72; FTD; Glucose Hypometabolism; RAN Translation.

Figure 1. The C9orf72-linked G₄C₂ repeat expansion disrupts brain energy balance.

Impact of C9orf72-linked G₄C₂ repeat expansion on brain energy balance. (A) Schematic depicting brain metabolite profiling experimental setup. (B) Principal component analysis (PCA) of the complete metabolomics panel. (C) Liquid chromatography-mass spectrometry (LC-MS) measurement of relative ATP concentrations, ATP:ADP ratios, and ATP:AMP ratios in C9-BAC animals versus WT animals. (D) Enrichment analysis of metabolite sets in C9-BAC animals versus WT animals. Enrichment ratio represents the number of metabolites within each metabolite set that are either increased (in blue) or decreased (in red) in the frontal cortex of C9-BAC versus WT animals. (E) Schematic of glycolysis pathway, with significantly altered glycolytic intermediates highlighted in blue text (decreased in C9-BAC animals) or red text (increased in C9-BAC animals). LC-MS measurement of relative concentrations of glucose-6-phosphate (G6P), glyceraldehyde-3-phosphate (GADP), and phosphoenolpyruvate (PEP) in C9-BAC versus WT animals. (F) LC-MS measurement of nicotinamide adenine dinucleotide (NAD⁺) and nicotinamide mononucleotide (NMN) concentrations in C9-BAC versus WT animals. All individual metabolite data are shown as median-normalized and log-transformed values (abbreviated as “Normalized conc.”). For all data, n = 7 C9orf72 BAC+ (C9-BAC) and 6 littermate control wild type (WT) animals. For all box and whisker plots, box edges denote upper and lower quartiles, horizontal lines within each box denote median values, whiskers denote maximum and minimum values, and shaded circles denote individual values for each animal. Student’s two-tailed t-test, *p < 0.05; **p < 0.01; ****p < 0.0001.

Figure 2. Glucose hypometabolism triggers accumulation of DPRs.

Cortical neurons transduced with (G₄C₂)₁₈₈ repeats are exposed to hypometabolic conditions. (A) Schematic of lentiviral RAN translation vector used for the experiments, containing a (G₄C₂)₁₈₈ repeat expansion with the 5’ flanking region of the C9orf72 gene (including exon 1a and intron 1) and a downstream GFP tag lacking an ATG start codon in frame with GA. The entire construct was driven by the human synapsin (hSyn) promoter. (B) Representative images depicting cellular localization patterns of RAN-translated GA-GFP aggregates. (C) Western blot analysis of GA-GFP levels in primary neurons using anti-GA antibody (RRID: AB_2728663) (NT = non-transduced; GFP = transduced with a GFP lentiviral vector (negative control); G₄C₂-GFP = transduced with RAN translation vector). (D) Schematic of experimental timeline from day in vitro (DIV) 0 to 10. (E) Fluorescent confocal imaging and quantification of DPR aggregate formation in primary neurons transduced with RAN translation vector, then incubated with either normo-glucose media (25 mM glucose + 0 mM 2DG) or media containing increasing concentrations of 2DG (n = 4). Mounted coverslips were imaged across >5 randomly selected fields of view/experiments. More than 30 cells were analyzed per condition. (F) qRT-PCR analysis of GFP mRNA levels in primary neurons transduced with either the RAN translation reporter vector, then incubated with either normo-glucose media or 10 mM 2DG-containing media (n = 4). (G) Schematic of lentiviral ATG translation vector used for the experiments, containing a 50 GA repeats encoded with alternative codons and a downstream GFP tag. (H) Fluorescent imaging and quantification of DPR aggregate formation in primary neurons transduced with ATG translation vector, then incubated with normo-glucose media or 10 mM 2DG-containing media (n = 3). (I) qRT-PCR analysis of GFP mRNA levels in primary neurons transduced with the ATG translation vector, then incubated with normo-glucose media or 10 mM 2DG-containing media (n = 3). For (E), one-way ANOVA with Dunnett’s test for multiple comparisons. For (F–I), student’s two-tailed t-test. All data are presented as mean ± standard error of the mean (SEM; n = 3–4 biological replicate). **p < 0.01, ****p < 0.0001, n.s. p > 0. Source data are available online for this figure.

Figure 3. Glucose hypometabolism activates the ISR in cultured neurons.

ISR activation is measured in cortical neurons exposed to 2DG. (A, B) RNA sequencing assessment of select transcriptomic changes in C9orf72 patient-derived i³Neurons (n = 2 individual i³Neuron lines with 2 separate differentiations per line) incubated in media containing 10 mM 2DG for 48 h versus those maintained in normal media, including significantly upregulated gene ontology (GO) pathways (A) and significantly upregulated individual integrated stress response (ISR) target transcripts (B; all p_adj < 0.05). (C) Immunofluorescence-based measurement and quantification of nuclear ATF4 expression level in MAP2-positive primary neurons incubated in media containing 0, 2.5, 5.0, or 10 mM 2DG for 48 h (n = 4). (D) Correlation between the number of GA-GFP aggregates per field of view (from Fig. 2E) and nuclear ATF4 expression level (from Fig. 3C) in primary neurons incubated with either normo-glucose or various concentrations of 2DG. Best-fit line and R² value represent linear regression analysis. (E) Western Blot analysis for phospho-eIF2α and eIF2α of cortical neurons treated with glucose or 2DG and relative quantification of the ratio between phospho-eIF2α and eIF2α protein expression (n = 3; m = 3 technical replicates). For (A), the values adjacent to each bar represent the number of altered genes in each GO pathway. The statistical test applied is the Fischer’s exact test with a Benjamini–Hochberg False Discovery Rate (FDR). For (B), bars represent median values, and individual dots represent individual replicates. For (C), one-way ANOVA with Dunnett’s test for multiple comparisons. Data are presented as mean ± SEM of biological replicates (n = 3). At least 30 neurons were analyzed/condition. *p < 0.05, **p < 0.01, ****p < 0.0001. For box and whisker plots, box edges denote upper and lower quartiles, horizontal lines within each box denote median values, whiskers denote maximum and minimum values, and shaded circles denote individual values for each replicate. Source data are available online for this figure.

Figure 4. DPR accumulation caused by glucose hypometabolism correlates with ISR activation and is blocked by inhibition of the GCN₂ kinase.

DPRs aggregation in cortical neurons is measured in hypometabolic conditions. (A) Schematic depicting the proposed mechanism through which nutrient deprivation increases RAN translation. A92 was used as a pharmacological inhibitor of GCN2 kinase activity. (B, C) Fluorescent confocal imaging and quantification of DPR formation in primary neurons transduced with RAN translation vector, then incubated with either normo-glucose media, 10 mM 2DG-containing media, or 10 mM 2DG-containing media with 5.0 μM A92, all in the presence of 0.1% DMSO (n = 4). (D) Quantification of DPR aggregates formation in primary neurons transduced with RAN translation vector, then treated with A92 concentrations ranging from 0.1 to 7.5 μM, all in the presence of 10 mM 2DG and 0.1% DMSO (n = 4). Best-fit line and IC₅₀ value were derived from non-linear regression analysis. For (C), one-way ANOVA with Dunnett’s test for multiple comparisons. Images were taken across 5 randomly selected fields. All data are presented as mean ± SEM (n = 4 biological replicates). ****p < 0.0001. Source data are available online for this figure.

Figure 5. Glucose deprivation is selectively toxic to C9orf72 patient-derived i³Neuron.

i³Neurons from patient and control cases are exposed to glucose deprivation and neuronal viability over time is recorded. (A) Schematic of the doxycycline-inducible neuronal differentiation cassette used to drive rapid differentiation of human induced pluripotent stem cells (hiPSCs) into i³Neurons from days post-induction (DPI) 0 to 27. (B) Timeline of differentiation into i³Neurons and subsequent glucose deprivation. (C) Live-cell longitudinal imaging of healthy control- or C9orf72 patient-derived i³Neurons cultured with glucose-deprived media over 6 days using DRAQ7 as a dead cell indicator. (D) Kaplan–Meier survival analysis of i³Neurons derived from either C9orf72 patients or healthy controls and maintained in glucose-deprived media (n = 3 i³Neuron lines per genotype with 3 independent differentiations per line). (E) Quantification of dot blot for GP performed on i³Neurons derived from either C9orf72 patients or healthy controls and maintained in glucose-deprived media. (F) Kaplan–Meier survival analysis of C9orf72 patient-derived i³Neurons maintained in either normo-glucose or glucose-deprived media and treated with either 2.5 μM A92 with 0.1% DMSO or 0.1% DMSO only as vehicle control (n = 3 i³Neuron lines per treatment with 3 independent differentiations per line). (G) Quantification of dot blot for polyGP performed on C9orf72 patient-derived i³Neurons maintained in either normo-glucose or glucose-deprived media and treated with either 2.5 μM A92 with 0.1% DMSO or 0.1% DMSO only as vehicle control (n = 3 i³Neuron lines per treatment with 3 independent differentiations per line). Kaplan–Meier log-rank survival test. *p < 0.05, **p < 0.01, ****p < 0.0001. Neurons were imaged live across >5 randomly selected fields of view/conditions. The same field of view was imaged daily. More than 100 neurons were tracked/condition/experiment. All data are presented as mean ± 95% CI (n = 3 biological replicates). For box and whisker plots, box edges denote upper and lower quartiles, horizontal lines within each box denote median values, whiskers denote maximum and minimum values, and shaded circles denote individual values for each replicate. Source data are available online for this figure.

Figure 6. 2DG treatment exacerbates metabolic stress and drives disease-related phenotypes in C9orf72 BAC transgenic mice.

Effects of 2DG treatment on metabolic stress and disease phenotypes in C9orf72 transgenic mice. (A) Schematic of experimental setup: C9-BAC animals were treated with 2-deoxyglucose (2DG) or saline (vehicle control) by i.p. injection. (B) RT-qPCR measurement of spinal cord mRNA levels of two ISR transcriptional targets (CHOP and GADD34) in C9-BAC animals acutely treated with various doses of 2DG or saline (n = 3–4 animals per condition). (C, D) LC-MS measurement of 2DG-6-P levels (C) and significantly altered metabolites (D) in the frontal cortex of C9-BAC mice following chronic weekly exposure to 4 g/kg 2DG versus saline (n = 7 animals per condition). Dark shaded lines indicate significance cut-offs (−log₁₀(p) > 1.3 and log₂(FC) > |1|). (E) Longitudinal assessment of inverted wire hang performance of C9-BAC animals chronically exposed to 4 g/kg 2DG or saline (n ≥ 5 animals/condition). The inset figure displays a longitudinal assessment of the 2DG treatment effects on the C9-BAC mice. Baseline measurements represent the initial state, followed by longitudinal measurements at 6 weeks post-treatment, illustrating the impact of 2DG intervention on C9-BAC mice over time. Y-axis in Log 2 scale. (F). Single molecule array (Simoa) measurement of 8 M urea-soluble GP levels relative to total protein levels in the spinal cord of saline-treated wild type animals (n = 3), saline-treated C9-BAC animals (n = 4), or 2DG-treated C9-BAC animals (n = 6). For (B), one-way ANOVA with Dunnett’s test for multiple comparisons. For (C, D), student’s two-tailed t-test. For (E), two-tailed t-test with Welch’s correction. For (F), one-way ANOVA. All data are presented as mean ± SEM (n > 3 biological replicates). *p < 0.05, ***p < 0.001, ****p < 0.0001. For box and whisker plots, box edges denote upper and lower quartiles, horizontal lines within each box denote median values, whiskers denote maximum and minimum values, and shaded circles denote individual values for each animal. Source data are available online for this figure.

Figure 7. Arginine-rich DPRs contribute to glucose hypometabolic stress.

Glucose uptake is measured in cortical neurons expressing DPRs. (A) Schematic of lentiviral DPR vectors and GFP-only (control) vector. (B) Representative images depicting cellular localization patterns of GFP-tagged DPRs in rodent primary neurons transduced with lentiviral vectors. (C) Luminescence-based measurement of glucose uptake (normalized to total protein and expressed in fold change) in primary neurons transduced with each of the DPR vectors or the GFP-only control vector (n = 4 biological replicates). (D) Seahorse extracellular flux assay measurement of extracellular acidification rate (ECAR; normalized to total protein) of primary neurons transduced with either the PR vector or GFP-only control vector (n = 3 biological replicates). (E) Immunofluorescence-based measurement and quantification of nuclear ATF4 expression level in MAP2-positive primary neurons transduced with either the PR vector or GFP-only control vector (n = 4 biological replicates). At least 30 neurons were analyzed/condition. For (C), one-way ANOVA with Dunnett’s test for multiple comparisons. For (D), multiple student’s two-tailed t-tests. For (E), student’s two-tailed t-test. All data are presented as mean ± SEM. *p < 0.05, **p < 0.01. Source data are available online for this figure.

Figure EV1. Analysis of brain metabolites of C9-BAC vs. WT mice.

(A) Schematic depicting the bacterial artificial chromosome (BAC) transgene used to drive expression of the entire human C9orf72 gene with a (GGGGCC)_100-1,000 repeat expansion in intron 1 (C9-BAC). Adapted from O’Rourke et al (2015). (B) LC-MS measurement of relative glucose concentrations in the frontal cortex of C9-BAC versus WT animals. (C) Schematic depicting the tricarboxylic acid (TCA) cycle pathway with all metabolic intermediates. LC-MS measurement of relative α-ketoglutarate concentrations in the frontal cortex of C9-BAC animals versus WT animals. (D) LC-MS measurement of the relative concentrations of four amino acids (isoleucine, cysteine, lysine, and ornithine) in the frontal cortex of C9-BAC vs. WT animals. (E) LC-MS measurement of relative NADP +, NADH, and NADPH concentrations in the frontal cortex of C9-BAC vs WT. All individual metabolite data are shown as median-normalized and log-transformed values (abbreviated as “Normalized conc.”). For box and whisker plots, box edges denote upper and lower quartiles, horizontal lines within each box denote median values, whiskers denote maximum and minimum values, and shaded circles denote individual values for each animal. Student’s two-tailed t-test, *p < 0.05, **p < 0.01. Source data are available online for this figure.

Figure EV2. Effect of 2DG on cellular respiration.

(A) RNA sequencing measurement of select downregulated gene ontology (GO) pathways in C9orf72 patient-derived i³Neurons incubated with 10 mM 2DG-containing media versus normo-glucose media for 48 h (n = 2 individual i³Neuron lines with 2 separate differentiations per line). (B) Seahorse extracellular flux assay-based measurement of extracellular acidification rate (ECAR) normalized to total protein and oxygen consumption rate (OCR) normalized to total protein in primary neurons incubated with either normo-glucose media or 10 mM 2DG-containing media for 48 h. Cells were sequentially treated with 1.5 μg/ml oligomycin (o.), 3 μM FCCP (f.), and 1 μM antimycin (a.) during the assay (n = 1 with 3 technical replicates). For (A), the values adjacent to each bar represent the number of altered genes in each GO pathway. The statistical test applied is the Fischer’s exact test with a Benjamini–Hochberg False Discovery Rate (FDR). For (B) data are presented as mean ± SEM of assay technical replicates. Source data are available online for this figure.

Figure EV3. Inhibition of ISR rescues DPR aggregation.

(A) Western blot analysis and quantification of ATF4 expression level relative to histone H3 expression level (as a loading control) in lumbar spinal cord tissue homogenates from either healthy control subjects or C9orf72-ALS patients (n = 3 subjects per genotype). (B) Live-cell imaging of human i³Neurons treated with either 0.1% DMSO or 5.0 μM A92 and 0.1% DMSO for 48 h, with corresponding quantification of survival proportions. DRAQ7 was used as a fluorescent dead cell indicator (n = 2 i³Neuron lines with 2 independent differentiations per line). Neurons were imaged live across 5 randomly selected fields of view/conditions. The same field of view was imaged daily. More than 100 neurons were tracked/condition/experiment. (C) Fluorescent confocal imaging and quantification of DPR aggregate formation in primary neurons transduced with RAN translation vector, then incubated with either normo-glucose media, 10 mM 2DG-containing media, or 10 mM 2DG-containing media with 0.5 μM or 5 μM ISRIB (all in the presence of 0.1% DMSO; n = 4). (D) Fluorescent confocal imaging and quantification of DPR formation in human i³Neurons transduced with RAN translation vector, then incubated with either normo-glucose media, 10 mM 2DG-containing media, or 10 mM 2DG-containing media with 2.5 μM A92 (all in the presence of 0.1% DMSO; n = 1). At least 30 neurons were analyzed/condition. For (A, B), student’s two-tailed t-test. For (C), one-way ANOVA with Dunnett’s test for multiple comparisons. All data are presented as mean ± SEM. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. Source data are available online for this figure.

Figure EV4. Effect of glucose deprivation in I³neurons.

(A) Longitudinal microelectrode array (MEA)-based measurement of spontaneous neuronal activity of i³Neurons (healthy control line) incubated in either normo-glucose or glucose-deprived media over a 48-h period (n = 3 biological replicates). Glucose deprivation was initiated at time = 0 h. (B) Seahorse extracellular flux assay-based measurement of extracellular acidification rate (ECAR) normalized to total protein and oxygen consumption rate (OCR) normalized to total protein of primary neurons immediately following incubation with either normo-glucose media or glucose-deprived media for 48 h (n = 1 with 3 technical replicates). Cells were sequentially treated 1.5 μg/ml oligomycin (o.) and 3 μM FCCP (f.) during the assay. (C) Immunofluorescence-based measurement and quantification of nuclear ATF4 expression level in MAP2-positive primary neurons immediately following incubation with either normo-glucose media or glucose-deprived media for 48 h (n = 4 biological replicates). Neurons were imaged live across >5 randomly selected fields of view/conditions. The same field of view was imaged daily. More than 100 neurons were tracked/condition/experiment. (D) RNA sequencing analysis of select individual ISR target transcripts (all p_adj < 0.05) in i³Neurons immediately following exposure to either normo-glucose or glucose-deprived media for 48 h (n = 2 i³Neuron lines with 2 differentiations per line). (E) Representative images of i³Neurons transduced with RAN translation vector and then cultured in either normo-glucose or glucose-deprived media for 48 h, with corresponding quantification of GA-GFP fluorescence intensity (n = 3 biological replicates). Neurons were imaged live across 5 randomly selected fields of view/conditions. The same field of view was imaged daily. More than 100 neurons were tracked/condition/experiment. (F) qRT-PCR measurement of mRNA levels of two ISR transcriptional targets (CHOP and GADD34) in i³Neurons immediately following incubation with either normo-glucose or glucose-deprived media for either 24, 48, or 72 h (n = 1). For (A), multiple student’s two-tailed t-tests. For (C–E), student’s two-tailed t-test. All data (except D and F) are presented as mean ± SEM. *p < 0.05, **p < 0.01, ***p < 0.001. Source data are available online for this figure.

Figure EV5. Glucose deprivation in control and C9orf72 derived I³neurons.

(A) Kaplan–Meier survival analysis of i³Neurons derived from either C9orf72 patients or healthy controls and maintained in glucose-deprived media (n = 3 i³Neuron lines per genotype with 3 independent differentiations per line). Line A, B, C. ***p<. (B) Kaplan–Meier survival analysis of C9orf72 patient-derived i³Neurons maintained in either normo-glucose or glucose-deprived media and treated with either 2.5 μM A92 with 0.1% DMSO or 0.1% DMSO only as vehicle control (3 independent differentiations per line). Line A. (C) Kaplan–Meier survival analysis of C9orf72 patient-derived i³Neurons maintained in either normo-glucose or glucose-deprived media and treated with either 2.5 μM A92 with 0.1% DMSO or 0.1% DMSO only as vehicle control (3 independent differentiations per line). Line B. (D) Kaplan–Meier survival analysis of C9orf72 patient-derived i³Neurons maintained in either normo-glucose or glucose-deprived media and treated with either 2.5 μM A92 with 0.1% DMSO or 0.1% DMSO only as vehicle control (3 independent differentiations per line). Line C. (E, F). Dot blot assessment of 8 M urea-soluble GP levels and relative total protein levels in i³Neurons derived from either C9orf72 patients or healthy controls maintained in either normo-glucose or glucose-deprived media and treated with either 2.5 μM A92 with 0.1% DMSO or 0.1% DMSO only as vehicle control. (n = two independent lines per genotype with 2 or 3 technical replicates per treatment) Kaplan–Meier log-rank survival test. *p < 0.05, **p < 0.01, ****p < 0.0001. Neurons were imaged live across >5 randomly selected fields of view/conditions. The same field of view was imaged daily. More than 100 neurons were tracked/condition/experiment. All data are presented as mean ± SEM (n = 3 biological replicates). *p < 0.05, **p < 0.01. Source data are available online for this figure.

Figure EV6. Effect of 2DG treatment in rodents.

(A) Enrichment analysis of significantly altered metabolite sets in the frontal cortex of C9-BAC animals immediately following chronic exposure to either 4 g/kg/week 2DG or saline (n = 7 animals per condition). Enrichment ratio represents the number of metabolites within each metabolite set that are either increased (on the left) or decreased (on the right). (B) Optical coherence tomography (OCT) retinal imaging of C9-BAC animals immediately following chronic exposure to either 4 g/kg/week 2DG or saline. Yellow arrows indicate hyperreflective foci (n = 2–3 animals per condition). (C) Longitudinal measurement of body weight during chronic exposure of C9-BAC animals to 4 g/kg/week 2DG or saline (n = 7 animals per condition). (D). Longitudinal assessment of inverted wire hang performance of wild-type littermate control animals chronically exposed to 4 g/kg/week 2DG (n = 2 animals) or saline (n = 3 animals). (E) Dot blot assessment and corresponding quantification of 8 M urea-soluble GP levels relative to total protein levels in the spinal cord C9-BAC animals treated with either saline (n = 3 animals) or 2DG (n = 6 animals). Bovine serum albumin (BSA) and spinal cord lysate from a wild-type animal were used as controls. (F) Western Blot analysis for p-eIF2α and eIF2α of mouse brain cortices treated with saline or 2DG. (G) Quantification of the ratio between pEif2α and Eif2α protein of the WB in Fig EV6f (n = 3 biological replicates). For (A), the statistical test applied is the Fischer’s exact test with a Benjamini–Hochberg False Discovery Rate (FDR). For (C, D), two-way ANOVA. For (E–G), two-tailed t-test with Welch’s correction. All data (except A) are presented as mean ± SEM. *p < 0.05, n.s. p > 0.05. For box and whisker plots, box edges denote upper and lower quartiles, horizontal lines within each box denote median values, whiskers denote maximum and minimum values, and shaded circles denote individual values for each replicate. Source data are available online for this figure.

Figure EV7. DPRs transduction in rat cortical neurons.

(A) qRT-PCR measurement of DPR and GFP-only vector lentiviral titers (n = 1 technical replicate). (B) Representative images of primary neurons transduced with DPR or GFP-only lentiviral vectors, then stained for MAP2. (C) Quantification of the percentage of MAP2-positive cells also positive for GFP from (B) (n = 1 with 3 technical replicates). (D) Validation of luminescent assay (time- and temperature dependence) for measurement of glucose uptake in primary neurons (n = 1 with 3 technical replicates). (E) Seahorse extracellular flux assay measurement of oxygen consumption rate (OCR normalized to total protein) of primary neurons transduced with either the PR or GFP-only vector (n = 3 biological replicates). Cells were sequentially treated with 1.5 μg/ml oligomycin (o.), 3 μM FCCP (f.), and 1 μM antimycin (a.) during the assay. All data (except A) are presented as mean ± SEM. For (C), one-way ANOVA with Dunnett’s multiple comparisons tests. For (E), multiple student’s two-tailed t-test. n.s. p > 0.05. Source data are available online for this figure.