CREB3 gain of function variants protect against ALS

Abstract

Amyotrophic lateral sclerosis (ALS) is a fatal and rapidly evolving neurodegenerative disease arising from the loss of glutamatergic corticospinal neurons (CSN) and cholinergic motoneurons (MN). Here, we performed comparative cross-species transcriptomics of CSN using published snRNA-seq data from the motor cortex of ALS and control postmortem tissues, and performed longitudinal RNA-seq on CSN purified from male Sod1^G86R mice. We report that CSN undergo ER stress and altered mRNA translation, and identify the transcription factor CREB3 and its regulatory network as a resilience marker of ALS, not only amongst vulnerable neuronal populations, but across all neuronal populations as well as other cell types. Using genetic and epidemiologic analyses we further identify the rare variant CREB3^R119G (rs11538707) as a positive disease modifier in ALS. Through gain of function, CREB3^R119G decreases the risk of developing ALS and the motor progression rate of ALS patients.

Fig. 1. Cross-species transcriptomic analysis prioritizes vulnerable cell-types in ALS.

a Experimental design displaying integration of single nuclei RNAseq from ALS patients and healthy controls, the human motor cortex and our database of mouse retrogradelly labelled CSN and control cells (mCSN and mCtl cells) from WT and Sod1^G86R male mice. b UMAP plot of the 116 clusters identified in the human M1 motor cortex (left) and expression of the mCSN and mCtl cells enriched genes in each cluster (right). c Experimental design of the cross-species DEG enrichment analysis wherein DEG identified in mCSN were intersected with each of the 30 DEG cell populations in human ALS, FTD and healthy controls. d Cluster dendrogram and heatmaps showing enrichment of mCSN and mCtl cells geneset in each of the 30 DE human cell populations (Cellular identity) as well as enrichment of mCSN DEG in all DE ALS or FTLD cell populations (Cellular vulnerability). We observed a significant enrichment of the mCSN DEG in human excitatory neurons (two-tailed hypergeometric test: Bonferroni’s adjusted *p < 4^e-04). Based on mCSN geneset enrichment and cross-species DE analysis, a cumulative score was calculated for each DE group and ranked to prioritize the most affected cell populations in ALS patients (Cellular rank). Created in BioRender. Dieterle, S. (2025) https://BioRender.com/d49e840.

Fig. 2. Cell-type specific consensus WGCNA identifies conserved regulatory networks in ALS.

a Experimental design of the WGCNA approach. b Histogram showing conserved signed correlation of WGCNA modules to ALS in prioritized cell types, corrected using the Benjamini-Hochberg method. Dashed lines highlight correlation FDR < 0.05. c UMAP gene network of the selected modules, each dot represents individual gene which size is proportional to its importance (signed kME). d Dot plot showing cross-species conservation of significant modules with strong (Zscore > 10) to moderate (5 < Zscore < 10) preservation. e GO analysis of the turquoise module with genes having signed kME (> 0.6) in the network (the red dashed line indicates an FDR < 0.05). f Violin plot showing turquoise eigengenes expression (two-way ANOVA: F_interaction group*cell = 4.88, **p < 0.01; Tukey-multiple comparisons: *p-adjusted = 0.037 and ***p-adjusted < 0.001). g Heatmap representing the turquoise module expression in healthy controls (HC) and ALS patients (one-tailed Wilcoxon rank sum test: *p-adjusted < 0.05 for significant cell-type). h GO analysis of the lightyellow module with genes having signed kME (> 0.6) in the network (the red dashed line indicates an FDR < 0.05). i Violin plot showing lightyellow eigengene expression (two-way ANOVA: F_interaction group*cell = 5.007, **p < 0.01; Tukey-multiple comparisons: ***p-adjusted < 0.001). j Violin plot showing turquoise eigengene expression in BA9 and cerebellum of ALS patients (two-tailed t-test: t = 6.74, ***p < 0.0001 and t = 1.99, p = 0.058, respectively) and in post-mortem human cortical neurons associated or not with TDP-43 loss-of-function LOF (two-tailed t-test: t = 3.20, p = 0.0079). k Turquoise module expression in Sod1^G86R mCSN (two-way ANOVA: F_group = 45.25, p < 0.001; Tukey post-hoc: WT vs Sod1^G86R at 30 d p = 0.070; 60 d p = 0.27; 90 d ***p < 0.001; 105 d p = 0.089). Boxplots show median and quartile distributions, upper and lower lines represent the 75th and 25th percentiles. l Representative images of puromycin incorporation in mCSN (Arrowheads point at L5 neurons with large CTIP2-positive nuclei) from 90d-old WT and Sod1^G86R animals. Scale bar = 20 µm. m Dot plot showing puromycin incorporation quantification in mCSN. Each dot represents one animal (n = 5 WT [163 neurons], 5 Sod1^G86R [199 neurons] at 30 d, and n = 8 WT [165 neurons], 8 Sod1^G86R [337 neurons] at 90 d). Mixed effect model followed by one-tailed Fisher’s LSD test; p = 0.9802 at 30 d and *p = 0.0271 at 90 d; mean +/− SEM). Created in BioRender. Dieterle, S. (2025) https://BioRender.com/d49e840.

Fig. 3. Integrative genetic analysis identifies CREB3 as protective factor in ALS.

a Experimental design to identify transcription factors (TFs) upstream of the WGCNA turquoise module which genes were intersected with TFs from ENCODE. Each TF was then investigated using an integrative approach combining WGCNA-based connectivity, expression profile of the TF-target genes and burden of rare variants. b Cluster dendrogram and heatmap showing TF-target gene expression in each of the 30 DE cell populations in ALS patients compared to controls. Row clustering of TFs based on their target gene expression identifies two clusters (blue and orange). The orange cluster (dashed line rectangle) was further prioritized based on a higher expression level of its target genes in L5-ET. The right-sided heatmap shows WGCNA-based TF connectivity in prioritized cell populations, association of missense rare variants in each TF with ALS risk, and target genes expression profile in mCSN at a presymptomatic (30–60 d) and symptomatic stages (90–105 d). The bottom heatmap shows the prioritized TF CREB3 expression profile in ALS and FTD patients compared to controls. c Quantile-quantile plot of the meta-analyzed gene burden of rare missense variants in a cohort of 1873 ALS patients and 3926 healthy controls showing CREB3 and ALS known genes. d Locus zoom plot showing the SNP (+/− 500KB) rs11538707 (R119G) association with ALS in the discovery cohort and the replication of 1873 ALS cases and 3926 healthy controls. Orange dashed lines shows the genome-wide significant SNP at a p-value < 5^e-08 and colored dots represent LD with the lead variant (red diamond). e Forest plot of gene burden association of CREB3 gene together with known ALS genes. Aggregation of rare missense variants in CREB3 confers a reduced risk of ALS (OR = 0.66 95%CI 0.51–0.87; Firth-logistic regression p = 2.9^e-03). Genome-wide RVBA association after Bonferroni correction 0.05/14245 = 3.51.e⁻⁰⁶. f Forest plot showing association of the missense variant rs11538707 (R119G) on CREB3 with ALS risk, compared to formerly identified associations (see Supplementary Data 25). Errors bars represent 95% confidence interval to the odds ratios. P-values were calculated using a linear-mixed model and represent uncorrected genome-wide association. Genome-wide significance is fixed at p < 5e⁻⁰⁸.

Fig. 4. CREB3-target genes expression as a resilience marker in mouse and human neurons.

a Violin plot showing increased CREB3 mRNA expression in pre- and symptomatic Sod1^G86R mCSN (edgeR-QL F-test, *FDR = 0.028) b Violin plot showing CREB3-target gene expression in pre- and symptomatic Sod1^G86R mCSN of (two-tailed permuted-p: ***p = 0.00019). The boxplots show median and quartile distributions, the upper and lower lines representing the 75th and 25th percentiles. c Heatmaps showing increased CREB3-target gene expression in the motor and frontal cortices of ALS and FTLD patients versus controls (*posthoc-Tukey significance in both tissues after Bonferroni correction 0.05/40 = *p < 0.001). d Representative images of Creb3 and Fezf2 mRNA expression in L5 of the motor cortex of 90d-old WT and Sod1^G86R mice (left, scale bar = 20 µm), and dot plot showing Creb3 probe integrated intensity quantification (right), each dot representing one neuron (98 neurons from n = 1 females and 4 WT males, and 112 neurons from n = 2 females and 3 Sod1^G86R males; data are presented as mean values +/− SEM; two-tailed nested i-test; *p = 0.0255). e Representative images of CREB3 and CTIP2 immunoreactivity in L5 of the motor cortex of end-stage Sod1^G86R mouse and WT littermate (left, scale bar = 20 µm), and dot plot showing CREB3 immunoreactivity quantification in CTIP2+ neurons (right), each dot representing one neuron (532 neurons from n = 2 females and 4 WT males, and 474 neurons from n = 2 females and 4 Sod1^G86R males; averaged age at perfusion = 102 d for WT, 103 d for Sod1^G86R; data are presented as means +/− SEM; two-tailed nested t-test; p = 0.4161). f Correlation between CREB3 protein levels averaged for each Sod1^G86R animal and their respective disease duration (n = 2 females and 4 Sod1^G86R males; one-tailed Pearson correlation, r² = 0.5753; *p = 0.0402). g Western blot revealing CREB3 protein in post-mortem motor cortex extracts from sporadic ALS patients and controls (left; Source data are provided as a Source Data file), and CREB3 protein level quantification (right; n = 3 controls and 3 ALS patients; data are presented as means +/− SEM; one-tailed t-test; *p = 0.0369). Created in BioRender. Dieterle, S. (2025) https://BioRender.com/d49e840.

Fig. 5. CREB3 gain of function is associated with a slower motor progression rate in ALS.

a Schematic of CREB3 regulon identification from RNAseq brain and blood encompassing CREB3-target genes identified through Chip-seq and co-regulated genes identified through weighted gene co-expression analysis. b Scatter plot showing positive correlation between CREB3 regulon activity in the motor cortex of ALS patients and survival (two-tailed Pearson R² = 0.28; **p = 0.0088). c Survival curves showing disease duration of ALS patients with high (~ 45 months) versus low (~ 30 months) brain CREB3 regulon activity (multivariate cox-ph survival: HR = 2.31 (95%CI:1.28–4.16); **p = 0.0053). d Scatter plot showing positive correlation between CREB3 regulon activity in the blood of ALS patients and survival (two-tailed Pearson R² = 0.337, t = 3.47; ***p = 0.00076). e Survival curves showing disease duration of ALS patients with high ( ~ 80 months) versus low ( ~ 58 months) blood CREB3 regulon activity (multivariate cox-ph survival: HR = 1.64 (95%CI:1.04–2.56); *p = 0.03). f ALS-FRS slope progression in rs11538707-A and rs11538707-G carriers. g Violin plot showing a slower motor progression rate in rs11538707-G carriers (mean ALS-FRS slope = −0.50 pts/months) compared to rs11538707-A carriers (mean ALS-FRS slope = −0.88 pts/months) (*two-tailed permuted-p = 0.033, Cohen’s d effect = 0.27). h Violin plots showing disease duration of rs11538707-G carriers compared to the rs11538707-A carriers (mean disease duration ~ 55.5 months versus ~ 43.9 month; *two-tailed permuted-p = 0.041, Cohen’s d effect = 0.36). Boxplots show median and quartile distributions, the upper and lower lines representing the 75th and 25th percentiles. i Genome browser Chip-seq track showing CREB3 binding on SEC24A promoter (light orange highlight). j Design of CREB3 and CREB3-R119G expression vectors and SEC24A-GLuc-ON reporter plasmid transfected in HEK293 cells (left), and violin plots (right) showing increased SEC24A promoter activity in CREB3-transfected cells compared to control empty vector (one-way Nested ANOVA: F_2–15 = 45, p < 0.0001; Bonferroni’s test: ***p < 0.0001), further enhanced by the R119G missense mutation (Nested ANOVA: F_2–15 = 45, p < 0.0001; Bonferroni’s test: **p = 0.0068). Created in BioRender. Dieterle, S. (2025) https://BioRender.com/d49e840.