A Reproducible ALS-Enriched Astrocytic Cross-State Transcriptional Axis Across Two Independent Human Spinal-Cord snRNA-seq Cohorts A DPP10/NRXN3/CD44/DCLK1/GLIS3-centered coordinated state bridging previously separable astrocytic transcriptional configurations

Astrocytes contribute to amyotrophic lateral sclerosis (ALS) through non-cell-autonomous and inflammatory mechanisms, yet disease-associated astrocytic heterogeneity remains difficult to summarize with discrete cluster labels alone. Here, we reanalyzed two independent public human spinal-cord single-nucleus RNA-sequencing (snRNA-seq) cohorts and tested whether a coordinated five-gene astrocytic transcriptional state centered on DPP10, NRXN3, CD44, DCLK1, and GLIS3 reproducibly distinguishes ALS from control tissue. The first cohort, GSE287257, comprised postmortem cervical spinal cord from eight individuals with ALS and four non-neurological controls. The second, GSE330130, comprised lumbar spinal cord multiome-derived snRNA-seq from 11 ALS and nine control donors. Analyses were performed at donor level to avoid cellular pseudoreplication, with exact label-permutation testing, false-discovery-rate correction, and leave-one-donor-out sensitivity analyses.In GSE330130, 30,024 author-annotated astrocytes showed a higher five-gene joint score in ALS than controls (mean 2.832 versus 2.224; difference +0.608; exact P=0.0117; FDR q=0.0164). Detection of all five genes within the same astrocytic nucleus was also increased (57.2% versus 30.8%; P=0.00122; q=0.00427), and DCLK1-DPP10 co-detection was strongly enriched (79.4% versus 54.4%; P=1.85×10^-4; q=0.00129). A control-standardized five-dimensional state-space analysis showed higher ALS cross-state strength (+0.124 versus -0.302; P=0.00566; q=0.00707) and reduced distance to an AS4-like-AS5-like axis (1.637 versus 2.273; P=0.00236; q=0.00394). Reverse replication in 20,490 stringently gated astrocytes from GSE287257 reproduced the increased cross-state strength (+0.176 versus -0.205; P=0.00808; q=0.0202), increased all-five-gene detection (64.5% versus 38.3%; P=0.00808; q=0.0202), and reduced distance to the five-dimensional axis (1.772 versus 2.212; P=0.0182; q=0.0303). The exact fraction of cells occupying a predefined central bridge region did not replicate (P=0.887), indicating that the robust feature is a coordinated continuous program rather than expansion of one discrete intermediate cluster.These results support a reproducible ALS-associated astrocytic cross-state transcriptional axis spanning DCLK1- and DPP10-associated configurations. The five genes are not individually novel ALS or astrocyte markers; rather, the candidate novelty lies in their coordinated joint state, its low-dimensional geometry, and its reproducibility across independent cervical and lumbar human ALS spinal-cord cohorts. Because the study is retrospective and transcriptomic, the axis should not yet be interpreted as a new lineage, a temporal AS4-to-AS5 transition, or a causal disease mechanism. Spatial, longitudinal, and functional perturbation studies are required to determine whether this state marks adaptive remodeling, pathogenic astrocyte dysfunction, or a mixture of both.