Neurodegenerative diseases can have long preclinical phases and insidious progression patterns, but the mechanisms of disease progression are poorly understood. Because quantitative accounts of neuronal circuitry affected by disease have been lacking, it has remained unclear whether disease progression reflects processes of stochastic loss or temporally defined selective vulnerabilities of distinct synapses or axons. Here we derive a quantitative topographic map of muscle innervation in the hindlimb. We show that in two mouse models of motoneuron disease (G93A SOD1 and G85R SOD1), axons of fast-fatiguable motoneurons are affected synchronously, long before symptoms appear. Fast-fatigue-resistant motoneuron axons are affected at symptom-onset, whereas axons of slow motoneurons are resistant. Axonal vulnerability leads to synaptic vesicle stalling and accumulation of BC12a1-a, an anti-apoptotic protein. It is alleviated by ciliary neurotrophic factor and triggers proteasome-dependent pruning of peripheral axon branches. Thus, motoneuron disease involves predictable, selective vulnerability patterns by physiological subtypes of axons, episodes of abrupt pruning in the target region and compensation by resistant axons.