Hereditary spastic paraplegia refers to rare genetic neurodevelopmental and/or neurodegenerative disorders in which spasticity due to length-dependent damage to the upper motor neuron is a core sign. Their high clinical and genetic heterogeneity makes their diagnosis challenging. Multigene panels allow a high-throughput targeted analysis of the increasing number of genes involved using next-generation sequencing. We report here the clinical and genetic results of 1550 index cases tested for variants in a panel of hereditary spastic paraplegia related genes analyzed in routine diagnosis. A causative variant was found in 475 patients (30.7%) in 35/65 screened genes. SPAST and SPG7 were the most frequently mutated genes, representing 142 (9.2%) and 75 (4.8%) index cases of the whole series, respectively. KIF1A, ATL1, SPG11, KIF5A and REEP1 represented more than 1% (> 17 cases) each. There were 661 causative variants (382 different ones) and 30 of them were structural variants. This large cohort allowed us obtaining an overview of the clinical and genetic spectrum of hereditary spastic paraplegia in clinical practice. Because of the wide phenotypic variability, there was no very specific sign that could predict the causative gene but there were some constellations of symptoms that were found often related to specific subtypes. Finally, we confirmed the diagnostic effectiveness of a targeted sequencing panel as a first-line genetic test in hereditary spastic paraplegia. This is a pertinent strategy because of the relative frequency of several known genes (i.e.: SPAST, KIF1A) and it allows identifying variants in the rarest involved genes and to detect structural rearrangements via coverage analysis, which is less efficient in exome data sets. It is crucial because these structural variants represent a significant proportion of the pathogenic hereditary spastic paraplegia variants (∼6% of patients), notably for SPAST and REEP1. In a subset of 42 index cases negative for the targeted multigene panel, subsequent whole exome sequencing allowed to reach a theoretical diagnosis yield of ∼50%. We then propose a two-step strategy combining the use of a panel of genes followed by whole exome sequencing in negative cases.