The identification of brain markers of suicidal risk is highly expected. However, neuroimaging studies have yielded mixed results, possibly due to phenotypic heterogeneity. In the present study, we addressed this issue using structural brain imaging. First, two independent samples of suicide attempters (n = 17 in Montreal, 32 in Jena), patient controls (n = 26/34), and healthy controls (n = 66/34) were scanned with magnetic resonance imaging. Groups were compared with FSL. We then reviewed the literature and run a GingerALE meta-analysis of 12 structural imaging studies comparing suicide attempters and patient controls with whole-brain analyses (n = 693). Finally, we explored the potential contribution of two variables previously associated with biological/cognitive deficits: a family history of suicide (FHoS), and the use of a violent suicidal means (VSM). Here, we added two groups of healthy first-degree biological relatives of suicide victims and depressed patients (n = 32). When comparing all suicide attempters and controls, very limited between-group differences were found in the two samples, and none in the meta-analysis. In contrast, a FHoS was associated with reduced volumes in bilateral temporal regions, right dorsolateral prefrontal cortex, and left putamen, several of these differences being observed across groups. VSM was associated with increased bilateral caudate (and left putamen) volumes. Some morphometric variations in cortico-subcortical networks may therefore be endophenotypes increasing the suicidal vulnerability, while others (notably in striatum) may modulate action selection. These results therefore confirm at the neural level two phenotypes at high lethal risk with a strong biological background, and uncover motives of heterogeneous findings in neuroimaging studies of suicidal behavior.