BackgroundBone marrow metastasis (BMM) is underestimated in gastric cancer (GC). GC with BMM frequently complicate critical hematological abnormalities like diffused intravascular coagulation and microangiopathic hemolytic anemia, which constitute a highly aggressive GC (HAGC) subtype. HAGC present a very poor prognosis with peculiar clinical and pathological features when compared with not otherwise specified advanced GC (NAGC). But the molecular mechanisms underlying BMM from GC remain rudimentary.MethodsThe transcriptomic difference between HAGC and NAGC were analyzed. Genes that were specifically upregulated in HAGC were identified, and their effect on cell migration and invasion was studied. The function of ACTN2 gene were confirmed by GC cell lines, bone-metastatic animal model and patients’ tissues. Furthermore, the molecular mechanism of ACTN2 derived-BMM was explored by multiple immunofluorescence staining, western blot, chromatin immunoprecipitation, and luciferase reporter assays.ResultsWe elucidated the key mechanisms of BMM depending on the transcriptomic difference between HAGC and NAGC. Five genes specifically upregulated in HAGC were assessed their effect on cell migration and invasion. The ACTN2 gene encoding protein α-Actinin-2 was detected enhanced the metastatic capability and induced BMM of GC cells in mouse models. Mechanically, α-Actinin-2 was involved in filopodia formation where it promoted the Actin filament cross-linking by replacing α-Actinin-1 to form α-Actinin-2:α-Actinin-4 complexes in GC cells. Moreover, NF-κB subunit RelA and α-Actinin-2 formed heterotrimers in the nuclei of GC cells. As a direct target of RelA:α-Actinin-2 heterotrimers, the ACTN2 gene was a positive auto-regulatory loop for α-Actinin-2 expression.ConclusionsWe demonstrated a link between filopodia, BMM and ACTN2 activation, where a feedforward activation loop between ACTN2 and RelA is established via actin in response to distant metastasis. Given the novel filopodia formation function and the new mechanism of BMM in GC, we propose ACTN2 as a druggable molecular vulnerability that may provide potential therapeutic benefit against BMM of GC.