Despite their similar genetic makeup, honeybee (A. mellifera) queens and workers show alternative morphologies driven by nutritional difference during the larval stage. Although much research have been done to investigate the causes of honeybee caste polymorphism, information at subcellular protein levels is limited. We analyzed queen- and worker-destined larvae mitochondrial proteome at three early developmental stages using combinations of differential centrifugation, two-dimensional electrophoresis, mass spectrometry, bioinformatics, and quantitative real time PCR. In total, 67, 69, and 97 protein spots were reproducibly identified as mitochondrial proteins at 72, 96, and 120 h, respectively. There were significant qualitative and quantitative protein expression differences between the two castes at three developmental stages. In general, the queen-destined larvae up-regulated large proportions of proteins at all of the developmental stages and, in particular, 95% at 72 h. An overwhelming majority of the queen larvae up-regulated proteins were physiometabolic-enriched proteins (metabolism of carbohydrate and energy, amino acid, and fatty acid) and involved in protein folding, and this was further verified by functional enrichment and biological interaction network analyses as a direct link with metabolic rates and cellular responses to hormones. Although wide-ranging mitochondrial proteomes participate to shape the metabolic, physiologic, and anatomic differences between the two castes at 72 h, physiometabolic-enriched proteins were found as the major modulators of the profound marking of this caste differentiation. Owing to nutritional difference, prospective queen larvae showed enhanced growth, and this was manifested through the overexpression of metabolic enzymes. Differently from similar studies targeting the causes of honeybee caste polymorphism, this subcellular level study provides an in-depth insight into mitochondrial proteins-mediated caste polymorphism and greatly improves protein coverage involved during honeybee caste determination. Hence, it is a major step forward in the analysis of the fundamental causes of honeybee caste pathway decision and greatly contributes to the knowledge of honeybee biology. In particular, the consistency between the 22 proteins and mRNA expressions provides us important target genes for the reverse genetic analysis of caste pathway modulation through RNA interference.