ABSTRACT Low-cost, refrigerator-stable COVID-19 vaccines will facilitate global access and improve vaccine coverage in low- and middle-income countries. To this end, subunit-based approaches targeting the receptor-binding domain (RBD) of SARS-CoV-2 Spike protein remain attractive. Antibodies against RBD neutralize SARS-CoV-2 by blocking viral attachment to the host cell receptor, ACE2. Here, a yeast-produced recombinant RBD antigen (RBD-L452K-F490W or RBD-J) was formulated with various combinations of aluminum-salt (Alhydrogel®, AH; AdjuPhos®, AP) and CpG 1018 adjuvants. We assessed the effect of antigen-adjuvant interactions on the stability and mouse immunogenicity of various RBD-J preparations. While RBD-J was 50% adsorbed to AH and <15% to AP, addition of CpG resulted in complete AH binding, yet no improvement in AP adsorption. ACE2 competition ELISA analyses of formulated RBD-J stored at varying temperatures (4, 25, 37°C) revealed that RBD-J was destabilized by AH, an effect exacerbated by CpG. DSC studies demonstrated that aluminum-salt and CpG adjuvants decrease the conformational stability of RBD-J and suggest a direct CpG-RBD-J interaction. Although AH+CpG-adjuvanted RBD-J was the least stable in vitro, the formulation was most potent at eliciting SARS-CoV-2 pseudovirus neutralizing antibodies in mice. In contrast, RBD-J formulated with AP+CpG showed minimal antigen-adjuvant interactions, a better stability profile, but suboptimal immune responses. Interestingly, the loss of in vivo potency associated with heat-stressed RBD-J formulated with AH+CpG after one dose was abrogated by a booster. Our findings highlight the importance of elucidating the key interrelationships between antigen-adjuvant interactions, storage stability, and in vivo performance to enable successful formulation development of stable and efficacious subunit vaccines.