Elemental stoichiometry of biomass is a focal point that connects different biogeochemical cycles. Yet, the mechanistic underpinnings of elemental stoichiometry are poorly quantified in many cases. We combined targeted and untargeted metaproteomics, Bayesian statistical modelling, and geochemical measurements to quantify the contribution of specific proteins to metal stoichiometry in natural populations of Southern Ocean diatoms. Our analyses indicate that a substantial amount of non-photosynthetic manganese (Mn) in diatoms in an Antarctic polynya can be attributed to superoxide dismutases (~0.7 mol Mn : mol Carbon; ~20% of the total cellular Mn quota). We then used cultures and proteomic profiling of the key polar diatom Fragilariopsis cylindrus to identify environmental controls on superoxide dismutases, and discovered that iron concentration has little influence on the abundance of two Mn superoxide dismutases, while Mn limitation induces the depletion of these Mn superoxide dismutases and an accompanying increase of nickel superoxide dismutase. Overall, we combined metaproteomic approaches to quantify proteomic composition and connected these measurements to metal-to-carbon ratios and their responses to metal availability. Because metal quotas are key parameters in some biogeochemical models, our approach provides a direct mechanism for informing ecosystem-scale models with molecular measurements.