Even though ZFC is a first-order theory, it can interpret properties that are not first-order in a different language. For instance, in the language of graphs, the property of a graph being finite is not first-order expressible. But in any model of ZFC, we can find a set corresponding to the set of all finite graphs.
Yet how can we know that every mathematical concept can be interpreted in ZFC? Could there not exist some property or object in the literature which can not be represented in a model of ZFC? Do we just work from the assumption that ZFC can interpret everything -- or are there good reasons to believe so?