Endogeneity among explanatory variables is the fundamental issue that instrumental variable methods and the first-stage regression of two-stage least squares are specifically designed to address, making the statement conceptually backward by definition.
The claim states that endogeneity is a problem in the first-stage regression of a 2SLS model. In econometric theory, endogeneity in the structural equation is precisely the reason 2SLS is used; the first-stage regression regresses the endogenous explanatory variable on the instruments to purge that endogeneity (yielding predicted values). Endogeneity is the problem the model solves, not a failure of the first stage itself. Because this is a definitional matter regarding the mechanics of instrumental variable estimation, no citation is required.