Many biomarkers have been epidemiologically linked with CKD; however, the possibility that such associations are due to reverse causation or confounding limits the utility of these biomarkers. To overcome this limitation, we used a Mendelian randomization (MR) approach to identify causal mediators of CKD. We performed MR by first identifying genetic determinants of 227 serum protein biomarkers assayed in 4147 participants of the Outcome Reduction with Initial Glargine Intervention (ORIGIN) trial who had early or prediabetes, and assessing the effects of these biomarkers on CKD in the CKD genetics consortium (=117,165; 12,385 cases) using the inverse-variance weighted (fixed-effects) method. We then estimated the relationship between the serum concentration of each biomarker identified and incident CKD in ORIGIN participants. MR identified uromodulin (UMOD) and human EGF receptor 2 (HER2) as novel, causal mediators of CKD (UMOD: odds ratio [OR], 1.30 per SD; 95% confidence interval [95% CI], 1.25 to 1.35; <5×10; HER2: OR, 1.30 per SD; 95% CI, 1.14 to 1.48; =8.0×10). Consistent with these findings, blood HER2 concentration associated with CKD events in ORIGIN participants (OR, 1.07 per SD; 95% CI, 1.01 to 1.13; =0.01). Additional exploratory MR analyses identified angiotensin-converting enzyme (ACE) as a regulator of HER2 levels (=0.13 per SD; 95% CI, 0.08 to 0.16; =2.5×10). This finding was corroborated by an inverse relationship between ACE inhibitor use and HER2 levels. Thus, UMOD and HER2 are independent causal mediators of CKD in humans, and serum HER2 levels are regulated in part by ACE. These biomarkers are potential therapeutic targets for CKD prevention.