Recently, the randomized phase‐II Triton study demonstrated that mesenchymal stromal cell (MSC) therapy facilitated early tacrolimus withdrawal in living donor kidney transplant recipients. The current sub‐study analyzed formation of de novo donor‐specific HLA antibodies (dnDSA) in the context of the degree of HLA eplet mismatches. At the time of protocol biopsy at 6 months, 7/29 patients (24%) in the MSC group and 1/27 patient (3.7%) in the control group had developed dnDSA. In the MSC group, all dnDSA were anti‐HLA‐DQ; two patients had anti‐DQ alone and five patients combined with anti‐class I, HLA‐DR or ‐DP. Despite excess dnDSA formation in the MSC‐arm of the study, the evolution of eGFR (CKD‐EPI) and proteinuria were comparable 2 years posttransplant. All dnDSA were complement‐binding and three patients had antibody‐mediated rejection in the protocol biopsy, but overall rejection episodes were not increased. Everolimus had to be discontinued in nine patients because of toxicity, and tacrolimus was reintroduced in six patients because of dnDSA formation. The HLA‐DQ eplet mismatch load independently associated with dnDSA (adjusted hazard ratio = 1.07 per eplet mismatch, p = 0.008). A threshold of ≥11 HLA‐DQ eplet mismatches predicted subsequent dnDSA in all 11 patients in the MSC group, but specificity was low (44%). Further research is warranted to explore HLA molecular mismatch load as a biomarker to guide personalized maintenance immunosuppression in kidney transplantation.