(T4C), rs6497476 (A4G), and rs4293393 (T4C). What led us to the awkwardness of mixing strands? We initially considered that, in many previous reports on UMOD, investigators were accustomed to referring to rs13333226 A/G (antisense strand), rs6497476 T/C (antisense strand), and rs4293393 C/T (sense strand), and we wanted to remain consistent with past nomenclature. But clearly we created more confusion than clarity. We hope that Figures 1a and b begin to clarify the strand issues. Q-2: The second question was about longer haplotype association analyses with urinary uromodulin level. At the end of the day, the question of highest interest here is whether genetic variants at the UMOD locus can predict UMOD protein excretion into the urine; i.e., is there a UMOD cis-QTL, especially in the proximal promoter? To answer this question, we typed the UCSD twin cohort for UMOD excretion (i.e., urine UMOD/creatinine), and then tested the effect of two UMOD promoter variants on the trait: rs13333226 (T4C, sense strand) and rs4293393 (T4C, sense strand). These two SNPs, in perfect linkage disequilibrium (LD) (R 2 ¼ 1.0), were both highly predictive (each P-value ¼ 2.01E À 06) of UMOD/creatinine, in an additive genetic model (C/CoT/CoT/T) (Table 1). Why did the UCSD twin cohort so readily yield this positive result? Although the UCSD twin data set was collected and phenotyped over the better part of a decade, aliquots of each urine sample were freshly frozen at À 70 1C and then all thawed and assayed at the same time for UMOD plate assay (Meso Scale Discovery with electrochemiluminescent detection, Gaithersburg, MD). Thus, the effects of any inter-assay variability would be minimized. In the Beijing cohorts, UMOD promoter single SNPs did not readily predict UMOD excretion, in either a hypertension study 2 or the current uric acid study. 3 To probe the possibility of more subtle UMOD marker-on-UMOD trait predictions, we turned to the alternative haplotype approach, 'tagging' the locus with SNPs to compose haplotypes of up to seven SNPs 430 kbp. Initially, this seemed reasonable, as the overall HapMap LD plot for the CHB (Chinese-Han-Beijing) population indicated stretches of LD (by D 0) spanning the entire UMOD locus, and up to B35 kbp (ref. 2; main paper, supplementary Figure S1 online), albeit with internal blocks wherein LD declines. However, both you (your Figure 1b) and we 2,3 have shown, by custom plots in HaploView, that as many as three LD blocks may span this region, diminishing the feasibility of an extended haplotype approach. In the end, the entire controversy may be rendered moot by the association of UMOD excretion with UMOD promoter variants in the UCSD twins (Table 1).
