The CLZ domain forms a trimer. a, Gel-filtration experiment with N-terminally HA-tagged A3C-28 (see Fig. 2b). See Methods for details. The Mr values of the peaks were estimated from their Stokes radii based on a standard curve (not shown). b, Cross-linking experiment on N-terminally HA-tagged A3C-28 or A3C-32 (CLZ-disrupted A3C-28, see Fig. 2b) with the cross-linker DSP. The faint >100K band in the gels seems to be non-specific. DTT, dithiothreitol. c, Cross-linking experiments on HA-tagged A1C-28 (Lys 574 to end of human CNGA1) and A2C-28 (Lys 555 to end of rat CNGA2). d, Native-gel electrophoresis. All constructs were N-terminally HA-tagged and detected with an anti-HA antibody. A3C-87 is A3C-28 with mutation L633 A, which disrupted the CLZ-mediated interaction without changing the charge of the construct. The position of A3C-87 indicates the monomer position of A3C-28. The two A3C-28 mutants, A3C-88 (K596E and R603E) and A3C-89 (D613K and E615K), had different mobilities in native-gel electrophoresis. When these two constructs are co-expressed, three bands are expected for a dimeric interaction (P2, PQ and Q2) and four bands for a trimer (P3, P2Q, PQ2 and Q3). Four bands were seen (indicated by arrowheads) when A3C-88 and A3C-89 were co-expressed. In the bottom panel, a fixed amount of A3C-88 was co-transfected with increasing amounts of A3C-89. Four bands can be clearly seen (arrowheads). e, Equilibrium analytical centrifugation. A 48-residue peptide corresponding to the CLZ domain (Lys 624 to Gly 671 of hCNGA3) was commercially synthesized and purified to more than 95% purity (Synpep, Dublin, CA). This peptide was acetylated at the N terminus and amidated at the C terminus, with a calculated Mr of 5,582. The peptide was subjected to analysis by equilibrium analytical centrifugation. A total of nine data sets were collected (0.5, 1.0 and 2.0 mg ml−1, each spun at 40,000, 45,000 and 52,000 r.p.m.). A representative data set (2.0 mg ml−1 centrifuged at 40,000 r.p.m.; open grey squares) and its fitting curves are shown in the upper panel, with residuals (difference between data and fittings) in the lower panel. The data are best fitted by an ideal monomer–trimer model (red) with a Mr of 5,468 (confidence interval 5,375–5,559) and an association constant of 1.43 × 1010 M−2 (confidence interval (3.68–71.7) × 109 M−2). A similar fit with a monomer–dimer model gave an unrealistic Mr in excess of 7,000. With Mr fixed at 5,582 (the calculated value), neither the monomer–dimer (black) nor the monomer–tetramer models (green) fitted well.