Volume 139, No. 5, May 7, 2012. Pages 371–388.
We have recently discovered that the current data acquisition software for analytical ultracentrifuges, ProteomeLab XL-A/XL-I Graphical User Interface version 6.0 by Beckman Coulter, deployed worldwide by the manufacturer since early 2011, introduces timestamp discrepancies into the scan data files (Rhyner, 2013; Zhao et al., 2013). As described in detail elsewhere (Zhao et al., 2013), the scan files contain entries for elapsed times that underreport the experimental times by 10% at rotor speeds of 50,000 rpm. Although reported scan times are irrelevant for the interpretation of sedimentation equilibrium (SE) experiments, in sedimentation velocity (SV), these discrepancies lead to overestimates of the derived sedimentation coefficients by 10%. We have shown that a corrective time dilation factor can be retrieved from the scan file timestamps of the Windows operating system, as implemented in current versions of the SEDFIT software.
A subset of the conventional SV data on GluA2 and enhanced green fluorescent protein (EGFP) in some of the studies we originally reported was collected with Beckman Coulter acquisition software version 6.0. Therefore, we have reanalyzed these data with accurate timestamps, leading to the following corrections:
(1) For the EGFP control experiments, the s-value at 20°C from timestamp-corrected conventional analytical ultracentrifugation (AUC) data are 2.61 ± 0.03 S (previously 2.87 ± 0.03 S), which is now reasonably consistent considering instrument-dependent errors in temperature calibration with the fluorescence detection optics (FDS)-derived value of 2.54 ± 0.02 S, and with the value of 2.62 ± 0.02 S determined using the absorbance detector in the same FDS instrument.
(2) Corrections were applied to the following GluA2 datasets shown in Fig. 3: The absorbance c(s) data shown in A and C, the sw isotherm for EndoH in B, as well as the absorbance-derived sw isotherm shown in D now have 10% lower s-values, with concomitant slight corrections to the reported binding constants. Studies on GluA3 shown in E and F were conducted before the installation of ProteomeLab version 6.0 data acquisition software and are unaffected. A corrected Fig. 3 is pro vided below. Notably, the gross difference in the limiting s-values for dimeric GluA2 in FDS and conventional detection is now removed (corrected Fig. 3, C and D). However, the best-fit unconstrained isotherm of the FDS-derived GluA2-FAM, with its high sw values for low concentrations and hydrodynamically unlikely ratio of best-fit dimer-to-monomer ratio of s-values, is unchanged. As a consequence, the last two sentences of the abstract no longer adequately summarize our results and should be deleted.
(3) After correcting for these timestamp discrepancies, we find that the GluA2-binding constants derived from sw isotherms from three experiments have only slightly different values, within the previously reported confidence intervals (CIs). Even though all sw values of these isotherms are uniformly 10% lower, which by itself should leave Kd values invariant, different best-fit values arise because hydrodynamic constraints were derived from hydrodynamic modeling and did not experience the same offsets. A corrected Table 1 is provided below. Overall, the range of best-fit Kd values for unlabeled GluA2 is unchanged, but the average of individual measurements at 20°C for all constructs becomes 9.4 nM, rather than the previously stated 7.1 nM (page 378). In addition, we discovered a previous inconsistency in the hydrodynamic constraints applied to the FDS-SV data when analyzed alone, in comparison with the constraints applied to conventional SV. Adjustment of the FDS-SV constraints to the range of 2.9 to 3.3 S for the monomer and 4.6 to 5.3 S for the dimer, respectively, leads to a FDS-derived Kd of the hydrodynamically constrained fit of 2.5 nM (95% CI; 0.4–7.8 nM), rather than 5.3 nM (95% CI; 3.0–14 nM) stated on page 386, confirming the consistency of FDS and conventional Kd values as previously concluded.
(4) As described on page 380 (Results) and page 386 (Discussion), we previously hypothesized that there may be an unrecognized instrument error causing a difference of 10% between data from the FDS-equipped instrument and the conventional instruments, for example, from possible temperature calibration errors. Accordingly, we previously applied ad hoc upward corrections to the FDS data and examined global fits of FDS and absorbance data, as shown in Figs. S2 and S3. These figures should now be deleted, as we have determined the source of the instrument difference, and a new, corrected replacement Fig. S2 substituted, where rather than increasing the FDS data the corrected absorbance data decrease by 10%.
In summary, our conclusions regarding the source of the 2,400-fold range of Kd values in the literature, which were the main focus of our paper, are unaffected by the accuracy of the scan data files generated by current Beckman Coulter data acquisition software for absorbance and interference systems (Zhao et al., 2013). In contrast, the mysterious large overall discrepancy between s-values determined with the FDS-equipped instrument and those determined with our set of conventional AUCs is resolved, but with one exception. The timestamp discrepancies introduced by the data acquisition software fully explain the 10% larger s-values measured in our control experiments with EGFP. Likewise, the same 10% overestimate applies to the s-values measured in conventional AUCs by absorbance optics as a control for the FAM-labeled, EndoH-digested GluA2 molecules. However, a surprising feature of the GluA2-binding isotherm determined by FDS is that the s-value of the GluA2 monomer is larger than predicted by hydrodynamics. At present, we do not know the cause of the discrepancy.