Don’t waste good methods on bad buffers and ambiguous data

Recently in PNAS, Liu et al. (1) presented single-molecule fluorescence resonance energy transfer experiments (sm-FRET) on the protein BBL (QNND-BBL) that claim to show there is only one populated state during folding rather than the two separate states we had previously, clearly observed (2). Liu et al. (1) stated about our work: “… we were unable to reproduce their results even at the most basic level. For instance, the equilibrium chemical unfolding curve of QNND-BBL at 279 K has C_m of approximately 2.4 M GdmCl and not 4 M GdmCl as reported before …” That claim is untrue. Liu et al. did not repeat our experiments but replaced our Mops buffer by phosphate, which dramatically changes pH during the experiment to much lower values where the protein is far less stable. Liu et al. (1) did their GdmCl denaturation in supposedly pH 7 phosphate buffer. It is well known (3) that anionic buffers are sensitive to salt concentration—and dianionic phosphate exquisitely so—whereas amine buffers are rather insensitive. Accordingly, we always use and recommend (4) amine buffers for GdmCl titrations. We performed our denaturation experiments in a Mops buffer, the pH of which at 279 K is constant at 7.35 ± 0.05 from 0 to 6 M GdmCl (its pH at 298 K is 7). We found the pH of the phosphate buffer used by Liu et al. dropped rapidly to below pH 6 and was pH 5.5 at the denaturation midpoint of the transition in Mops buffer (Fig. 1A). They thus progressively lowered the pH during the denaturation transition by one to two units below that of ours. The stability of BBL drops rapidly with decreasing pH below 7 (Fig. 1B) (5), and so Liu et al. (1) found a lower artifactual midpoint (C_m). We actually reported one experiment in phosphate buffer at pH 5.8 at 2.5 M GdmCl (ref. 2 and figure S3 in ref. 2), which shows the protein to be largely denatured, unlike for the titrations in Mops buffer with C_m = 4 M (figure S1 in ref. 2), consistent with the Liu et al. data.

Fig. 1.

Experimental data for QNND-BBL. (A) GdmCl titration of BBL in different buffers. The panel is taken from Liu et al. (1) and relabeled to show the pH values measured by us. The blue circles are data obtained in phosphate buffer by Liu et al. (1) and the red circles are our own data obtained in Mops (2). The pH value for Mops buffer was in the range 7.3–7.4 across all GdmCl concentrations. (B) The influence of pH on the free energy of denaturation, ΔG_D-N, of BBL at 298 K, from ref. 5. (C) sm-FRET bursts obtained under following conditions: 279 K, pH 7 in Mops buffer, binning time of 200 μs, and the same settings for the confocal microscope as in ref. 2. (D–F) sm-FRET histograms at 0, 3, and 4 M GdmCl, respectively, with binning time of 200 µs, threshold of 70 at 279 K.

Liu et al. (1) also changed the protocols for sm-FRET to conditions that we consider unsuitable. For example, they decreased the resolution of the histogram in some cases by increasing the bin size from 0.03 to 0.05 for different figures and even in the same figure (e.g., figure 3 of ref. 1). It is difficult to evaluate the quality of their experiments because insufficient raw data are presented (e.g., the single-molecule bursts and number of events in histograms). However, it is clear in some cases that there are insufficient events for accurate analysis. In figure S9 of Liu et al. (1), with a large binning size of 0.05 and low threshold of 20, there were only 50–60 events for the peak. Those conditions make it very difficult to resolve different populations. In response to Liu et al. (1), we repeated our experiments and found our previous results reproducible with two clear peaks (Fig. 1 C–F).

The problem with sm-FRET experiments is that if only one peak is observed, the interpretation is ambiguous: the data are consistent either with a single state or two states that are rapidly interconverting or inadequate resolution of the data. However, if two separate peaks are observed, then there are at least two species definitely present and mechanisms involving just one state are eliminated. We stand by our previous experiments—which are reproducible—and their interpretation.