Technical Comments

Response to Comment on “Innovative scattering analysis shows that hydrophobic disordered proteins are expanded in water”

See allHide authors and affiliations

Science  31 Aug 2018:
Vol. 361, Issue 6405, eaar7949
DOI: 10.1126/science.aar7949

Abstract

Best et al. claim that we provide no convincing basis to assert that a discrepancy remains between FRET and SAXS results on the dimensions of disordered proteins under physiological conditions. We maintain that a clear discrepancy is apparent in our and other recent publications, including results shown in the Best et al. comment. A plausible origin is fluorophore interactions in FRET experiments.

Using our new molecular form factor (MFF), we analyzed small-angle x-ray scattering (SAXS) data for three intrinsically disordered proteins (IDPs) and found that upon a shift from 6 to 0 M guanidinium chloride (Gdn), there was a mild decrease in radius of gyration (Rg) (1). For these 124- to 334-residue IDPs with amino acid sequences typical of foldable proteins, the value of Rg decreased by 20 to 28% in water. As predicted from scaling laws, the denatured states of smaller proteins such as ubiquitin and protein L (76 and 64 amino acids, respectively) will contract less (17 and 15%, respectively). Notably, approximately half of this contraction occurs below 1 M Gdn, beyond the limit of many prior kinetic studies. For example, for ubiquitin, only an 8% decrease (~2 Å) in Rg is expected from 6 to 1 M Gdn. This mild and nonlinear decrease in Rg explains why many previous SAXS studies of small proteins did not observe a statistically significant decrease in Rg (26)—a conclusion we highlighted in figure 4A of our publication (1).

In contrast, Förster resonance energy transfer (FRET) studies have observed much more contraction, up to a 50% decrease in Rg in the absence of denaturant, even for small proteins (3, 7, 8). Furthermore, the majority of this contraction occurs above 1 M Gdn (Fig. 1). These findings are different from those observed using SAXS and are noticeable in figure 1 of Best et al. (9) (compare green and black points at lowest denaturant concentration). Hence, our methods and those of Best et al. concur that in the absence of denaturant, SAXS analysis returns Rg values ~15% larger than FRET studies when each data set is analyzed independently. In our publication, we hypothesized that this discrepancy could be due to the addition of dyes necessary for FRET studies (1). In contrast, Best et al. provide no physical explanation and instead rely on a joint analysis of both types of data that only reduces the appearance of the discrepancy.

Fig. 1 SAXS and FRET exhibit different denaturant dependence for two proteins, R17 and ACTR, studied by Best and co-workers.

All primary data are from (22). Open and solid red circles are two models applied to FRET data (reweighted simulations and a SARW, respectively), as presented in (22). Open and solid black circles are SAXS data fit using reweighted simulations [as presented in (22)] and our MFF (1), respectively. The black line is the expected SAXS denaturant dependence of Rg taken from (1), assuming that 2 M urea is equivalent to 1 M Gdn.

An alternative, chain length–independent method to compare results from SAXS and FRET studies is to compare the Flory exponent (ν in the relationship RgNν, where N = protein length). Values of ν equal to 0.33 or 0.60 correspond to a globule or a self-avoiding random walk (SARW), respectively. In a theta solvent, ν = 0.50 and intrachain interactions are equally as favorable as solvent-chain interactions; this value defines the boundary between good and poor solvents. For the three foldable sequences we examined using SAXS, ν in water is 0.54. In contrast, using FRET, Hofmann et al. (10) found that ν for four foldable sequences in water ranged from 0.4 to 0.51. Because the range of ν between a globule and a SARW is only 0.27, a difference of 0.1 units represents 38% of the entire range. Hence, to argue that both methods are in agreement if the values of ν are near ½ is imprecise.

This discrepancy is not limited to the specific proteins selected by us or by Hofmann et al. (10). We reanalyzed available SAXS data for other IDPs with sequences typical of folded proteins (1117) using our new MFF procedure and compared these results to ν extracted from available FRET studies of disordered proteins with foldable sequences (Fig. 2A). When measured by SAXS, ν is typically above 0.54, whereas ν from FRET studies is typically below the theta condition of 0.50. Note that our Fig. 2A should, in principle, match figure 2B of Best et al. However, figure 2B of Best et al. does not include results from two of the most collapsed proteins that Fuertes et al. (11) studied by FRET: R98 and NSP. The FRET and SAXS values for these proteins differ significantly (ν = 0.44 versus 0.56 for R98 and 0.49 versus 0.60 for NSP, respectively). Moreover, Best et al. plot results for other proteins (e.g., Csp, hCyp, and R15) in ~1 M denaturant rather than the value extrapolated to 0 M denaturant, as they calculated in figure 2A and reported in their original study [Hofmann et al. (10)]. These differences misleadingly minimize the discrepancy between the SAXS and FRET results. Plotting instead results for all proteins with sequences typical of folded proteins in the absence of denaturant, as in our Fig. 2A, reveals clear discrepancies between results from these methods.

Fig. 2 SAXS and FRET show inconsistent solvent qualities for IDPs in water.

(A) Trends of hydrophobicity (Kyte-Doolittle scale) versus ν in water for SAXS data of foldable protein sequences fit to our MFF for data from our recent study (1) (black circles), Fuertes et al. (11) (black squares), and other studies (1217) (black triangles). Also shown are results from FRET studies calculated as in (10) for data from (10) (red circles) and (11) (red squares). The red trend line for FRET results is from (10). The black trend line is the best fit to the SAXS data shown. At the top is a histogram of hydrophobicity of representative proteins in the PDB [dataset from (1)] with sequences in the FRET-inferred poor solvent region highlighted in red. (B) SAXS profiles from (11) for unlabeled and labeled versions of the NLS protein fit using our MFF. Solid lines span data points used for fitting. Error bars are the SD appropriate for Poisson counting statistics. (C) PEG results from (8) remain unchanged with improved analysis. Small-angle neutron scattering (SANS) data (black) are fit to our MFF; FRET data (red) are fit assuming a SARW distribution and normalized to high denaturant.

Figure 2 of Best et al. emphasizes the sole exception to ν > 0.5 in our SAXS data: the P domain, a hydrophobic IDP region of low sequence complexity whose collapse correlates with stress granule formation (18). Although this finding indicates that some hydrophobic, nonfoldable sequences can collapse, it does not speak directly to the discrepancy between SAXS and FRET results for protein-like sequences.

Having established that discrepancies remain between results of SAXS and FRET studies, we considered possible origins. As we discussed (1), Chan and others have proposed that complications in converting FRET efficiency to Rg could account for some of this difference (11, 1921). The other obvious issue is the presence of fluorophores.

Schuler and co-workers have found that some fluorophore pairs influence FRET signals but maintain that the Alexa 488/594 pair is suitable (22). Their all-atom molecular dynamics simulations, however, reported a 10% contraction for an IDP with versus without fluorophores in 1 M urea (23). In the absence of denaturant, this dye effect would presumably be even larger. Recently, Fuertes et al. conducted SAXS measurements on five proteins with and without Alexa 488/594 (11). In all cases, addition of fluorophores changed the SAXS profile. From the change in scattering between the labeled and unlabeled proteins, the authors calculated meaningful differences (νunlabeled – νlabeled = Δν = 0.08, 0.03, 0.03, –0.02, –0.04). Reanalysis of these data with our MFF (1) finds Δν = 0.09, 0.06, 0.03, –0.02, –0.08 (Fig. 2B). To address an additional concern raised by Best et al., we reanalyzed published polyethylene glycol (PEG) scattering data (8) with our MFF and confirmed the published observation that whereas FRET-measured PEG labeled with Alexa 488/594 contracts in lower denaturant, PEG without labels does not contract (Fig. 2C). Hence, significant discrepancies exist between FRET and scattering techniques even when using fluorophores considered suitable.

We are gratified that Best et al. agree that water should be considered to be a good solvent for the denatured state of most foldable sequences (ν > 0.5). However, an analysis of the FRET results alone would lead one to believe the opposite (ν < 0.5). The presence of fluorophores in FRET studies is a likely origin of this difference.

References and Notes

Acknowledgments: We thank H. S. Chan, R. Best, W. Zhang, L. Lapidus, K. Plaxco, E. Martin, and A. Holehouse for useful discussions. The analysis used in this and our original publication are available from https://github.com/sosnicklab/SAXSonIDPs. A web server to fit SAXS data using our MFF is available at http://sosnick.uchicago.edu/SAXSonIDPs. Funding: Supported by NIH grants GM055694 (T.R.S., K.F.F.), GM097573 (P.L.C.), GM103622 and 1S10OD018090-01 (T. C. Irving), T32 EB009412 (T.R.S.), T32 GM007183 (B. Glick), and T32 GM008720 (J. Picirrilli) and by NSF grants GRF DGE-1144082 (J.A.R.) and MCB 1516959 (C. R. Matthews). Use of the Advanced Photon Source was supported by the U.S. Department of Energy under contract DE-AC02-06CH11357. Author contributions: All authors contributed to and approved the manuscript. J.A.R. performed the data analysis. Competing interests: None.
View Abstract

Navigate This Article