Synchrotron-based small-angle X-ray scattering of proteins in solution

Abstract

With recent advances in data analysis algorithms, X-ray detectors and synchrotron sources, small-angle X-ray scattering (SAXS) has become much more accessible to the structural biology community. Although limited to ∼10 Å resolution, SAXS can provide a wealth of structural information on biomolecules in solution and is compatible with a wide range of experimental conditions. SAXS is thus an attractive alternative when crystallography is not possible. Moreover, advanced use of SAXS can provide unique insight into biomolecular behavior that can only be observed in solution, such as large conformational changes and transient protein-protein interactions. Unlike crystal diffraction data, however, solution scattering data are subtle in appearance, highly sensitive to sample quality and experimental errors and easily misinterpreted. In addition, synchrotron beamlines that are dedicated to SAXS are often unfamiliar to the nonspecialist. Here we present a series of procedures that can be used for SAXS data collection and basic cross-checks designed to detect and avoid aggregation, concentration effects, radiation damage, buffer mismatch and other common problems. Human serum albumin (HSA) serves as a convenient and easily replicated example of just how subtle these problems can sometimes be, but also of how proper technique can yield pristine data even in problematic cases. Because typical data collection times at a synchrotron are only one to several days, we recommend that the sample purity, homogeneity and solubility be extensively optimized before the experiment.

Figure 1. A basic SAXS setup

A collimated, monochromatic X-ray beam incident on the sample generates scattered X-rays, which are imaged by a detector. The transmitted beam is usually blocked by a beamstop, resulting in a shadow in the image. The scattering vector, q, describes the change in direction of the elastically scattered X-rays, and is roughly parallel to the detector face in the small-angle approximation. The images are integrated to yield scattering intensity as a 1D function of q. The scattering intensity can also be expressed as a function of s = q/2π, which is equivalent to “resolution” or d-spacing in crystallography. Because q and s are often used interchangeably in the literature, the exact usage should be explicitly defined in any publication with SAXS data.

Figure 2. Size exclusion chromatogram of HSA

Multiple peaks are observed, indicating that the HSA solution is a monomer-oligomer mixture prior to this purification step.

Figure 3. Typical scattering from protein solution (red) and the corresponding buffer (gray)

The scattering profiles are offset by a constant positive value. Note that apart from the sharp upturn at low q, the background scattering is largely featureless.

Figure 4. Examples of good and bad buffer subtractions

(a) Here, the scattering from HSA solution is subtracted with that of a matched buffer (red), a different buffer at another pH (blue), and a similar buffer with glycerol added (green). Note that proper subtractions should lead to positive intensity values that, for a well-folded protein, approach near zero at high q. In these examples, buffer mismatch leads to extreme over-subtraction at low q and unphysical, negative intensities or artificially high, positive intensities at high q. It is important to note, however, that distortions of the scattering profile at low q can be a real consequence of inter-particle interactions. Similarly, less compact forms of proteins will not exhibit a sharp intensity fall-off at high q. Thus, confidence in buffer subtractions is absolutely essential for proper interpretation of data. (b) The scattering behavior at high q is enhanced when plotted as Kratky curves (same colors as in (a)).

Figure 5. Comparison of HSA with and without the final purification step

(a) The scattering profiles of unpurified HSA (red curves, from bottom to top: 1.9, 3.8, 6.5 mg/mL) and purified HSA (black curves, from bottom to top: 1.9, 3.8, 6.5, 15.4 mg/mL) show no signs of aggregation, which would appear as upturns in the lowest q portion of a profile. (b) When plotted as Guinier curves, the absence of upturns at low q is clear. In addition, the effect of the final purification step is evident: for a given concentration, the slope and vertical intercept of unpurified HSA are greater, consistent with the presence of species larger than a monomer. Finally, concentration effects can be seen, particular in the case of purified HSA. With increasing concentration, a greater downward curvature is observed at low q, consistent with inter-particle repulsion.

Figure 6. Example of Guinier analysis

A line (red) is fitted to the low q region of a Guinier curve (black), such that maximum q to be included in the fit is 1.3/R_g or less. The linearity of the fitted region is determined by the flatness of the residuals (green). R_g is derived from the slope, and I(0) is derived from the vertical intercept.

Figure 7. Concentration effects on R_g and I(0)

(a) For both purified (black) and unpurified (red) HSA, the R_g values from Guinier analysis show a decreasing linear trend with increasing protein concentration. The error bars represent standard deviations from linear fits to the Guinier curves in Fig. 6. This trend is caused by distortions in the scattering profiles at low q that arise from inter-particle repulsion. (b) Over the 1.9 – 6.5 mg/mL range, I(0) shows a linear relationship with protein concentration for both purified (black) and unpurified (red) HSA. Deviations from a linear trend can be indicative of inter-particle effects or changes in oligomerization state.

Figure 8. The pair distance distribution function, P(r), is sensitive to concentration effects and sample polydispersity

Here, PRIMUS was used to automatically generate the P(r) of purified HSA (solid curves, from red to blue: 1.9–15.9 mg/mL) and unpurified HSA at 1.9 mg/mL (dotted curve). Inter-particle repulsion leads to negative regions at high r, which leads to an artificially low D_max. The presence of aggregates or higher order oligomers extends P(r) at high r, leading to an artificially high D_max.

Figure 9. Comparisons to a crystal structure

(a) Low q data from 1.9 mg/mL purified HSA was merged with higher q data from 3.8 mg/mL purified HSA (shown in gray with standard errors). A smooth curve was fitted (black) to the data in GNOM that yielded a well-behaved P(r) curve. Ten bead models were reconstructed in DAMMIF, which were aligned and averaged in DAMAVER with no rejections and a normalized spatial discrepancy of 0.874 ± 0.032. The fit of a typical DAMMIF model (cyan) shows good agreement to the data as does the theoretical scattering profile of a crystal structure (magenta). (b) Following averaging in DAMAVER, the “damstart” model was used for a round of refinement in DAMMIN, yielding the final SAXS envelope (gray), shown superimposed with the crystal structure (magenta).