**2. Results**

#### *2.1. Isolated LrtA Was Involved in a Self-Association Equilibrium in Solution*

We first tried to elucidate the oligomerization state of the protein to identify the protein-concentration range where we must characterize the conformational stability of the protein.

To map the hydrodynamic properties of LrtA we used three hydrodynamic techniques: DOSY-NMR (diffusion ordered NMR spectroscopy), DLS (dynamic light scattering), size exclusion chromatography (SEC), glutaraldehyde cross-linking, and lifetime fluorescence measurements. Furthermore, we tried to measure the self-association of LrtA by using isothermal titration calorimetry

(ITC), but in all attempts, protein precipitated at the concentrations required to carry out the experiments. It is important to pinpoint the differences among the different hydrodynamic techniques used in this work. With NMR, we shall obtain information about the low-molecular weight species, whose overall rotational tumbling is very fast. By using DLS, we shall obtain information about the hydrodynamic parameters, assuming a spherical shape, for all the species (high or low molecular weight) present in solution, and we shall be able to see whether those hydrodynamic parameters are protein-concentration-dependent. By using SEC, we shall be able to monitor the elution volume of LrtA, which will depend on the molecular weight and the shape of the molecule, but that volume could be also affected by possible interactions with the column. Finally, by lifetime fluorescence measurements, we shall determine how the decay of the electronic excited states can be affected by: (i) the presence of conformational isomers; (ii) energy transfer among the eight Tyr residues in LrtA; or (iii) even transient electronic effects in collisional quenching [12,13]. To elucidate the oligomerization state of the protein, we also tried to carry out *T*<sup>2</sup> echo measurements estimating the averaged correlation time of the species present in solution from the amide region. However, at the times used (2.9 ms and 400 μs) in our echo experiments, most of the amide peaks disappeared, and only the proton resonances of the His ring (around 8.5 ppm) could be clearly measured, yielding a very long value for the *T*2, unreliable to estimate the mobility of the backbone of the polypeptide chain.

The DOSY-NMR measurements at pH 8.0 yielded a translational diffusion coefficient (*D*) with a value of (7.2 ± 0.2) × <sup>10</sup>−<sup>7</sup> cm2 <sup>s</sup>−<sup>1</sup> (Figure S1A). By taking into account the hydrodynamic radius, *<sup>R</sup>*S, of dioxane (2.12 Å), and its *<sup>D</sup>* under our conditions ((6.8 ± 0.2) × <sup>10</sup>−<sup>6</sup> cm2 <sup>s</sup>−1), the *<sup>R</sup>*<sup>S</sup> (Stokes radius) estimated for LrtA was 20 ± 2 Å. We can compare this value with that theoretically determined for a polypeptide with the length of LrtA. The *R* value for an unsolvated, ideal, spherical molecule can be estimated from [14]: *R* = <sup>3</sup> 3*MV*/4*NAπ*, where *NA* is Avogadro's number, *M* is the molecular weight (22.717 kDa) and *V* the specific volume of LrtA (0.729 mL/g). The calculated *R* for LrtA is 18.7 Å, but since the hydration shell is 3.2 Å wide [15], the hydration radius would be 21.9 Å, which is similar to the *R*<sup>S</sup> from DOSY-NMR. On the other hand, it has been shown that the *R*<sup>S</sup> of a folded spherical protein can be approximated by [16]: *RS* = (4.75 ± 1.11)*N*0.29, where *<sup>N</sup>* is the number of residues; in a 197-residue-long protein such as LrtA, this expression yields 21 ± 6 Å, similar to that determined by DOSY-NMR experiments. Therefore, by the DOSY-NMR measurements, we are only detecting a monomeric globular species of LrtA. In fact, the 1D 1H-NMR spectrum of LrtA at pH 8.0 (with 500 mM NaCl) (Figure S2) corresponds to that of a well-folded protein with dispersed peaks in the methyl and amide regions; interestingly enough, the spectrum had down-field shifted Hα protons (between 5.0 and 6.0 ppm), suggesting the presence of residues involved in β-strands. It is interesting to note that, as possible higher-order molecular species could not be observed in the NMR spectrum due to their molecular weight (and therefore signal broadening), the presence of the majority of the amide protons belonging to possible disordered regions in the protein (see below, 2.4.) would not be observed in the amide region due to: (i) solvent hydrogen-exchange at pH 8.0; (ii) conformational exchange broadening; or (iii) overlapping with the signals from the well-folded region (as indicated by the largest increase of intensity around 8.3 ppm, Figure S2B). In addition, the alkyl resonances belonging to possible polypeptide disordered regions would be hindered by the rest of the methyl groups of the protein in the up-field shifted region. On the other hand, the 1D 1H-NMR spectrum at pH 4.5 showed a smaller intensity and a poorer signal-to-noise ratio (due probably to the precipitation during sample preparation) than the spectrum at pH 8.0. In addition, the spectrum also showed broader peaks (and less intense, as it is evident by comparing Figure S3B with Figure S2B), and the absence of well-dispersed signals in the amide and methyl regions (Figure S3), as shown, for instance, by the lack of peaks around 0.3 and 0.5 ppm, which appeared at pH 8.0. The broader peaks at pH 4.5 (when compared to pH 8.0) could be due to the presence of uni-molecular conformational-exchange equilibria or, alternatively, to the presence of self-associated species. Thus, to elucidate whether at pH 4.5 there were concentration-dependent equilibria, we carried out far UV CD experiments at protein concentrations of 4.5 and 9.8 μM (in protomer units); these experiments (Figure S4) showed that the

molar ellipticity and the shape of the spectra were protein-concentration-dependent. Therefore, these results suggest that the conformation of the protein and its self-associated features were different at the two pH values of 4.5 and 8.0.

Second, we measured the hydrodynamic features of LrtA by using DLS at several protein concentrations. Two peaks were identified in the size distribution analysis for the concentration of 68 μM (in protomer units): the first peak with *R*<sup>S</sup> = 39 ± 5 Å that accounts for the 97.5% of the protein in the solution and a second peak with *R*<sup>S</sup> = 409 ± 200 Å corresponding to a small amount of aggregates (Figure 1A). Taking into account that particle scattering intensity is proportional to the square of the molecular weight, a small percentage (in this case 2.5%) of protein aggregates dominates the intensity distribution, which can be misleading, and therefore, the results in Figure 1A are shown as size distribution by volume instead of intensity. That *R*S, obtained from the first peak, corresponds to a molecular weight of 81 kDa, by using an empirical mass vs. size calibration curve in the instrument software. The experiments at different LrtA concentrations indicate that the *R*<sup>S</sup> (obtained from the volume peak measurements of the first observed peak) varied with the protein concentration. These findings suggest the presence of a self-associated equilibrium at pH 8.0 (Figure 1B) involving the protein, as also described in other oligomeric proteins [17]. It is interesting to note that these oligomeric species should not be observed in NMR due to their large molecular weights [15].

**Figure 1.** Hydrodynamic measurements of LrtA by dynamic light scattering (DLS) and size exclusion chromatography (SEC): (**A**) DLS measurements of the hydrodynamic radius RS of LrtA as a function of the percentage of the volume peak at 68 μM concentration (in protomer units). (**B**) Variation of the calculated *R*<sup>S</sup> with LrtA concentration (in protomer units). Error bars are standard deviations from the fitting to a spherical shape. (**C**) SEC chromatogram of LrtA at 97 μM (in protomer units) at pH 8.0 (50 mM Tris) in 0.7 M NaCl; the arrows at the top indicate (from left to right) the elution volumes of blue dextran (7.1 ± 0.1 mL), albumin (12.1 ± 0.1 mL; 63.7 kDa), and bovine RNase A (15.1 ± 0.1 mL; 15.7 kDa) (the errors are standard deviations of three independent measurements). Chromatogram was baseline-corrected by UNICORN 5.01 software (GE Healthcare), and therefore, the origin of the sharpening observed in the peaks. Experiments were carried out at 25 ◦C.

Third, we also detected the presence of oligomeric species using glutaraldehyde cross-linking, which can react as monomer, but also as a heterogeneous polymer, involving accessible lysine residues. Our results (Figure S5) indicate that in the presence of a final concentration of 1% gutaraldehyde, there were dimers (which appear close to the band of the protein marker at 48 kDa) and other high-molecular-weight species with molecular weights larger than 210 kDa, at the top of the SDS-PAGE (sodium dodecyl sulfate polyacrylamide gel electrophoresis) gel lanes. These high molecular weight species could be due to the presence of cross-linked dimeric species.

Next, we used SEC in a Superose 12 10/300 GL, in buffer pH 8.0 (50 mM Tris) with 0.7 M NaCl to elucidate whether the protein behaved as an oligomer. We used such high concentrations of NaCl to avoid, as much as possible, any kind of protein-column interactions (as those we have observed to occur with other kind of matrix columns, see below Section 4). The protein markers used to calibrate the column were also loaded in the same buffer. At loading concentrations of 97 μM (in protomer units) of LrtA, the protein eluted as several peaks (errors are standard deviations from three independent measurements) at 7.3 ± 0.2, 8.5 ± 0.1, 9.8 ± 0.2, 11.5 ± 0.1, and 14.7 ± 0.2 mL (Figure 1C). These peaks, especially that at 7.3 mL, indicate that the protein behaved as an oligomer, with molecular weights larger than those of albumin (63.7 kDa) and bovine RNase A (15.7 kDa), although other higher-order molecular species of LrtA were present in solution. The peak at 14.7 mL could be due to the monomeric species, which was observed under these conditions. The other oligomeric species could be assigned to hexamers (11.5 mL) and dodecamers (9.8 mL), whereas the other two could be due to the presence of aggregates (as those species detected in DLS, see above); however, it is important to indicate that some of the peaks could be also due to protein-column interactions even in the presence of high NaCl concentration.

Finally, we measured the fluorescence lifetimes of LrtA at different protein concentrations. The experimental decay of the total protein fluorescence was best fit to bi-exponential functions (Table 1, Figure S6), and thus, two lifetimes were observed; attempts to fit the experimental data to more than two exponentials led to an increase in the χ2. At any of the protein concentrations, the shortest lifetime corresponded to the largest amplitude (a1), and it did not change with the protein concentration. Interestingly enough, the longest lifetime (as well as its amplitude, a2) was concentration-dependent (Table 1). Furthermore, the <τ> showed also a concentration-dependence: going from a value of 6 ns (at the smallest concentration) to 1 ns at 98 μM. It is well-known that the intrinsic fluorescence lifetime of the first excited electronic singlet state does not change, but due to various quenching processes, changes in the environment around the fluorophores or even conformational changes in the molecular species, the measured lifetime is different from the intrinsic one [12,13]. Thus, even ruling out a possible fitting of the data to a monomer ↔ oligomer equilibria because we are observing the lifetimes of eight Tyr residues, each of them with a different environment, we can conclude that the concentration-dependence observed in the <τ> (Table 1) should have its origin in association-dissociation events.

**Table 1.** Fluorescence lifetimes of LrtA (100 mM, phosphate buffer (pH 8.0), with 500 mM NaCl) at 25 ◦C a.


<sup>a</sup> Errors are from fitting to a bi-exponential function.

It could be thought that the detected self-associated species could be random-oligomers; although we cannot rule out the presence of aggregates (from the DLS, SEC, and glutaraldehyde cross-linking results), however, there are at least three pieces of evidence suggesting that the self-associated species do not oligomerize un-specifically: (i) the DLS results show a linear dependence with the concentration (Figure 1B); (ii) the protein-dependent, and almost exponential, variation of the life-times; and, (iii) the presence of bands at particular molecular weights in the glutaraldehyde experiments. Therefore, those results, together with experiments from (GdmCl) chemical denaturations (see below, Section 2.3.), must be due to the presence of self-associated equilibria, involving the regions around some of the eight Tyr residues.

#### *2.2. LrtA Acquired a Native-Like Conformation Between pH 6.0 and 9.0*

We analyzed the structure of LrtA at varying pH to find out in which interval the protein acquired a native-like conformation. To this end, we used several biophysical techniques, namely, intrinsic and ANS fluorescence, CD and NMR. We used intrinsic fluorescence to monitor changes in the tertiary structure around its eight Tyr residues. Furthermore, ANS fluorescence was used to monitor the burial of solvent-exposed hydrophobic patches. We acquired far-UV CD spectra to monitor the changes in secondary structure. Finally, we acquired 1D 1H NMR spectra that show the presence of secondary and tertiary structure at physiological pH (see above, Section 2.1.). These spectra indicate (Figures S2 and S3) that the secondary and tertiary structures of the protein at pH 4.5 and 8.0 were completely different.

#### 2.2.1. Fluorescence

Intrinsic Steady-State Fluorescence and Thermal Denaturations—The fluorescence spectrum of LrtA at physiological pH showed a maximum at 308 nm, as expected for a polypeptide chain with fluorescent Tyr residues. The pH-dependence of the intrinsic <1/λ> showed two transitions (Figure 2A, left axis, filled circles). The first transition finished at pH 6.0, but we could not determine its p*K*<sup>a</sup> due to the absence of an acidic baseline. This transition was probably due to the titration of some of the seventeen Glu and/or twelve Asp residues of the LrtA sequence [18,19], which can alter the environment around some of the Tyr residues. However, we cannot rule out that it could be also due to the titration of some of the eight naturally-occurring His residues, taking place at an unusually low p*K*a. The second transition occurred at basic pH, starting at pH > 9.0, but, in this case, we could not determine the p*K*a due to the absence of a baseline at the highest pH values. This transition was probably due to the titration of at least some of the eight Tyr residues in the sequence. Therefore, the changes observed in fluorescence as the pH was changed could be due to titrations of specific residues around the Tyr residues, or alternatively, to conformational changes involving those fluorescent amino acids.

Thermal denaturations of LrtA were carried out at several pH values with a protein concentration of 9.8 μM, in protomer units. At pH values larger than 6.0, we observed an irreversible broad transition (Figure 2C, left axis, blank circles). Below pH 6.0, we did not observe any sigmoidal behavior, and we did not observe any sigmoidal transition at pH 13.0 either (Figure S7A). It could be argued that as fluorescence is intrinsically temperature-sensitive [12,13], we are not monitoring the denaturation of the protein. However, it must be kept in mind that fluorescence temperature sensitivity is linear (as observed at low pH values, Figure S7A; or in the native and unfolded baselines of the curve shown in Figure 2C), but it is not sigmoidal as observed in the denaturations at pH 7.0, with a midpoint around 45 ◦C (Figure 2C), or at pH 8.4 (Figure S7A). We also carried out experiments at LrtA concentrations of 5 μM, in protomer units (Figure 2C), and denaturation was also irreversible. Therefore, irreversibility was not associated with the amount of protein used during thermal denaturations.

ANS-Binding—At low pH, the ANS fluorescence intensity at 480 nm was large and decreased as the pH was raised (Figure 2A, right axis, blank squares), suggesting that LrtA had solvent-exposed hydrophobic regions. We could not determine the p*K*a of this titration due to the absence of an acidic baseline. The burial of solvent-exposed hydrophobic residues was complete at pH 6.0, as it happens with the transition observed by following the intrinsic fluorescence (see above). Since ANS reports on burial of hydrophobic surface, and therefore it monitors conformational changes, we must conclude that the protein had structural changes at acidic pH values; then, the variations monitored by intrinsic fluorescence (see above) at acidic pH must be associated with conformational changes due to the protonation of Asp and Glu residues (or the other amino acids described above).

In conclusion, our results indicate that, at low pH values, LrtA had solvent-exposed hydrophobic regions.

Solvent-Exposure of Tyr Residues Monitored by Iodide and Acrylamide Quenching—We carried out quenching experiments at pH 3.0, 7.0, and 11.0, because these are the three regions where we observed a different intrinsic fluorescence behavior of LrtA (Figure 2A). We used two quenching agents because of the charge effects probably occurring at extreme pH value with I−. We have assumed that the fluorescence lifetimes of the self-associated protein (for a fixed protein concentration) did not change in the whole pH interval. The *K*sv values for KI and acrylamide in the absence of denaturant were smaller than those measured in other proteins containing only Tyr residues [20,21] (Table 2). As a general trend, the *K*sv values of LrtA in the presence of acrylamide were smaller at acidic pH values than at physiological or basic ones; these differences could be due to the presence of higher-order self-associated species at the acidic pH values (as suggested by the ANS results (Figure 2A) and the CD data at low pH (Figure S4)). Furthermore, these results indicate that the structure of LrtA underwent some conformational changes at acidic pH (in agreement with results from intrinsic and ANS fluorescence, Figure 2A, and the NMR results, Figures S2 and S3). In the presence of GdmCl (guanidine hydrochloride), the *K*sv values were larger (either in KI or acrylamide) than those in the absence of denaturant (Table 2), suggesting that Tyr residues were more solvent-exposed.

**Figure 2.** pH-denaturation of LrtA: (**A**) Intrinsic (left axis, filled circles) and ANS (right axis, blank squares) fluorescence of LrtA, as the pH was modified. (**B**) Changes in the [-] at 222 nm as the pH was varied (filled circles). Inset: far-UV CD spectrum of LrtA at 100 mM phosphate buffer (pH 8.0), with 500 μM NaCl at 25 ◦C. (**C**) Thermal denaturations followed by intrinsic fluorescence (left axis, blank circles) at pH 7.0 and 5 μM (in protomer units) of LrtA, and raw ellipticity at 222 nm at pH 7.0 and 9.8 μM, in protomer units (right axis, filled circles).


**Table 2.** Quenching parameters for LrtA under several conditions at 25 ◦C.

<sup>a</sup> Not determined due to protein precipitation.
