Analysis of Site-specific Histidine Protonation in Human Prolactin^,

Abstract

The structural and functional properties of human prolactin (hPRL), a 23 kDa protein hormone and cytokine, are pH dependent. The dissociation rate constant for binding to the extracellular domain of the hPRL receptor increases nearly 500-fold over the relatively narrow and physiologic range from pH 8 to 6. As the apparent midpoint for this transition occurs around pH 6.5, we have looked towards histidine residues as a potential biophysical origin of the behavior. hPRL has a surprising number of nine histidines, nearly all of which are present on the protein surface. Using NMR spectroscopy, we have monitored site-specific proton binding to eight of these nine residues and derived equilibrium dissociation constants. During this analysis, a thermodynamic interaction between a localized triplet of three histidines (H27, H30 and H180) became apparent, which was subsequently confirmed by site-directed mutagenesis. After consideration of multiple potential models, we present statistical support for the existence of two negative cooperativity constants, one linking protonation of residues H30 and H180 with a magnitude of approximately 0.1, and the other weaker interaction between residues H27 and H30. Additionally, mutation of any of these three histidines to alanine stabilizes the folded protein relative to the chemically denatured state. Detailed understanding of these complex protonation reactions will aid in elucidating the biophysical mechanism for pH dependent regulation of hPRL’s structural and functional properties.

Prolactin (PRL) is a 23 kDa protein hormone and a member of the family of hematopoietic cytokines, which includes erythropoietin, granulocyte colony-stimulating factor, interleukin-6, and many others, but is most closely related both evolutionarily and functionally to human growth hormone (hGH) and placental lactogen (hPL) (1). Proteins in this family share a common four α-helical bundle structural topology and recognize a conserved family of type 1 cytokine receptors (2). In humans, PRL is secreted by pituitary lactotrophs under hypothalamic regulation, where it circulates in the bloodstream and acts distally as an endocrine hormone. Human PRL (hPRL) is also synthesized in many extrapituitary tissues, including breast, prostate, and the female reproductive tract, where it appears to act locally to regulate cellular growth and differentiation (3;4). Closely associated with this autocrine/paracrine function as a local growth factor or cytokine, hPRL has been implicated in the growth and development of human cancers arising from the same tissue sites (5-8). Interestingly, a number of post-translational modifications to PRL have been identified, including phosphorylated (9-13) and proteolytically-cleaved (14-17) variants, which appear to largely counter the pro-tumorigenic activities of the wild-type (WT) hormone.

We recently described a dependence of the structural and functional properties of hPRL on solution pH (18). The structural stability of the recombinant protein decreases from 7.6 kcal/mol at pH 8 to 5.6 kcal/mol at pH 6. More striking is the greater than 500-fold decrease in the equilibrium association constant for the extracellular domain of the PRL receptor over this same relatively narrow and physiologic pH range. The biologic consequences of these effects are currently undescribed, but many possibilities exist. A pH dependent conformational change and subsequent aggregation reaction have been postulated to drive packaging of the newly expressed hormone into the secretory granules of pituitary lactotrophs (19;20). The mechanisms for generating the phosphorylated and proteolytically-cleaved variants of hPRL are not well understood, but could be regulated by pH dependent destabilization of the protein. Proteolytic cleavage of hPRL by the endocytic enzyme cathepsin D has been proposed (21-25). As cathepsin D is activated by acidic environments around pH 5 - 6, perhaps concomitant structural destabilization of hPRL increases its susceptibility to proteolytic cleavage in a synergistic manner. Similarly, as phosphorylation of hPRL at S179 is complicated by the complete burial of its side chain within the helical bundle, acid-induced destabilization of the protein may allow phosphorylating enzymes access to the residue. Lastly, the biology of hPRL must be impacted by the dramatic decrease in receptor-binding affinity occurring between the physiologic pH range of around 7.3 - 7.5 to slightly more acidic values between 6.0 - 6.5. A similar shift in pH can be seen in extracellular fluid of malignant versus surrounding tissues (26), implying decreased receptor occupancy of hPRL in hypoxic tumors. As well, after endocytosis of hPRL-receptor complexes, the subsequent decrease in endosomal pH should induce dissociation of the hormone from the receptor.

We have begun to explore the biophysical basis for the above pH dependent regulation of the structural and functional properties of hPRL. As the observed perturbations occur largely over a pH range from 6 - 8, with an apparent midpoint to the transition around pH 6.5, protonation of the imidazole side chains of His residues are the most likely origin of the behavior. hPRL contains a surprising number of nine histidines, comprising 4.5% of its amino acid composition, in comparison to the lower 2% incidence of histidines in a survey of modern proteins (27). The closely related proteins hGH and hPL contain three and seven His residues, respectively. Displayed on the tertiary structure of hPRL (28;29) in Figure 1, His imidazoles are largely dispersed throughout the protein surface, with the exception of a single localized cluster of H27, H30 and H180, which forms part of the expected high affinity receptor-binding site (29). To better understand the protonation reactions occurring in hPRL around pH 6.5, we have measured site-specific pH titrations for each of the observable histidines using NMR spectroscopy at near physiologic solution conditions. Although a majority of His residues titrated according to a classic “Henderson-Hasselbach” model, the localized triplet of H27, H30 and H180 displayed more complex behavior. Given the likely importance of these residues in regulating pH dependent receptor-binding, a detailed analysis of their protonation reactions is warranted and presented here.

Figure 1.

Backbone ribbon diagram for the tertiary structure of hPRL (PDB code 1RW5, first member of the ensemble shown) with histidine side chains highlighted in red and labeled with their respective residue number.

Experimental Methods

Preparation of isotope-labeled recombinant proteins for NMR spectroscopy

Wild-type (WT) human prolactin (hPRL) was recombinantly expressed in BL21 DE3 Escherichia coli and purified from inclusion bodies as previously described (30). Expression vectors for the H27A, H30A and H180A hPRL site-directed mutants of hPRL were obtained from Dr. Charles Clevenger (Northwestern University) and protein was similarly prepared. Uniform ¹³C and ¹⁵N protein-labeling was achieved using growth in M9 minimal media, appropriately enriched with 1 g of ¹⁵N-NH₄Cl and 3 g of U-¹³C₆-glucose (Cambridge Isotope Laboratories, Inc., Andover, MA) per liter of bacterial culture. Selective ¹⁵N histidine-labeled WT and H180A hPRL were prepared by transformation of their expression vectors into a histidine auxotroph strain of BL21(DE3) E. coli developed (31) and generously provided by Dr. David Waugh (National Cancer Institute, Bethesda, MD). Similar M9 minimal media was supplemented with ¹⁵N histidine (Cambridge Isotope Laboratories, Inc.) and all other unlabelled amino acids (Sigma-Aldrich, Inc., St. Louis, MO) based on a previous protocol (32). After transformation, the auxotrophic bacteria were plated onto LB-enriched agar containing 0.2 g/L ampicillin and incubated overnight at 37°C. A single isolated colony was used to inoculate 50 ml of the minimal media, also with 0.2 g/L ampicillin, and again incubated overnight at 37°C. An aliquot of this culture was used to inoculate one liter of minimal media supplemented with 0.1 g of each unlabeled amino acid and 0.02 g of unlabeled histidine. After growth of this culture at 37°C until its absorbance at 600 nm reached ∼0.8, 0.1 g of ¹⁵N-histidine was added and allowed to grow for another ten minutes before inducing recombinant protein expression by the addition of IPTG to a final concentration of 200 mM. Finally, ¹⁵N-histidine hPRL was refolded and purified from inclusion bodies as described above with yields between 50 and 100 mg of purified protein.

For NMR spectroscopy, hPRL was concentrated to 0.6 mM (unless otherwise noted) and exchanged into 25 mM potassium-phosphate buffer at pH 7.5, 25 mM NaCl., 5% ²H₂O, 1 mM NaN₃ and 1 μM of the protease inhibitors leupeptin, pepstatin and PMSF. A series of samples at different pH values, typically within a range of 4 - 8, were prepared for ¹³C¹⁵N-enriched WT, H27A, H30A, H173A and H180A-hPRL and for ¹⁵N histidine-labeled WT and H180A-hPRL. The samples were put in susceptibility-matched NMR tubes (Shigemi, Inc., Allison Park, PA) for low volume samples and the pH was adjusted to the desired values with 10 - 50 μl of 0.1 M HCl or 0.1 M NaOH. The pH adjustment was performed adding HCl or NaOH drop by drop, waiting two minutes after each addition to let the pH reach equilibrium. Special care was taken in the preparation of lower pH samples to avoid protein precipitation and the sample was centrifuged if precipitation did occur. The resulting pH of each sample was determined at room temperature before and after NMR data collection. The pH was measured in the NMR tube with an Orion glass micro-pH meter combination electrode (Ag/AgCl) (Thermo Electron Corporation, Waltham, MA, model 9826BN) using an Accumet AP61 pH meter (Fisher Scientific, Pittsburgh, PA).

NMR spectroscopy

All NMR experiments were collected on a Varian INOVA 600 MHz spectrometer using a 5 mm triple resonance probe equipped with triple-axis pulsed magnetic field gradients and utilized pulse sequences from the Varian BioPack User Library (Varian Inc., Palo Alto, CA). Histidine side-chain resonance assignments were determined using ^lH-¹³C HSQC and ¹H-¹⁵N HSQC NMR experiments acquired at 35 °C. The ^lH-¹³C HSQC spectra were recorded with a spectral width of 9000.9 Hz and 2048 total (real plus imaginary) points in the ¹H dimension, a spectral width of 4100 Hz and 256 quadrature (real plus imaginary) increments in the ¹³C dimension, signal-averaged over 128 transients, and with a relaxation delay of 1 sec. For the His-tautomeric ¹H-¹⁵N HSQC NMR experiments, the BioPack library “gNhsqc” pulse-sequence was modified as previously described (33) to allow GARP ¹³C decoupling during the evolution of the ¹⁵N chemical shift and spectra were recorded with a spectral width of 8000 Hz and 1024 total (real plus imaginary) points in the ¹H dimension, a spectral width of 2500 Hz and 128 quadrature (real plus imaginary) increments in the ¹⁵N dimension, signal-averaged over 256 transients, and with a relaxation delay of 1 sec. In order to detect long-range couplings between imidazole-ring ¹⁵N nuclei and the ¹H_δ1 and ¹H_ε2 nuclei, the INEPT relaxation delay was adjusted to correspond to an effective ²J_NH coupling constant of 22 Hz, which represents a compromise between the theoretical value of 5 - 10 Hz (absolute value) and losses in signal-to-noise due to transverse (R₂) relaxation (33). The spectra were recorded for the WT and mutant hPRL at pH values ranging from 3.97 to 8.30. All NMR spectra were processed using NMRPipe (34), with subsequent display and analysis in Sparky (35). Chemical shifts were referenced indirectly to external DSS at 0.00 ppm, with heteronuclear dimensions referenced using their relative gyromagnetic ratios (36).

Curve fitting and statistical analysis

NMR spectra were recorded at many pH values between pH 4.0 and 8.5 for WT and mutant variants of hPRL. Signs of protein precipitation were noticeable below pH 4.0. Equilibrium thermodynamic parameters were derived for eight of the nine histidines in hPRL from the pH dependencies of their NMR chemical shifts. Using the Scientist 3.0 Software Package (Micromath Research Inc., St. Louis, MO), the observed, pH-dependent chemical shifts (δ_obs) were fitted to a simple linear combination of the fully protonated (δ_AH+) and deprotonated (δ_A) states weighted by their relative populations:

$${\delta }_{\mathit{obs}}={\delta }_{{\mathit{AH}}^{+}}{f}_{{\mathit{AH}}^{+}}+{\delta }_A{f}_A$$ (Eq. 1)

For each of the independently titrating residues (H46, H59, H138, H173 and H195),

$$f_{A{H}^{+}}=\frac{{10}^{p{K}_a-pH}}{1+{10}^{p{K}_a-pH}}$$ (Eq. 2)

$$f_A=1-f_{A{H}^{+}}$$ (Eq. 3)

For the triplet of coupled titration sites (H27, H30 and H180) modeled in Figure 2, a binding polynomial (37) was written as

$$Z=1+K_{27}H+K_{30}H+K_{180}H+K_{27}{K}_{180}{H}^2+c_{27,30}{K}_{27}{K}_{30}{H}^2+c_{30,180}{K}_{30}{K}_{180}{H}^2+c_{27,30}{c}_{30,180}{K}_{27}{K}_{30}{K}_{180}{H}^3$$ (Eq. 4)

where, H = 10^-pH and K_x = 10^pK_a,X. The fractions of protonated and unprotonated states with respect to any single residue “X” were expressed as

$$f_{X{H}^{+}}=\frac{d\left(\ln Z\right)}{d\left(\ln K_X\right)}=\frac{dZ}{d{K}_X}\frac{K_X}{Z},\text{and}$$ (Eq. 5)

$$f_X=1-f_{X{H}^{+}}$$ (Eq. 6)

Figure 2.

Thermodynamic model for protonation of the imidazole rings of H27, H30, and H180 in hPRL, where the superscript denotes the state of each residue. The linkage coefficients and acid dissociation constants describing each deprotonation reaction are shown above their respective arrows.

A variety of possible thermodynamic linkages were considered for the triplet of H27, H30 and H180, detailed in Results section below. Fixing the two cooperativity constants (c_27,30 and c_30,180) at unity during global fitting of the data simulated an independent titration model for these three residues. Alternatively, allowing either or both of the cooperativity constants to float during fitting evaluated possible thermodynamic linkages between either residues H27 and H30 alone, residues H30 and H180 alone, or both pairs of residues simultaneously.

For the statistical analysis presented in Tables 2 and 3 of the Results, F-statistics were calculated as follows;

$$F=\frac{SS{E}_1-SS{E}_2}{P_2-P_1}/\frac{SS{E}_2}{n-P_2}$$ (Eq. 7)

where SSE₁ and SSE₂ represented the sum-squared errors (or residuals) between the experimental data and their best fit values for the two models under comparison. The more simplistic model #1 was considered nested within (or a special case of) the more complex model #2. The number of fitting parameters for each model was represented by P₂ and P₁, where P₂ must be greater than P₁, and n was the number of independent experimental data points used to fit both models. Note that by definition the more complex model provides a better fit to the data (i.e. SSE₁ > SSE₂) and the F-statistic tests whether this improvement in the fit was statistically significant by comparison to corresponding “critical” values derived from the F-distribution, based on a desired probability threshold and the appropriate degrees of freedom in the comparison, in this case (P₂ - P₁) and (n - P₂). Threshold probability values in Table 3 for each calculated F-statistic were calculated in Excel (Microsoft Corporation, Seattle, WA) using the FDIST command.

Table 2.

Summary of the “summed squared errors” (SSEs) and corresponding statistical parameters for each of the binding models considered. SSEs are separately calculated for each independent variable along with the total error

	(1) Independent	(2) c27,30	(3) c30,180	(4) c27,30 & c30,180	N (data pts.)
H27 1Hε1 SSEs	0.024742	0.008757	0.024742	0.016949	34
H30 1Hε1 SSEs	0.087695	0.014108	0.015538	0.012907	34
H180 1Hε1 SSEs	0.058582	0.058582	0.014959	0.018397	34
Total SSEs	0.17102	0.081448	0.05524	0.048253	102
No. of Params	9	10	10	11

Table 3.

F-statistic calculations comparing each binding model with corresponding probability values in parentheses.¹

H27 Hε1	(2) H27-H30 Interacting	(3) H30-H180 Interacting	(4) H27-H30 & H30-H180 Interacting
(1) All Independent	43.8 (7.6 × 10-7)	Equivalent2	5.3 (0.013)
(2) H27-H30 Interacting			Rejected3
(3) H30-H180 Interacting			10.6 (0.004)

H180 Hε1	(2) H27-H30 Interacting	(3) H30-H180 Interacting	(4) H27-H30 & H30-H180 Interacting
(1) All Independent	Equivalent2	70.0 (1.4 × 10-8)	25.2 (1.6 × 10-6)
(2) H27-H30 Interacting			50.2 (3.2 × 10-7)
(3) H30-H180 Interacting			Rejected3

H30 Hε1	(2) H27-H30 Interacting	(3) H30-H180 Interacting	(4) H27-H30 & H30-H180 Interacting
(1) All Independent	125.2 (5.2 × 10-11)	111.5 (1.7 × 10-10)	66.6 (2.7 × 10-10)
(2) H27-H30 Interacting			2.1(0.16)
(3) H30-H180 Interacting			4.7 (0.041)

All Data	(2) H27-H30 Interacting	(3) H30-H180 Interacting	(4) H27-H30 & H30-H180 Interacting
(1) All Independent	101.2 (1.7 × 10-16)	192.8 (2.66 × 10-24)	115.8 (9.9 × 10-26)
(2) H27-H30 Interacting			62.6 (5.8 × 10-12)
(3) H30-H180 Interacting			13.2 (4.7 × 10-4)

¹The first three boxes represent each individual dependent variable used during fitting and the last considers all data.

²No improvement in fitbetween models.

³More complex model has poorer fit.

Monitoring urea unfolding of hPRL using intrinsic Trp fluorescence

A 9.8 M urea (American Bioanalytical, Natick, MA, ultra-pure urea) stock solution was prepared and its concentration confirmed from refractive index measurements using a hand-held refractometer (Atago U.S.A., Bellevue, WA, model R5000). A series of 24 buffered urea solutions were prepared in quadruplicate in a 96 deep-well tray using an automated reagent dispenser (Thermo Electron Corporation, Waltham, MA, model Multidrop DW) controlled via the serial port of a personal computer. All urea solutions contained 25 mM NaCl and were buffered to their reported pH with 25mM potassium phosphate buffer. All pH measurements were taken with an Accumet AP61 hand-held pH meter (Fisher Scientific, Pittsburgh, PA) and Orion glass micro pH combination electrode (Ag/AgCl) (Thermo Electron Corporation, Waltham, MA, model 9826BN). Protein solutions were manually transferred into the deep-well plates containing the buffered urea solutions using a multi-channel pipette to a typical final concentration of 6 μM, after which the deep-well plates were sealed with adhesive aluminum film, manually tumbled to stir and allowed to equilibrate for a minimum of two hours at room temperature (23 °C) prior to fluorescence readings. Protein/urea solutions were manually transferred to quartz cuvettes in a Cary Eclipse fluorescence spectrophotometer (Varian Instruments, Walnut Creek, CA) employing both temperature control and stirring with micro stir bars. Fluorescence measurements were carried out with an excitation wavelength of 280 nm with a 20 nm slit width and detection at 325 nm with a 5 nm slit width.

Results

Assignment of Histidine imidazole ring chemical shifts

^lH-¹³C HSQC and ¹H-¹⁵N HSQC spectra of uniformly ¹³C¹⁵N- and ¹⁵N-His labeled hPRL recorded at several pH values provided the data for the titration curves of the ¹H, ¹³C and ¹⁵N nuclei in the histidine imidazole rings. The assignment of each histidine ¹H(¹³C)_ε imidazole chemical shifts signals in the ^lH-¹³C HSQC spectra was achieved using uniformly ¹³C,¹⁵N-labeled His to Ala hPRL mutants and pre-existing assignments available in the BMRB (accession codes: 5599 and 6643). These ¹H_ε chemical shift values from ^lH-¹³C HSQC spectra allow us to assign unequivocally the ^I5Nε ^I5N_δ and ¹Hδ histidine chemical shifts in the ¹H-¹⁵N HSQC hPRL spectra. The assigned chemical shifts are given as Supporting Information and also have been deposited in the BioMagResBank (http://www.bmrb.wisc.edu), BMRB accession number 15773.

Identification of the histidine tautomeric states

Previous investigators have recognized how histidine side-chains have distinctive ¹⁵N NMR chemical shifts depending on the protonation state of the imidazole ring (33;38). In the neutral state, the unprotonated nitrogen is located at higher NMR chemical shifts (up to 80 ppm) compared to its protonated counterpart. As the neutral imidazole ring becomes protonated with lowering of solution pH, the chemical shift of the originally unprotonated nitrogen migrates to progressively lower values. Figure 3 displays how these differences in chemical shifts, along with the relative intensities of the ²J_NH-coupled ¹⁵N_δ1-¹H_ε1 and ¹⁵N_ε2-¹H_δ2 crosspeaks in a ¹H-¹⁵N HSQC NMR spectrum, allow the identification of the His tautomeric state and correct chemical shift assignment of these four nuclei. In a neutral imidazole ring, depending on the location of the nitrogen-bonded hydrogen, two tautomeric states are possible. The most common tautomeric state (N_ε2-H) has a pattern as shown in the upper left of Figure 3. The ¹H_ε1 correlates with two ¹⁵N frequencies, one at ∼167.5 ppm corresponding to the protonated nitrogen and one at ∼249.5 ppm corresponding to the unprotonated nitrogen. The ¹H_δ2 correlates only with a nitrogen frequency corresponding to one protonated nitrogen as the ³J(¹⁵N_δ1-¹H_δ2) coupling is too small (1 - 2 Hz) to result in an observable cross-peak. This pattern enables the unambiguous assignment of the nuclei and at the same time specifies which nitrogen is protonated. In the less common N_δ1-H tautomeric state shown in the upper right of Figure 3, the ¹⁵N_δ1 is protonated and the ¹⁵N_ε2-¹H_δ2 cross-peak is located at 249.5 ppm for the ¹⁵N chemical shift.

Figure 3.

Top: Schematic diagram showing the nomenclature used to describe the two possible tautomeric states of the neutral imidazole ring of histidines, termed N_ε2-H and N_δ1-H, and the expected pattern of long-range (²J_NH) correlations in the ¹H-¹⁵N HSQC NMR spectrum [adapted from Pelton et al., 1993]. Bottom: ¹H-¹⁵N HSQC spectrum of ¹⁵N His-labeled hPRL in 25 mM potassium phosphate buffer at pH 7.5 and 25 mM NaCl collected at 35°C. Histidine cross peak patterns are indicated for each residue. Tautomeric assignments for each residue are listed under the representative spectra. The tautomeric states of H57, H97, H138 and H195 are not assigned.

Shown at the bottom of Figure 3 is the ¹H-¹⁵N HSQC spectrum collected on ¹⁵N His-labeled WT-hPRL at pH 7.5, with individual histidine spin systems noted. By simple inspection of their patterns we conclude that, at pH 7.5, H27, H46, H59, H173 and H180 are in the more common N_ε2-H tautomeric state; whereas, H30 adopts the less common N_δ1-H tautomeric state. Spin systems for H59, H97, H138 and H195 were either missing or incomplete, precluding assignment of their tautomeric states. Although the ¹H_δ2 and ¹H_ε1 nuclei detected in this experiment do not undergo rapid exchange with solvent, they are susceptible to line broadening from rapid amide hydrogen exchange elsewhere in the imidazole ring or potential conformational exchange processes, which we presume are responsible for the missing resonances. Lastly, Figure 4 shows superposed ¹H-¹⁵N HSQC spectra of ¹⁵N His-labeled WT-hPRL ranging from pH 4.5 to 8.0. The changes in ¹H and ¹⁵N chemical shifts versus pH are generally linear, consistent with perturbations due to only a single pH-dependent reaction, with the dramatic exception of the ¹⁵N_δ1 nucleus of H30. In the spectral expansion shown for WT hPRL, the H30 ¹H_ε1-¹⁵N_δ1 crosspeak follows a highly curved trajectory, indicating a dependence on more than one pH-dependent effect. We hypothesize that, given the close relationship between the H30 and H180 imidazole rings in the tertiary structure of hPRL, the H30 ¹⁵N_δ1 nucleus must detect protonation of both residues approximately equally. This was confirmed by reversion to a linear trajectory in ¹⁵N His-labeled H180A hPRL, which can be seen in the second expansion at the bottom right of Figure 4.

Figure 4.

Overlaid ¹H-¹⁵N HSQC NMR spectra of ¹⁵N-his labeled WT and H180A PRL from pH 4.5 to 8.0. The ¹H-¹⁵N chemical shift changes with pH generally follow linear trajectories as illustrated by the black trace, with low pH peaks at the start of the trajectories in the upper left. Spectral expansions are shown at the bottom depicting the curved trajectory for the H30 ¹H_ε1-¹⁵N_δ1 crosspeak in WT hPRL (left), which reverts to a linear trajectory in H180A hPRL (right).

pH-dependent titration of His NMR chemical shifts for WT, H27A, H30A, H173A and H180A-hPRL

Figure 5 plots the ¹H_ε1 and ¹³C_ε1 NMR chemical shifts against solution pH for eight of the nine His residues in hPRL. Titrations are superposed for both the WT protein and multiple His to Ala mutants in order to assess the effect of the mutations on the other residues. Not shown are the analogous pH titration curves for the five additional imidazole ring nuclei (i.e., ¹⁵N_ε2, ¹⁵N_δ1, ¹H_δ2, and ¹H_ε2), which were also collected for a majority of the residues. The ¹H_ε1 and ¹³C_ε1 nuclei provided the largest number of data points and the most complete titration curves. For the five residues shown at the top of the figure, there is very good agreement of the titration data across the large number of independent NMR samples, which represent different protein preparations, labeling schemes and even mutants of hPRL. We compared simultaneous fitting of all six nuclei together, versus ¹H_ε1 and ¹³C_ε1 together, and versus ¹H_ε1 chemical shifts on their own and, for all residues, there were no significant differences in the pKa values obtained.

Figure 5.

The ¹H_ε1 and ¹³C_ε1 NMR chemical shifts are plotted against solution pH for eight of the nine His residues in hPRL. Titrations are superposed for both WT and multiple His to Ala mutants of hPRL.

Four of the nine His residues in hPRL fit well to the simplest model of a single, non-interacting pH titration site. As well, in all of the His to Ala mutations investigated, pH titrations for these residues overlapped well with the WT protein. Hence, we believe that H46, H138, H173 and H195 titrate in classical manner, independently from other residues. Representative, fitted titration curves for the ¹H_ε1 and ¹³C_ε1 nuclei of H195 are shown at the top of Figure 6. Summarized in Table 4, the fitted pKa values for H46, H138, and H195 fall within the typical range of 6.0 - 6.5; whereas, the pKa for H173 is unusually low at 5.0. This residue is buried in tertiary structure of hPRL. Protonation of H173 would result in the highly unfavorable burial of an uncompensated positive charge, most likely resulting in disruption of the local, favorable structural interactions in order to allow hydration of the charged residue. Therefore, the neutral state of H173 is effectively stabilized, evidenced by its more acidic pKa.

Figure 6.

Experimental NMR chemical shifts and best-fit theoretical curves for histidine imidazole ring nuclei in WT hPRL as a function of pH. Panels (a) and (b), respectively, display ¹H_ε1 and ¹³C_ε1 NMR chemical shifts for H195 fitted to a classic, non-interacting model of protonation. In panel (c), ¹H_ε1 chemical shifts for residues H30 (blue), H180 (mauve) and H27 (red) are shown with the best-fit curves for both the fully independent model (dashed line, c_27,180 = c_30,180 = 1, model #1 in Table 1) and the fully interacting model (solid line, c_27,30 = 0.42 and c_30,180 = 0.14, model #4 in Table 1) for each residue, colored to match the experimental data.

Table 4.

Summary of the best fit binding parameters and the corresponding 95% confidence intervals

Value	Best Fit	95% Confidence Interval
H46 pKa	6.56	6.51-6.58
H59 pKa	5.84	5.71-5.97
H138 pKa	5.82	5.77-5.87
H173 pKa	4.99	4.91-5.07
H195 pKa	5.91	5.87-5.96
H27 pKa	6.64	6.49-6.79
H30 pKa	6.00	5.67-6.33
H180 pKa	5.96	5.83-6.09
c27, 30	0.42	0.03-0.82
c30, 180	0.14	0.02-0.27

In contrast to the above four independently titrating residues, our data suggest a strong interaction between H27, H30 and H180. Firstly, inspection of Figure 5 shows how mutation of either H30 or H180 strongly perturbs the pH titration curve of the other. Note that an interaction was already suspected based upon the unusual pH dependent trajectory of H30’s ¹⁵N_δ1 resonance in Figure 4. Similarly, titration curves for H27 and H30 in Figure 5 also show small, but significant, perturbations when the other is mutated; in contrast, H27 and H180 do not appear to have a significant direct interaction. Additional, independent evidence for these direct interactions can be derived from an analysis of their individual pH titration curves. In Table 1, we compare the results of simultaneous fitting the ¹H_ε1 titration curves for all three residues according to the thermodynamic model depicted in Figure 2, but considering four separate situations: (1) independent titration of all three residues, (2) linkage between H27 and H30 only, (3) linkage between H30 and H180 only, and (4) simultaneous linkage between both pairs of residues. At the bottom of Figure 6, best fit curves for the fully independent (1) and the fully interacting (4) models are compared to the experimental data.

Table 1.

Summary of fitting the ¹H_ε1 NMR chemical shifts for the triplet of H27, H30 and H180 according to the model of linked protonation sites in Figure 2

Model	H27 pKa	H30 pKa	H180 pKa	c27,30	c30,180
(1) All Independent	6.58 (6.36 - 6.80)	5.28 (5.06 - 5.51)	5.86 (5.67 - 6.05)	1 (fixed)	1 (fixed)
(2) H27-H30 Interacting	6.66 (6.48 - 6.84)	6.10 (5.87 - 6.34)	5.85 (5.73 - 6.00)	0.07 (0.02 - 0.13)	1 (fixed)
(3) H30-H180 Interacting	6.58 (6.45 - 6.71)	5.70 (5.54 - 5.85)	5.96 (5.82 - 6.10)	1 (fixed)	0.10 (0.01 - 0.20)
(4) H27-H30 & H30-H180 Interacting	6.64 (6.49 - 6.79)	6.00 (5.67 - 6.33)	5.96 (5.83 - 6.09)	0.42 (0.03 - 0.82)	0.14 (0.02 - 0.27)

Visually, it is evident from Figure 6 that the ¹H_ε1 titration curves for H30 and H180 do not fit well to the simplest model of independent protonation. The need for a more complex model to describe the titration of H27, H30 and H180 can be more rigorously supported by the use of the F-statistic. Calculation of F-statistics requires the residual sum-squared errors (SSEs) between the experimental data and their best fit values along with the degrees of freedom during fitting, which is derived from both the number of fitting parameters as well as the number of independent pieces of data. In order to more accurately describe the true degrees of freedom during the comparison of our three titration models, we have chosen to include only the ¹H_ε1 NMR chemical shift data, as the data from this nucleus is the most complete for this triplet of residues. Elsewhere, we have simultaneously fit multiple NMR chemical shifts derived from a single sample (e.g. ¹H_ε1 and ¹³C_ε1) to derive individual pKa values for each residue. Although inclusion of multiple chemical shifts for each residue should not degrade the accuracy of the best-fit value, it does distort estimation of the statistical errors by over-estimating the number of truly independent pieces of data. In these experiments, the largest source of random error likely derives from the preparation of the individual NMR samples and, to a lesser degree, the separate acquisition of each NMR spectrum. These errors are contained equally in each of the chemical shift values acquired from each NMR spectrum collected on a single sample and, thus, cannot be considered truly independent measurements. This lack of independence is evidenced by a strong correlation between the residual errors (i.e. the difference between the experimental and fitted values) for each of the six nuclei in a single residue (not shown). By restricting our analysis to a single NMR chemical shift per residue, we are largely able to avoid this over-estimation of the statistical degrees of freedom and be more confident in the applicability of the F-statistic to identify and reject overly complex and statistically non-significant models. However, because we have chosen to globally fit the NMR chemical shifts for all three residues simultaneously (necessary for consideration of thermodynamic linkage between residues), we cannot avoid utilizing three independent variables from each NMR spectrum. Therefore, we have chosen to determine SSEs (Table 2) and F-statistics (Table 3) for each of the individual, dependent variables as well as their total. Neither approach is ideal, as the statistics for the isolated dependent variables ignore the compromises occurring during global fitting of the three residues simultaneously and, as well, the statistics for the overall (total) fit likely overestimates the number of independent data points.

Beginning at the top of Table 3, F-statistics for the isolated H27 ¹H_ε1 dependent variable support the inclusion of thermodynamic linkage between H27 and H30, despite the small visual difference between the fitted curves for H27 in Figure 4. Note that the additional inclusion of linkage between H30 and H180 did not improve the fit for H27 when considered on its own, which should be expected. Similarly, when considering only the quality of fit for the H180 ¹H_ε1 dependent variable, the F-statistics clearly support the presence of negative cooperativity between H30 and H180, but consideration of linkage between H27 and H30 does not improve agreement for H180. The situation for the H30 ¹H_ε1 dependent variable is more complex. The quality of fit for H30 is improved in all cases by the addition of thermodynamic linkage between either pairs of residues (see Table 2) and the corresponding F-statistics are significant at a probability >95% for all comparisons except for one (see Table 3). Lastly, when all of the experimental data is considered together inclusion of both thermodynamic linkage parameters greatly improves the quality of fit and the corresponding F-statistics are highly significant in all cases, although the degree of statistical confidence is likely overestimated by the lack of complete independence between the data for each residue. Regardless, when all of the evidence is considered in aggregate, the existence of negative cooperativity between H27 and H30 and also between H30 and H180 is clear. The quality of fit for H27 and H30 depend on their linkage and similarly the quality of fit for H30 and H180 depend on their linkage. This statistical analysis is supported by the deviation in the NMR chemical shift titration curves for each residue in the His to Ala mutants shown in Figure 5.

We present in Table 4 our current, best assessment of the thermodynamic parameters describing the protonation reactions of 8 of the 9 His residues in hPRL. Note that although there can be no doubt about the existence of negative cooperativity between residues H27, H30 and H180, we cannot define the individual cooperativity constants with precision. In our various attempts to globally fit the selected NMR data to various models, these constants displayed the greatest variability in their best fit values. However, interestingly, their product (c_27,30*c_30,180) remained roughly constrained to the order of ∼ 0.1. This conservation was likely required to properly account for the reduced slope of the transition seen in the H30 ¹H_ε1 titration curve compared to the non-cooperative case. Although our data appears to lack statistical power to precisely define their individual values, there is sufficient evidence to conclude that the negative cooperativity between H30 and H180 is stronger than that for H27 and H30. First, the H30 ¹⁵N_δ1 NMR chemical shift in Figure 4 appears to depend on titration of both the H30 and H180 imidazole rings. Secondly, a more dramatic change is seen in the titration curves of both H30 and H180 when either is mutated as compared to H27 and H30. This conclusion is further supported by the closer relationship between H30 and H180 in the tertiary structure of hPRL compared to H27 and H30. Therefore, we have chosen to include the best fit values for these cooperativity constants in Table 4 as their relative magnitudes are in agreement with our assessment. However, uncertainty in their precise values is appropriately reflected in their relatively large 95% confidence limits.

Throughout the above description, we have purposely avoided discussion of the results for residues H59 and H97. In all of the above described NMR spectra of WT hPRL, NMR signals for the imidazole ring of H97 are either extremely weak or unobservable. Interestingly, when H59 was mutated to Ala (in order to confirm its assignment) a new strong NMR signal was observed, which was assigned as belonging to H97. Although not in direct contact, H59 and H97 are nearby structurally and both are part of a small cluster of aromatic residues between the two long loops and the 2^nd and 4^th helices. The most likely explanation for our results is that in WT hPRL the H97 imidazole NMR signals undergo chemical exchange on a timescale similar to their difference in NMR chemical shifts, most likely due to conformational mobility within this aromatic cluster. Mutation of H59 to Ala would be expected to structurally loosen these interactions and, thereby, increase the mobility of H97 to a faster exchange regime where it becomes observable. We also notice that the titration curve for H59 displays slight, systematic deviations from a model of independent titration similar to H30 and H180. We are currently investigating the structural and thermodynamic linkage between H59 and H97, which will be detailed in a future publication.

Contribution of selected histidines to global unfolding of hPRL

The individual contribution of each histidine within the interacting triplet to the global stability of hPRL was investigated using chemical denaturation. Figure 7 plots normalized, intrinsic Trp fluorescence across a wide range of urea concentrations for WT, H27A, H30A and H180A-hPRL. Approximately two-state global unfolding of the proteins is indicated by the sudden and sharp transitions between native-like and denatured protein fluorescence. Shifts in the global unfolding transitions to lower or higher urea concentrations indicate decreases or increases in the global unfolding free energies (ΔG_unf), respectively. The experimental data have been fit to a theoretical description of unfolding as previously described (18) to derive quantitative descriptions of [urea]_mid, the urea concentration at the midpoint of the transition, and the ΔG_unf for each protein. The data clearly indicate that the presence of each individual imidazole ring for residues H27, H30 and H180 is overall destabilizing towards protein folding.

Figure 7.

Chemical denaturation curves of WT, H27A, H30A and H180A-hPRL are presented in order to compare their global stabilities. Normalized, intrinsic Trp fluorescence is plotted as a function of urea concentration and the data fitted to theoretical curves, from which are derived the displayed [urea]_med and ΔG_unf values.

Discussion

Proteins often display pH dependence to their structural stability and functional behavior. Such pH dependence is ultimately mediated by the thermodynamics of protonation reactions of ionizable groups, which are located primarily in the side-chains of Asp, Glu, His, Arg and Lys residues. The pH dependence of any equilibrium between protein structural states (and thus functional states) can be completely described by detailing all the changes in protonation equilibrium constants (i.e. the pKa values) for all the individual ionizable groups. NMR spectroscopy is uniquely suited to describing site-specific protonation and has been routinely utilized to derive residue-specific pKa values (39-49), during which the identification of thermodynamic linkage between protonation sites is not uncommon. We recently discovered dramatic pH dependence to both the structural stability and receptor-binding affinity of hPRL, over the relatively narrow physiologic pH range of 6 - 8 (18). As the initial step to detailing the molecular basis of this pH dependence, we have analyzed the site-specific (or microscopic) titration curves for 8 of the 9 His residues in hPRL. In doing so, we describe a novel approach for combining a statistical analysis of NMR titration curves with site-directed mutagenesis to identify thermodynamic linkage, or cooperativity, between the protonation reactions of individual residues.

The protonation equilibrium constants (pKa values) for the histidines in hPRL all fall in a relatively typical range from 5.8 to 6.6, with the exception of H173 whose unusually acidic pKa of 5.0 is likely the result of the complete burial of its imidazole ring. The unusual and somewhat novel finding is the presence of thermodynamic linkage between the protonation reactions of H27, H30 and H180, corresponding to an overall perturbation in the free energy of protonation of approximately 1.4 kcal/mole. We considered the possibility of hydrogen bonding between these coupled imidazole rings, but this was not supported by their positioning in either the NMR (29) or the X-ray (28) structure of hPRL. Shown in Figure 8, H27 and H30 are not close enough to support hydrogen-bonding (> 3 Å) and H30 and H180 are stacked vertically in a staggered manner consistent with favorable aromatic stacking interactions. Therefore, based upon this positioning, it appears more likely for the negative cooperativity between protonation reactions to be governed primarily by unfavorable electrostatics or, possibly, interference with aromatic stacking.

Figure 8.

Positions of the H27, H30 and H180 imidazole rings in the X-ray crystallographic structure of a hPRL variant (PDB code: 2Q98).

The thermodynamic interactions between the H27, H30 and H180 described here are likely to contribute towards the biological function of hPRL, based on both their presumed structural location within the high affinity receptor-binding site (29) and an evolutionary argument regarding their unfavorable contribution to protein stability. First, although the structure of hPRL bound to its receptor has not been described, the related protein hGH also binds to the hPRL receptor (but only in the presence of Zn²⁺) and a high resolution structure of this complex has been reported (50). H27 and H30 from hPRL are conserved in hGH. Structurally, these homologous residues are located within the binding interface and directly coordinate the Zn²⁺ cation required for receptor recognition. H180 is homologous to an Asp in hGH that is also located within the interface but does not interact directly with the Zn²⁺ cation. Site-specific mutation of H27, H30 or H180 has been shown to decrease either the receptor-binding affinity or the mitogenic effects of hPRL (51), further supporting their expected location within the high affinity receptor-binding site. We also report here that single-site mutation of either H27, H30 or H180 in hPRL stabilizes the folded protein relative to its chemically denatured state. This is an unusual finding as a majority of amino acids in proteins appear to be evolutionarily optimized towards their contribution to global stability (52). Conformational strain associated with conserved residues, such as seen here, is often an indicator of functional importance. Although the current results do not clearly define the structural basis of the previously reported pH dependent behavior (18), they do form a foundation for ongoing studies on the contribution of specific His residues to the structural stability and receptor-binding properties of hPRL. In the future, we hope that by directly measuring site-specific changes in pKa values associated with receptor-binding and protein unfolding, a better understanding of the biophysical basis of hPRL’s pH dependent behavior will ultimately result.

The abbreviations used are

hPRL

human prolactin

hGH

human growth hormone

hPL

human placental lactogen

WT

wild type

NMR

nuclear magnetic resonance

Asp

aspartate

Glu

glutamate

Lys

lysine

Arg

arginine

His

histidine

Ala

alanine

DSS

2,2-Dimethyl-2-silapentane-5-sulfonate sodium salt

PPM

part per million

HSQC

heteronuclear single quantum coherence

2D

two-dimensional

INEPT

insensitive nuclei enhanced (by) polarization transfer

SSE

sum-squared error

IPTG

Isopropyl β-D-1-thiogalactopyranoside

PMSF

phenylmethylsulphonyl fluoride

GARP

Globally Optimized Alternating Phase Rectangular Pulse