Enhancing Sensitivity of Nuclear Magnetic Resonance in Biomolecules: Parahydrogen-Induced Hyperpolarization in Synthetic Disulfide-Rich Miniproteins

Abstract

Hyperpolarization of small peptides by parahydrogen-induced polarization (PHIP) to increase the sensitivity of nuclear magnetic resonance (NMR) techniques is well established, while its application to larger biopolymers is still a mainly unexplored area. A particular challenge is the presence of folding-essential disulfide bridges. They tend to form metal complexes, thus hampering catalytic hydrogenation, a prerequisite for PHIP. We applied the PHIP technique to enhance NMR signal intensity in cystine-knot miniproteins—highly ordered peptide architectures covalently stabilized by three disulfides. To achieve PHIP, we introduced an l-propargyl tyrosine label at different positions in three synthetic open-chain variants of a natural trypsin inhibitor MCoTI-II. For the folded cystine knot, we observed NMR signal enhancements of up to 499 in methanol, 307 in a D₂O–methanol mixture, and 964 for the cysteine-bearing reduced precursor. Trypsin inhibition assays elucidated that introducing a PHIP label into the terminal regions is preferable to alterations within the functional loop to preserve bioactivity. Substitution of the native tyrosine resulted in the highest bioactivity. A drastic reduction in PHIP enhancement was observed in the presence of trypsin due to slower hydrogenation, conditioned by the accessibility of the label within an enzyme–inhibitor complex.

Introduction

NMR spectroscopy is an efficient analytical tool to understand the structural and dynamic characteristics of materials. However, its sensitivity is limited due to the small energy differences in the nuclear spin states. To overcome this limitation, different hyperpolarization methods are used, which strongly enhance the NMR signal. One of the most cost-efficient, easiest, and least demanding but very effective hyperpolarization techniques is parahydrogen-induced polarization (PHIP), which allows for signal enhancement (SE) via hydrogenation of unsaturated labels with parahydrogen gas.^1,2 In the standard PHIP experiment, sometimes called hydrogenative (h-PHIP) to distinguish it from the nonhydrogenative SABRE (Signal Amplification By Reversible Exchange)^3,4 variant, hyperpolarized compounds are generated from unsaturated precursors. In contrast, in SABRE, hyperpolarization occurs by transfer of polarization by cross-relaxation inside a reversibly formed transition metal complex between the catalyst, parahydrogen, and a substrate. Because of its cost efficiency and versatility, PHIP of bioactive compounds is a fast-developing field with potential application to larger and larger targets of biomedical and biological relevance.⁵⁻⁷

PHIP hyperpolarization has been used as a magnetic resonance imaging contrast mechanism for metabolic imaging to control tumor therapy^8,9 with pyruvate,¹⁰⁻¹⁴ which is crucial to energy metabolism, and fumarate,^9,15 which takes part in the citric acid cycle being the most prominent examples. Furthermore, hyperpolarized compounds can help to study biological reactions and are applied in drug development (see review⁷ and references therein). The SABRE technique has been applied to hyperpolarization of molecules containing nitrogen heterocycles like pyridyl, pyridazyl, pyrimidyl, or similar groups,¹⁶ as well as amino acids¹⁷ and small oligo peptides with up to 38-fold ¹H SEs.^18,19 Besides parahydrogen as the polarization source, another widely applied technique is DNP, which can be applied to large molecules or even whole cells.²⁰ While mostly applied in solid-state NMR,²¹ dissolution DNP with hyperpolarized water has achieved SEs of up to 300 in protein hydrolysates.²²

While there are certain peptides with naturally occurring unsaturated bonds suitable for PHIP,²³ most peptides require some sort of labeling. A number of labeling strategies for standard PHIP experiments have been applied to peptide molecules to date. Thus, the PHIPable moiety has been installed in the backbone,²⁴⁻²⁶ which, however, poses a challenge for the overall conformation and folding of a peptide. Alternatively, unsaturated elements have been introduced in the side chains, among them allyl or alkyne groups.^27,28 In the side-arm hydrogenation (PHIP-SAH) by Reineri et al., the reacted group has been cleaved from the target molecule after polarization transfer.²⁹ As the latter two strategies preserve the conformational freedom of the backbone of peptides or proteins, they are, in general, better suited for the design of bioactive architectures.

Non-natural amino acids with unsaturated bonds in the side chains are powerful PHIP markers that have been applied as building blocks for neurotransmitters and proteins.^5,26,30 In our previous work, the sunflower trypsin inhibitor SFTI-1 appeared as the first successfully PHIP-hyperpolarized biologically active peptide.²⁸ Later, PHIP was applied to a number of bioactive peptide targets, e.g., to antiplatelet aggregation inhibitors³¹ or enzyme inhibitors.^32,33 Among different unsaturated labels, l-propargyl-modified tyrosine (Figure 1, upper right corner) demonstrated the best results, allowing its easy incorporation into peptides upon chain assembly on solid support and high accessibility as an Fmoc-protected synthetic building block.^28,30 Thus, introducing this label into the cyclic oligopeptide SFTI-1 built of 14 amino acids with one disulfide bridge led to a factor of 70 in SE with 50% para-enriched hydrogen in ALTADENA-type experiments²⁸ and up to 1200 times SE with 90% para-enriched hydrogen in an automated PASADENA setup in 1D spectra.^33,34 Moreover, the SE provided enough sensitivity for 2D single-shot ultrafast NMR spectra of micromolar solutions. In addition, installing l-propargyl tyrosine at a noncrucial position of the octapeptide octreotide containing 8 amino acids and one disulfide bridge resulted in up to a 2000-fold signal increase under preservation of bioactivity.³²

Figure 1.

Central panel: schematic representation of a parent open-chain variant of miniprotein oMCoTI. Roman numerals count cysteines, while Arabic numerals follow the numbers of amino acids in the chain. 1, 5, and 27 are positions for labeling in the peptide chain, also marked with stars. Yellow lines resemble the three respective disulfide bonds of cystine knots. Outside panel: synthetic oMCoTI variants (magenta) modeled to include the PHIP marker (cyan) in complex with trypsin (gray), MP-1P_ox, MP-5P_ox, and MP-27P_ox. Abbreviations are given as follows: MP—miniprotein, P and A—propargyl and allyl substituents at the side chain of the PHIP marker, respectively, red—linear precursor, and ox—folded miniprotein. The hydrogenation reaction with the help of parahydrogen (pH₂) leads to the attachment of two hydrogens to a triple bond of the propargyl tyrosine (top right), which are denoted H^A to H^C on the allyl tyrosine.

In this study, we focus on miniproteins that display increasing structural complexity and build a class of molecules between oligopeptides and full-size proteins. As a larger and structurally more demanding class of protease inhibitors of medical relevance, we study PHIP label members of the family of the so-called cystine-knot miniproteins, or knottins (inhibitor cystine knots). Several knottins have already been engineered to bind to medically relevant targets,³⁵⁻⁴⁷ among them miniprotein MCoTI-II derived from Momordica (M.) cochinchinensis, a natural peptide of 34 amino acids possessing a characteristic three-disulfide pattern that has a structure of a pseudoknot, which gives this miniprotein exceptional stability.⁴⁸ While MCoTI-II reveals a conserved structural core, the surface-exposed loops possess high flexibility in terms of primary structure. Substitution of surface-exposed residues generates tailor-made compounds for potential diagnostic and therapeutic applications.^35,36,49,50

The open-chain variant of MCoTI-II⁵¹ (oMCoTI, “o” indicates open-chain) lacks a macrocyclic backbone and the respective N- and C-terminus connecting loop of five amino acids. However, it still shows a high affinity for the enzyme trypsin in the low nanomolar to picomolar range. Thus, we decided to skip head-to-tail macrocyclization for synthetic simplicity. Miniproteins allow one to choose different positions for the installation of a label due to the flexibility of their loosely tight loop, which makes these molecules valuable scaffolds for polypeptide engineering and drug design.⁵²⁻⁵⁴

In the present work, an open-chain variant of a natural cyclic miniprotein McoTI-II was decorated with l-propargyl tyrosine-based labels at three different positions to yield different variants of oMCoTI. The three substrates designed based on oMCoTI with disulfide bridges marked in yellow and the PHIP label in cyan are shown in Figure 1. They were synthesized by solid-phase peptide synthesis (SPPS), where propargyl-bearing labels were installed in the amino acid sequence in the form of Fmoc-protected building blocks. We investigated whether the disulfide-bridged scaffold remains intact upon installation of the PHIP label and whether the biological activity of miniproteins is preserved between those labeled in the functional loop and at the N- and C-terminal positions of a peptide. It was also important to assess how folding affects the PHIP procedure. To that end, the linear reduced precursors were examined and compared with fully folded miniproteins.

The crucial step in PHIP hyperpolarization is the catalyzed hydrogenation with a parahydrogen molecule. In order to achieve high SEs, the turnover of this reaction should be as high as possible. For this reason, in studies of organic reactions, where biocompatibility is not an issue, usually deuterated organic solvents, such as, e.g., pure methanol-d₄ or acetone-d₆, are employed. They combine fast hydrogenation catalysis with high H₂ solubility.

In the case of biocompatible systems, the problem of toxicity is more important than reactivity; therefore, aqueous solvents have to be employed. Thus, to achieve a compromise between reactivity and toxicity, in this initial study, we employed the mixtures of methanol-d₄ with D₂O as the solvent for the investigation of the PHIP hyperpolarization of the labeled miniprotein. In previous work, we also demonstrated the feasibility of mixtures with less toxic ethanol-d₆ and D₂O.³² Alternatively, several methods have been developed to replace the organic solvent with water, such as phase transfer or rapid evaporation by spray-flash distillation. These treatments can additionally remove the catalyst from the solution.^55,56 Though they require more specialized equipment, these steps are essential to increase the biocompatibility of the hyperpolarized agents.

In this manuscript, first, we describe the strategy, design, and synthesis of the PHIP-labeled miniproteins, as well as the validation of their bioactivity. Then, we report the results of the hyperpolarization experiments in biocompatible solvents on the solitary miniprotein and on the miniprotein interacting with its target protein. We show that significant NMR SEs can be achieved despite the presence of a compact cystine-rich scaffold.

Materials and Methods

Synthesis of Modified Miniproteins

All miniproteins were assembled by a combination of automated and manual Fmoc-based solid-phase peptide synthesis (Fmoc-SPPS) on a RAM polymeric support, followed by oxidative folding and chromatographic isolation (for the detailed procedures, refer to Section 4 in the ESI and the respective Figure S1).

Evaluation of Bioactivity by Trypsin Inhibition Assays

Kinetic curves were recorded by monitoring the proteolytic degradation of the chromogenic substrate Boc-QAR-pNA by bovine trypsin in the presence of the miniproteins, followed by the calculation of an apparent inhibition constant K_i^app (for the detailed procedure, refer to the ESI, Section 6).

NMR and PHIP Experiments

NMR experiments were performed in a 11.7 T OXFORD 500 MHz magnet, equipped with a Bruker AVANCE III HD spectrometer. The parahydrogen enrichment was performed with a parahydrogen generator from Advanced Research Systems Inc., comprising a DE204A cryostat and an ARS 4HW compressor. The cryostat was cooled to 30 K. More than 95% para-enriched hydrogen was delivered into an NMR sample tube placed inside the magnet at 7 bar and 25 °C.

The PHIP experiments were conducted under PASADENA² conditions, employing an automated setup for the reaction control and timing of the measurements.^34,57 As a hydrogenation catalyst, the rhodium complex [Rh(dppb)(COD)]BF₄ was used at three different concentrations of 0.45, 0.9, and 1.8 mM from a stem solution of 3 mg/mL (4.14 mM) in MeOD-d₄.

The concentration of the miniprotein was chosen to be 0.5 mM in MeOD-d₄ and in mixtures of 40% D₂O in MeOD-d₄. The miniprotein was dissolved directly in MeOD, whereas for experiments with trypsin, the latter was first dissolved in pure D₂O and subsequently combined with the miniprotein and catalyst solutions. For the samples with 20% D₂O in MeOD-d₄, the concentrations of miniprotein and trypsin were 0.22 mM.

When the hydrogen gas was bubbled through the sample solution, there had to be a 2–3 s delay between the end of the reaction and the start of the measurement for the hydrogen bubbles not to disturb it, facilitating a homogeneous sample.

In a typical PHIP experiment, first, an averaged spectrum at thermal polarization was collected in 32 scans as a baseline, and then, one scan for the PHIP spectrum was acquired, and after complete relaxation of the hyperpolarization, an averaged spectrum over 32 scans after the hydrogenation reaction with PHIP was recorded (for more details, refer to the ESI, Section 5).

To acquire the PHIP reaction kinetics, the measurements were repeated with the same sample in a pseudo-2D experiment until full conversion of the substrate was achieved. A single reaction step comprised 15 s of parahydrogen insertion (bubbling), followed by a short delay of 3 s, 3 s of data acquisition, and another short relaxation delay of 1 s. For samples with 20% D₂O, the bubbling time was reduced to 5 s and the delay until acquisition to 2 s.

T₁-times for the allyl protons of the hydrogenation product were determined by inversion recovery experiments.

Results and Discussion

Design and Synthesis of Labeled Miniproteins

l-propargyl-modified tyrosine is an efficient site-selective PHIP marker for peptides. It is compatible with standard SPPS protocols and promotes very strong hyperpolarization in the resulting allyl protons. In order to find the proper site within the sequence in view of both the maintenance of biological activity and PHIP efficiency (Figure 1 and Table 1), three mutants of the trypsin inhibitor miniprotein oMCoTI with the l-propargyl PHIP marker at different positions were synthesized and examined for bioactivity and PHIP hyperpolarization.

Table 1. K_i Values of Miniproteins for Inhibition Assays against Trypsin.

synthesized MP	Ki [nM]
MP-1Pox	60.6
MP-1Aox	302.5
MP-5Pox
MP-5Aox
MP-27Pox	7.7
MP-27Aox	2.5

In the MP-27P variants (Figure 1 and Table 1), a native tyrosine (Tyr) located at the C-terminal loop was substituted with l -propargyl tyrosine; in MP-1P, the label was placed N-terminally in order to cause as little steric hindrance as possible, and in MP-5P, it was introduced at the P1 position of the inhibitory loop. In addition, the variants of the hydrogenated mutants containing l-allyl tyrosine at the mentioned positions were synthesized as references in the PHIP experiments.

Evaluation of Bioactivity

The influence of the PHIP marker on bioactivity was evaluated by bioassays. Cystine knots derived from M. cochinchinensis possess inhibitory activity against trypsin, and this activity is associated with correct folding.^58,59 The highest inhibitory activity against trypsin was detected for the MP-27A_ox variant (Figure 2 and Table 1).

Figure 2.

Results of inhibition assays of miniproteins toward trypsin. The variant MP-27_ox shows by far the strongest inhibition of trypsin activity, followed by the variant MP-1_ox, while MP-5_ox demonstrates no inhibition.

With a K_i^app of 2.5 and 7.7 nM for propargyl-/allyl-bearing counterparts MP-27P_ox/A_ox, respectively, their activity was in good accordance with the inhibition ability of an open-chain MCoTI-II variant, possessing a K_i^app of 0.3 nM.⁵¹ In contrast, the MP-5P variant bearing the PHIP label within the inhibitor loop did not show any measurable activity against trypsin due to the lack of a basic amino acid at the P1 position.^60,61 The miniprotein variant MP-1P, where the PHIP label is located at the N-terminus (MP-1P_ox/A_ox), showed inhibition of trypsin in the double-digit nanomolar range, clearly demonstrating that this bulky, hydrophobic substituent affects the bioactivity even when it is located not directly in the functional loop but close to it.

PHIP Activity and SEs

Since the employed catalyst performs only in organic media, deuterated methanol (MeOD-d₄) was used as the main solvent. It is polar enough to dissolve both peptides and miniproteins. The folding of the oxidized forms of the knottins is not affected by the organic solvent as their structure is stabilized by the interlocked cystine-knot motif.⁶² For later experiments, in conjunction with trypsin, however, pure methanol was not a suitable solvent since trypsin can only tolerate up to 70% of methanol.⁶³ Instead, we used mixtures of up to 40% D₂O and MeOD-d₄.

To achieve the highest possible SE, different conditions with varying catalyst concentrations and reaction times were explored. The peptide concentration stayed usually fixed at 0.5 mM, and the catalyst concentration was 0.45, 0.9, and 1.8 mM. The reaction time varied between 10 and 25 s. Previous work showed that a higher catalyst concentration generally led to a faster hydrogenation reaction, which in turn facilitated a stronger SE due to a larger amount of simultaneously hyperpolarized molecules before the polarization relaxation to thermal equilibrium with T₁.³² A reaction that is too fast, however, can result in complete turnover and decay of the hyperpolarization before the measurement starts. It was important therefore to find the best adjustments between the catalyst concentration, the reaction velocity, and T₁ times.

The T₁ times are summarized in Table 2. H^B and H^C are the hydrogens added during the hydrogenation reaction (Figure 1, top right). In general, H^C has the longest T₁ time in all variants; however, it also shows a complicated multiplet signal that, in our experience, often reduces its apparent signal intensity due to increased overlap in the antiphase PHIP signal. Due to cross-relaxation effects, the proton H^A shows only weak SEs.⁶⁴ Therefore, the SEs listed below (Table 3) are mostly related to proton H^B, which generally shows the highest signal intensities and enhancement factors ε in our PHIP experiments. In some cases, H^C shows a higher ε value despite the partial overlap of antiphase signals (Table 3).

Table 2. T₁ Times at 11.7 T for the Allyl Protons of the Hydrogenation Products.

Miniprotein	T1(HC 6.05 ppm) [s]	T1(HA 5.40 ppm) [s]	T1(HB 5.25 ppm) [s]
MP-1Aox	3.32 ± 0.42	2.26 ± 0.14	3.16 ± 0.26
MP-5Aox	2.72 ± 0.36	2.48 ± 0.22	2.21 ± 0.15
MP-5Ared	3.57 ± 0.55	2.52 ± 0.13	2.74 ± 0.12
MP-27Aox	3.74 ± 0.39	2.89 ± 0.14	2.96 ± 0.19

Table 3. Maximum SEs of H^B at Different Catalyst Concentrations Dissolved in MeOD-d₄ at Their Respective Optimal Reaction Time of 15–25 s.

sample	SE ε HB 5.25 ppm	SE ε HC 6.05 ppm	reaction time [s]	catalyst concentration [mmol]	peptide concentration [mmol]
MP-1Pox	33.9		25	0.45	0.5
	110.1		20	0.9	0.5
	201.4		15	1.9	0.5
MP-5Pox	476.6	498.6	10	1.0	0.22
	194.9		20	0.9	0.5
MP-5Pred	704.0	963.6	20	0.9	0.5
	75.6		25	0.45	0.5
MP-27Pox	131.1		20	0.9	0.5
	89.6	184.9	20	1.8	0.5

Figure 3 presents the results of a PHIP measurement. Two dispersion signals typical for the employed PHIP marker at about 5.25 and 6.1 ppm (Figure 3b) correspond to the two protons added to the alkyl group of the PHIP marker. This process reduces each propargyl variant MP-XP_ox toward the respective allyl MP-XA_ox, with X marking positions 1, 5, or 27. These two corresponding signals are hardly seen in the thermal spectrum after hydrogenation (Figure 3c), as their intensity is quite low, and they partly overlap with other signals. In order to determine an ε value, a thermal spectrum of pure MP-XA_ox at the same concentration was also recorded (Figure 3d), where the allyl signals were more apparent.

Figure 3.

¹H NMR spectra at 11.7 T (500 MHz) of oMCoTI (MP-27P_ox) at (a) thermal polarization (TP) before PHIP, (b) PASADENA PHIP spectrum, and (c) thermal equilibrium after the hydrogenation reaction with PHIP. The insets show the allyl signals with vertical dashed lines to guide the view. For comparison, panel (d) shows a pure allyl spectrum (MP-27A_ox) at TP. Note that thermal allyl signals in panel (c) overlap with background signals. Other antiphase signals in panel (b) stem from catalyst-bound parahydrogen and the catalyst ligand cyclooctadiene. The receiver gain (rg) of the PHIP spectrum in panel (b) was decreased to avoid receiver overload.

The highest ε of the folded miniproteins with 476.6 for H^B and even 498.6 for H^C was achieved for MP-5P_ox at 4.5 equiv of catalyst and a very short reaction time of 10 s. Note that the peptide concentration here was reduced to achieve a higher catalyst ratio. The same composition was later employed for the experiments in 20% D₂O. MP-5P_ox showed a higher ε of 194.9, even at a medium catalyst concentration, at 1.8 equiv, which was higher than that of the other variants at the same catalyst concentration. The higher reactivity of MP-5P_ox might be explained by better accessibility of the exposed P1 position of the binding loop to the catalyst. This modification, however, comes with the trade-off of impaired bioactivity.

Generally, a higher catalyst concentration ensures faster performance, shorter reaction times, and higher SE factors. This was indeed the case for MP-5P_ox and MP-1P_ox but not for MP-27_ox (see Table 3). Hydrogenation of MP-1P_ox resulted in a maximum ε = 201.4 at 3.8 equiv of catalyst and 15 s reaction time. For MP-27P_ox, the maximum ε of 131.1 for H^B was reached at a middle concentration, 1.5 equiv of catalyst, and 20 s reaction time. At a higher catalyst concentration, the enhancement for H_B was lower than that at a medium concentration. However, for the other hydrogen atom H^C, an even higher SE than for H^B of 184.9 was achieved at the highest catalyst concentration and a 20 s reaction time.

In order to understand the effects of miniprotein fold on the PHIP reactivity, we compared the enhancements of folded MP-5P_ox and unfolded linear MP-5P_red. The latter exhibited a 2–5 times higher SE, which is an indication of better accessibility of the PHIP marker due to the higher motional flexibility of the peptide backbone. Also, the variant MP-5P_red had the highest ε on the signal for H^C (963.6). The much longer T₁ time of H^C compared to H^B might outweigh the loss of signals due to overlapping antiphase signals. This was also the case for MP-27A_ox.

Different approaches, equipment, and conditions make it difficult to compare our resulting SE with those of other PHIP publications, provided they actually reported it. However, within our group, the presented methodology has been established and improved over time, allowing for a comparison of SE in bioactive peptides using side-chain hydrogenation of propargyl tyrosine. Compared to previous work³² with the octapeptide octreotide, up to 2056 enhancement has been achieved, utilizing the same conditions of >95% pH₂ at 7 bar. In the same study, the monomeric marker achieved an SE of up to 6200. With an SFTI peptide consisting of 14 amino acids, we achieved up to 1200 with 90% pH₂ at 3 bar.³³ All of these experiments were conducted under PASADENA conditions. The latter example was previously also hyperpolarized with 50% pH₂ at 2 bar in ALTADENA experiments achieving an enhancement of 74.²⁸ From this series, first, we can conclude that a higher percentage of pH₂ enrichment and the change from manual ALTADENA to automated PASADENA and increase in hydrogen pressure drastically improved the SE. Second, the SE shows inverse proportionality to the length of the amino acid sequence. Indeed, when comparing octreotide with MCoTI (MP-5P_ox), the 3.6 times increase in amino acid count leads to a 4 times lower SE. When compared with SFTI, the 2 times increase in length leads to a 2.4 times decrease in enhancement. Nevertheless, the values of 499–964 achieved in this work still correspond to enormous time-saving factors in the order of 10⁵–10⁶.

PHIP of Enzyme–Inhibitor Interactions

The major aim of our study was to determine whether the hyperpolarized inhibitor could be utilized in binding studies with its natural target enzyme trypsin. As shown above in Figure 2, the variants MP-27 and MP-1 demonstrated a tight-binding behavior to trypsin, while MP-5 did not bind.

We found that the best way to get trypsin into solution was to dissolve it first in pure D₂O and then add it to the mixture of the inhibitor MP variant with the catalyst in MeOD-d₄. 40% D₂O was the best solvent composition at the chosen peptide concentration of 0.5 mM.

Prior to the addition of trypsin, we studied PHIP in aqueous media, assessing the kinetics of the reaction. The resulting curves of the signal intensity against the reaction time in 40% D₂O are plotted in Figure 4. Without trypsin, the variant MP-5P_ox showed the fastest performance with the maximum intensity of the PHIP signal after the first 20 s of reaction. The variants MP-1P_ox and MP-27P_ox reacted slower, with maxima after about 40 s. Interestingly, when comparing SEs in methanol and 40% D₂O/methanol mixture at the same catalyst concentration (Tables 3 and 4), MP-1P_ox and MP-5P_ox showed a decrease in their SE factors, which could be attributed to the higher viscosity of the solutions that impaired mixing with parahydrogen gas and slowed down the reaction. MP-27P_ox, on the contrary, increased its SE factor, suggesting improved accessibility of the PHIP marker in the more polar solvent, thus facilitating faster performance despite the increased viscosity.

Figure 4.

Absolute integrals of H^B in the PHIP experiments plotted against the reaction time for MP-1 (a), MP-5 (b), and MP-27 (c) in 40% D₂O and MeOD-d₄ with trypsin (red) and without trypsin (cyan). For those configurations where several samples were measured, the results are shown as an area plot. Single samples are plotted as lines. The addition of trypsin leads to diminished signal intensities in all samples, indicating unspecific interactions most likely due to increased viscosity.

Table 4. Maximum SEs of H^B during the Reaction Kinetics of Different MP Variants, Preparations with 40% or 20% D₂O in MeOD-d₄, Without and in the Presence of Trypsin (Reaction Time in Brackets).

sample	0.5 mM MP + 1.8 mM catalyst + 40% D2O		0.22 mM MP + 1 mM catalyst + 20% D2O
	no trypsin	0.5 mM trypsin	no trypsin	0.22 mM trypsin
MP-1Pox	104.8 (39 s)	14.4 (60 s)	28.2 (44 s)	11.9 (32 s)
MP-5Pox	122.1 (18 s)	18.5 (40 s)	306.9 (29 s)	66.1 (29 s)
MP-27Pox	221.5 (39 s)	14.7 (60 s)	136.5 (20 s)	65.6 (32 s)

After trypsin addition, these curves reflect a pronounced reduction of the signal intensities by about 1 order of magnitude. This could be attributed to either interaction of the miniprotein with trypsin or an additional increase in viscosity upon trypsin addition. The PHIP intensity and SEs of MP-5P_ox with trypsin in 40% D₂O were slightly higher than those of the other two variants (also see Table 4). This could be explained by the lack of inhibitory activity of the MP-5 variant, hence no binding to trypsin, and therefore higher accessibility to the catalyst compared to the other variants tightly bound to the active site of an enzyme.

To determine whether the observed reduction of SE upon trypsin addition was a specific binding effect or solely an increase in viscosity, the measurements were repeated at compositions with lower concentrations, specifically 0.22 mM for both the peptide and trypsin. The amount of D₂O required to dissolve trypsin was also reduced to a concentration of 20%. All of these changes led to a lowered viscosity of the samples. The catalyst concentration was proportionally decreased to 1 mM.

For the low-concentrated samples, MP-27_ox showed faster kinetics than MP-5_ox (Figure 5, cyan); however, the latter displayed higher SE values without trypsin (Table 4). The addition of trypsin again resulted in a reduced PHIP intensity. However, the reduction was not as prominent as at higher concentrations of trypsin and 40% D₂O. The SE values of MP-27_ox and MP-5_ox in the presence of trypsin were rather similar, namely 65.6 and 66.1, respectively. These values were 3 to 4 times higher than those in the 40% D₂O solution. This indicated that the reduction in PHIP intensity was most probably caused by alterations in viscosity upon the addition of reaction components.

Figure 5.

Absolute integrals of H^B in the PHIP experiments plotted against the reaction time for MP-1 (a), MP-5 (b), and MP-27 (c) in a mixture of 20% D₂O in MeOD-d₄ with trypsin (red) and without trypsin (cyan). For those configurations where several samples were measured, the results are shown as an area plot. Single samples are plotted as lines. After trypsin addition, samples MP-5 and MP-27 still show PHIP activity. In contrast, sample MP-1 shows no activity at all due to blocking of the reactive site by trypsin.

MP-5 and MP-1 showed a clear tendency to higher SE in methanol compared to aqueous media. However, for MP-27_ox, the situation was the opposite as it showed higher SEs, along with increasing water content in PHIP mixtures, probably due to the higher structural similarity with the native parent miniprotein.

The most important finding was that for the first time, we were able to distinguish between the bound and unbound states of a protease inhibitor with the PHIP experiment. Indeed, MP-5_ox and MP-27_ox showed smooth hydrogenation of a triple bond upon the addition of trypsin, suggesting easy accessibility of both labels to the catalyst. MP-5_ox did not bind trypsin due to the lack of P1 lysine at the binding site, leaving enough space for the catalyst to dock and react. Vice versa, the label of MP-27_ox was far away from the binding site and not sterically hindered and therefore addressable. In contrast, the propargyl site at position 1 in MP-1_ox was shielded from the catalyst by the target enzyme due to its proximity to the binding site and was not active when the inhibitor was bound to its target. Hence, no reaction product was observed with trypsin addition, as indicated by the flat intensity profile in Figure 5a.

Summary and Outlook

In this work, we succeeded to apply PHIP for the molecules of biological relevance that go beyond small peptides. With three disulfide bridges, the investigated knottins display a higher level of structural complexity. We demonstrated synthetic accessibility of PHIP-labeled cystine-knot miniproteins on an example of three open-chain variants of trypsin inhibitors from M. cochinchinensis bearing l-propargyl tyrosine label. A vast number of disulfide-rich miniproteins have been reported to date, all of them sharing the combination of the stabilizing three (or more)-disulfide knot and flexible loops.⁵³ In view of the PHIP procedure, a typical representative of disulfide-rich miniproteins was examined, and the obtained results also likely apply to other members of this family.

The PHIP labels were installed at three different positions within the peptide chain: N-terminally, C-terminally, and at the P1 position of the inhibitor loop, which generally ensures biological activity of this class of peptide. Two variants, MP-1_ox with an N-terminal label and MP-27_ox with a native tyrosine replaced by the label, showed binding affinity to trypsin at a low nanomolar range. The third variant, MP-5_ox, in which the label replaced lysine at the P1 position in the binding loop, expectedly lacked binding affinity. The retained bioactivity in the first two variants confirms the correct folding. In our previous studies, we have shown that engineering of knottins yields miniproteins that are able to bind/inhibit targets of therapeutic relevance.^40,65 The general folding of these miniproteins was, nevertheless, very similar to that of the parent miniproteins. Another way to improve the binding is to incorporate backbone cyclization. This would be another synthetic step upon chemical assembly but might increase the resulting binding affinity. In addition, the combined recombinant and chemical way of synthesis could be taken into consideration.³⁹

All three variants expressed high reactivity upon the PHIP experiments in the organic solvent MeOD-d₄, yielding SEs ε of up to 964 for the reduced variant MP-5_ox and 498.6 for the folded miniprotein. The reaction in solvent mixtures of MeOD with D₂O led to maximum ε values of 221.5 (MP-27_ox) and 306.9 (MP-5_ox) in 40 and 20% D₂O, respectively. The highly bioactive MP-27_ox, which differs from the parent oMCoTI miniprotein only by the presence of a triple bond at the side chain of tyrosine 27, obviously favored a more polar environment as the SE in 40% D₂O was even higher than in pure methanol. The MP-5P_ox variant achieved the highest enhancements in pure methanol and 20% D₂O due to the more exposed PHIP marker, facilitating faster reaction times.

We found that SE in peptides is rather obvious—inversely proportional to the length of the amino acid sequence. Indeed, when comparing octreotide with MCoTI, the 3.6 times increase in amino acid count leads to a 4 times lower enhancement of the signal. When compared with SFTI, the 2 times increased length leads to a 2.4 times decrease in enhancement. Comparing folded and unfolded MCoTI in this work, we see a factor of 2, an increase in enhancement of the unfolded one. While increasing the peptide size reduces the observed reactivity and SE, the values of 499–964 achieved here still correspond to enormous time-saving factors in the order of 10⁵–10⁶.

To the best of our knowledge, we were the first to apply hydrogenative PHIP to bioactive peptide molecules of the size of 29 amino acids. In order to circumvent diminishing SE with increasing peptide size, the experimental conditions can be improved in view of concentrations, temperature, pH, and gas pressure. Also, numerous PHIP markers could be introduced simultaneously.⁶⁶

PHIP experiments in the presence of trypsin showed a reduction in SE of typically 1 order of magnitude due to slower hydrogenation reactions. Nevertheless, the achieved ε values were still in the range between 10 and 70, which corresponded to shortened measurement times by factors 100–5000.

The observed signal reduction upon enzyme addition is primarily caused by the high sensitivity of the PHIP reaction toward sample viscosity, as demonstrated by the experiments at different solvent compositions. Furthermore, an unspecific interaction of trypsin with the miniproteins and the catalyst could also mask any effect due to the specific binding of the miniproteins to trypsin. To circumvent this issue, lower amounts of the peptides and D₂O in the solvent were utilized. The lack of SE on a marker close to the binding site of trypsin demonstrated for the first time that PHIP can be used as a means of detecting bioactivity or localizing bioactive sites in biomolecules.

In conclusion, this approach can be used to investigate chemical transformations of complex peptide biomolecules and get insights into their bioactivities, as well as for further development of the more natural ways of working with catalysts in water solutions in order to cross the gap to medical applications. Furthermore, it can help to study biological reactions, i.e., between proteases and their inhibitors, and could be applied in drug development.

Moreover, it is possible to add the PHIP marker to larger proteins of interest by different methods, i.e., upon recombinant production of proteins^67,68 or by posttranslational modifications applying transglutaminase, sortase, or other bond-forming enzymes and furnishing the proteins with the respective recognition sequences for addition of the marker to the biomolecule of interest.^69,70 We also look forward to continuing to widen the application horizon of PHIP to larger proteins such as affibodies, which cannot be synthesized chemically.

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/jacs.4c11589.

• Detailed description of materials and methods and synthesis and characterization of compounds (PDF)

Author Present Address

^§ Department of Chemistry, North Carolina State University, 851 Main Campus Drive, Raleigh, North Carolina 27606, United States

The authors declare no competing financial interest.