Studies on the selectivity of proline hydroxylases reveal new substrates including bicycles.

Studies on the substrate selectivity of recombinant ferrous-iron- and 2-oxoglutarate-dependent proline hydroxylases (PHs) reveal that they can catalyse the production of dihydroxylated 5-, 6-, and 7-membered ring products, and can accept bicyclic substrates. Ring-substituted substrate analogues (such hydroxylated and fluorinated prolines) are accepted in some cases. The results highlight the considerable, as yet largely untapped, potential for amino acid hydroxylases and other 2OG oxygenases in biocatalysis.

Introduction

Hydroxylated amino acids are common starting materials for the synthesis of pharmaceuticals and agrochemicals. They are also intermediates in the biosynthesis of natural products with interesting biomedicinal properties, including many antimicrobials [1], [2], [3]. Recent work has also expanded the set of identified hydroxylated protein residues, which, although not as large as that of hydroxylated amino acids, is extensive [1].

The ferrous iron- and 2-oxoglutarate (2OG)-dependent dioxygenases (ODDs) possibly catalyse the widest range of oxidations of amino acids and proteins of any identified enzyme family [1], [2], [3]. Prolyl hydroxylation of collagen was the first identified reaction catalysed by 2OG oxygenases and occurs to give C-3- and C-4-hydroxyprolyl protein products. Analogous prolyl hydroxylations of many collagen-like proteins have been identified [1]. Trans-C-4-prolyl hydroxylation of the hypoxia inducible transcription factor plays a key role in the adaptive response to hypoxia in animals [4]; related modifications have been identified in lower eukaryotes and prokaryotes [5], [6]. Trans-C-3-prolyl hydroxylation also occurs in a ribosomal protein (RPS23) in eukaryotes ranging from yeasts to humans [7], [8]. Interestingly, the yeast, but not the human, ribosomal prolyl hydroxylase catalyses a second hydroxylation of the same residue to give a dihydroxylated product [7].

In connection with the work on prolyl hydroxylation aimed at defining regio- and stereo-selectivities, we are interested in exploring the selectivity of proline (1) hydroxylases and using them to prepare standards for amino acid analyses, associated with the functional assignment of protein mono and di-hydroxylases. Aside from their use in preparing standards for structure validation, 2OG oxygenases have considerable potential as industrial scale biocatalysts, as demonstrated by the use of a recombinant l-proline (1) trans-4-hydroxylase (trans P4H)[9].

Hydroxyprolines and modified (e.g. by methylation) prolines are present in many peptide-based natural products [10], [11], [12]. In some cases, the proline-hydroxylations occur prior to peptide-synthetase-catalysed oligomerisation [10], [11], [12]. Following from pioneering labelling [10], [11] and purification studies [13], recombinant forms of proline-hydroxylases (PHs) have been produced (Fig. 1) [1], [14], [15], [16], [17], [18], [19], [20], [21], [22], [23], [24], [25], [26], [27]. Crystallography has revealed that the PHs contain a distorted, double-stranded, beta-helix (DSBH) fold, a His-X-(Asp/Glu)…His iron-binding motif, and a cosubstrate-binding mode, all of which are characteristic of a 2OG oxygenases [28], [29], [30]. Proline (1) hydroxylation occurs with retention of stereochemistry [31].

Fig. 1

(a) Products of proline hydroxylase (PH) reactions using the natural substrate (l-proline, 1) [(2S,3S)-cis-3-hydroxy-l-proline (2), (2S,4S)-cis-4-hydroxy-l-proline (3), and (2S,4R)-trans-4-hydroxy-l-proline (4)], and (b) the stoichiometry of the proline-hydroxylase reactions.

Several studies are reported on the selectivity of PHs – these reveal they can accept 3,4-dehyroproline to give epoxides [14], [20], [32], some methylated proline derivatives [23], [25], and azetidine- and piperidine C-2-carboxylates [20], [22], [26], [27], [32] to give hydroxylated products with different stereo- and regio-chemistries. Here, we report further studies on the substrate and product selectivities of three recombinant PHs: l-proline cis-3-hydroxylase type I (cisP3H) [15], [20], [33], l-proline cis-4-hydroxylase (cisP4H) [21], [26], [27], and l-proline trans-4-hydroxylase (transP4H) [17], [18]. The results reveal PHs can catalyse the hydroxylation of substrates of varied ring sizes, substituted prolines, sizes, N-substituted prolines, including fluorinated and hydroxylated rings, and, importantly, bicyclic rings systems.

Results and discussion

Substrate analogue studies

Recombinant forms of the three bacterial PHs were produced in Escherichia coli using standard procedures (see Supplementary Information): cisP3H (from Streptomyces spp. [strain TH1]) [15], [20], [33], cisP4H (from Sinorhizobium meliloti) [21], [26], [27], and transP4H (from Dactylosporangium sp.) [17], [18]. We used LC/MS and 1H/19F NMR assays (1H: 500/700 MHz; 19F: 470 MHz) to monitor substrate turnover. Using the natural l-proline (1) substrate, the three recombinant PHs efficiently (>95% substrate conversion) produced hydroxylated products (2, 3, and 4) with the reported diastereoselectivities under standard conditions (Fig. 1) [15], [18], [20], [32], [33], as shown by LC/MS (using authentic standards) (Fig. 1). In these and subsequent reactions substrate hydroxylation was accompanied by turnover of 2OG to succinate. We then tested a range of substrate analogues for activity with the three PHs.

Substrate analogue studies using substrates with different ring sizes

Consistent with literature reports, cisP3H and cisP4H catalyse the hydroxylation of (2S)-l-azetidine-2-carboxylic acid (5) to give (2S,3R)-cis-3-hydroxy-l-azetidine-2-carboxylic acid (6) (Fig. 2a) [14], [20], [26], [32]. We did not observe transP4H-catalysed hydroxylation of (2S)-l-azetidine-2-carboxylic acid (5). With (2S)-l-aziridine-2-carboxylic acid (Table S6 – Supporting Information), we did not observe hydroxylation to give a detectable product, though we cannot rule out oxidation to give an unstable product.

Fig. 2

Products generated by proline hydroxylases using substrate analogues with different ring sizes (4-, 6- and 7-membered rings). In this and subsequent schemes all PH catalysed reactions are coupled to the oxidation of 2OG /O2 to succinate/CO2. Note the precise structures of some products are unassigned. See Supplementary Information for details of incubations, structure assignments and (approximate) yields (Table S7).

With (2S)-l-pipecolic acid (7), all 3 PHs generated singly hydroxylated products, as reported (Fig. 2b) [20], [22], [25], [26], [27], [32]. However, with cisP4H, in addition to the reported (2S,3S)-cis-3-hydroxy-l-pipecolic acid (8) and (2S,5S)-cis-5-hydroxy-l-pipecolic acid (9) products, we detected a dihydroxylated product (Fig. 2b), (2S,3R,5S)-3,5-dihydroxy-l-pipecolic acid (10). With cisP3H and transP4H, (2S,3S)-cis-3-hydroxy-l-pipecolic acid (8) and (2S,5S)-trans-5-hydroxy-l-pipecolic acid (11) were observed (Fig. 2b), respectively. Interestingly, the 7-membered ring substrate analogue, (2S)-l-azepane-2-carboxylic acid (Azp) (12), was also hydroxylated, albeit in lower yields (~2.5–30%) (Fig. 2c). The 7-membered ring product observed in the highest yield was with cisP3H, i.e. (2S,3R)-cis-3-hydroxy-l-azepane-2-carboxylic acid (13), as measured by 1H NMR. LC/MS-analysis also implied that cisP3H reactions produced, albeit at relatively low levels, an additional mono-hydroxylated product, as well as a dihydroxylated product (17). With cisP4H, LC/MS detection showed that three hydroxylated products were generated, one of which had the same retention time as 13 (from the cisP3H reactions), suggesting this product was likely 13; the other two cisP4H products (15 and 16) were likely C5/C4-hydroxylated species. Reactions catalysed by transP4H produced a single product, which was characterised by NMR and assigned as (2S,5R)-trans-5-hydroxy-l-azepane-2-carboxylic acid (14). The observation of dihydroxylated products suggested that more complex substrates might be accepted by the PHs.

Substrate analogue studies using N-functionalised substrates

The results indicate that substrates require a (2S)-amino acid to be accepted by the PHs, consistent with the literature [13]. Thus, we did not observe any hydroxylation of substrate analogues with the (2R)-stereochemistry, N-acyl groups, or carboxylic acid derivatives (e.g. esters/amides functionality) (see Supplementary Information). We did, however, detect PH-catalysed hydroxylation of N-methylated substrates, including (2S)-N-methyl-l-proline (N-Me-Pro) (18) and (2S)-N-methyl-l-pipecolic acid (N-Me-Pip) (22). LC/MS analysis of reactions with cisP3H and cisP4H manifested mono-hydroxylated products; NMR characterisation showed that these were (2S,3R)-cis-3-hydroxy-N-methyl-l-proline (19) and (2S,4S)-cis-4-hydroxy-N-methyl-l-proline (20), respectively (Fig. 3a). Interestingly, in the case of transP4H, in addition to the major assigned (it was not isolated to high purity for definitive assignment by NMR) product (2S,4R)-trans-4-hydroxy-N-methyl-l-proline (21), LC/MS provided evidence for low-level production of a product with the same retention time as 20 (suggesting it was 20) (Fig. 3a). With (2S)-N-methyl-l-pipecolic acid (22), the major assigned products (23, 24, 25) are analogous to those seen using a (2S)-pipecolic acid (7) substrate (Fig. 3b). In the case of cisP3H, LC/MS provided evidence for a dihydroxylated product, proposed to be (2S,3R,5S)-3,5-dihydroxy-N-methyl-l-pipecolic acid (26). Tested substrates with larger N-substitutions, such as (2S)-N-benzyl-l-proline (Table S6 – Supporting Information), were not accepted by the PHs.

Fig. 3

Products generated by proline hydroxylases using N-methylated substrates analogues: the products appear analogous to those produced with the unmethylated analogues (Fig. 1). See Supplementary Information for details of incubations, structure assignments, and (approximate) yields (Table S7). (? = provisionally assigned structure).

Substrate analogue studies using C-functionalised substrates

Recent studies have shown that trans-3-methyl-l-proline [23] and cis-3-hydroxy-l-proline [26] are PH substrates, suggesting that other ring-substituted proline analogues may be substrates. LC/MS analyses showed that (2S)-2-methyl-l-proline (2-Me-Pro) (27) was a substrate for all three PHs tested (Fig. 4a). LC/MS analysis showed that the cisP3H and transP4H reactions produced singly hydroxylated products, likely (2S,3R)-cis-3-hydroxy-2-methyl-l-proline (28) and (2S,4R)-trans-4- hydroxy-2-methyl-l-proline (30), respectively.

Fig. 4

Products generated by proline hydroxylases using ring-substituted substrate analogues: note alkylated, hydroxylated, and fluorinated substrate analogues are substrates. See Supplementary Information for details of incubations, structure assignments, and (approximate) yields (Table S7). (? = provisionally assigned structure).

LC/MS assays suggested that cisP4H produces two hydroxylation products: one with the same retention time as the cisP3H product (28) (suggesting it is the same) (‘minor’ product), and one assigned as (2S,4S)-cis-4-hydroxy-2-methyl-l-proline (29) (‘major’ product), by NMR (Fig. 4a). LC/MS-assays indicated that a substrate analogue with a larger C-2-functionality, i.e. (2R)-2-(4-methylbenzyl)-l-proline, was not hydroxylated, likely due to steric factors (Table S6 – Supporting Information).

LC/MS-analyses showed that a C-4-alkynylated substrate analogue, (2S,4R)-trans-4-(prop-2-yn-1-yl)-l-proline (31), was a substrate for cisP3H and cisP4H (but not transP4H) (Fig. 4b); in both cases, a single-hydroxylation product was detected, assigned by NMR as (2S,3R,4S)-3-hydroxy-4-(prop-2-yn-1-yl)-l-proline (32) (Fig. 4b).

Proline analogues with C-5-/C-6-carboxyl functionalities were tested as PH substrates by LC/MS assays; (2R,5S)-5-carboxy-l-proline, (2R,5S)- 5-carboxy-N-methyl-l-proline, and (2R,6S)-6-carboxy-l-pipecolic acid were not PH substrates (Table S6 – Supporting Information). Substrate analogues with a C-5-carboxymethyl functionality were also tested as substrates, i.e. cis- and trans-5-(carboxymethyl)-l-proline (c-CMP and t-CMP); these compounds were also not PH-substrates (Table S6 – Supporting Information). However, the ‘carboxymethyl ester’ of t-CMP, trans-5-(2-methoxy-2-oxoethyl)-l-proline (‘Me-t-CMP’) (33), was found to be a cisP3H substrate by LC/MS assay; NMR characterisation showed that the product was (2S,3R,5S)-3-hydroxy-5-(2-methoxy-2-oxoethyl)-l-proline (34) (Fig. 4c). This result is interesting, because a CMP is an intermediate during the biosynthesis of the carbapenem antibiotics [34], [35], [36].

C-4-Fluorinated proline analogues were also assayed as PH substrates using LC/MS and NMR detection, i.e. cis-4-fluoro-l-proline (35) and trans-4-fluoro-l-proline (38). LC/MS and 1H/19F NMR data showed that 35 was a substrate of cisP3H and cisP4H (giving the same product), as well as transP4H (Fig. 4d); NMR characterisation showed that the cisP3H product was (2S,3S,4R)-4-fluoro-3-hydroxy-l-proline (36); LC/MS and NMR analyses suggest the transP4H product is a ketone, 4-oxo-l-proline (37). In the latter case, we propose that transP4H hydroxylates the C-4-position of cis-4-F-Pro to produce an unstable intermediate, (2S,4S)-4-fluoro-4-hydroxy-l-proline, which fragments releasing fluoride. Analogous reactions have been observed previously with: (i) the human HIF prolyl hydroxylase, PHD2, which catalyses hydroxylation of a ‘cis-4-fluorinated-prolyl’ substrate analogue to produce a ‘4-keto-prolyl’ product [37]; and (ii) γ-butyrobetaine hydroxylase (BBOX), which hydroxylates (3S)-3-fluoro-γ-butyrobetaine (GBBF) to give 3-keto-γ-butyrobetaine following spontaneous fluoride release [38], [39]. LC/MS and 1H/19F NMR analysis showed that trans-4-F-l-proline (38) was a substrate for cisP3H and cisP4H; cisP3H catalyses production of one hydroxylation product, which was assigned by NMR as (2S,3S,4S)-4-fluoro-3-hydroxy-l-proline (39); cisP4H generates two products, i.e. (2S,3S,4S)-4-fluoro-3-hydroxy-l-proline (39) and 4-oxo-l-proline (37) (Fig. 4e). A synthetic sample of the ‘4-ketone’ product, 4-oxo-l-proline, was also tested as a PH substrate (with LC/MS and NMR assays), and was found to be inactive as a PH substrate.

Hydroxyprolines were also tested as PH substrates, initially assaying by LC-MS. In the case of (2S,3S)-trans-3-hydroxy-l-proline (40), transP4H produced one product, assigned as (2S,3R,4S)-3,4-dihydroxy-l-proline (41) by NMR (Fig. 4f). In the case of cis-4-hydroxy-l-proline (3), cisP3H produced one hydroxylation product, which was assigned as (2S,3S,4R)-3,4-dihydroxy-l-proline (42) by NMR (Fig. 4g). LC/MS evidence suggested that, with (2S,4S)-cis-4-hydroxy-N-methyl-l-proline (20), cisP3H produces a single hydroxylation product, likely (2S,3S,4R)-3,4-dihydroxy-N-methyl-l-proline (43) (Fig. 4h). In the case of (2S,4R)-trans-4-hydroxy-l-proline (4), three products (44, 45, and 46) were detected (by LC/MS) for both cisP3H and cisP4H reactions (Fig. 4i). (Note that compared with the substrate, these products have mass increments of +32, +14, and +14 Da, respectively); one product is provisionally proposed to be a ‘C-3/C-4-dihydroxylated’ product, (2S,3S)-3,4,4′-trihydroxyproline (44) (i.e. the ‘hydrated’ form of a hydroxyketo species), and the other two products are proposed to be isomers of a C-3/C-4-hydroxyketo species (45, 46).

Substrate analogue studies using bicyclic substrates

The fact that N-methylated and carbon-backbone-functionalised substrates are accepted by the PHs suggested to us that bicyclic substrates may also be accepted by them. Whilst the bridged-ring systems of (1R,3S,4S)-2-azabicyclo[2.2.1]heptane-3-carboxylic acid (Table S6 – Supplementary Information) were not hydroxylated by the PHs, hydroxylation was observed with substrates containing two fused, 5-membered rings (fused at the C-4- and C-5-postions of proline (1)) (Fig. 5). With (2S,3aS,6aS)-octahydrocyclopenta[b]pyrrole-2-carboxylic acid (47), cisP4H and transP4H produced different mono-hydroxylated products (by LC/MS) (48 and 50, respectively) (Fig. 5a). However, in the case of cisP3H, two hydroxylation products were detected (by LC/MS) (Fig. 5a); the major product (48) had the same retention time as the cisP4H product (suggesting they are the same). NMR characterisation showed that the hydroxylation sites were at the C-3 and C-4 positions of the starting material (47) (i.e. on the ‘proline’ and ‘cyclopentane’ rings, respectively); analysis showed that the major product was (2S,3aS,4R,6aS)-cis-4-hydroxy-octahydrocyclopenta[b]pyrrole-2-carboxylic acid (48), and the minor product was (2S,3R,3aR,6aS)-cis-3-hydroxy-octahydrocyclopenta[b]pyrrole-2-carboxylic acid (49). 48 represents the first report of a PH-catalysed hydroxylation of a ‘non-proline’-type ring position. Interestingly, this product is favoured over the ‘cis-3-hydroxy’ product analogue (49), which usually predominates in cisP3H reactions (Fig. 1a). The regioselectivity of this reaction may reflect the ‘butterfly-like’ conformation of this substrate (47) (Fig. 6), resulting in the cyclopentane ring being closer to the ferryl intermediate with cisP3H and cisP4H than the ‘proline’ ring; hence, the C-4 position is preferentially hydroxylated over the C-3 position.

Fig. 5

Products generated by proline hydroxylases using bicyclic substrate analogues. See Supplementary Information for details of incubations, structure assignments, and (approximate) yields (Table S7).

Fig. 6

Proposed correlation between bicyclic ring confirmations and proline hydroxylase selectivity of reactions using (a) (2S,3aS,6aS)-octahydrocyclopenta[b]pyrrole-2-carboxylic acid (47) and (b) (2S,3aR,6aR)-octahydrocyclopenta[b]pyrrole-2-carboxylic acid (51).

The ring-junction stereoisomer of the (2S,3aS,6aS)-octahydrocyclopenta[b]pyrrole-2-carboxylic (47) acid substrate, i.e. (2S,3aR,6aR)-octahydrocyclopenta[b]pyrrole-2-carboxylic acid (51), was also a PH substrate by LC/MS assays (Fig. 5b); two hydroxylation products were detected for cisP3H (52, 53), and one for transP4H (54). NMR characterisation assigned the cisP3H products as (2S,3R,3aS,6aR)-cis-3-hydroxy-octahydrocyclopenta[b]pyrrole-2-carboxylic acid (52) (major) and (2S,3aR,4R,6aR)-cis-4-hydroxy-octahydrocyclopenta[b]pyrrole-2-carboxylic acid (53) (minor) (Fig. 5b). Thus, in this case, the anticipated, ‘cis-3-hydroxy’ product predominates, likely in part because the fused cyclopentane ring is ‘trans’ to the carboxylic acid group and is directed away from the ferryl intermediate with cisP3H, making the ‘natural substrate’ C-3 proton the closest available for abstraction (Fig. 6).

cisP3H also catalyses hydroxylation of (2S,3aS,7aS)-octahydro-1H-indole-2-carboxylic acid (55), which has fused 5- and 6-membered rings (Fig. 5c). In a manner analogous to the (2S,3aR,6aR)-octahydrocyclopenta[b]pyrrole-2-carboxylic acid (51) reaction, cisP3H catalysed hydroxylation of the cyclohexane ring of 55 to give the ‘cis-4-hydroxy’ product, (2S,3aS,4R,7aS)-cis-4-hydroxy-octahydro-1H-indole-2-carboxylic acid (56) (by NMR). A compound containing two fused, 6-membered rings, decan-hydroisoquinoline-3-carboxylic acid (57), was a substrate of cisP3H and transP4H (by LC/MS); two hydroxylation products were detected for cisP3H and one for transP4H (we were unable to assign the stereochemistry of these by NMR) (Fig. 5d).

Substrate analogue studies using unsaturated substrates

Unsaturated proline analogues were then tested as proline hydroxylase substrates. (2S)-3,4-dehydro-l-proline (60) was a substrate of cisP3H, cisP4H, and transP4H by LC/MS analysis (Fig. 7). Consistent with previous studies, cisP3H and cisP4H produced (2S,3S,4R)-cis-3,4-epoxy-l-proline (61) [20], [26], [32], and transP4H produced (2S,3R,4S)-trans-3,4-epoxy-l-proline (62) [14], [32] (Fig. 7a). (2S)-4,5-dehydro-l-pipecolic acid (63), was also a hydroxylation substrate of cisP3H by LC/MS analysis (Fig. 7b). However, because the product of this reaction was observed to coelute with the starting material (63) during purification, and NMR characterisation for the product was inconclusive, it is unclear whether an ‘epoxide’ was formed [i.e. (2S,4S,5R)-cis-4,5-epoxy-l-pipecolic acid (64)], or an allylic alcohol [e.g. (2S,3S)-cis-3-hydroxy-4,5-dehydro-l-pipecolic acid (65)].

Fig. 7

Products generated by proline hydroxylases using unsaturated compounds. See Supplementary Information for details of incubations, structure assignments, and yields (Table S7).

Conclusions

2OG dependent amino acid hydroxylases have considerable value as industrial catalysts [9]. Our results substantially expand the biocatalytic potential of pH catalysed hydroxylations, including with the demonstration that certain N-alkylated amino acids and bicyclic ring systems are accepted as PH substrates. The products produced, including di- and tri-hydroxylated prolines should be useful in ongoing work concerning functional assignment studies on 2OG dependent protein hydroxylases [6], [7], [8].

The observation that substrate analogues with four, six, and seven membered ring containing substrate analogues undergo PH catalysed oxidations, with dihydroxylated products being produced in some cases, is of interest from a biocatalytic perspective (Fig. 4 and 5). Some substituted prolines, including N-methylated prolines, were also found to be PH substrates giving triply functionalised pyrrolidine rings (Fig. 4). However, N-acylated or (2R) – amino acids were not PH substrates (Table S6).

The observation of two products with some bicyclic substrate/PH combinations is mechanistically interesting, suggesting that the ferryl intermediate in 2OG-dependent PH catalysis [1] may be positioned approximately equidistant between different potentially oxidised C—H bonds (Fig. 6). However, a mechanism involving hydrogen abstraction at one site, followed by hydrogen atom transfer and hydroxylation at a second site cannot be ruled out. There is a possibility of such processes occurring in 2OG oxygenase catalysis in carbapenem antibiotic biosynthesis [34]. It is possible that such processes occur in 2OG oxygenase catalysis involving protein/nucleic oxidations.

2OG oxygenases have been shown [1], [2], [3] to be amenable to structure based protein engineering to produce enzymes with improved product selectivities with unnatural substrates [1], [9]. Although this work is at an early stage with PHs, it has been reported that a directed-evolution approach has led to a modified cisP4H (with 3 substitutions) with improved activity with respect to hydroxylation of l-pipecolic acid to give (2S,5S)-5-hydroxy-l-pipecolic acid [27]. The relaxed substrate/product selectivities of PHs, and some, but not all, other 2OG oxygenases (including some catalysing protein modifications, e.g. factor inhibiting HIF) suggests that they may be suitable for ‘late stage modifications’ of valuable chemicals; following the initial identification of a desired product, its oxygenase catalysed production may be optimised by protein engineering. There would, thus, appear to be considerable, as yet largely unexplored, potential for (engineered) PHs, and other 2OG oxygenases, in biocatalysis.

Materials and methods

DNA encoding for Streptomyces sp. (strain TH1) cisP3H type I, Sinorhizobium meliloti cisP4H, and Dactylosporangium sp. transP4H (with 5′-NdeI and 5′-HindIII sites) was synthesised by Life Technologies’ GeneART, and codon-optimised for E. coli expression. The PH genes were then subcloned into a Takara pCOLDTM I expression vector to produce N-terminally His6-tagged PHs. Sequence-verified clones (by The Gene Service, Source BioScience, Oxford, UK) were transformed into Stratagene BL21-Gold (DE3) competent cells for protein production. Reagents, casting equipment, and electrophoresis tanks for SDS-PAGE were obtained from Bio-Rab Laboratories; molecular-mass markers (Thermo PageRuler Plus) were from Thermo Scientific. Reagents for PH reactions were from Sigma-Aldrich and Fluka. Full details are given in the Supplementary Information.

Declaration of Competing Interest

There is no conflict of interest.