Water-assisted and protein-initiated fast and controlled ring-opening polymerization of proline N-carboxyanhydride

Abstract The production of polypeptides via the ring-opening polymerization (ROP) of N-carboxyanhydride (NCA) is usually conducted under stringent anhydrous conditions. The ROP of proline NCA (ProNCA) for the synthesis of poly-L-proline (PLP) is particularly challenging due to the premature product precipitation as polyproline type I helices, leading to slow reactions for up to one week, poor control of the molar mass and laborious workup. Here, we report the unexpected water-assisted controlled ROP of ProNCA, which affords well-defined PLP as polyproline II helices in 2–5 minutes and almost-quantitative yields. Experimental and theoretical studies together suggest the as-yet-unreported role of water in facilitating proton shift, which significantly lowers the energy barrier of the chain propagation. The scope of initiators can be expanded from hydrophobic amines to encompass hydrophilic amines and thiol-bearing nucleophiles, including complex biomacromolecules such as proteins. Protein-mediated ROP of ProNCA conveniently affords various protein-PLP conjugates via a grafting-from approach. PLP modification not only preserves the biological activities of the native proteins, but also enhances their resistance to extreme conditions. Moreover, PLP modification extends the elimination half-life of asparaginase (ASNase) 18-fold and mitigates the immunogenicity of wt ASNase >250-fold (ASNase is a first-line anticancer drug for lymphoma treatment). This work provides a simple solution to a long-standing problem in PLP synthesis, and offers valuable guidance for the development of water-resistant ROP of other proline-like NCAs. The facile access to PLP can greatly boost the application potential of PLP-based functional materials for engineering industry enzymes and therapeutic proteins.


INTRODUCTION
Proline (Pro) is the only proteinogenic amino acid bearing a secondary amine, which results from a circular side chain that loops back and reconnects with the backbone nitrogen. This pyrrolidine ring creates a sterically hindered nitrogen and constrains the conformation of both the Pro and its preceding amino acid residue. Not surprisingly, the Pro-Pro junction is even more restricted; as an extreme case, poly-L -proline (PLP) is a well-known rigid 'molecular ruler' that exists either as all-cis right-handed type I (PPI) helices in common organic solvents or as all-trans left-handed type II (PPII) helices in aqueous solution [1,2]. In naturally occurring proteins, proline-rich regions (PRRs) in PPII helices play important roles in regulating protein-protein and protein-nucleic acid interactions, signaling, mechanical elasticity, transcriptional activation, immune response, etc. [3][4][5][6]. In materials science, there is also a rapidly growing interest in prolineor hydroxyproline-derived polymers [7][8][9]. PLP derivatives have also demonstrated utility as mimics of antifreeze protein [10], building blocks for hierarchical self-assembly [11], templates for controlling Natl Sci Rev, 2022, Vol. 9, nwac033 nanoparticle growth [12], gelators [13], molecular rulers [14], antimicrobials [15] and cell-penetrating agents [16]. Despite the aforementioned progress, research to unlock the full application potential of PPII-helical PLP has been handicapped by the difficulties in synthesizing well-defined, high-molecular-weight (M n ) PLP, presumably due to the high steric hindrance and conformational constraint of Pro. This is evident even in native biosynthesis, where translation of Pro-rich sequences often induces ribosome stalling [17,18]. In chemical synthesis, the construction of a Pro-Pro junction via solid-phase peptide synthesis (SPPS) or native chemical ligation (NCL) is known to be inefficient and thus impractical for extending an oligoproline beyond 20 repeating units [19,20]. A number of early literature reports showed the ring-opening polymerization (ROP) of proline N-carboxyanhydride (ProNCA) to be a viable alternative, but the requirement of a strong base such as sodium methoxide inevitably resulted in racemization [21]. An additional caveat is that the M n of PLP in these early studies was estimated solely from viscosity. Progress in the ROP of ProNCA has been achieved in several recent studies using primary or secondary amines in anhydrous pyridine, dioxane or acetonitrile (ACN) under high vacuum or a continuous nitrogen flow (Fig. 1a) [13,22,23]. Previously, the best results were reported by Gkikas et al., who attained a M n up to 13 × 10 3 g/mol and dispersity (Ð = M w /M n ) of 1.23 based on size exclusion chromatographic (SEC) analysis in water/ACN (80/20) [22]. However, polymerization with amine-based initiators usually requires more than a week to reach 60%-80% monomer conversion in organic solvents. During the reviewing process of this work, Kramer achieved relatively faster PLP synthesis (1-2 h typically) via the use of highly efficient organometallic catalysts [24]. However, the M n and Ð of the obtained PLPs were not stringently characterized using SEC due to solubility issues. Moreover, a major flaw of all current methods is that PLP would precipitate prematurely in organic solvent in the form of PPI helices. As a result, time-consuming dialysis would be needed after the reaction to obtain watersoluble, PPII-helical PLP. Understandably, the inefficient synthetic method has proved to be a bottleneck for the biomedical and material application of PLP.
Controlled methodologies for the ROP of Ncarboxyanhydrides (NCAs) have bloomed in the past decade [25][26][27][28][29][30][31][32][33][34][35]. While the majority of these systems require stringent anhydrous conditions, in recent years the focus has been shifting toward achieving ROP in open vessels with aqueous solution [36][37][38][39]. To name a few, Cheng [37], Heise [38] and Lecommandoux and Bonduelle [39] have demonstrated the feasibility of controlled ROP of NCAs at the water-oil interface, in oil-in-water emulsion, or even in pure aqueous solutions. These pioneering works opened up a new avenue for developing water-tolerant ROP of NCAs. However, until now, aqueous-phase ROP of NCAs that generates PLP or other water-soluble polypeptides remains elusive.
Herein, we report the production of well-defined PLP with predictable M n s up to 18.7 × 10 3 g/mol and Ð in the range of 1.1-1.2, via the unexpected amine-initiated, water-assisted ROP of ProNCA in mixed ACN/H 2 O (Fig. 1b). In sharp contrast to previous methods, which typically require weeks to obtain PPII-helical PLP, we accomplished complete monomer conversion to product without induction in <5 min, sometimes even within one minute. Density functional theory (DFT) calculations of the model chain propagation reaction revealed that water played an as-yet-unreported role in assisting both the nucleophilic attack and hydrogen migration, which reduced the overall energy barrier of the polymerization by 7.1 kcal/mol. We also demonstrated that our method could be applied to the synthesis of protein-PLP conjugates via the site-specific grafting-to or randomly labeled grafting-from approach. Importantly, the PLP modification dramatically enhanced the stability of DHFR (dihydrofolate reductase) under extreme conditions while maintaining its enzymatic activity. Moreover, PLP modification prolonged the circulation time and mitigated immunogenicity of a therapeutic protein in a way that outperformed the well-established PEGylation. Together, these results highlight the enormous potential of PLP as a fully 'natural' and degradable polypeptide for industrial and biomedical applications.

RESULTS
Highly pure ProNCA was prepared using a moisture-resistant method that our lab recently developed [40], which significantly simplified the conventional procedure that involves laborious workup. The application of this method also improved the yield to 72%, up from 30% reported previously. We began our study by first performing primary amine-initiated ROP of ProNCA in dry ACN in a glovebox. Kinetic characterization suggested, rather unexpectedly, that the reaction began with an initial stage of fast polymerization with monomer conversion rapidly reaching ∼30% in 30 min, followed by a second stage in which another 10% of the monomer was slowly consumed over 2 days ( Fig. 1c and d, blue curves). SEC analysis of the PLPs obtained from different reaction time points consistently showed bimodal peaks with no significant increase in M n after 30 min of polymerization (Fig. 1e, blue curve, and Fig. S1). Because the precipitation of PLP occurred shortly after the addition of initiator (Fig. 1f), we hypothesized that this sequestered the growing PLP chain from the organic phase (i.e. ACN), leading to slow, uncontrolled polymerization that was the hallmark of the second stage. We further hypothesized that the above problem could be solved by adding water to the reaction to form a homogenous mixture. We thus tested eight common organic solvents, each mixed with water (v/v = 1/1) to serve as the solvent for the ROP of ProNCA. To our great surprise, polymerization in ACN/H 2 O was extraordinarily fast ( Fig. 1c and d, red curves), seemed to be well-controlled (Fig. 1e, red curves) and gave a clear solution without precipitation (Fig. 1f). Remarkably, the reaction consumed almost all monomer in <2 min at 10 o C, as evidenced by the generation of visible CO 2 bubbles immediately after the initiator was added ( Fig. 1g and supplementary video). A slightly slower ROP was observed in tetrahydrofuran/H 2 O, whereas all other solvent combinations led to bimodal/broader SEC traces (Fig. S2).
Subsequent reaction optimization in ACN/H 2 O yielded the following findings. First, unimodal SEC traces could be obtained at all tested temperatures in the range of 0 to 65 o C, with an inverse correlation between M n and temperature (Fig. S3). Second, pH variation between 5.0 and 9.0 seemed to have negligible effect on the M n (Fig. S4). Third, an initial monomer concentration ([ProNCA] 0 ) above 50 mg/mL was necessary to ensure both a fast reaction and good M n control (Fig. S5). Fourth, an H 2 O content of 40%-60% was optimal (Fig. S6); insufficient H 2 O would not completely dissolve PLP, whereas too much H 2 O appeared to be detrimental to M n control, likely due to the increased reaction competition from monomer hydrolysis. Fifth, ionic strength (i.e. concentration of NaCl) did not play a significant role in regulating the distribution of M n (Fig. S7).
We next characterized the reaction kinetics, M n , Ð, end group and chain extension to examine whether the ROP was fully living/controllable (Fig. 2). The polymerization was found to follow first-order kinetics ( Fig. 2a and   at various monomer-to-initiator molar ratios ([ProNCA] 0 /[I]). The chain propagation rate constant (k p ) was estimated to be 1.95-4.00 M -1 s -1 . Of note, primary amine-initiated ROP of NCAs produced k p values typically in the range of 10 -3 -10 -2 M -1 s -1 [27,41,42]. Varying the feeding [ProNCA] 0 /[I] ratio led to a linear increase in the M n of the PLP products , which were within 5% deviation from the theoretical values ( Fig. 2c and Table 1 showed a small tail at the lower M n region, suggesting the M n control reached its upper limit. The linear correlation between M n and the degree of monomer conversion was characteristic of chain-growth polymerization (Fig. 2d). The livingness of the ROP was further confirmed by the facile chain extension from an in-situ-generated PLP macroinitiator (Fig. 2e). Matrix-assisted laser desorption/ionization-time of flight (MALDI-TOF) mass spectrometric analysis of the PLP product from benzyl amine-mediated ROP at a [ProNCA] 0 /[I] ratio of 25/1 revealed a well-defined end group fidelity, where all peaks could be assigned to the molecular formula of BnNH-Pro n -H + Na + and none belonged to the chain initiation from hydrolyzed proline (Fig. 2f). Compared to the conventional method, which requires time-consuming dialysis to induce an incomplete PPI-to-PPII transition of the PLP product (  (Fig. 2g), whereas the ROP of D -ProNCA and DL -ProNCA furnished right-handed PPII- helical poly-D -proline (PDP) and disordered poly-DL -proline (PDLP) products, respectively (Fig. 2g).
The enhanced solubility of PLP in water was certainly a contributing factor to the faster ROP of ProNCA in ACN/H 2 O. However, a significant rate increase was also observed when the content of water was only 20% during our condition optimization (Fig. S6), at which the PLP was still insoluble. Moreover, the ROP was markedly accelerated even when only 2% H 2 O was added to ACN (Fig. S11). Collectively, these results indicated that the contribution of water must involve additional mechanisms apart from solubility improvement and an increase in solvent polarity/dielectric constant. We first ruled out the possible contribution of ProNCA hydrolysis based on two facts: (i) the MALDI-TOF spectrum displayed only the expected end group (Fig. 2f), which suggested very little initiation by hydrolyzed proline, and (ii) ProNCA remained intact for >4 min under the ROP conditions without an amine initiator (Fig. S12), which was longer than the time required for reaching complete ROP. We next disapproved the existence of cooperative polymerization, a self-accelerating mechanism that was elegantly demonstrated by Cheng and co-workers, in the ROP of NCAs, to produce αhelical polypeptides [42]. No induction period that corresponded to the nucleation stage, a well-known kinetic feature of cooperative polymerization, was observed in any of our kinetic studies ( Fig. 2a and  b). Moreover, there was no significant difference in the polymerization rate or M n control between the ROP of L-, D-and DL -ProNCA, regardless of whether the reaction was initiated by benzyl amine (Table 1, entries 3, 6 and 7) or the macroinitiator PLP (Fig. S13).
Interestingly, careful kinetic study revealed that the ROP was faster in 2% H 2 O/ACN than in 2% D 2 O/ACN (Fig. S11). This isotope effect implied that water molecules might be involved in the ratedetermining step (RDS), likely through hydrogen bonding or facilitating proton shift. We tested this hypothesis by conducting a detailed DFT study of a model chain-propagation reaction (Fig. 3) using (S)-N, N-dimethylpyrrolidine-2-carboxamide (SM) to represent the reactive chain end. It was found that the reaction in pure ACN proceeded as a threestage process that consisted of nucleophilic addition, intramolecular proton shift and decarboxylation. The transition state TS2a showed the highest activation Gibbs free energy of 26.6 kcal/mol, leading to the surprising finding that the proton shift, rather than the nucleophilic attack, was the RDS for the ROP of ProNCA in pure ACN ( Fig. 3 and Figs S14-15). In the presence of H 2 O, however, the same model reaction above was found to follow an alternative mechanistic route that could be roughly divided into four stages, including nucleophilic addition, ring-opening, intramolecular proton shift and decarboxylation. A key deviation from the ROP in anhydrous ACN was that H 2 O, both as a hydrogenbonding donor and acceptor, helped arrange SM and ProNCA in proximity to each other, which facilitated the nucleophilic addition ( Fig. 3 and Figs S16-17, TS1c). The activation Gibbs free energy of TS1c was 19.5 kcal/mol, which was the highest energy barrier for the whole process. Importantly, hydrogen bonding to water molecules stabilized the amide bond of the zwitterionic intermediate Int4c in a trans conformation, and facilitated the subsequent proton shift, leading to the final, spontaneous decarboxylation of Int5c. Overall, the presence of H 2 O shifted the RDS toward  nucleophilic addition, and dramatically lowered the overall energy barrier of the chain propagation by 7.1 kcal/mol, which could increase the rate of ROP by up to five orders of magnitude (Fig. 3).
To explore the scope of initiators, diethyl amine, p-methyl phenyl amine, p-methyl phenyl mercaptane and complex water-soluble biomolecules such as glucosamine (Fig. 4a) were tested and all mediated fast ROP of ProNCA under the optimized conditions. Among these initiators, p-methyl phenyl mercaptane, p-methyl phenyl amine and glucosamine afforded polymers with the predicted M n , whereas the M n of the diethyl amine-based polymer appeared to be slightly higher than expected ( Table 1, (Fig. 4b), which was almost identical to the expected M n . MALDI-TOF spectrometry and 1 H NMR spectroscopic analyses of these PLPs all revealed well-defined end groups (Figs S18-21). Notably, native chemical ligation (NCL) of MPT-PLP 50 with Cys-EGFP (enhanced green fluorescent protein bearing an N-cysteine) yielded the site-specific PLP-EGFP conjugate in 73% yield under mild conditions (Fig. 4c and d) [8,43]. This result was remarkable as proline alkyl thioester was well-known for its low reactivity for NCL.
Encouraged by the robustness of the ROP at various pH and salt concentrations (Figs S4 and S7), we further examined whether amine-bearing proteins could be utilized as macroinitiators to generate protein-PLP conjugates in situ (Fig. 5a) [44][45][46][47][48][49][50][51]. The protein-initiated ROP of various NCAs, in-Natl Sci Rev, 2022, Vol. 9, nwac033  cluding ProNCA, was previously explored in the 1950s and 1960s [52,53]. Here, we decided to revisit this topic thanks to the ultra-fast kinetics in mixed ACN/H 2 O. Gratifyingly, 10-min ROP of ProNCA at 10 o C in an ACN/phosphate buffer saline (PBS) (v/v = 1/1, pH 7.4) solution of EGFP led to the efficient formation of EGFP-PLP conjugates based on sodium dodecyl sulfate polyacrylamide gel elec-trophoresis (SDS-PAGE) analysis. When the feeding concentration of ProNCA was increased from 1.0 to 50 mg/mL, the products, appearing as smears, gradually shifted to higher molecular weights, with the upper limit exceeding 180 kDa (Fig. 5b). The average molecular weight of the EGFP-PLP conjugates was close to the theoretically predicted value, and no free EGFP was observed in the crude reaction mixture. These results were consistent with quantitative initiation efficiency and controlled PLP growth under these conditions. Of note, the fluorescence of EGFP was well maintained throughout the ROP and purification (Fig. 5b, inset). The ROP of PLP using dihydrofolate reductase (DHFR) as initiator yielded a product profile with a similar molecular weight distribution (Fig. 5c). Moreover, the resulting three DHFR-PLP conjugates exhibited almost identical enzymatic activity as untreated wt-DHFR (Fig. S23). Remarkably, one DHFR-PLP conjugate was found to retain ∼80% or 96% of the enzymatic activity of the native protein after 12 h of ethanol treatment or 10 min of heat shock at 80 o C. In contrast, wt-DHFR retained only 20% and 65% of the catalytic activity after the same ethanol and heat treatment, respectively (Fig. 5d). Inspired by these results, we next tested whether the growth of PLP on a therapeutic protein could improve its pharmacological properties. Asparaginase (ASNase) is a first-line enzyme-based anticancer drug for acute lymphatic leukemia and several aggressive subtypes of lymphoma, but is known for its notoriously poor pharmacokinetic profiles and strong immunogenicity, which necessitates polymer conjugation [54][55][56]. For this, we prepared an ASNase-PLP conjugate via the same grafting-from approach described above. Both SDS-PAGE and SEC characterizations confirmed that the ASNase-PLP conjugate was successfully synthesized and had a relatively uniform size ( Fig. 6a and b). In vitro, the purified ASNase-PLP conjugate showed a halfinhibition concentration (IC 50 ) of 4.3 ng/mL on the human NK/T lymphoma cell line NKYS (Fig. 6c), showing only a slight decrease in potency compared to its wild-type counterpart (IC 50 :2.4 ng/mL, no statistical significance). In vivo, ASNase-PLP injected via tail vein exhibited a significantly prolonged blood elimination half-life (t 1/2β ) of 26 h in SD rats, which was 18 times that of wt ASNase (Fig. 6d). More interestingly, in mice implanted with ∼300 mm 3 NKYS tumors, biweekly treatment of ASNase-PLP almost completely inhibited the tumor growth, and eventually led to its shrinkage, while wt ASNase produced almost no antitumor effect at the same dose and frequency of administration (Fig. 6e). Ki-67 staining of the tumor tissues confirmed that only ASNase-PLP treatment led to significant tumor For the immunization, SD rats (n = 4 for each group) were subcutaneously infused with wt ASNase, ASNase-PEG or ASNase-PLP (200 U/kg) once every week for totally four weeks. Antisera were drawn every week before each injection. For the ELISA assay, the plates were coated with (g) wt ASNase, a PLP-interferon conjugate (to detect anti-PLP antibodies) and (h, i) a PEG-interferon conjugate (to detect anti-PEG antibodies). Data are represented as mean ± SD. P values are determined by ANOVA (ns = no significant difference; * P < 0.05; * * P < 0.01; * * * P < 0.001).
killing (Fig. 6f). All mice exhibited remarkable tolerance of the drug conjugate at the applied dose based on the tracking of body weight (Fig. S24) and histological assessment of major organs (Fig. S25). These results vividly showed that extending the blood halflife of ASNase could significantly improve its antitumor efficacy in vivo. ASNase is a bacterial enzyme with extremely high immunogenicity, and the related side effects cannot be completely eliminated in clinical practice even through PEG modification. More worrisome is the hapten nature of PEG itself, which induces the production of anti-PEG antibodies after coupling with ASNase, which in turn leads to hypersensitivity and an accelerated blood clearance (ABC) effect. To interrogate whether PLP conjugation was a better alternative, we injected the rats with ASNase-PLP or ASNase-PEG subcutaneously once a week and collected antisera each time be-fore the injection. Enzyme linked immunosorbent assay (ELISA) analysis revealed that the titer of anti-ASNase IgG in the ASNase-PLP antisera harvested at week four was <400, compared to 12 800 and 102 400 for the antisera of ASNase-PEG and wt AS-Nase, respectively (Fig. 6g). Moreover, the levels of anti-polymer IgM and IgG in the ASNase-PLP antiserum were also significantly lower than in ASNase-PEG antiserum (Fig. 6h and i).

DISCUSSION AND CONCLUSIONS
PPII helices are among the most prevalent structural and functional motifs in proteins. Synthetic PPII-helical PLP can emulate the unique structural and functional roles of PPII helices, ultimately leading to the creation of novel biochemical materials.
However, conventional methods for PLP synthesis are slow and inefficient, require stringent moisturefree conditions, allow very little control of M n and Ð, and often result in premature product precipitation as PPI helices. In the current study, we demonstrated that controlled and dramatically accelerated ROP of ProNCA could be easily achieved by simply adding water as a co-solvent (Fig. 1). Notably, to obtain accurate M n and Ð information of the PLP, we employed an aqueous SEC system equipped with both multi-angle light scattering (MALS) and refractive index detectors, which allowed us to acquire the dn/dc value of the PLP and calculate the absolute M n . One notable limitation of many previous methods is that the scope of initiators was restricted to hydrophobic amines. This limitation was easily overcome in this study for the obvious reason that water is a good solvent for many small molecules and biomacromolecules. Indeed, we successfully expanded the scope of initiators to hydrophilic amineand thiol-bearing nucleophiles, including complex biomacromolecules such as proteins (Figs 4-6).
Moisture-tolerant ROP of NCAs has gained considerable attention in recent years with many notable advances. The unique feature of our work is that water played multiple beneficial roles in our reaction system rather than being a detrimental factor. Key contributing factors to the excellent reaction control that we observed included enhanced solubility of PLP in water and the unexpected waterassisted acceleration of the ROP, which allowed it to kinetically outcompete the monomer hydrolysis. DFT results revealed the crucial mechanistic role of water molecules in facilitating a hydrogen shift that lowered the energy barrier of the chain propagation by a substantial ∼7.1 kcal/mol (Fig. 3). However, it should be noted that at [ProNCA] 0 /[I] ratios higher than 200/1, the hydrolysis of ProNCA became significant.
Here, we showed that PLP was effective in protecting the conjugated protein from denaturation under harsh conditions such as high temperature and exposure to organic solvents. Interestingly, the PLP-grafted EGFP fully preserved the EGFP fluorescence and DHFR catalytic activity. There are three plausible reasons for this result. First, the fast kinetics of ROP allowed product generation within 10 min, and thus minimized the detrimental effects posed by exposure to the organic solvent. Second, it appeared that the growth of PLP on the surface of DHFR enhanced the resistance to organic solvent dramatically (Fig. 5d). Third, the substrates of the DHFR-catalyzed reaction were small molecules, whose diffusion and binding to the catalytic pocket was generally insensitive to the steric hindrance of PLP. Although this grafting-from approach might not be suitable for proteins whose biological functions are dependent on lysine(s) in the active pocket, it could still find broad application in a variety of proteins whose functions are lysine independent. For instance, our results showed that PLP modification substantially improved the pharmacokinetic properties of ASNase while largely preserving its catalytic activity, which eventually led to enhanced in vivo antitumor efficacy. Most interestingly, our results unambiguously proved that PLP could not only reduce the immunogenicity of the protein covalently attached to it, but its antigenicity per se was also lower than PEG. These results indicate that PLP, which is neutral in charge, biodegradable and completely natural, holds great potential as an outstanding alternative to the non-degradable PEG, whose clinical limitations have been increasingly recognized in recent years [54,60,[62][63][64][65].
Overall, this work provides a simple means of solving a longstanding problem in PLP synthesis and offers valuable guidance for the development of water-resistant ROP of other NCAs. The availability of PLP could facilitate our understanding of the biophysical and biological roles of PRRs in natural proteins and greatly boost the application potential of PLP-based functional materials for engineering of industrial enzymes and therapeutic proteins.