Accelerating the end-to-end production of cyclic phosphate monomers with modular flow chemistry

The biocompatibility, tunable degradability and broad functionalities of polyphosphoesters and their potential for biomedical applications have stimulated a renewed interest from Chemistry, Medicinal Chemistry and Polymer Sciences. Commercial applications of polyphosphoesters as biomaterials are still hampered because of the time and resource-intensive sourcing of their corresponding monomers, in addition to the corrosive and sensitive nature of their intermediates and by-products. Here, we present a groundbreaking challenge for sourcing the corresponding cyclic phosphate monomers by a different approach. This approach relies on the use of continuous flow technologies to intensify the end-to-end preparation of cyclic phosphate monomers with a semi-continuous modular flow platform. The applied flow technology mitigates both safety and instability issues related to the more classical production of cyclic phosphate monomers. The first flow module allows safe synthesis of a library of cyclic chlorophosphite building blocks and features in-line 31P NMR real-time monitoring. After optimization on the microfluidic scale, this first module is successfully transposed toward mesofluidic scale with a daily throughput of 1.88 kg. Downstream of the first module, a second module is present, allowing the quantitative conversion of cyclic chlorophosphites with molecular oxygen toward chlorophosphate derivatives within seconds. The two modules are concatenable with a downstream semi-batch quench of intermediate chlorophosphate with alcohols, hence affording the corresponding cyclic phosphate monomers. Such a continuous flow setup provides considerable unprecedented advantages to safely and efficiently synthesize a library of versatile high value-added cyclic phosphate monomers at large scale. These freshly produced monomers can be successfully (co)polymerized, using either batch or flow protocols, into well-defined polyphosphoesters with assessed thermal properties and cytotoxicity.


Introduction
With major roles as building blocks of the genetic information through deoxy-and ribonucleic acids and as cell energy currency in metabolic processes through adenosine triphosphate, polyphosphoester (PPE) structures are ubiquitous in Biology. 1 Contrastingly to polyesters (i.e. based on carboxylic acid esters), synthetic PPEs have only found scarce industrial applications such as ame retardants. [2][3][4] However, PPEs have recently gained a renewed interest from the academic community in light of their potential biomedical applications 5 such as drug delivery systems [6][7][8][9][10][11][12] and tissue engineering. [13][14][15][16][17] Synthetic PPEs are especially appealing for their tunable (bio)degradability and biocompatibility properties. 2,18 A convenient synthetic method to obtain PPEs involves the ring opening polymerization (ROP) of cyclic phosphate monomers (CPMs), which enables the preparation of high molecular weight materials with narrow polydispersities. 19 Among the various CPMs, 5-membered ring monomers are currently more popular compared to their 6-membered ring counterparts. The ring strain of the 5-membered CPMs makes them more reactive during polymerization, yet also more sensitive to hydrolysis, and prone to efficient organocatalyzed ROP. 2 In sharp contrast to classical aliphatic polyesters, the physicochemical properties of PPEs, such as hydrophobicity, 20,21 can conveniently be adjusted through straightforward modications of the pendant side-chain on the pentavalent phosphorous during the upstream monomer preparation. Moreover, introduction of a suitable functional pendant side-chain also enables postpolymerization modications 22,23 and introduction of bioactive molecules, 24,25 which further expands the scope of applications of PPEs.
One of the most signicant challenges met during the production of PPEs using ROP strategies actually relates to the sourcing of CPMs (Fig. 1). The preparation of such monomers is usually performed by a resource-and time-intensive 3-step procedure. The preparation of CPMs starts from PCl 3 and a suitable diol (1a) to yield a cyclic chlorophosphite adduct 2a with the concomitant release of 2 equiv. of gaseous HCl. Next, the chlorophosphite intermediate is further oxidized into its cyclic chlorophosphate counter-part 3a. The CPM is obtained aer a nal functionalization using an adapted alcohol. 18,26 This process involves the use of hazardous PCl 3 as starting material, corrosive and sensitive intermediates (chlorophosphi(a)tes 2a and 3a) and produces large amounts of HCl as a by-product. These circumstances lead to signicant safety concerns upon scalability trials.
To address these issues that challenge upscaling a modular semi-continuous ow platform dedicated to the production of CPMs has been developed. The platform allows to produce CPMs with unprecedented versatility, efficiency and scalability starting from bulk chemicals. The modular ow platform features the concatenation (i.e., the connection within a same uninterrupted system) of 3 chemical steps and mitigates the inherent safety hazards related to production of large amounts of by-product HCl, as well as to handling of corrosive intermediates and reactants. Moreover, this approach also signicantly reduces the overall process time for production of chlorophosphi(a)te intermediates in a continuous fashion (batch: 5 h to 4 d; this ow protocol: <2 min). This study also demonstrates that model CPMs are successfully polymerized through either batch or continuous ow protocols. The absence of cytotoxicity of selected PPE samples is also assessed, as well as their thermal properties. This contribution unlocks the production of CPMs and consequent PPEs, and thus opens new avenues to access novel biomaterials and large-scale production of industrially relevant PPEs by enabling their safer and efficient end-to-end production.

Results and discussion
Continuous ow preparation of cyclic chlorophosphites and analogs The preparation of cyclic chlorophosphites in batch typically relies on the method of Lucas et al., 27 which involves the careful addition of a diol to a solution containing PCl 3 with the subsequent release of HCl (2 equiv.). Even with appropriate care during the addition process, reckless release of HCl gas raises signicant safety concerns upon scale-up. Taking these hazards into account prompted us to adapt the synthesis of chlorophosphites under continuous ow conditions for enabling a smoother and more controllable generation of HCl gas, as well as a fast and efficient production of cyclic chlorophosphites and analogs.
The rst exploratory optimizations involved the preparation of model compound 2a through the addition of neat ethylene glycol (1a) to a concentrated solution of PCl 3 in CH 2 Cl 2 at room temperature. The outlet of the reactor was directly connected to a closed surge with a ow of an inert gas to continuously ush gaseous HCl out of the reactor effluent toward an alkaline aqueous trap. Starting from a 1 M PCl 3 solution, reducing the residence time to 1 min allowed complete conversion of PCl 3 ( Table 1, entries 1-4). Next, to maximize the output, the concentration of the PCl 3 solution was increased (entries 5 and 6) with an appropriate adjustment of the counter-pressure to avoid HCl degassing inside the coil reactor. A starting 3 M solution (entry 5) maintained complete conversion and excellent selectivity, whilst affording a 76% isolated yield aer fractional distillation under reduced pressure. A very high hourly productivity of 14.8 g was obtained using a 1 mL internal volume coil reactor, despite its minimal footprint. Further increasing the concentration of the starting solution (entry 6) led to a biphasic regime and to the appearance of major sideproducts; thus, the concentration was set to 3 M for the PCl 3 solution as a reference for the following experiments. A variety of solvents were assessed for the production of 2a, yet leading to similar (entries 7, 9 and 10) or lower (entry 8) yields. 2-Methyltetrahydrofuran (MeTHF) emerged as a more sustainable alternative to CH 2 Cl 2 , as it is biobased and more favorable than CH 2 Cl 2 according to the CHEM21 selection guide. 28 MeTHF also supersedes THF in terms of health and environmental scoring. In-line 31 P NMR was conveniently implemented downstream the reactor for real-time reaction monitoring, hence accelerating the optimization process as well as improving process safety (see ESI, Section 2.5.1 †). Implementation of in-line monitoring indeed contributes to ensure that hazardous PCl 3 is fully converted and that no major impurities are present, while avoiding time-consuming high-eld NMR analysis that could lead to a degradation of the crude sample containing a sensitive chlorophosphite. Chlorophosphite 2a was detected as a sole peak ($168 ppm) during in-line monitoring while the absence of PCl 3 ($219 ppm) was conrmed. No other peaks were observed by 31 P NMR. With a highly productive and compact continuous ow system that mitigates the hazardous preparation of 2a in hand, the scope of the procedure was next extended to a variety of cyclic chlorophosphites and analogs 2. Starting from various 1,2-diol and derivatives 1 (including potentially biobased derivatives 1a-g, j, l, m) [29][30][31][32][33] allowed to rapidly generate a library of chlorophosphites bearing a range of backbone modulations (Fig. 2). Good to high isolated yields were obtained for most examples, except for 2e for which the nal vacuum distillation led to signicant decomposition (see ESI, Section 2.4.1 †). Switching from a 1,2-diol to a 1,2-dithiol (1h) accessed dithiaphospholane derivative 2h in 58% of isolated yield, which was reported for the functionalization of nucleosides 34,35 and the preparation of phospholipid analogs. 36 With diverse applications as a phosphitylation agent for the preparation of glycosyl donors 37 or the derivatization of lignin samples for characterization by 31 P NMR, 38-40 the preparation of 2i was next envisioned. Exploring the reaction with hindered 1,2diol such as pinacol (1i) led to a major loss of selectivity and precluded the isolation of the product. According to the prior Art, 41 highly sterically congested pinacol-like diols are associated with a much slower formation rate toward the corresponding cyclic chlorophosphites, hence allowing for competitive formation of a biphosphite derivative. The latter, in presence of HCl, underwent dealkylation eventually leading to the formation of 2 equivalents of cyclic H-phosphonate. Attempts to avoid formation of such side-products relied on addition of a base to the reaction mixture; however, in order to avoid extensive precipitation of hydrochlorides upon continuous ow operation, organic bases known to form ionic liquids under protonation, such as 1-methylimidazole, were privileged. 42 Minor adaptations from the original continuous ow system were implemented with the selection of acetonitrile as a solvent to allow solubilization of the ionic liquid (see ESI, Section 2.5.2 †). Chlorophosphite 2i was isolated in 66% yield using the adapted system accordingly, leading to a hourly productivity of 7.3 g. With such an adaptation, preparation of more demanding building blocks 2j-m was foreseen (Fig. 2). The six-membered ring dioxaphosphinane 2j was successfully produced in the presence of 1methylimidazole, leading to a 53% isolated yield. This derivative is used in the preparation of various six-membered CPMs. 2,18 Oxathiaphospholane 2k and oxazaphospholidine 2i were produced in average to high yields, thus even further broadening the scope. Finally, the synthesis of the diazaphospholidine derivative 2m was developed but required additional adaptations to avoid clogging of the continuous ow system. Switching to 1,8diazabicyclo [5.4.0]undec-7-ene (DBU) as a base and lowering the concentration of the starting feed solutions prevented precipitation of solid residues, therefore enabling the isolation of 2m in a moderate yield (43%). This continuous ow protocol offers a compact solution for on-demand production of versatile cyclic P(III) adducts while enabling a gram-scale production in minutes and mitigating the safety issues related with synthesis of such derivatives.

Scalability of the continuous ow preparation of cyclic chlorophosphite 2a
The optimized procedure for the generation of cyclic chlorophosphites in a microuidic environment bears signicant potential for further scalability trials. 43,44 To this end, the preparation of 2a was transposed in a commercial glass mesouidic continuous ow reactor (see ESI, Section 2.4.3 †). The reactor setup consisted in two glass uidic modules (FMs), uidically connected in series (5.4 mL total internal volume). The production of 2a was conveniently handled with the controlled release of the HCl gas by-product through a dome-type back-pressure regulator (4 bar) downstream the reactor. The effluent was collected in a closed vessel under a continuous ow of argon to ush out HCl toward an alkaline aqueous trap. The mesouidic scale was associated with an isolated yield of 74% and a daily production of 1.88 kg under safe conditions (STY ¼ 348 kg per L per day).
Cyclic phosphate monomers (CPMs) production in a semicontinuous ow platform The concatenation of the following steps toward functionalized CPMs was envisioned. The oxidation of cyclic chlorophosphites is typically carried out with a stream of gaseous molecular oxygen which is bubbled in an organic solution containing the chlorophosphite. The reaction time needed for this batch procedure typically varies between 8 h to 4 days. 8,26,45 Continuous ow oxidation of the model compound 2a was optimized using molecular oxygen. The lab-scale continuous ow setup included the upstream generation of 2a, which was further reacted with oxygen in a second PFA coil reactor (see ESI, Section 2.5.3 †). Quick analysis of the effluent by a benchtop 31 P NMR allowed to qualitatively control the conversion and selectivity of the reaction. A signicant improvement of the selectivity was observed upon using a coil reactor with a smaller internal diameter and an arrowhead-type micromixer system. The mixing efficiency appeared as a much powerful leverage than residence time for the oxidation step, indeed, the slug ow regime of the gas-liquid system combined with a medium counter-pressure (15 bar), a smaller internal diameter of the coil reactor and a specic arrowhead micromixer (see Table S1 in the ESI †) leads to an improved interfacial area, mixing and solubility of oxygen within the liquid phase and, thus, to high mass transfer and improved performances. Quantitative conversion with 59% of selectivity toward 3a was obtained under 65 C within 21 s of residence time in the presence of 4 equiv. of molecular oxygen. Despite these very promising results, chlorophosphate 3a rapidly degraded aer collection. It was therefore decided to directly implement the last condensation reaction to access CPMs by directly collecting the reactor effluent in a surge containing an alcohol and a base. Such a fully concatenated protocol would not only enable the neutralization of the remaining HCl gas, but also the reaction of the intermediate chlorophosphate to form the desired CPM directly from PCl 3 and a diol. It was indeed foreseen that avoiding the handling of intermediate chlorophosphi(a)tes 2 and 3 would signicantly increase process safety and efficiency. Using the optimized conditions for the preparation of 2a and 3a (see Table 1, entry 10 and ESI, Table S3, entry 8 †), a semi-  Table 1, entry 10) or a base-involving (see ESI, Section 2.5.2 †) procedure. continuous ow system was developed (Fig. 3). The upstream generation of 2a allowed its direct and quick conversion to 3a and a nal functionalization toward a CPM in the semi-batch reactor using an adapted alcohol in presence of pyridine. This exible process allowed generation of various CPMs from bulk chemicals by conveniently switching the alcohol present in the collection surge. Two model monomers EEP and BEP were isolated in less than a day starting from PCl 3 and 1a using this procedure and required only one nal purication step. Moreover, the overall isolated yields obtained (39% and 49% for EEP and BEP, respectively) over the 3 chemical steps outperform the typical output achieved in batch (17-21%). 8,46 The concatenation under continuous ow indeed contributes to reducing the number of purications required and has thus a positive impact on the overall yield. Minor traces of a cyclic H-phosphonate were noticed in samples of EEP and BEP (see ESI, Section 2.5.4 †) potentially highlighting a partial hydrolysis of 2a in the system. 47 Starting directly from puried 2a led to a high purity monomer free of H-phosphonate impurities through a 2-step semi-continuous ow procedure (see ESI, Section 2.5.5 †).
A novel EClMEP monomer bearing an ethoxy pendant group, additionally to a chloromethyl functionality linked to the cyclic backbone, was successfully isolated using an adapted procedure (see Fig. 3 and ESI, Section 2.4.5 †). The 2-step process started from puried 2g, which allowed to avoid side-products from the synthesis of the chlorophosphite and, subsequently, to obtain high purity EClMEP in 70% of isolated yield over 2 chemical steps. A batch procedure for the preparation of the monomer EClMEP was additionally developed (see ESI, Sections 2.5.6-2.5.8 †).

Batch and ow (co)polymerization of cyclic phosphate monomers (CPMs)
The development of a semi-continuous ow system for the convenient and safer preparation of CPMs paves the way toward a better accessibility to PPEs as source for potential ameretardants or biomedical applications at lab-and industrialscale. To demonstrate the potential of this methodology, EEP was rst selected as a reference monomer to assess its polymerizability. EEP ROP was rst carried out in a batch reactor using DBU and a thiourea derivative (TU) as a well-established organocatalytic system 48 in the presence of tetraethylene glycol (TEG) as an initiator ( Table 2, entries 1-3). The EEP monomer produced through the semi-continuous ow platform was successfully polymerized accordingly (entry 1) with a good control of the macromolecular parameters. A well-dened polymer TEG-PEEP homopolymer with a narrow dispersity (Đ ¼ 1.14) and an average molar mass of 13 600 g mol À1 close to the targeted value (12 000 g mol À1 ) was obtained. The small molar mass difference was attributed to unavoidable trans-esterication side-reactions at high monomer conversion. 48 Polymerization reactions can also benet from continuous ow systems, leading to lower dispersity and improved reproducibility by taking advantage of better thermal management and mass transfer. 49 The homopolymerization of EEP was further explored under microuidic conditions (entries 2 and 3) with inspiration from the study of Junkers et al. 23 The outlet of the reactor was directly collected in cold Et 2 O, triggering precipitation of PPE. A residence time of 10 min at 0 C (entry 2) led to a molar mass of 5300 g mol À1 and a low dispersity (Đ ¼ 1.06). Interestingly, increasing the residence time to 20 min (entry 3) led to an increase of molar mass (M n NMR ¼ 9400 g mol À1 ) while keeping a low polydispersity index (Đ ¼ 1.11). The molecular weight distributions were monomodal for TEG-PEEP homopolymers obtained under ow conditions, whereas a shoulder toward high molecular weight was observed for the TEG-PEEP sample obtained under batch conditions (see ESI, Sections 2.8.1-2.8.3 †).
Design of novel block or gra copolymers based on the association of poly(ethylene oxide) (PEO) and PPEs have already demonstrated their potential for applications such as drug delivery systems 9,24,25,50-52 and hydrogels. 53,54 High reproducibility of continuous ow procedures is particularly appealing for preparation of such biomaterials to face demanding regulatory policies required for potential commercialization phases.
Using a PEO-based macroinitiator (MeO-PEO-OH, 5000 g mol À1 ) in conjunction with a DBU/TU catalytic system and based on the batch study of Jérôme et al., 20 the polymerization of EEP was carried out toward the formation of a MeO-PEO-b-PEEP diblock copolymer under continuous ow conditions (entry 4). Performing the reaction at 0 C within a residence time of 20 min led to the targeted M n of 7000 g mol À1 toward a well-dened block copolymer (Đ ¼ 1.05). Cytotoxicity assays of MeO-PEO-b-PEEP on bovine broblast cells and human umbilical vein endothelial cells revealed that no cytotoxic effects were observed for a concentration of up to 0.1 mg mL À1 of copolymer (see ESI, Section 2.5.12 †). These results are in agreement with data collected in a previous study regarding PEO-PPE diblock copolymers, 55 which paves the way for a potential biomedical application.
Following the encouraging results collected for the polymerization of EEP, the polymerization of EClMEP (obtained from 2g, see Fig. 3) was investigated for the production of novel PPEs. Indeed, the presence of a chloromethyl functionality on the polymer backbone will rst modulate the physicochemical properties of the corresponding poly(EClMEP) and additionally enable further post-polymerization derivatization as similarly described in literature for poly (3-caprolactone). 56,57 Moreover, the introduction of a chloromethyl moiety directly linked to a carbon of the polymer main-chain could render a more stable functionality aer derivatization over elimination by hydrolysis compared to a more conventional functionalization through the pendant side-chain linked to the phosphorus atom which can be degraded under basic or acidic conditions. 58 For this purpose, homopolymerization of EClMEP monomer was carried out in a batch process with similar conditions used for EEP. Unfortunately, these conditions are not adapted for the polymerization of EClMEP and did not lead to the formation of the expected homopolymer. This might be explained by a higher steric hindrance compared to the EEP monomer but also by the potential formation of a secondary alcohol at the u chain-end aer the initiation step, which could be insufficiently reactive to polymerize further with a second EClMEP. Nevertheless, random copolymerization of EClMEP with EEP using benzyl alcohol (BzOH) as an initiator was successful when the molar ratio of EClMEP in the co-monomer feed was lower or equal to 0.50 (Table 3, entries 1 and 2). BzO-PEEP-co-PEClMEP copolymers of two different compositions and molar mass were prepared. Monomodal molecular weight distribution with low dispersity were obtained for both copolymers.
Regarding the effects of the chloromethyl functionality on the physicochemical properties of both BzO-PEEP-co-PEClMEP copolymers as compared to a BzO-PEEP homopolymer reference (entry 3), a limited inuence on the value of the glass transition temperature (T g ) of the copolymers was observed. Both copolymers are amorphous with a T g varying between À70 and À60 C. Concerning the thermal stability, measured by TGA  (see ESI, Sections 2.8.5-2.8.7 †), the presence of the chloromethyl moiety modies the degradation prole of BzO-PEEPco-PEClMEP copolymers with an increase of the percentage of combustion residues at 600 C compared to BzO-PEEP. These results pave the way for potential applications of these new copolymers as ame-retardant materials which are currently investigated in a follow-up study.

Conclusions
A semi-continuous ow modular platform is described and allows convenient and safer preparation of CPMs directly from bulk chemicals. The rst module was dedicated to the preparation of cyclic chlorophosphite building blocks with high productivities despite a small footprint system. This procedure allowed to efficiently mitigate the safety concerns inherently associated with the preparation of CPMs, while beneting from real-time 31  The preparation of a model cyclic chlorophosphite was transposed to pilot-scale with a mesouidic reactor upon minor parameters adjustments leading to a daily productivity of $2 kg. Next, the concatenation of the rst module was implemented with (a) a downstream continuous ow oxidation using molecular oxygen and (b) a subsequent nal functionalization using an alcohol in a semibatch reactor. The oxidation step toward a cyclic chlorophosphate was intensied by taking advantage of a high mixing efficiency of the continuous ow system leading to a residence time of 21 s to achieve a quantitative conversion of the chlorophosphite. Finally, both cyclization and oxidation modules were telescoped with a downstream semi-batch reactor to conveniently perform the nal functionalization toward CPMs using an adapted alcohol. The fully integrated semicontinuous ow platform allowed the production of CPMs while mitigating signicant safety hazards. Next, a model CPM, EEP, obtained from the semi-continuous ow platform was successfully (co)polymerized by batch and continuous ow procedures leading to well-dened and biocompatible PPEs. Finally, EClMEP, a novel CPM bearing a chloromethyl functionality was successfully copolymerized leading to new PPEs with promising thermal properties. With the development of a semi-continuous ow platform for the efficient end-to-end production of CPMs, we forecast an accelerated and easier access to PPEs and their industrial applications by enabling a safer, more convenient and scalable preparation of the source monomers.

Data availability
All experimental supporting data are provided in the ESI. †

Author contributions
RM planned, executed, optimized the experiments in batch and ow for the preparation of CPMs and wrote the manuscript. RM and RR performed the polymerization experiments in batch and ow. RR executed the assessment of thermal properties of PPEs. NMSVDA and DGMM are responsible for the biocompatibility studies and related cytotoxicity assays and provided feedback to the manuscript. CJ advised and proofread the manuscript. JCMM supervised the project, advised, and corrected the manuscript.

Conflicts of interest
There are no conicts to declare.