Semiautomated synthesis of sequence-defined polymers for information storage.

Accelerated and parallel synthesis of sequence-defined polymers is an utmost challenge for realizing ultrahigh-density storage of digital information in molecular media. Here, we report step-economical synthesis of sequence-defined poly(l-lactic-co-glycolic acid)s (PLGAs) using continuous flow chemistry. A reactor performed the programmed coupling of the 2-bit storing building blocks to generate a library of their permutations in a single continuous flow, followed by their sequential convergences to a sequence-defined PLGA storing 64 bits in four successive flows. We demonstrate that a bitmap image (896 bits) can be encoded and decoded in 14 PLGAs using only a fraction of the time required for an equivalent synthesis by conventional batch processes. Accelerated synthesis of sequence-defined polymers could also contribute to macromolecular engineering with precision comparable to natural precedents.

INTRODUCTION

DNA stores genetic information as a sequence of four monomers having different pendant nucleobases constituting a polymer chain. Chemical synthesis of DNA based on the solid-phase synthesis enables encoding of nonbiological information at a far greater density than any existing magnetic, optical, and electronic media without a periodic rewriting owing to the deterioration of media over storage (–). Well-established sequencing technologies can efficiently decode the information stored in DNA (, ). Consequently, in recent decades, DNA has been pursued as a macromolecular medium to store digital information, which could circumvent the impending shortage of storage capacity due to the accelerated production of digital information in the age of Internet and mobile devices (, ). However, information encoding in DNA primarily relies on the repetitive addition of individual monomers to define nonbiological sequences. Considering that one repeating unit of DNA of an average molecular weight of 309 Da stores less than 2 bits (), this encoding based on repetitive monomer additions consumes time and cost and, thus, remains a substantial obstacle for realizing large-scale information storage in molecular media (, ).

Sequence-defined synthetic polymers can serve as an alternative to DNA as information-storing molecular media because of their efficient and low-cost synthesis through streamlined chemical reactions with simpler monomers at the expense of losing biological functions such as replication and enzyme-based sequencing (–). We recently reported that the sequence of a copolyester built with two monomers, representing 0 and 1, could be unambiguously defined by a step-economical pathway, that is, the cross-convergent approach (). Compared to the solid-phase synthesis for sequence-defined macromolecules (, ), the cross-convergent method facilitates encoding a binary code up to the length of up to 256 bits as a sequence of monomers with a minimal number of synthetic steps and without using a large excess of reagents (). The copolyester, poly(l-lactic acid-co-glycolic acid) (PLGA), resulting from cross-convergent synthesis using l-lactic acid and glycolic acid as the building blocks can store 1 bit per 60 Da of mass, which is a twofold greater density than that of DNA. Theoretically, 2 g of sequence-defined PLGA can store the entire information that is currently stored in data centers worldwide (~2500 exabytes) ().

Despite step-economical synthesis of sequence-defined polymers using the cross-convergent method, realizing large-scale information storage in PLGA mandates the synthesis of a multitude of PLGA chains. For example, the encoding of 106 bit (1 Mbit) of information requires 15,625 sequence-defined PLGA chains built with 64 monomers (64-mer). Here, we demonstrate that this challenge could be overcome through automated and programmable synthesis of PLGA. Continuous flow chemistry to synthesize sequence-defined polymers could offer several advantages over the repetition of traditional batch processes such as integrating individual reactions in a single operation, reducing reaction time, easy scale-up synthesis, and massively parallel operation (–). This process can handle the complexity involved in the generation of aperiodic sequences without resorting to the repetition of batch reactions by manually selecting the required building blocks.

RESULTS AND DISCUSSION

Synthesis of sequence-defined tetrads using continuous flow chemistry

The cross-convergent strategy for sequence-defined polymers consists of two distinct stages (Fig. 1). In a divergent stage, two monomers, L and G representing 1 and 0, respectively, were coupled to form four sequenced dimers (00, 01, 10, and 11) with orthogonal protecting groups; tert-butyldimethylsilyl (TBDMS) for the hydroxyl group (O terminus) and benzyl (Bz) for the carboxylic acid end group (C terminus). Using a typical batch process, these dyads, that is, the dimers covering all possible permutations of 2-bit signals, were cross-converged to generate 16 tetramers (tetrads) to encode all possible binary sequences via orthogonal deprotections and esterification. This divergent stage introduces complexity in synthesis as the encoding of tetrameric sequences requires cross-convergences between four dyads, thereby resulting in the exponential increase in the number of permutations of L and G. To synthesize all tetrads by conventional batch processes, 48 individual reactions with purifications over 16 days should be performed.

Fig. 1.

Cross-convergent synthesis of sequence-defined copolymers.

(A) Cross-convergent synthesis of a tetramer, LLLL. A dimer is converted to tetramer via orthogonal deprotection and esterification. (B) Synthetic strategy for sequence-defined polyester. In divergent stage, all possible tetramers are generated by cross-convergent synthesis of dimers. In convergent stage, sequence-defined polyester with 2 repeating units is produced by n times successive cross-convergent synthesis of tetramers. Magenta and blue circles represent glycolic acid and l-lactic acid units.

To convert the divergent stage to a single continuous flow, we devised a semiautomated process that can generate all sequence-defined tetrads in a single continuous flow setup (Fig. 2A). The dyads of L and G having TBDMS and Bz protecting groups were synthesized by batch reactions on a large scale (>50 g). As a test reaction, we programmed the flow setup to synthesize a tetrad 1111 using a dyad 11 as a building block. The process started with the selective injection of 11 (100 mg, 0.8 M in CH2Cl2) out of four building block solutions to both deprotection lines at a flow rate of 0.1 ml min−1 for 10 min through a Teflon capillary (diameter, 0.5 mm), which was regulated by a programmable six-way syringe connected to a computer. For the deprotection of the O terminus of 11, the injected solution was mixed with trifluoroborane etherate (BF3∙Et2O, 4 M in CH2Cl2) through a T-mixer. The synchronization of flows in the T-mixer was achieved by adjusting the length of capillaries. This mixture was allowed to react in the reaction loop (the volume Vrl of 2 ml), followed by the passage through an in-line module for the extraction of the resulting solution with water (fig. S7). Similarly, for the deprotection of the C terminus, the 0.8 M solution of 11 in tetrahydrofuran (THF) was injected to mix with the 2.4 M solution of triethylsilane (Et3SiH) in THF, which was subsequently passed through a column of palladium catalyst (Pd/C, 10%) equipped with a degasser to conduct hydrogenation (fig. S8). Last, the deprotected dyads were synchronously combined and mixed with a solution of 1-ethyl-3-(3-dimethylaminopropyl)carbodiimide (EDC) to perform esterification in the reaction loop (Vrl = 6 ml) (fig. S9). Thereafter, the coupled tetrad 1111 was collected using a fraction collector, which was subsequently purified using a 10-min run on an automated column chromatography instrument using silica column. After purification, 83 mg of pure 1111 (60% yield) was obtained after a full flow of the dyad solutions to the coupled tetrad for 35 min (fig. S10). This modest coupling yield was attributed to the presence of impurities from the hydrogenation with Et3SiH. However, upon addition of the off-line purification step after hydrogenation, the coupling yield from the continuous flow process was comparable to or higher than the yield of the corresponding batch reaction.

Fig. 2.

A continuous flow process to synthesize sequenced tetrads in a programmable fashion.

(A) Schematic illustrations of the continuous flow process for the preparation of sequence-defined tetramers. Synchronous deprotection of selected dyads and subsequent coupling are performed. All permutations of tetrads are generated by a single continuous flow. (B) 13C NMR spectra of 16 tetramers. Blue, green, yellow, and red squares (l-lactic acid) or circles (glycolic acid) represent the first, second, third, and fourth repeating units, respectively.

For serial synthesis of all possible tetrads, we performed 16 convergence steps to synthesize all tetrads in a gram scale in a single continuous flow with an independent feed of four dyads to the deprotection reactor lines through programmable syringes controlled by the software. Each convergence process consisted of 50-min reaction flow and 10-min purge with two pure solvents to clean up the lines. After the flow, the synthesized tetrads were individually collected in 16 test tubes in a fraction collector. The reaction time for the synthesis of all 16 tetramers in a single flow was 16 hours. After off-line purification, the yields of the tetrads were in the range of 50 to 61% (0.94 to 1.20 g) (table S4).

Sixteen tetrads could be unambiguously sequenced by 13C nuclear magnetic resonance (NMR) spectroscopy. In a 13C NMR spectrum, four peaks corresponding to the α carbon of each repeating unit of 1111 appear at the chemical shift of 68.1, 68.7, 69.0, and 69.4 parts per million (ppm), which correspond to the L residues from the O terminus to C terminus. Similarly, the peaks corresponding to four α carbon of G units of 0000 appear at the chemical shift of 61.3, 60.2, 60.7, and 61.1 ppm. With these assignments as references, all tetrads consisting of L and G could be sequenced by 13C NMR (Fig. 2B and fig. S11) without using mass spectrometry (MS). We also note that this spectroscopic sequencing of oligo(l-lactic acid-co-glycolic acid) (OLGA) by NMR could be applied to 8-mers without reading errors (fig. S12).

Encoding process to sequence-defined polymers via continuous flow synthesis

To demonstrate the potential of the continuous flow process for accelerated synthesis of a set of sequence-defined polymers, we encoded a low-resolution bitmap data (896 bits) converted from a photograph in a series of 64-mer PLGA chains (Fig. 3). Each PLGA of the 14-chain library contained fragmented data (56 bits) with an 8-bit address at the O terminus for the chain identification by sequencing. In a divergent stage, the reaction time for the synthesis of each tetrad was adjusted by setting the flow time of the required dyads for the coupling according to the relative occurrence of the tetrad in the sequences of a series of PLGA chains corresponding to the encoded information (fig. S13). After the divergent stage, the resulting library of tetrads was cross-converged to encode information in 64-mer PLGA through four successive couplings using the product of the previous convergence as the building blocks (). Each coupling process, which required a flow time of 30 min, was followed by off-line purification using an automated silica column chromatography for up to 32-mers and preparative size exclusion chromatography (prep-SEC) for polymers with higher molecular weights. The total process time for the synthesis of 64-mer PLGA from tetrads was 8 hours including off-line purifications after each convergent process. Last, we repeated the cross-convergence to complete the synthesis of 14 PLGA chains.

Fig. 3.

Storing information in sequence-defined polymers.

Process to encode a bitmap image in a multiple of sequence-defined PLGAs. Original image is converted to bitmap image, and bitmap image is converted to binary code. Binary information is divided into several polymer chains with 64 repeating units by continuous flow synthesis. Sequence-defined PLGA is composed of 8 bits of address and 56 bits of fragmented data. Fourteen synthesized sequence-defined PLGAs are represented by bar code. Red and blue rectangles represent glycolic acid and lactic acid unit.

The resulting 64-mer PLGAs were unambiguously characterized by 1H NMR, analytical SEC, and matrix-assisted laser desorption/ionization–time-of-flight (MALDI-TOF) MS (Fig. 4). The mass analysis of the PLGA chains provides the number of L and G repeating units constituting the polymer chain since the total number of repeating units was fixed to 64. In addition, 1H NMR analysis revealed the identity of the first repeating unit from the O terminus of PLGA as the peak corresponding to the α proton of the L unit in the vicinity of Si atom typically appears at 4.39 or 4.43 ppm, and the G unit appears at 4.32 to 4.36 ppm (fig. S14). Considering this information, the complete sequences of the PLGA chains could be decoded by tandem MS using MALDI-TOF MS. The fragmentation pattern under the condition used for the tandem mass sequencing revealed that the G units were fragmented with an adjacent L. In contrast, the L was fragmented as a single repeating unit (fig. S15). Reading two series of decreasing masses of the fragments (O to C terminus and C to O terminus) from the parent mass value, the entire sequence of the PLGA was decoded without any error by a single tandem mass spectrum (Fig. 4D).

Fig. 4.

Characterization of sequence-defined polymers.

(A) 1H NMR spectrum of sequence-defined PLGA. A total of 4.3 to 4.5 ppm of region indicates the first repeating unit from O terminus. (B) Summarized table of the information of sequence-defined PLGA. NLA indicates the number of the lactic acid units. X indicates the first repeating unit from O terminus. (C) MALDI-TOF mass spectrum of PLGA chain 5 (theoretical mass, 4532.13 Da; experimental mass, 4532.63 Da). (D) Tandem mass spectrum of PLGA chain 5 (parent ion, 4532.63 Da). The sequence-defined PLGA is decoded by reading two series of decreasing masses of fragments (lactic acid and glycolic acid units have 72 and 58 molecular weight). Blue and orange boxes represent the sequence from O terminus and C terminus. m/z, mass/charge ratio.

Decoding process of sequence-defined polymers by tandem MS

As indicated by the mass analysis of the PLGA library, the combination of two monomers at a fixed chain length (64 repeating units in this study) frequently introduces duplication of molecular weights of PLGAs, despite different encoded information. For example, the masses of chains 1 and 11 were identical at 4504.6 Da with the same composition of L and G units but different arrangement. The equimolar mixture of these PLGAs only showed a single peak in MALDI-TOF mass spectrum; thus, both the chains were indistinguishable in the mass spectrum. This requires each PLGA chain to be stored separately, which complicates the storage and decoding of the PLGA library. We performed tandem mass sequencing of a mixture of chains 1 and 11 (1:1, w/w) under the same condition as that for single-chain sequencing (fig. S16). The resulting mass spectrum revealed two independent series of peaks, which were identified by reading the sequence of the first eight residues from the O terminus, the 8-bit address implemented as a chain identifier in PLGA. The assignment of these series of mass peaks allowed us to sequence chains 1 and 11 simultaneously from one tandem mass spectrum.

This simultaneous mass sequencing of multiple PLGA chains sharing molecular weights allowed bundling of the multiple chains, encoding different information, together as a piece of plastic (Fig. 5A). The information stored in the bundle could be decoded by MS without separating the mixture to individual chains. The MALDI-TOF mass spectrum of this plastic showed a set of 10 peaks, indicating the mass duplications between PLGA chains (Fig. 5B).

Fig. 5.

Decoding of stored information in the sequence-defined PLGAs.

(A) Photograph of polymer resin mixed of 14 sequence-defined PLGAs. (B) MALDI-TOF mass spectrum of the polymer resin. Each peak corresponds to the molecular weight of 14 polymers. (C) Tandem mass spectrum of a parent ion (PLGA chains 3 and 6, 4546.86 Da). Both fragments generated from chains 3 (green diamond) and 6 (violet diamond) are shown in the spectrum. Each stored information can be decoded by reading 8-bit address codes (Si-GGLLLLLL, chain 3; Si-GLLGLLLL, chain 6). Decoded information is represented by bar code. Fifty-six–bit fragmented code corresponds to the bitmap image.

This mixture of 14 chains was subjected to simultaneous mass sequencing by targeting each mass peak including four duplications of the masses of PLGA chains. The mass sequencing decoded the entire information distributed among the 14 PLGA chains (Fig. 5C). Our demonstration suggested that the encoding and decoding of digital information distributed in a multitude of sequence-defined PLGAs could be streamlined via the efficient on-demand synthesis of polymer chains using continuous flow chemistry and the collective sequencing without handling individual polymer chains ().

Macromolecular engineering of synthetic polymers through accelerated synthesis

The continuous flow process to construct sequence-defined copolyesters shown in this work could also serve as a platform for the molecular engineering of polymers at a comparable precision exhibited by biopolymers such as proteins. Because of the vast amount of information generated by aperiodic sequences of monomers, accelerated synthesis of prototype polymers having different sequences should be synthesized for screening their properties depending on the specific information encoded as composition and arrangement of constituting monomers (, ). For example, we synthesized a series of stereoregular polyhydroxyalkanoates (PHAs) using diastereomeric building blocks composed of (R)- or (S)-enantiomers of 3-hydroxybutanoate (3HB) and 4-hydroxypentanoate (4HP). Four tetrameric building blocks were built from the dimers of 3HB-alt-4HP having two stereocenters by a single continuous flow process described in this work (fig. S17). The convergence of these building blocks resulted in the synthesis of monodisperse poly(3HB-alt-4HP) having two stereoregular repeating units (up to 32 units) alternating each other (fig. S18). Sequence-defined PHAs without molecular weight distribution have not been synthesized by conventional polymerization or biosynthesis using microbes (–).

The screening of their structural, optical, and physical characteristics by 13C NMR, circular dichroism spectroscopy, and differential scanning calorimetry of a series of stereoregular PHAs revealed the differences between erythro-diisotactic and threo-disyndiotactic poly(3HB-alt-4HP)s without molecular weight distribution (figs. S19 and S20). These differences in properties arising from the difference in stereochemical regularity persisted when the PHAs became polydisperse in molecular weight, which were prepared by the condensation polymerization of tetrameric building blocks (figs. S21 and S22). We envisage that the massively parallel synthesis of sequence-defined polymers through continuous flow chemistry should contribute to the molecular engineering of synthetic polymers for solving important technical and societal challenges such as the development of molecular media for information storage, degradable polymers, biomedical materials, and delivery of biomacromolecules.

MATERIALS AND METHODS

Materials

l-Lactic acid (≥98%), trifluoroborane etherate, Pd/C (10%), 4-(dimethylamino)pyridine, and N,N′-diisopropylcarbodiimide were purchased from Sigma-Aldrich and used without further purification. Glycolic acid (≥98%) and 1-(3-dimethylaminopropyl)-3-ethylcarbodiimide hydrochloride were purchased from Tokyo Chemical Industry and used without purification. Dichloromethane (DCM) and THF were distilled over CaH2 under N2. Legato 101 syringe pump was purchased from KD Scientific. Cadent 3 syringe pump was purchased from IMI Norgren. SEP-10 was purchased from Zaiput Flow Technologies. Perfluoroalkoxy tubing (1/16″ outside diameter/0.02″ and 0.03″ inside diameter) was purchased from Revodix. Omnifit EZ column was purchased from Diba Industries Inc. Gastorr AG-42-01 was purchased from GL Sciences. CF-2 fraction collector was purchased from Spectrum Chemical Mfg. Corp.

Synthesis of sequence-defined polymers using continuous flow synthesis

In divergent stage, 4 mmol of dyad (0.8 M, 5 ml) was injected to the deprotection of TBDMS or Bz module for 50 min. Subsequently, flow lines were washed with pure solvents for 10 min. The fraction collector allowed to receive produced tetrad into separated glass tube. Sequence generation and washing cycle were repeated 16 times. Collected tetrads were purified with automated column chromatography as eluent of hexane/ethyl acetate mixture. In convergent stage, three-way syringes were used for injecting sequence-defined OLGA protected with TBDMS and Bz groups (LAGA and LAGA, 0.8 to 0.1 M and 30 to 40 min according to the molecular weight). Crude product having twice repeating units (LAGA) was purified by column chromatography or prep-SEC.

Synthesis of stereoregular PHAs

Diastereomeric building block protected with TBDMS and Bz group was dissolved in DCM. The solution was cooled to 0°C, followed by addition of trifluoroborane etherate (4 equiv.). The reaction mixture was stirred at room temperature for 4 hours. The reaction was quenched with saturated NaHCO3 and diluted with water. The separated organic layer was dried over MgSO4, and the solvent was removed under reduced pressure. The resulting mixture was purified by column chromatography. Pd/C (10%, 0.05 eq.) was added to hydroxyl compound dissolved in ethyl acetate. The suspension was purged with hydrogen gas at room temperature for 3 hours. Upon completion of the reaction, the catalyst was removed through passing Celite cake and washed with ethyl acetate. The solvent was removed under reduced pressure. The resulting diastereomeric compound having O terminus and C terminus was dissolved in DCM. The solution was cooled to 0°C, followed by addition of N,N′-diisopropylcarbodiimde (1.5 eq.) and 4-(dimethylamino)pyridinium 4-toluenesulfonate (DPTS; 0.2 eq.). The solution was stirred at room temperature for 6 hours. The polymer was purified by precipitation in methanol.

Characterization

1H and 13C NMR spectra were on a Varian INOVA 500-MHz NMR spectrometer using CDCl3 as solvent. Gel permeation chromatography (SEC) was performed on an Agilent 1260 Infinity equipped with a PLgel 5-μm MIXED-D column and differential refractive index detectors. THF was used as an eluent with a flow rate of 0.3 ml min−1 at 35°C. A polystyrene standard kit (Agilent Technologies) was used for calibration. Automated column chromatography was performed on a Biotage Selekt flash chromatography purification system equipped with a Sfär silica column cartridge. n-Hexane and ethyl acetate were used as eluent. SEC was performed on a Recycling Preparative HPLC (high-performance liquid chromatography) (LC-9260 NEXT, Japan Analytical Industry) system equipped with JAIGEL-2.5HR/2HR columns and a differential refractometer. Chloroform (J.T.Baker) was used as an eluent with a flow rate of 10 ml min−1. Before injection, the solution was filtered through a polytetrafluoroethylene syringe filter (Whatman; 0.2-μm pore). The SEC was performed under a cycling mode until the coinciding peaks were separated. The desired fraction was collected using a fraction collector.

MALDI-TOF mass analysis of polymers

Molecular weights of sequence-defined PLGAs and their fragments were measured on a Bruker Ultraflex TOF/TOF mass spectrometer equipped with a smartbeam 2 (Nd:YAG laser) at 2000 Hz (MALDI-MS) or 1000 Hz (MALDI-MS/MS). For MALDI-MS analysis, the instrument was operated in a positive reflector mode with voltage for ion source 1 (20 kV), ion source 2 (17.65 kV), lens (8.4 kV), reflector 1 (21.2 kV), and reflector 2 (10.65 kV). Voltage for ion source 2, lens, and reflector 2 is raised up to 17.75, 8.8, and 10.8 kV depending on the molecular weight of a polymer. External calibration was based on peptide and protein (ProteoMass Peptide/Protein MALDI-MS Calibration Kit, Sigma-Aldrich). Tandem MS was performed in positive reflector mode with acceleration voltages for ion source 1 (7.62 kV), ion source 2 (6.8 kV), lens (3.6 kV), reflector 1 (29.5 kV), reflector 2 (13.9 kV), LIFT 1 (19.00 kV), and LIFT 2 (2.85 kV) using no gas option. The precursor ion was used as internal calibration. For MALDI and Tandem MS analysis, 2-(4-hydroxyphenylazo)benzoic acid was used as a matrix, and sodium fluoroacetate was used as a cationizing agent. A polymer sample, matrix, and cationizing agent were dissolved in THF at 5, 30, and 2 mg ml−1, respectively, and these solutions were mixed in 1:1:1 to 1:5:1 ratio depending on the molecular weight of the analyte. One microliter of the mixed solution was spotted on a MALDI plate and dried in the air. PLGAs were fragmented into ai fragment containing the original alpha group (TBDMS) and a new C terminus and yi fragment having the original omega group (Bz ester) and a new alkene terminus via 1,5-H rearrangement. For sequencing, the mass difference between adjacent fragments (72 Da for L residue and 58 Da for G residue) was used.