Molecular Encryption and Steganography Using Mixtures of Simultaneously Sequenced, Sequence-Defined Oligourethanes.

Molecular encoding in abiotic sequence-defined polymers (SDPs) has recently emerged as a versatile platform for information and data storage. However, the storage capacity of these sequence-defined polymers remains underwhelming compared to that of the information storing biopolymer DNA. In an effort to increase their information storage capacity, herein we describe the synthesis and simultaneous sequencing of eight sequence-defined 10-mer oligourethanes. Importantly, we demonstrate the use of different isotope labels, such as halogen tags, as a tool to deconvolute the complex sequence information found within a heterogeneous mixture of at least 96 unique molecules, with as little as four micromoles of total material. In doing so, relatively high-capacity data storage was achieved: 256 bits in this example, the most information stored in a single sample of abiotic SDPs without the use of long strands. Within the sequence information, a 256-bit cipher key was stored and retrieved. The key was used to encrypt and decrypt a plain text document containing The Wonderful Wizard of Oz. To validate this platform as a medium of molecular steganography and cryptography, the cipher key was hidden in the ink of a personal letter, mailed to a third party, extracted, sequenced, and deciphered successfully in the first try, thereby revealing the encrypted document.

Introduction

Sequence-defined polymers (SDPs) have shown potential as dense and durable information storage media.[1] These macromolecules, often referred to as digital polymers, store information at the molecular level in the form of a defined and absolute monomer sequence (i.e., primary structure).[2−4] Encoding information at the molecular level can be used to surmount some drawbacks of conventional storage devices, such as durability, longevity, and excessive spatial occupation.[1] Such macromolecular information storage is now well established with artificial DNA biopolymers, which have been shown to store and retrieve significant amounts of information.[5−9] Two strengths of using DNA for information storage are the ability to replicate and retrieve data,[10] as well as the ability to exploit rapid advances in sequencing, such as Next-Gen methods[5,11] and nanopore technology.[12−14] Alongside DNA, advances in the synthesis and sequencing of abiotic SDPs have improved the information storage capabilities of these macromolecular systems,[15−30] with commensurate advances toward this goal seen with multicomponent reactions[31−33] and small molecule strategies.[34−36] However, significant advances are still needed to rival the effective storage capacity of nucleic acids using SDPs, let alone silicon-based data storage.[29,37]

Typically, only small pieces of digital information (e.g., a few bytes, a word) can be stored in a single abiotic SDP due to limitations in the ability to synthesize long sequences and challenges in decoding the primary structure of these macromolecules. Frequently, the sequencing/read-out of the encoded information proves to be the most difficult aspect of molecular information storage.[37] This challenge in sequence deconvolution is directly at odds with the parameters required to increase the storage capacity of digital polymers, namely: (1) chain length and (2) number of unique monomers (bits/monomer, i.e., information density). As one increases the chain length and monomer pool, the capabilities of a given sequencing platform to correctly deconvolute the sequence information can be quickly reached. That being said, chain lengths and storage capacities in single molecules are continually increasing (128 bits in an aperiodic copolyester and 144 bits in a poly(phosphodiester)).[22,24]

The most common sequencing methodologies for abiotic SDPs utilize tandem mass spectrometry (MS/MS)[15−17,19−22,25,27,28,38] or more recently pseudo MS3 sequencing,[24,39] LC/MS,[40] and mutated nanopores.[41,42] In each of these cases, subtle, or even dramatic, variations in the monomers become difficult to elucidate, and as a result, examples of multiply functionalized macromolecules for data storage are limited.[16,23−25,33,43]

Nearly all abiotic SDPs so-far applied to information storage purposes use short oligomeric structures, inherently limiting the data stored per molecule. While storing information in multiple oligomers offers the advantages of simplifying the synthesis and sequencing methodologies, it requires spatial organization of the oligomers for proper read-out. For example, self-assembled layers and mixtures,[34,44] crystals,[31] or well plates[43,45] have been used. Generally, every oligomer must be analyzed independently to prevent overcomplication of the MS/MS spectra. Methodologies to analyze mixtures of information carrying macromolecules will effectively increase the information density of a given sample and are needed to achieve a truly dense storage medium.

To complement and improve upon many of the strategies discussed above, we recently described a chain-end depolymerization sequencing methodology for sequence-defined oligourethanes (OUs) that utilizes a thermally induced intramolecular cyclization to iteratively remove the terminal monomer (Scheme ). This allows each truncated OU to be characterized by simple liquid chromatography–mass spectrometry (LC/MS), forgoing MS/MS protocols and greatly simplifying deconvolution.[46] Here, we report the use of this chain-end depolymerization to deconvolute a complex mixture of eight different 10-mer OUs as a means to increase the information density per sample. We demonstrate the storage and retrieval of a 256-bit cipher key held within a single mixture of the OUs, which to our knowledge is the most information stored in a single sample of abiotic SDPs to date, and we apply it to molecular cryptography and steganography.

Scheme 1

Intramolecular, Chain-End Depolymerization for Sequencing via a 5-exo-trig Cyclization Removes One Monomer at a Time, Allowing for Controlled Sequencing of Multiple Oligomers Simultaneously without Interoligomer Cross-Reactivity

Results and Discussion

We posit that a mixture containing X n-mers has the same effective storage capacity as a single X·n-mer. However, to achieve effective sequencing of such a complex mixture, it would be required that each oligomer be distinguishable from one another and that no intermolecular cross-reactivity occurs during the sequencing event. The oligomers need to be distinguished not only to correctly sequence them, but also to know their proper ordering such that the stored information can be accurately reconstructed. One well-established methodology for distinguishing species in a complex mixture using mass spectrometry is isotopic labeling.[47] Frequently in proteomics, stable isotope labels are utilized to identify otherwise equivalent peptides or peptide fragments. Isotope-ratio mass spectrometry can go so far as to measure the relative abundance of isotopes in a given sample.[48] Frölich et al., described the use of halogen-based isotope tags to successfully retrieve 64.5 bits of information from three hexamers, sequenced simultaneously by MS/MS.[25] In another example, Holloway et al. reported the use of ESI-MS and haolgen-based labels to identify the initial position of different monomer “digits” in thermoreversibly scrambled 1,2,4-trizoline-3,5-dione-indole oligomers.[58] Building upon these previous works, as well as our own sequencing paradigm, we hypothesized it would be possible expand on isotopic labels previously used to differentiate sequence-defined oligomers in complex mixtures, and sort the oligomers according to these labels to store more significant amounts of information.

Isotopic Labeling and Isotopologue Design

To store 256 bits of information, we chose to encode a cipher key in hexadecimal (base-16) in a mixture of eight 10-mer OUs (8 of the 10 monomers encode information, see below). In base-16, each monomer provides a storage density of 4 bits per monomer, thus 32 bits per 10-mer, and overall, 256 bits in the sample. Thus, 16 monomers and 8 unique isotope labels entirely distinct from one another, that are chemically persistent throughout the sequencing, were designed. The monomers were synthesized by protio-reduction[49] or deutero-reduction[43] of commercially available canonical and noncanonical amino acids (Scheme S1 and Scheme S2, Figures S1 and S2). Simple deuteration at the α-methylene of each monomer effectively doubles the monomer pool without changing the complexity of the synthesis or sequencing chemistries.[43] Reaction with 4-nitrophenyl chloroformate converted each monomer to an activated carbonate in good yield (Scheme S3). All 16 monomers are shown in Figure S2.

Next, we sought eight unique and differentiable isotope labels. In this design criteria, each label would be placed on the N-terminal monomer, making it persistent and present on each precursor ion throughout the sequencing process (Scheme ). Chlorine exists as two stable isotopes, 35Cl and 37Cl, with relative abundances of 75.76% and 24.24%, respectively.[50,51] Thus, this natural abundance of the stable isotopes is observed in the mass spectra of monochlorinated organic molecules by an increase in the intensity of the M + 2 peak, relative to this natural abundance (an approximately 3:1 ratio of the M to the M + 2 peaks; Figure S3). The same is observed for bromine, which also exists as two stable isotopes, 79Br and 81Br (50.69% and 49.31% abundance, respectively; an approximately 1:1 ratio of the M to the M + 2 peaks; Figure S3).[50] By substituting a molecule with two chlorines, we observe intensification of the M + 2 and the M + 4 peaks, generally resulting in a 9:6:1 ratio relative to the molecular ion (Figure S3). Similarly, two bromines generally give a 1:2:1 ratio relative to the molecular ion peak (Figure S3). Monomers with one chlorine, two chlorines, one bromine, and two bromines were synthesized and isolated (Figure S4). To generate four other isotope labels distinct from each halogen tag, we altered the isotopic composition of the N-terminal urethane monomer, creating isotopologues.

Isotopologues are molecules with the same chemical formula and bonded arrangement of atoms, but one or more atoms have been replaced by an isotope with a different number of neutrons than the parent molecule (i.e., hydrogen vs deuterium).[52] By altering the deuteration level of the N-terminal amino alcohol, we can alter the relative proportions of the M + 2 and the M + 3 peaks to give unique and identifiable mass spectra. Thus, protio-reduced and deutero-reduced monomers were combined at various stoichiometries to attain specific ratios of isotopologues. As a test, three tetramers were synthesized (Scheme S4) using our previously reported methodology.[43] The final coupling step for each oligomer used a stoichiometric mixture of deuterated and nondeuterated monomers (1:3, 1:1, 3:1 ratio; see Figures S5–S7). By LC/MS, the unique isotopologue mixtures could be observed and easily distinguished (Figure ).

Figure 1

Mass spectra of isotopologues A1, A2, and A3. The ratio of M to the M + 2 peaks is directly correlated to the stoichiometry of nondeuterated to deuterated oligomers.

As may be expected, the high-resolution mass spectra reveals that the ratio of deuterated to nondeuterated oligomers is the same as deuterated to nondeuterated monomers (Figures S8–S10) used during synthesis. Further, as may be predicted, applying ratios of 3:1 and 1:1 nondeuterated to deuterated monomers (A1 and A2) created isotope patterns mimicking the calculated chlorine and bromine isotope patterns (Figure S11). However, a 1:3 ratio (A3) created an entirely new pattern. Thus, we sought to create mixtures of isotopologues that would provide unique mass spectra distinguishable from one another, as well as being distinguishable from the halogen tags. In this regard, the nondeuterated to deuterated ratios implemented include 1:0, 5.6:1, 1:2, and 1:5.25 (ca., 0%, 15%, 67%, and 84% deuterated, respectively; Figures S12–S14). The entire suite of isotope labels includes one tag with no label, three isotopologues, and four halogen-based labels (Figure and Figure S4).

Figure 2

Isotope tags using isotopologues and halogen tags to provide specific “fingerprints” for each oligourethane via distinct and predictable isotope patterns.

Molecular Data Encryption

To synthesize and encode a 256-bit cipher key in oligourethanes, an encryption (Mol.Encrypter) and decryption script (Mol.Decrypter) capable of generating and utilizing the key was developed (see Github link, Supporting Information). This encryption and decryption software is based on the Advanced Encryption Standard (AES), the standard adopted by the U.S. Government, which is a publicly accessible symmetric block cipher approved by the National Security Agency (NSA).[53] AES utilizes a symmetric-key algorithm, meaning the key is used to both encrypt and decrypt the data, and is ideal for protecting data at rest. A 256-bit cipher key is considered impenetrable by brute force or exhaustive key searches. Thus, storing the cipher key within a molecular medium would offer a layer of security akin to the use of a hardware security module, as key storage is a considerably important part of cryptography.

A 256-bit string (the cipher key) was generated and used to encrypt a document containing the novel The Wonderful Wizard of Oz by L. Frank Baum (1900). The 256-bit key was then converted to hexadecimal according to the Unicode standard (Figure S15).[54] Within the encoding algorithm, the hexadecimal characters were arbitrarily assigned to each of the 16 monomers (Table S1). With each hex character representing four bits, the hexadecimal string was 64 characters long. Thus, as mentioned previously, we encoded the 64 characters across eight 10-mer OUs. In this encoding scheme, each oligomer contained eight information containing characters, one indexing monomer (at the O-terminal), and the N-terminal isotope tag. The indexing monomer (phenylalaninol) provides a reading frame with which to begin sequence deconvolution, allowing the analyst to determine the isotope pattern present in the mass spectra between two precursor ions of the same oligomer. Further, utilizing the same resin with the same monomer and loading provides an aspect of quality control and uniformity across the syntheses.

Moreover, extremely important to our strategy of simultaneous sequencing is knowing the proper order of the eight 10-mers. Just as the 16-monomers were arbitrarily assigned hexadecimal characters, the mass tags of Figure were arbitrarily assigned to the first, second, third... to eighth oligomer that would read-out the 256-bit code in the proper order. In other words, identification of the isotope tag is required for accurate sorting of the eight hexadecimal strings to rebuild the original cipher key.

The eight oligomers (B1–B8) were successfully synthesized on the solid phase, in parallel, in a fritted 96-well plate (Scheme S5, structures in the Supporting Information). The N-terminal amine was capped with a long-wavelength chromophore, 4-fluoro-7-nitrobenzofurazan to monitor the chain-end self-sequencing and aid in sequence deconvolution. B1–B8 were cleaved from the solid phase, purified, and characterized by high-resolution MS (Figures S16–S23, Table S2).

Simultaneous Sequencing

For the simultaneous sequencing and decoding, 500 nmol of each oligomer (B1–B8) was combined in a single sample (Figure S24). The oligomers were dissolved in DMSO with 0.01 M cesium carbonate (Cs2CO3) and heated to 70 °C in an incubated shaker and sampled for LC/MS at designated time points (Figure , Figure S25), monitoring the depolymerization over time. At the 0 min time-point, only the eight starting oligomers were observed, with each truncated iteration being formed as the reaction progressed until nearly full consolidation at the 1-mer for all eight SDPs at 550 min. Figure succinctly shows the complexity of the sequencing reaction. In one sequencing experiment, as the oligomers simultaneously and orthogonally depolymerize, we cleanly and clearly observe the existence of all 80 discrete precursor ions by negative mode ESI-MS (Figure S26, Table S3), as well as the 16 unique cyclized oxazolidinones (observed in positive mode) across the 8 time points. It is worth noting that these oligomers were found to ionize strongly under an acetonitrile gradient with at least 10–50 mM ammonium acetate buffer, attributed to the suppression of undesired salt adducts in ammonium acetate buffers.[55] Furthermore, during the time-point sampling of the sequencing reaction, a separate LC/MS sample was prepared wherein each time-point was combined. By high-resolution MS, the precursor ions of all 80 corresponding oligomers, as well as the 16 oxazolidinones, were observed in a single high-resolution LC/MS sample (Figures S37–S132). This highlights the power of using chain-end depolymerization rather than tandem MS for reading of the stored information.

Figure 3

Concurrent sequencing of oligomers B1–B8 in DMSO with Cs2CO3. Reaction was heated to 70 °C and sampled at designated time intervals by LC/MS.

Moreover, because of the chain-end depolymerization mechanism, simple visual inspection of the low-resolution LC/MS data was sufficient for complete sequence deconvolution. Prior to high-resolution confirmation, utilizing the isotopic fingerprints, the authors (SDD and CDW) were each able to correctly identify and sort the 80 precursor ion masses into a templated spreadsheet (Tables S3 and S4). Each author singlehandedly and independently analyzed and identified the 80 unique masses, with CDW never having seen the structures of the oligomers prior to their analysis. To aid in mass sorting and sequence deconvolution, CDW was provided a series of guidelines (see included Powerpoint, Supporting Information). Critical to rebuilding the cipher key, both authors were able to successfully identify all eight of the stable isotope tags, which were used for the correct ordering of the hexadecimal character string. Figure shows the low-resolution LC/MS data for three of the eight oligomers in detail, denoting how the isotope tags were capable of distinguishing oligomeric species (all eight oligomers are shown in Figures S27–S34). Finally, the spreadsheet containing the precursor ion masses was fed into the decryption algorithm, which then converted the sequencing information into the original 256-bit cipher key. This 256-bit cipher key was able to decrypt the document containing the novel The Wonderful Wizard of Oz, successfully revealing the encrypted information.

Figure 4

LC/MS traces of three of the eight information containing oligomers. As observed, the stable isotope tag imparts a unique mass spectrum onto each ion, allowing for each mass to be sorted and assigned to the appropriate oligomer. The mass differences between ions are then calculated and correlated back to the monomer that was cyclized and cleaved, revealing the information stored within the macromolecule.

A few noteworthy observations were made during the data analysis and sequence deconvolution. The first is that, as a given cyclization event occurred, the resulting oligomer would lose one nitrogen atom from its molecular formula. Thus, the nitrogen rule[56] could be utilized as a tool to aid in mass sorting: as an oligomer sequences, its nominal mass will alternate from an even to an odd number as the molecular ion alternates from an even to odd number of nitrogen atoms (as seen in Figure ). Furthermore, when the oligomers are longer (i.e., 10-mer or 9-mer), the exact identity of a given isotope tag can obscured by the contributions of naturally abundant isotopes (13C is 1.107%, 15N is 0.366%, and 18O is 0.204%).[57] Thus, the M + 1, M + 2, M + 3, etc. peaks are intensified with respect to the amount of carbon, nitrogen, and oxygen present in the molecule, with this effect amplified as the number of atoms in each oligomer increases. Of course, this atomic composition varies between each oligomer (B1–B8) but also between a given oligomer as it depolymerizes. The isotope pattern of an oligomer will incrementally decrease in complexity with each cyclization event until arriving at a very simple and distinguishable pattern at the 1-mer (as seen in Figure ). As such, if an isotope tag cannot be identified at the 10-mer, the mass spectra can be compared and matched (for example, comparing the similar patterns of 10-mer and 9-mer) until the isotope tag is identified at the smaller truncated oligomers. Likewise, knowing the indexing monomer prior to sequencing allows immediate identification of the 9-mer, which, upon comparison with the 10-mer, provides a baseline MS with which to begin identifying and sorting the masses. Important to this encoding paradigm, oligomers of the same mass can have differing retention times, which, when analyzed, produce their respective and unique extracted ion chromatograms (EIC) (Figure ). Upon analysis of the EICs, these two oligomers with the exact same base peak m/z were easily distinguished due to the differences in the resulting spectra provided by their respective isotope tag.

Figure 5

Comparison of the MS of two oligomers (B3–B8-mer and B5–B6-mer) with the same mass but different isotope patterns and retention times, allowing for differentiation.

Independent Validation, Steganography

A medium of information storage is only as good as its ability to accurately and easily confer the information being stored. We believed our sequencing paradigm to be robust and simple and therefore easily decodable by a third-party without any prior experience implementing these sequencing and deconvolution protocols. For this, a real-world demonstration of this platform in cryptography and steganography was performed using our molecular 256-bit cipher key. First, the molecular cipher (500 nmol of each oligomer, B1–B8) was dissolved in isopropanol and combined with glycerol and soot, making a writable ink (see the Steganography section in the Supporting Information). The oligourethane “ink” mixture was placed in an emptied ballpoint pen, which was then used to handwrite a letter to our coauthor J.F.R. (Figure S35). The molecular cipher key, embedded in a standard piece of printer paper was mailed from Austin, Texas to Lowell, Massachusetts where it was extracted with dichloromethane and concentrated. The third-party collaborators (J.F.R. and R.A.S.) were given a set of discrete instructions on the reaction set up and sequence deconvolution (Note S1, see Supplemental PowerPoint). In their very first attempt, the collaborators were able to correctly sequence the eight oligomers and identify the precursor ions (Table S5, Figure S36), which were entered into the decryption algorithm, decrypting the file and revealing the document containing The Wonderful Wizard of Oz.

Conclusions

Herein, we have described a robust method for the simultaneous synthesis and sequencing of multiple discrete macromolecules with the use of isotope tags to increase the information storage capacities of sequence-defined abiotic oligomers. The isotope tags made each oligomer distinguishable from one another, such that sequence deconvolution was possible even in a mixture containing up to 96 unique molecules. Important to note, all 96 discrete molecules were characterized by high-resolution LC/MS in a single sample. Two programs, Mol.Encrypter and Mol.Decrypter, were developed and utilized to encode and decode documents using a 256-bit cipher key, following the Advanced Encryption Standard. This molecular medium of information storage was able to store and deliver the 256-bit cipher key, which we believe to be the most information ever stored in single sample of abiotic sequence-defined macromolecules. The molecular cipher key and Mol.Decrypter were used successfully to decrypt a document containing The Wonderful Wizard of Oz. In an example of molecular steganography, a sample of discreetly hidden oligourethanes embedded in “ink” was mailed to a third party, where the sequencing, structure deconvolution, and cipher key recovery were independently performed. This paradigm has significant potential for widely accessible molecular information storage and encryption. Future iterations will look to robotically automate the writing and reading processes, furthering its accessibility and practicality for real-world applications.