A compact cassette tape for DNA-based data storage

Abstract

DNA with high storage density can serve as an alternative storage medium to respond to the global explosion of data growth and become a powerful personal storage memory if an integrated compact device can store and handle large-scale data. Here, we incorporate a DNA cassette tape with 5.5 × 10⁵ addressable data partitions (addressing rate up to 1570 partitions per second), a DNA loading capacity of 28.6 mg per kilometer, and deposit-many-recover-many (DMRM) features per partition to flexibly manage large-scale storage data and achieve hundreds of years of data preservation. We develop a compact DNA cassette tape drive and verify its functionality by randomly depositing incomplete images into the data partition, demonstrating a completely automated closed-loop operation involving addressing, recovery, removal, subsequent file deposition, and file recovery again, all accomplished within 50 min. Last, the complete image is restored by next-generation sequencing and decoding. DNA cassette tape provides a strategy for fast, compact, large-scale DNA-based cold or warm data storage.

A DNA cassette tape system with high-density partitions enables fast addressing, DMRM features, and long-term data preservation.

INTRODUCTION

With the explosive growth of data in the present day, nonvolatile memory based on semiconductors has reached the limits of Moore’s Law, and new media are necessary to store unbelievably large amounts of data (1–4). DNA has the potential to become the next-generation information storage medium due to its high storage density (about 455 Exabytes/g) and long-term storage time without electrical maintenance (5–9). The general process of existing DNA-based data storage typically includes data writing, storage, indexing, modifying, and reading. Among them, high-throughput DNA synthesis (data writing) and high-throughput DNA sequencing (data reading) technologies benefit from the rapid development of biochemical research and have already been commercialized semi-automatically on instruments (4, 10). Although DNA synthesis costs are still very high, it is expected to be used on a large scale in the next decade (11). However, no existing DNA data storage device has yet achieved the fundamental storage functionalities and robust data management capabilities comparable to commercial storage systems, which will be essential when that time arrives.

Advanced storage devices have strong expansibility and versatility and share a key feature: They enable data interactions at a single physical location through optical, electrical, or magnetic mechanisms, facilitating efficient data management and manipulation by supporting file storage, retrieval, repeated access, deletion, and replacement. As DNA data storage technology advances, DNA is frequently combined with material substrates to enhance data retention duration (12–15), reduce polymerase chain reaction (PCR) bias (12, 16, 17), simplify operational processes (18–22), and provide indexing capabilities (12, 16). These are typically prepared in particle form to provide physical partitions that distinguish between files. However, they are different in material composition, structural form, and fabrication methods, which makes integration different. Furthermore, establishing the storage system and addressing target files and repeated operations on these files require constant switching between manual and instrumental processes, which is labor-intensive and results in a notable waste of consumables (table S5). There has been exciting progress in introducing microfluidic chips to integrate the process of DNA data storage, such as DNA synthesis and sequencing (23), DNA retrieval process (24–26), and DNA encapsulation/decapsulation process (8, 9); however, the versatility of channel-based microfluidic chips is poor due to their complex channel designs and extensive pump-valve structures, face challenges in integrating additional functions and processes. Digital microfluidic chips require locating the target DNA file on a glass substrate and manually installing it during extraction (25), which leads to poor scalability for large-scale data processes. In addition, some materials [e.g., polydimethylsiloxane (26) and glass (24)] are used to prepare microfluidic chips due to their excellent processing properties. The low space utilization of substrate materials leads to low DNA load (table S5). Thus, a material system has not been developed that realizes the entire DNA storage process and enables compact data storage with completely automated file operations.

Furthermore, although DNA data storage is traditionally considered suitable for cold data storage (7, 12, 16, 27), a long-standing challenge is to design a platform that enables rapidly loading DNA files, high-speed file addressing, multiple files recovery, multiple files replacement, and the long-term stability of DNA data and could even handle warm data and serve as a personal big data storage hub. Here, we propose a DNA-based storage device with a form factor similar to that of a magnetic tape, named “DNA cassette tape” (Fig. 1A). DNA cassette tape (DNA tape) is composed of a polyester-nylon composite membrane with excellent mechanical properties and has a large number of hydrophilic and hydrophobic spaces to form barcode patterns. The barcodes are continuously arranged on DNA tape by laser inkjet printing, generating 5.45 × 10⁵ addressable data partitions per 1000-m tape and achieving an addressing rate of 1570 partitions/s. In addition, the deposit-many-recover-many (DMRM) feature can independently operate on each data partition of DNA tape. Zeolitic imidazolate frameworks (ZIFs) as a protective layer can be synthesized on DNA tape in situ for long-term data storage, realize rapid encapsulation within 10 s, and decapsulation within 10 min on DNA tape. Last, we developed a compact DNA cassette tape drive for DNA tape (Fig. 1B), which can perform file addressing, decapsulation, encapsulation, recovery, removal, and redeposition operations on DNA files quickly and automatically.

Fig. 1. Schematic of DNA cassette tape and DNA cassette tape drive.

(A) Using barcode patterns for optical file addressing and thereby creating physical partitions on DNA tape. Each physical partition has a unique address and supports the DMRM function. The ZIFs layer protects the encoded DNA and can be quickly generated and removed before and after DNA recovery. (B) DNA cassette tape drive with completely automatic operation and file management system is used with DNA tape.

RESULTS

Physical data partitions and file addressing on the DNA tape

A polyester-nylon composite tape with barcode patterns was prepared by inkjet printing, cutting, and rolling (Fig. 2A, top, and details in Materials and Methods). On the tape, synthetic DNA was deposited in the white areas (named “space” of the barcode) and was separated by the black areas (named “bar” of the barcode) (Fig. 2A, middle). The fibrous structure is the main structure of the space area, which is hydrophilic and can be used to load synthetic DNA (fig. S16). The bar area is filled with black ink and cross-linked polydimethylsiloxane (PDMS), which is hydrophobic and serves as a data partition barrier to ensure the DNA data in different partitions would not be disturbed (Fig. 2A, bottom, and fig. S16). Barcode patterns not only create physical partitions but also generate addressable information. Such as the folder name or codes could be generated with the Code-128 barcode generator and be used as addressing information on the DNA tape. A folder name can be directly encoded into a barcode and searched as the simplest mapping method (similar to “Absolute Path” in the computer system). Files within the folder are stored in the barcode space and sequentially numbered. For example, “JK Li” can be encoded into a barcode and as the folder, and the addresses of two files within the folder, “Lanterns.png” and “Family party.avi” will be represented as “JK Li_5” and “JK Li_22,” respectively, where “JK Li” is the folder name, and “5” and “22” are the 5^th and 22^nd spaces within the “JK Li” barcode (Fig. 2B). In addition, barcodes can also be generated from certain “codes” (similar to “Symbolic Links” in the computer system), which can be addressed more efficiently using advanced algorithms (28, 29). For example, “SUSTech Address.txt” and “List of employees.pdf” both belong to the category of documents related to the institution of Southern University of Science and Technology, and their respective link addresses are “28483_2” and “u55_7.” By establishing symbolic links, these files can be grouped for advanced file classification (Fig. 2B), such as similarity searches (30) and artificial intelligence (AI)–based categorization (31). In the following descriptions, unless otherwise specified, the method of file mapping will be “absolute path.”

Fig. 2. Physical partitions and file-addressing on DNA tape.

(A) Roll of DNA tape with a barcode pattern (Code-128) and a size of 5 mm by 15 m (top). Distribution of DNA files on DNA tape (middle). Using dyes with different colors to simulate files in different partitions (bottom). Scale bars, 5 mm. (B) Barcodes are continuously distributed on the tape, and the files generate an address assignment symbol of “Decoding result and Sequence number (DR_SN) of Space” by the mapping relationship. (C) Composition of Code-128 barcodes (top) and the proportion of space area used to store files (bottom). The black dotted line is the average value of all characters in area proportion. (D) According to the encoding rules of Code-128, the maximum number of physical partitions generated with different numbers of data characters in barcodes. (E) Number of physical partitions produced by DNA tapes with different aspect ratios and lengths. (F) Two-level file addressing process. The barcode reader (left) and photoelectric sensor (right) position the barcode and the space where the target file is located. (G) The readable rate of the primary positioning at different rotation speeds and radius of gyration is 0.025 m during the test (n = 10). The groups are different quantities of data characters in one barcode. Data are presented as median values ± SEM.

To evaluate the proportion of the storage area partitioning with barcodes, we counted the area ratio of the space area in the four main components (comprising start, data, check, and stop character) of Code-128 from the Code-128 encoding table (table S1). The space area within the three start characters (start A, B, and C) accounts for ~57.6%, the space area within the data and check characters accounts for ~50%, and the space area within the end character accounts for ~45.5%, so for any composed barcode, the space area (data storage area) accounts for ~50%. (Fig. 2C and text S1). Every time a data character is added, three DNA file addresses that can deposit DNA will be added (each character is composed of three bars and spaces) and generate many barcodes by rearranging and combining. Thus, the physical addresses generated by the barcode are infinite (Fig. 2D and calculation details in text S1). The number of physical partitions is related to the density of barcodes and the length of the DNA tape. When the barcode with an aspect ratio of 10 is used on the DNA tape, a tape of length 1000 m could generate 545,400 data partitions of files (Fig. 2E and calculation details in text S1).

The linear tape file system (LTFS) of commercialized magnetic tape divides the tape into two parts: one part stores the index (metadata), and the other contains the actual data. By structuring the data in a more accessible way and leveraging caching, LTFS notably reduces the time required to locate and access files on tape (32, 33). Same as the magnetic tape, file addressing on DNA tape was sequentially recognized by rotating the tape at high speed. The target file was addressed in the two-level addressing, which was the target barcode addressing and the space of the target barcode addressing (Fig. 2F). In the addressing process, the primary addressing occupies most of the file addressing time because it has to make a preliminary selection of a large number of files, and the secondary addressing can be completed within 2 s. To test the primary addressing rate, we used barcodes with different numbers of data characters and gradually increased the rotation speed of the motor to achieve 100% reading accuracy at a maximum of 2400 rpm (10 data character-based barcode, and the radius of gyration during the test is 0.025 m) and obtained a file addressing rate of 1570 files/s (Fig. 2G and calculation details in text S2). The file addressing speed is limited by using a barcode reader based on a complementary metal-oxide-semiconductor image sensor, which can realize about 45 barcodes/s. Nevertheless, compared with using the quick response (QR) code for DNA file indexing (17), the barcode can achieve 100% readability, which is 10 times faster than the QR code at a faster movement rate (text S3). Although the modified primers allow efficient addressing (12, 16), the cost of fluorescently labeled synthetic primers is high (table S2), and the single-stranded DNA (ssDNA) attached to beads is prone to hydrolysis over long-term archival storage. Using a barcode structure for addressing can avoid the need for synthesizing DNA and be realized for long-term use (fig. S17). These results suggested that the barcode patterns not only provided a way to fast addressing but also created millions of reliable data partitions of files on DNA tape.

DNA file deposition and storage capacity on the DNA tape

We chose the polyester-nylon composite membrane with carboxyl groups as the base material of DNA tape, which had a high tensile modulus and a low flexural modulus and was widely used in reverse line bolt hybridization technology (34, 35). We bonded 20-nt amino-modified ssDNA to the space areas of the DNA tape by 1-ethyl-3-(3-dimethylaminopropyl) carbodiimide (EDC) coupling reactions and formed a blank tape mode (Fig. 3A, left). To deposit DNA files, the adapter of the synthetic DNA files hybridized with the handle of the blank tape by Watson-Crick base pairing and extended stable double-stranded DNA (dsDNA) by WarmStart DNA polymerase (Fig. 3A, right).

Fig. 3. DNA file deposition and storage capacity on the DNA tape.

(A) Schematic of the DNA deposition process. The blank tape is prepared by coupling DNA handles. The synthetic DNA pool and DNA polymerase mixed solution (DNA files) are loaded into the physical partition of DNA tape to complete the DNA data deposition by hybridization with handles and polymerase extension. (B) Two fluorescently labeled ssDNA with AF488 (x) and Cy7 (y) are used to simulate the different files. x and y are introduced in the space area with or without the handle. Scale bar, 5 mm. (C) Fluorescence signal is measured after adding x and y to the handle-bonded DNA. (i) Raw tape, (ii) AF488 channels, (iii) Cy7 channels, and (iv) merge. Scale bar, 5 mm. (D) Coupling efficiency of the DNA handle under different ambient humidity (n = 3). Data are presented as median values ± SEM. (E) For different tape path lengths, the deposition rate of files corresponds to the deposition time for each specific file. (F) The recovery yield after DNA file deposition and redeposition on tape (with or without handles) in different deposition times (n = 3). The “+” symbol represents the DNA recovered after multiple washing steps to remove non-specifically adsorbed DNA, while the “-” symbol indicates the DNA recovery without washing steps. (G) DNA load capacity under different amounts of DNA deposits on DNA tape (n = 3). Data are presented as median values ± SEM. (H) Relationship between the length of DNA tape and storage capacity. The orange-red dots represent the extrapolated storage capacity based on the actual files stored in this work. The purple dots indicate the maximum storage capacity extrapolated from the DNA loading limit on the DNA tape, which is calculated in text S4.

We aim to produce blank tapes at room temperature through soaking to simplify the preparation process and explore the potential of large-scale production. As proof of principle, we selected a 4-cm-long region on the DNA tape for testing. The tape is soaked in 16% (w/v) EDC for 10 min to obtain EDC-activated tape. Subsequently, 1 μl of handle solution (100 μM) is loaded onto the physical partitions of the DNA tape using a fully automated pipetting system and incubated for 5 min to complete the preparation of blank tape (text S9). We used hybridization assays with two fluorescently labeled ssDNA to evaluate the file fixation with and without handles after washing with 2× saline sodium phosphate EDTA (SSPE)/0.1% SDS solution, which effectively removes the unfixed nucleic acid (35). Compared to the areas with handles, the fluorescence signal disappeared in the areas without handles after washing with 2× SSPE/0.1% SDS solution 10 times, indicating that handles are successfully coupled to DNA tape and can be hybridized with ssDNA (Fig. 3B). In addition, because of the hydrophobic barrier, the different fluorescently labeled ssDNA could be deposited to neighboring spaces without interference, and single partition could deposit different fluorescently labeled ssDNA (Fig. 3C). Theoretically, the number of fixed strands (original handle strands) equals the number of its complementary strands. To calculate the number of handles on DNA tape, we obtained their complementary strands by denaturing and analyzing them by quantitative PCR (qPCR) (fig. S18). We analyzed the coupling efficiency of the handle on the tape under different humidity conditions and found that higher evaporation rates (lower humidity) increase the amount of handle coupled to the tape (Fig. 3D), which is attributed to the enhanced reaction rate due to rapid liquid concentration (36, 37). This enables blank DNA tape to be quickly prepared in large quantities in a standard humidity (~50% relative humidity, RH).

Once large-scale production of oligo pools becomes practical, the bidirectional nature of DNA tape movement, combined with the pipetting system, would streamline the handling process and enhance the scalability and automation of DNA file deposition on the tape. DNA tape is stretched to expose the tape path for file deposition. DNA files are printed onto one end of the tape, and the tape maintains a constant linear velocity, transporting the DNA files to the other end to complete the deposition process (Fig. 3A, middle and right). The linear velocity is adjusted on the basis of the desired file deposition time, ensuring the files pass through the tape path. Thus, shorter deposition times and longer tape path lengths allow for faster DNA file deposition. From the perspective of theoretical verification, we calculated the relationship between tape path length and deposition rate under three different file deposition times. When the deposition time was 5 min, a tape path length of 50 cm achieved a deposition rate of 0.5 files/s. When the deposition time was reduced to 1 min, a tape path length of 10 cm achieved the same deposition rate of 0.5 files/s, and a tape path length of 50 cm increased the deposition rate to 2.3 files/s (Fig. 3E). It is worth noting that achieving this speed requires different oligo pools with reaction mixture to be continuously deposited onto the DNA tape at a short intervals of time. However, this speed exceeds the upper limit of current commercial automated pipetting systems, indicating that customized hardware may be required to adapt these systems for DNA tape integration (text S9). Next, we explore the impact of deposition time on file recovery. Considering the sensitivity of biochemical reaction systems to concentration, we selected a high-humidity environment (80% RH and 60°C in the artificial climate chamber) to minimize liquid evaporation during DNA deposition, and 10 ng of DNA was deposited. After deposition, DNA was recovered for subsequent qPCR analysis. To evaluate DNA loss during the deposition process, we tested the amount of DNA deposited on the DNA tape before and after washing to remove nonspecific binding interference. We then assessed the impact of varying deposition times on DNA deposition efficiency. When handles were coupled on the tape, the recovery yields for the direct recovery were 97.5, 122.2, and 109.4% after deposition times of 1, 5, and 10 min, respectively. For the recovery after washing, the recovery yields were 22.4, 19.7, and 17.2% after 1, 5, and 10 min of DNA deposition, respectively. In contrast, when handles were not coupled on tape, the recovery yields for the direct recovery were 79.7, 60.8, and 57.4% after 1, 5, and 10 min of DNA deposition, respectively. For the recovery after washing, the recovery yields dropped to 0.056, 0.089, and 0.189% after the same deposition times (Fig. 3F). This indicates that most of the DNA files were successfully fixed on the tape through hybridization and extension process, making it less likely to detach, with a DNA recovery rate of 19.8%. Moreover, the recovery yields of the original pool remained stable after multiple rounds of file recovery (discussed in detail in the next section). Conversely, DNA deposited on the tape without interaction force decreased in quantity after washing, losing ~10,000 times the original amount. In addition, the amount of DNA fixed to the substrate did not increase notably with longer deposition times, so we deposited for 1 min for the next. Thus, it can be estimated that depositing 10,000 DNA files on the DNA tape would take 27.8 min (10,000/2.3/60).

To assess the maximum DNA loading capacity on the DNA tape, we deposited increasing concentrations of DNA on the DNA tape. When the DNA concentration in the droplet reached 120 μM, the DNA loading on the tape plateaued at a capacity of 5 × 10¹¹ copies/mm³ (76.3 g/m³) and reached a saturated state (Fig. 3G). However, the recovery rate decreases as the DNA concentration approaches the plateau phase. Thus, increasing physical redundancy might be necessary to accommodate ultra-large files. Another advantage of DNA tape is that it can extend in one direction and be wound into a compact device, allowing for storing as much data as possible. As a proof of concept, we deposited 156.6 kb of data in a single partition and successfully recovered it (figs. S19 and S20). Based on this data, we estimated that DNA tape could store 74.7 GB of actual data per kilometer (Fig. 3H and calculation details in text S4). Excluding known theoretical losses, such as the physical space taken up by barcodes and the DNA losses during deposition and recovery operations, we estimate that the theoretical maximum storage capacity could be as high as 362 Petabytes of DNA tape per kilometer (Fig. 3H and calculation details in text S4).

DMRM system on the DNA tape

For synthesized oligo pools, file reading should not consume the original pools to reduce the cost of DNA synthesis and improve the data fidelity (12, 38). For pools with a large amount of data, the system should support the removal and replacement of individual files to economic costs, thereby becoming a robust personal memory solution. To achieve this function, we encoded two digital files (one text file and one image file) using the DNA Fountain (100 nt) (39), added a restriction enzyme cutting sequence (4 nt) for file removal at the 3′ end of the sequence, and added the primer sequence (20 nt) for complementary base pairing with the handle and PCR at both ends of the sequence (Fig. 4A).

Fig. 4. DMRM on the DNA tape.

(A) Design of the encoded-DNA sequence. (B) Schematic diagram of multiple DNA file recoveries, removals, and redeposition on DNA tape. (C) After 10 file recoveries and redeposition, the number of DNA copies is isolated from tape each time (n = 3). Data are presented as median values ± SEM. (D) Oligo coverage frequency distribution for each time after 10 file recoveries (n = 3). Data are presented as median values ± SEM. (E) Error rate analysis of 10 times file recoveries, “Mutation (RCC)” is the mutation error rate after RS Code error correction. (F) After 10 file redepositions of 10 different sequences, the ratio of subsequent files to old files is analyzed by NGS.

Oligo pools were deposited on the DNA tape using the method described previously (details in Materials and Methods). The partition-bonded DNA released ssDNA to extract DNA files by soaking the target partition in NaOH solution (0.15 M) and sequencing the denaturation solution to complete the file reading. The partition-bonded ssDNA (named the template strand) could be restored to dsDNA by extension. Repeating the above operation can achieve multiple DNA file recoveries on the DNA tape (Fig. 4B). After 10 file recoveries on a data partition (5 mm by 1 mm) of the DNA tape, the DNA copies remained roughly constant (Fig. 4C). In addition, we sequenced the results of each one of the 10 file recoveries by next-generation sequencing (NGS), the bias profile [the probability function of perfect calls per million (pcpm) reads and the frequency distribution] of DNA copies was unchanged throughout, and the average coverage per oligo was about 35 pcpm (Fig. 4D). In the 10 file recoveries, the substitution error rate was higher than the insertion error and the deletion error for 10 file recoveries, and after the Reed-Solomon (RS) Code was corrected, the substitution error rate notably lowered (Fig. 4E). The files were successfully decoded and read in each of the ten file recoveries (movie S1).

To realize file removal and redeposition, we added a restriction endonuclease restriction recognition sequence in the DNA sequence to achieve multiple DNA data removals and redeposition of the DNA tape. Two design principles of the restriction enzyme sequence were as follows: (i) the cutting site should be at the beginning of the 5′ end of the recognition sequence to ensure that each digestion process could leave a complete handle sequence, and (ii) the restriction endonuclease recognition sequence should be as short as possible to reduce the cost of DNA synthesis. Thus, we selected the Mbo I, which had a short recognition sequence (5′-GATC) and could retain the handle on DNA tape after cleaving. After digestion with Mbo I and the denaturation process, the files in a specific partition could be removed, and the deposition of the subsequent DNA file could be completed through the steps of the DNA deposition process (Fig. 4B and fig. S21). It is worth noting that the 5′ end adapter sequence of the subsequent file needs to be updated. Otherwise, it will produce some sequences of old files through PCR, thus interfering with the decoding of the subsequent file (fig. S5). We used the first 10 sequences encoded with the SUSTech Address.txt file to represent 10 different files and tested the efficiency of the file redeposition. After 10 files redeposition on a data partition (5 mm by 1 mm) of the DNA tape, the DNA copies remained roughly constant (Fig. 4C). In addition, to evaluate the proportion of the subsequent files after each redeposition operation, we sequenced the subsequent files after redeposition, recovery, and PCR. Approximately 99.9% of the original data was replaced by subsequent data (Fig. 4F). These results all indicated that the DMRM system could stably operate within each data partition on the DNA tape. In addition to enabling the multiple removal and redeposition of files on DNA tape, we have developed a linear coding algorithm that allows for text editing operations such as substitution, insertion, and deletion directly on the tape (text S5). Compared with some DNA-based Read-Only memory (17, 40) or with some single-use nanoparticles (12, 13, 15, 18), which typically compromise the scaffold material (such as the micro-disks or the chromosomes of yeast) when file replacement is carried out, DNA tape supports the DMRM system on a single partition, which can perform multiple high-fidelity DNA file recoveries and redeposition of files at high resolution.

Long-term storage and rapid recovery of data on the DNA tape

For the long-term preservation of DNA data and fast multiple recoveries, an encapsulation method with strong protection and fast decapsulation is essential. DNA can be stored in bone for hundreds of years (41), but the DNA phosphate backbone can be broken by hydrolysis within a few days under normal conditions, resulting in data that cannot be read by PCR amplification (14, 42, 43). We coated the ZIFs protective layer on the DNA tape by adding Zn(NO₃)₂ solution to the system, in which DNA tape soaked in the 2-methylimidazole (2-Melm) solution (Fig. 5A). Zn ions adsorbed to the DNA phosphate backbone through coordination and synthesized ZIF structure in situ (44), which had been shown to protect biological macromolecules from chemical damage effectively (45, 46). After the protective layer of ZIFs was coated, the barcode patterns on the ZIF-coated DNA tape [encapsulated DNA tape (E-DNA tape)] were still very clear and did not affect the addressing process (Fig. 5B), and the tape became hydrophobic (fig. S22).

Fig. 5. Long-term storage and rapid recovery of data on the DNA tape.

(A) Schematic of the rapid encapsulation of the DNA tape. (B) Photograph of DNA tape before and after coating the ZIF protective layer. (C) Resistance to DNase I degradation with increasing the concentration of Zn ions and 2-Melm (C_{Zn ions}:C_2-Melm is 1:10) on DNA tape (n = 3). Data are presented as median values ± SEM. (D) DNA content of DNase I–free groups (gray) and DNase I–treated groups (purple) (n = 3). Pure DNA with about 5 × 10⁷ copies and DNA tape with 5 × 10⁷ copies. Data are presented as median values ± SEM. (E) Amount of DNA loss on D-DNA tape and E-DNA tape at a fixed humidity (50% RH) and different temperatures (60°, 65°, and 70°C) for 3 weeks (n = 3). Data are presented as median values ± SEM. (F) Oligo coverage frequency distribution of the DNA files extracted from D-DNA tape and E-DNA tape after 0, 3, and 6 weeks at various harsh conditions (50% RH, 70°C). Two PCR amplifications were performed on the DNA tape groups at week 3 and week 6. “*” indicates that the data is not applicable (N.A.), due to sequencing failure. (G) Half-life of the DNA stored in D-DNA tape and E-DNA tape extrapolated according to Arrhenius equation and compared to literature data on DNA storage in silica (13), fossil (41), and commercial magnetic tape (48). Extrapolated DNA half-life at various temperatures is the dotted line. (H) Schematic diagram of fast decapsulation, extracting files, and encapsulation on DNA tape. 1 is the decapsulation process, 2 is the file extraction and restoration process, and 3 is the encapsulation process. (I) DNA recovery yield after cyclic DRE processing (n = 3). (J) XPS spectra of Zn 2p on E-DNA tape in removing and synthesizing the ZIF protection layer process.

The thickness of the ZIFs protective layer on the DNA tape could be adjusted by adding different concentrations of zinc ions and 2-Melm (fig. S23). To evaluate the protective effect of ZIF, we digested decapsulated DNA tapes (D-DNA tape) and E-DNA tapes with deoxyribonuclease (DNase) I, an enzyme that disrupts the DNA phosphate backbone (47). As the thickness of the ZIFs protective layer increased, the resistance of DNA molecules was enhanced to DNase I digestion. The protective ability of ZIFs did not increase notably when the Zn ion concentration was greater than 30 mM, and ~48% of the DNA molecules were lost (Fig. 5C). In contrast, the copies of unprotected DNA and D-DNA tapes were reduced to 0.01% (Fig. 5D). Furthermore, we performed accelerated aging tests and sequencing analysis by placing D-DNA tape and E-DNA tape at 50% humidity and different temperatures (60°, 65°, and 70°C) for 3 weeks due to the protective effect of ZIFs, the numbers of DNA copies on E-DNA tape decayed slowly, remaining at 1.7% after 3 weeks (70°C). In contrast, the number of DNA copies on D-DNA tapes could not be detected at 70°C after 3 weeks (Fig. 5E). In addition, the files were successfully read on E-DNA tape at 70°C after 6 weeks (movie S1), and the average coverage per oligo dropped from 34.2 to 29.6 pcpm, indicating that the quality of the deep copy was substantially worse than the initial oligos of the tape. Meanwhile, the files on the D-DNA tape experienced two PCR amplifications and could be sequenced after 3 weeks (70°C), and the average coverage per oligo dropped from 33.8 to 26.4 pcpm. While the files on the D-DNA tape could not be sequenced after PCR amplifications twice at the 6 weeks (70°C), indicating that the data had been lost (Fig. 5F). We calculated the half-life of the DNA tape by the Arrhenius equation and fitted the curve of temperature vs. half-life. Although the decay rate of DNA in E-DNA tape is higher than DNA encapsulated in silicon (13) and fossil (41), it extends the data retention time by 40.5 times (20 °C) compared to D-DNA tape. In addition, E-DNA tape exhibits a decay rate similar to that of LTO tape (48). Combining the results, we calculated that the E-DNA tape could store data for more than 345 years at room temperature (20°C) and about 20,000 years at Changbai Mountains (Fig. 5G and calculation details in text S6).

The protective layer of ZIFs on DNA tape could be quickly removed (~15 s) using a citrate phosphate buffer (pH = 4.5) for DNA file recovery and redeposition, and DNA could also be quickly encapsulated (10 min) for long-term data storage (Fig. 5H). We characterized the encapsulated and decapsulated processes of the ZIFs protective layer by x-ray photoelectron spectroscopy (XPS). The spectrum of the Zn 2p on the DNA tape completely disappeared at 15 s, and the ZIFs particle completely disappeared on the DNA tape (fig. S23). Moreover, the spectrum of the Zn 2p recovered to the initial state about 10 min after adding Zn ions and 2-Melm (Fig. 5I). We tested the impact of the encapsulation-decapsulation process on DNA recovery yield from the DNA tape by observing the amount of DNA recovered after each decapsulation-recovery-encapsulation (DRE) cycle. After conducting five DRE cycles on the DNA tape, the DNA recovery yield remained consistent at 26.2% with an SD of 9.9%. Approximately 70% of the DNA loss was attributed to nonspecific adsorption during the deposition process (Fig. 5J). This indicates that the DRE process did not affect the DNA recovery yield. Compared with the silicon encapsulation system (encapsulation time more than 4 days) (12–14), our work could switch the file operation state and long-term preservation state more quickly and ensures that the substrate material remains intact, allowing for DMRM than the silicon encapsulation system, which laid the foundation for building an integrated and compact DNA storage device.

Miniaturized and completely automated DNA cassette tape drive for integrated data storage

Just as the latest commercial tape drives, which perform file operations (such as reading and rewriting, and so forth) and management (such as file indexing and directory) on the inserted cassette tape (33, 49), we also designed and prepared a DNA tape drive for DNA tape, which was roughly the same size as a commercial tape drive (Fig. 6A). DNA tape drive was composed of four parts (micro-computer, motor system, addressing system, and liquid delivery system) and the preset program in the micro-computer controlled the other three systems to automate the file addressing, DNA file recovery, DNA file removal, and DNA file redeposition of files (Fig. 6B and text S7). The DNA tape was rolled up and stowed into a cassette, which contained a “head” that was used to press the target file into a micro-reaction chamber for subsequent file manipulation (Fig. 6C and text S7). Insert the DNA cassette tape into the DNA tape drive and operate on the screen to complete the fast addressing and manipulating of the files.

Fig. 6. Miniaturized and automated DNA tape drive for rapid file manipulation.

(A) DNA cassette tape and DNA tape drive held in hands. Scale bar, 5 cm. (B) The internal structure of the DNA tape drive. (C) Internal structure of the DNA cassette tape (top) and the “head” is used to recover the DNA files (bottom). (D) Three DNA image files and one text are deposited to the four addresses (c7_2, c8_6, e4_5, and b6_8, respectively), where SUSTech Address.txt is considered the false location to deposit. The text file is erased, and Puzzle 3.png is redeposited to recover the complete lantern image. (E) Micro-reaction chamber performs six operations on a single physical partition. (F) Operations of DNA file recoveries and redeposition. (G) Time axis of the operation of the DNA tape drive, which automatically recovers three DNA image files and redeposits one DNA text file (the colors on the timeline represent the different steps, the white bar for addressing, the pink bar for decapsulating, the yellow bar for extracting files, the gray bar for recovering, the purple bar for encapsulating, the green bar for erasing files, the blue bar for redepositing DNA files, and the red bar for washing the DNA tape), and the histograms display the pcpm of the DNA files recovery. The oligo numbers of SUSTech Address.txt have been corrected to the Puzzle 3.png in the histogram analysis. (H) DNA tape drive successfully recovers Lantern.png.

To evaluate the robustness of the DNA tape drive, we set four picture files (four puzzle pieces pictures can form a Chinese lantern picture) and randomly allocated four data partition addresses on the DNA tape (Puzzle 1.png, c7_2, Puzzle 2.png, c8_6, Puzzle 3.png, e4_5, and Puzzle 4.png, b6_8). If the DNA tape drive has errors in any process (such as errors in the file addressing, errors in extracting the target file, failure of the liquid supply system, etc.), then some or all of the four puzzle pieces pictures will not be recovered, which will result in “the Chinese lantern” picture being unrestorable, same as the jigsaw puzzle. In the DNA deposition process, we deposited the SUSTech Address.txt into the “e4_5” address to make a false location for the deposit. To successfully store and recoveries the “Lantern.png,” we had to redeposit the SUSTech Address.txt into the “Puzzle 3.png” (Fig. 6D). We prepositioned the six reagents that were related to encapsulation and decapsulation, DNA file recovery, and DNA redeposition process in the syringes of the DNA tape drive and programmatically controlled the fluid entry and exit within the micro-reaction chamber (Fig. 6E). A typical DNA file recovery goes through four steps (decapsulation, files recovery, template recovery, and encapsulation) and takes about 25 min, while a typical DNA file redeposition goes through six steps (decapsulation, file removal, file redeposition, file recovery, template recovery, and encapsulation) and takes about 50 min (Fig. 6F). Last, three DNA image file recoveries and one DNA image file redeposition were completed in about 150 min, all four images had normal sequencing profile distribution and the SUSTech Address.txt sequencing profile disappeared after the redeposition process (Fig. 6G), and last, the lantern images were successfully recovered (Fig. 6H). It is worth noting that the above process can be reduced to 47 min if the recovery operation and the synthesis of protection layers are performed uniformly on the data partition after the DNA data recovery (movie S1).

DISCUSSION

In this work, we established a DNA-based data storage cassette tape model, which has the following advantages and disadvantages:

First, the rolled configuration of the DNA tape efficiently maximizes the spatial utilization of the material, enabling portability and extending the number of available areas and storage capacity by increasing its length. By transforming a two-dimensional plane into a three-dimensional structure, DNA tape allows continuous extension in one direction while maintaining a compact volume, enhancing the density of available area per unit volume. In addition, by incorporating barcode patterns without expanding the overall volume of the roll, DNA tape achieves addressable physical partitions with a density of 364 partitions/cm³. This partition density is 30 times higher than that of the digital microfluidic (DMF) chip (table S5 and text S1) despite each partition on the DNA tape using ~4 mm², which is four to five times the area of each spot (a radius of ~1 mm) deposited on the DMF chip. However, the partition density on DNA tape is lower than that of current particle-based storage (table S5). Nevertheless, when attempting to store and manipulate 10⁵ files, particle-based storage technology requires semi-automated operation within 615 liters of physical space (table S5); DMF chips can achieve the same within 22.92 liters of physical space, although this requires manual handling to move the glass slides and assemble them into DMF chips (table S5). In contrast, DNA cassette tape system can operate fully automatically within a physical space of 6.85 liters. In addition, using more complex barcodes on the DNA tape (such as code-49 or code-16k), akin to adding more tracks on magnetic tape, or creating more precise physical partitions with addressing patterns [e.g., using lithography technology (50)], could further increase partition density. However, this would require more advanced optical recognition systems or higher control resolution “heads” than this work for file addressing and manipulation.

Second, the DNA tape allows for convenient integration of functionalities and devices on the tape path, and through motorized operation, it can perform specific tasks in designated regions. One example is DNA deposition, which can be integrated into a tape system. Each file deposition process is faster (takes ~1 min for a single file, the theoretical limit achieves 2.3 files/s in continuous conduction mode) within an acceptable loss compared to some DNA deposition methods of solvent evaporation (24), electrostatic interaction (12, 15, 51), and coupling (8, 16) (fig. S24), and the tape rotation allows for the continuous deposition of DNA files. While inkjet methods have been used for DNA synthesis and deposition on two-dimensional chips (24, 52), increasing data storage capacity by simply raising dot density is challenging. Expanding data capacity without altering dot density typically requires enlarging the chip size. DNA tape overcomes this by converting the two-dimensional surface into a three-dimensional structure, allowing for infinite extension in one direction while maintaining a compact volume, thus linearly increasing data storage capacity and shortening the deposition time.

Another example is the ability to perform various biochemical processes, such as DNA deposition, DMRM, removal, and encapsulation/decapsulation, with high precision in a small, confined space by incorporating micro-reaction chambers. Unlike current microfluidic-based technologies for DNA storage (8, 9, 24–26), which require complex channel designs, micro-fabrication, and chip assembly to manipulate the liquid into a specific solid-phase region for subsequent mixing, reactions, or achieving specific functions. In contrast, our approach allows for manipulating the solid phase into the liquid for reaction by using a motor to rotate and a simple mechanical “head” to move up and down. This setup immerses the tape into the liquid and uses hydrophobic barriers on the tape to ensure limited diffusion of the liquid on the tape. This setup enables these enzymatic reactions (file removal) and polymerization (file redeposition) to be performed in designated solid regions and the generation (encapsulation) and removal of micro-nanostructures (decapsulation) and is universal for other liquid-solid operations. In our current setup, although the “head” can only remove and replace data in one partition at a time, the linear encoding method allows us to demonstrate the continuous operation of the “head” to process multiple partitions, enabling precise text editing within files (text S5).

In addition, for long-term use of the tape, the DNA files on the tape will only remain readable if the indexing system and the tape are at least as long-term chemically stable as the DNA. We observed that the barcode patterns and nylon layers did not detach or disappear (text S5), and the remaining stress of the tape was more than 99% after accelerated aging (text S8). Thus, it enables the long-term loading of DNA files. Another concern for the long-term use of the tape is wear and breakage. Although the surface roughness of a specific area on the tape decreased after 100 wear cycles, there was no notably detachment of the encapsulation layer (text S8). However, for high-speed file addressing and frequent file access, a more wear-resistant coating may be needed to ensure that the encapsulation layer continues to protect the DNA. Moreover, the DNA tape exhibits a comparable maximum force for fracture to commercial data storage tapes (Dell LTO Ultrium 3) (text S8). Even in the event of accidental breakage, the DNA tape can be rejoined using transparent adhesive tape to maintain the mechanical integrity, barcode readability, and file accessibility required for continued operation (text S8). For long-term storage, the tape casing should be made of oxidation- and corrosion-resistant materials such as titanium or stainless steel to prevent disintegration of the casing.

Our approach not only enhances operational flexibility but also supports the scalability of storage systems by allowing modular integration of additional components for different functions. For example, adding a barcode reader can easily introduce the file addressing system to the cassette tape system. The file addressing rate on DNA tape (1570 files/s, using a 10-character barcode) rivals Boolean search methods (12) (1000 files/s with multiple copies per file) and achieves a smaller storage space than desktop-level DNA data storage through non-oligonucleotide synthesis methods (table S5). If a physical partition is filled with data (2 mm by 5 mm by 0.159 mm by 3.23 by 10¹² bits/mm³ = 5.14 × 10¹² bits, corresponding to 0.59 TB), the rate of file addressing can reach about 917 TB/s (0.59 TB × 1570 files/s) and ~ 9 × 10⁵ times faster than the Boolean search method (~1 GB/s). In addition, the most advanced barcode reader based on the laser can achieve faster recognition speed (which supports 148 symbols of code-128, 1300 barcodes/s, barcode reader BL-1300 series, Keyence), which can theoretically reach a file addressing speed at 577,200 files/s (details in text S2). It is important to note that the file addressing rate refers to the time required to locate the physical position of the DNA file, excluding subsequent DNA recovery, amplification, and sequencing operations. Thus, when reading data in parallel, the file retrieval time on the DNA tape needs to be cumulatively increased. This may necessitate adding more “heads” to increase the throughput of file recovery. Because of the aforementioned characteristics, different functional modules can be easily integrated into a compact device, creating a fully automatic DNA cassette tape drive.

Last, DNA cassette tape offers several cost-saving advantages in data addressing and multiple recoveries (tables S2 and S4) and serves as a scalable and customizable compact memory solution. The barcode information embedded within the DNA tape is flexible, allowing for the customization of files stored on the tape. This flexibility enables a collection of DNA cassette tapes to be mapped to different customers with unique catalogs, making it an adaptable storage system for diverse user needs. Furthermore, we developed a compact DNA tape drive to achieve a fully automated closed-loop operation for DNA files, including deposition, addressing, recovery, repeated deposition, and recovery, excluding DNA synthesis and sequencing. Nonetheless, it is essential to acknowledge that the large-scale, low-cost synthesis of oligo pools and their integration into storage systems remains a substantial barrier to the practical implementation of DNA data storage. Practical solutions for large-scale input of oligo pools into DNA cassette tape systems may require adaptation to novel DNA data-generation technologies, development of customized pipetting systems, or in situ synthesis of DNA sequences to increase the bandwidth of molecular data transportation to improve the bottleneck of DNA input (text S9). Encouragingly, we have validated the feasibility of phosphoramidite-based DNA synthesis and sequencing-by-synthesis on the tape (fig. S25). Because of the modular nature of DNA tape, our next step is to implement a portable, end-to-end DNA data storage device.

MATERIALS AND METHODS

Materials and characterization

Nylon C membranes were obtained from PALL Corporation. EDC and sodium bromide were obtained from Sigma-Aldrich (Shanghai, China). Sodium hydroxide (NaOH), acetic acid, sodium bicarbonate (NaHCO₃), zinc nitrate hexahydrate [Zn(NO₃)₂.6H₂O], citric acid, and disodium hydrogen phosphate (Na₂HPO₄) were obtained from Aladdin (Shanghai, China). SDS and 2-methylimidazole were obtained from Macklin (Shanghai, China). A 1 × tri(hydroxymethyl)aminomethane (Tris)-ethylene diamine tetraactic acid (EDTA) buffer solution (TE, pH 7.4) was obtained from Bioroyee (Beijing, China). A 20× SSPE buffer solution (pH 7.4) was obtained from OKA (Beijing, China). Ultrapure water was obtained by an ultrapure water system (PURELAB Ultra FLC00006307, ELGA). All DNA oligo pools were synthesized by Twist Bioscience. All primers and gene handles were obtained from Sangon Biotech (Guangzhou, China). TaKaRa LA Taq DNA polymerase and deoxynucleotide triphosphate (dNTP) Mixture were obtained from Takara Biotech (Beijing, China). DNase I and Bst 2.0 WarmStart DNA polymerase were obtained from New England Lab. QIAquick Gel Extraction Kit was obtained from QIAGEN (Shanghai, China). PureLink Quick PCR Purification Kit, Qubit dsDNA HS, and BR Assay Kit were obtained from Thermo Fisher Scientific (Shanghai, China). PowerUp SYBR Green was obtained from Applied Biosystem (Shanghai, China). Amino-terminated polyethylene glycol was obtained from Chongqing Yusi Pharmaceutical Technology Co. Ltd. (Chongqing, China).

The scanning electron microscope images were taken using SU8220 (Hitachi, Japan). The contact angle images were taken using a contact angle tester (DSA25, KRUSS, Germany). The atomic force microscope (AFM) images were taken using Dimension Icon (Bruker, Germany). The optical images were taken using a smartphone (iPhone 14 Pro, Apple, China). The XPS data were taken using K-Alpha (Thermo Fisher Scientific, USA). Mechanical tensile test is performed on a universal testing machine (CMT5205, SUST, China).

Generating the DNA file sequences

A text file (0.11 KB, the SUSTech Address; fig. S19) and four image files (51.6, 24.8, 46.9, 33.3, and KB, the lantern image; fig. S19) were encoded using DNA Fountain software with the following parameters: --m 3 --gc 0.05 --rs 5 --delta 0.05. These parameters ensure that the generated sequences contain up to three contiguous identical bases, a CG content of about 50 ± 5%, and 5 B of RS code combined with 16 B of data information within each oligo. DNA file sequences were generated with a length of 100 nt (text file, 41 oligos, four image files, 5010 oligos, 2476 oligos, 4476 oligos, and 4391 oligos) first, added a 4-nt restriction endonuclease restriction recognition sequence at the 3′ end, and added 20-nt adapter sequences at both ends for DNA file deposition, recovery, amplification, and qPCR. The primer sequences are shown in table S3. The text file was synthesized by column-based chemical synthesis (Sangong, China), and image files were synthesized by inkjet printing technology (Twist, USA).

DNA sequencing and decoding

After obtaining the solution after DNA file recovery, the solution was adjusted to neutrality by acetic acid solution, followed by PCR (a 25-μl system, 11.5 μl of H₂O, 1 μl of template, 4 μl of 100 μM forward primer, 4 μl of 100 μM reverse primer, 2.5 μl of reaction buffer, 2.5 μl of dNTP, and 0.25 μl of LA Taq DNA polymerase) to amplify the file sequence, and the PCR procedure was as follows: (i) 95°C for 2 min, (ii) 95°C for 30 s, (iii) 60°C for 20 s, (iv) 72°C for 15 s, (ii to iv) repeated 30 times. After amplification, files were sequenced on the DNBSEQ-T7 platform (MGI, Shenzhen, China). The raw sequencing data were trimmed with cutadapt using the following parameters: --discard-untrimmed -m 21. Reads were aligned to template sequences using Burrows-Wheeler Aligner with default parameters, and the quality of the reads ≥ Q30 was used for analysis and decoding. The processed sequencing results were decoded for DNA decoding using the DNA Fountain software.

Quantification of the DNA

Synthetic DNA pools were first resuspended to 10 ng per μl in water and kept at 4°C until further use. We used qPCR (Thermo Fisher QuantStudio 7) to quantify DNA to draw standard curves using the synthetic DNA pools as templates. First, the synthesized DNA pools were diluted 100 times as PCR templates, and the procedure was performed as described previously. Next, the amplified DNA was measured using the Qubit Assay Kit (Invitrogen) and used as a known concentration. The known concentration DNA solution was diluted in a gradient, and a 50-μl qPCR system was configured (13 μl of H₂O, 1 μl of the sample, 8 μl of 1 μM forward primer, 8 μl of 1 μM reverse primer, and 20 μl of PowerUp SYBR Green Master Mix). qPCR was performed with the following thermal protocol: (i) 50°C for 2 min, (ii) 95°C for 2 min, (iii) 95°C for 15 s, (iv) 60°C for 15 s, (v) 72°C for 60 s, (iii to iv) repeated 30 times. The unknown DNA concentrations were obtained using qPCR, and the cycle threshold (Ct) values obtained were brought into the fitted standard curve to obtain the unknown concentration.

Fabrication of the DNA cassette tape

The nylon C membranes were cut to 20 cm by 15 cm with a paper cutter (deli) and taped onto A4 printing paper using transparent tape. Next, print the preset barcode pattern onto the nylon membranes using an inkjet printer (HP, M281fdw). A mixture of PDMS and the curing agent (10:1) was prepared and applied evenly to the bar area of the barcode by a syringe and completing the preparation of the hydrophobic barrier on the DNA tape. Next, the nylon membranes with the barcode pattern were cut into 20 cm by 0.5 cm strips by a paper cutter and attached first and last using double-sided adhesive tape (3M) to form an extra-long tape. The gene handle solution was prepared by mixing 100 μM gene handle solution with 750 mM NaHCO₃ in a 1:2 ratio and introduced into the space area of the DNA tape for 10 min, which had been immersed in 16% (w/v) fresh EDC solution for 10 min. The DNA tape was placed at 4°C until it is used.

To prepare encapsulated DNA tape, DNA tape was first immersed in the 2-Melm solution (700 mM), and then an equal volume of zinc nitrate solution (70 mM) was introduced into the system and allowed to wait for 10 min. After three washes, the encapsulated DNA tape was obtained. For decapsulating on DNA tape, 0.2 M NaHPO₄ solution and 0.1 M citric acid solution were mixed in the ratio of 8.8:11.2 to obtain a buffer (pH = 4.5). The encapsulated DNA tape was immersed in the buffer for 15 s, washed three times with water to remove the protective layer of ZIFs, and used for subsequent experiments. The DNA tape cassette was custom-made by computer numerical control machining.

DNA synthesis and sequencing on DNA tape

To realize DNA synthesis using the phosphoramidite chemistry on DNA tape, we first immersed the tape in 16% (w/v) EDC for 10 min and reacted with amino-terminated polyethylene glycol solution (with 750 mM NaHCO₃ in a 1:2 ratio) for 1 hour. After modifying the surface of the DNA tape with hydroxyl groups, the DNA tape was cut into 2 mm by 2 mm pieces and placed into DNA synthesis columns (DS0025, Biocomma Biotechnology Co. Ltd., Shenzhen, China), and used a DNA synthesizer (YB-192S, Yibo Biotechnology Co. Ltd., Shanghai, China) to synthesize 40-nt oligonucleotides (5-ATAAATGACCTGCCGTGCAATGGCTCATTTCACAATCGGT-3) following standard DNA synthesis protocols. Each cycle of the synthesis process included four steps: deblocking, coupling, capping, and oxidation. The deblocking reagent used was 5% (v/v) trifluoroacetic acid (TFA) in acetonitrile. The oxidation reagent consisted of 0.1 M iodine dissolved in a 9:1 pyridine/acetic acid solution (v/v). The coupling reagent was 0.5 M tetrazole in acetonitrile. The capping reagents were obtained from Suresyn Co. Ltd. (Shenzhen, China). Each synthesis cycle includes deblocking (120 s), washing (120 s), coupling A/T/C/G (480 s), washing (120 s), capping (80 s), washing (120 s), oxidizing (80 s), and washing (120 s) steps. Subsequently, we performed in situ PCR amplification on the synthesized DNA. The primers used for the PCR are listed in table S3.

File addressing rate test

We configured the DNA tape with 10 consecutive barcodes containing identical data characters. In addition, we established groups of barcodes with 2 (a0-a9), 4 (ab01-ab09), 6 (abc001-abc009), 8 (abcd0001-abcd0009), and 10 (abcde00001-abcde00009) data characters. By setting different rotational speeds within the range of 0 to 2000 rpm, we obtained sampling results via the terminal and calculated the accuracy of barcode reading. Each experiment was repeated three times, and the calculation formula is as follows

$$\begin{array}{c}\mathit{\text{Readable rate}}(\%)\hfill \\ =\frac{\mathit{\text{The number of accurately read barcodes}}}{\mathit{\text{The total number of barcodes}}}\times 100\%\hfill \end{array}$$

File deposition on DNA tape

To deposit DNA data on the DNA tape, the configured DNA file solution (1 μl, consisting of synthetic DNA file (0.2 ng/μl), 0.05 mM dNTP, 8 U of Bst 2.0 WarmStart DNA polymerase, 10× reaction buffer, and 6 mM MgSO₄) was introduced through a syringe onto the physical partitions of the DNA tape and reacted at 60°C, 80% RH for 1 min to complete the deposition process of the DNA file.

Multiple file recovery and redeposition on DNA tape

The target file to be recovered from the DNA tape was immersed in 20 μl of NaOH (0.15 M) for 5 min, and the alkaline liquid was collected to complete the file recovery operation. The collected alkaline liquid was adjusted to neutral by adding 5 μl of 1 × TE (pH = 7.4) and acetic acid solution and then sequencing and decoding to complete the file recovery. The area which had been read on the DNA tape could be restored to the initial dsDNA state by adding recovery solution (A 50-μl system, including 1 μl of 10 μM primer, 1 μl of 2.5 mM dNTP, 1 μl of Bst 2.0 WarmStart DNA polymerase, 5 μl of reaction buffer, 3 μl of MgSO₄, and 39 μl of H₂O). After each step, the tape was washed five times with 2× SSPE/0.1% SDS solution and 1× TE buffer solution, respectively. We used the Puzzle 1.png file and repeated the above operation 10 times for multiple DNA file recovery experiments to collect the alkaline liquid and perform subsequent sequencing analysis.

The target DNA file to be redeposited on the DNA tape was immersed in 40 μl of reaction mix (4 μl of Mbo I endonuclease, 8 μl of reaction buffer, and 28 μl of H₂O) for 15 min at 37°C, followed by the addition of NaOH (0.15 M) for 5 min to complete the file removal. For subsequent DNA file deposition, the area to be redeposited on the DNA tape was immersed in 50 μl of reaction mix [including 1 μl of subsequent file (10 ng/μl), 1 μl of 2.5 mM dNTP, 1 μl of Bst 2.0 WarmStart DNA Polymerase, 5 μl of reaction buffer, 3 μl of MgSO₄, and 39 μl of H₂O] for 10 min at 60°C to complete the DNA file redeposition. After each step, the tape was washed five times with 2× SSPE/0.1% SDS solution and 1× TE buffer solution, respectively. We used the first 10 sequences of SUSTech Address.txt and repeated the above operation 10 times for multiple redeposition experiments. The DNA file recovery operation was performed before each redeposition to collect the alkaline liquid from the extracted file and perform subsequent sequencing analysis.

Stability test of DNA tape

The DNA files were deposited onto DNA tape by the above method and cut to 1 mm–by–5 mm size. Then, the DNA tapes were placed in 2-Melm solutions of different concentrations (10, 100, 300, 500, 700, and 1000 mM). Different concentrations of Zn(NO₃)₂^.6H₂O (1, 10, 30, 50, 70, and 100 mM) was added dropwise and reacted for 10 min to form different thicknesses of ZIFs protective layer, followed by adding DNase I (19.5 μl of H₂O, 0.5 μl of 1 M MgSO₄, and 1 μl of DNase I) to them for digestion for 30 min. After decapsulated, DNA files were extracted using the method described above, and the concentration and quality of DNA files were analyzed using qPCR and MGI sequencing, respectively. For the DNA control group, 0.5 μl of MgSO₄ (1 M) and 1 μl of DNase I were added to 19.5 μl of DNA files (~3.2 × 10⁷ copies/μl), and the DNA tape control group did not experience synthesizing the protective layer.

For accelerated aging experiments, DNA tape and encapsulated DNA tape were placed at 50% humidity and different temperatures (60°, 65°, and 70°C) for 3 weeks as the previous research described (13, 17, 53). The 50% humidity environment was maintained by placing a saturated sodium bromide solution in a gas-tight container. DNA copies at different time points were obtained using the DNA file recovery method and qPCR.

Automated recovery and redeposition of the DNA files

We divided Lantern.png into four images (Puzzle 1.png, Puzzle 2.png, Puzzle 3.png, and Puzzle 4.png) and deposited them randomly into the physical data partition of DNA tape (the address assignment symbols of the four images are c7_2, c8_6, e4_4, and b6_9, respectively) using the DNA file deposition method described before. During the deposition process, SUSTech Address.txt had been intentionally deposited to the file address of Puzzle 3.png, and the removal of SUSTech Address.txt and the redeposition of the Puzzle 3.png file had been completed by the automatic operation of the DNA tape drive, which lastly completes the reading of Lantern.png after sequencing. The decapsulation solution mixture (1 ml, pH = 4.5, weak acid buffer), file recoveries solution mixture (1 ml, 0.15 M, NaOH), file recovery solution mixture (a 1-ml system, including 20 μl of 100 μM primer, 20 μl of 2.5 mM dNTP, 20 μl of Bst 2.0 WarmStart DNA Polymerase, 100 μl of reaction buffer, 60 μl of MgSO₄, and 780 μl of H₂O), file erasing solution mixture (A 400-μl system, 40 μl of Mbo I endonuclease, 80 μl of reaction buffer, and 280 μl of H₂O), subsequent file solution mixture [a 1-ml system, 20 μl of subsequent file (10 ng/μl), 20 μl of 2.5 mM dNTP, 20 μl of Bst 2.0 WarmStart DNA Polymerase, 100 μl of reaction buffer, 60 μl of MgSO₄, and 780 μl of H₂O], and encapsulation solution [500 μl, 70 mM Zn(NO₃)₂^.6H₂O and 500 μl, 700 mM 2-Melm, two different liquids in syringes were connected using a three-way valve] were first loaded into the liquid supply chamber of the DNA tape dive.

We designed an interactive interface and control program in Unity software to perform file manipulation in the DNA tape drive. The control program can control the barcode recognition camera, laser locator, and motor for primary and secondary positioning. Puzzle 1.png, Puzzle 2.png, SUSTech Address.txt, and Puzzle 4.png were pre-entered into the system directory of the DNA tape dive. Select the Puzzle 1.png, Puzzle 2.png, and Puzzle 4.png in the file directory on the touch screen display and set the recovery parameters as follows: Step 1 is the flow of the decapsulation solution mixture (liquid supply flow rate and time: 40 μl/min, 15 s, discharge flow rate and time: 40 μl/min, 15 s), step 2 is the flow of the file recoveries solution mixture (liquid supply flow rate and time: 160 μl/min, 12 s; discharge flow rate and time: 0 μL/min, 0 s; duration: 288 s) through the microfluidic duct into the micro-reaction chamber, step 3 is the flow of the file recovery solution mixture (liquid supply flow rate and time: 160 μl/min, 12 s; discharge flow rate and time: 0 μl/min, 0 s; duration: 588 s; temperature: 60°C) through the microfluidic duct into the micro-reaction chamber, step 4 is the flow of the encapsulation solution (liquid supply flow rate and time: 160 μl/min, 12 s; discharge flow rate and time: 0 μl/min, 0 s; duration: 588 s) through the microfluidic duct into the micro-reaction chamber. The “erase data” were selected from e4_4 and the subsequent file Puzzle 3.png to e4_4 was deposited on the screen display to complete the DNA file redeposition operation.

The DNA redeposition parameters are set as follows: Step 1 is the flow of the decapsulation solution mixture (liquid supply flow rate and time: 40 μl/min, 15 s; discharge flow rate and time: 40 μl/min, 15 s) through the microfluidic duct into the micro-reaction chamber, step 2 is the flow of the file erasing solution mixture (liquid supply flow rate and time: 160 μl/min, 12 s, discharge flow rate and time: 0 μl/min, 0 s; duration: 888 s, temperature: 37°C) through the microfluidic duct into the micro-reaction chamber, step 3 is the flow of the subsequent file solution mixture (liquid supply flow rate and time: 160 μl/min, 12 s; discharge flow rate and time: 0 μl/min, 0 s; duration: 588 s; temperature: 60°C) through the microfluidic duct into the micro-reaction chamber, step 4 is the flow of the file recoveries solution mixture (liquid supply flow rate and time: 160 μl/min, 12 s, discharge flow rate and time: 0 μl/min, 0 s; duration: 288 s) through the microfluidic duct into the micro-reaction chamber, step 5 is the flow of the file recovery solution mixture (liquid supply flow rate and time: 160 μl/min, 12 s; discharge flow rate and time: 0 μl/min, 0 s; duration: 588 s; temperature: 60°C) through the microfluidic duct into the micro-reaction chamber, step 6 is the flow of the encapsulation solution (liquid supply flow rate and time: 160 μl/min, 12 s; discharge flow rate and time: 0 μl/min, 0 s; duration: 588 s) through the microfluidic duct into the micro-reaction chamber. Afterward, the DNA tape drive automatically completed the setup steps and collected the target file into a liquid collection chamber for subsequent sequencing and decoding.

Supplementary Materials

The PDF file includes:

Texts S1 to S9

Figs. S1 to S25

Tables S1 to S5

Legend for movie S1

References

Other Supplementary Material for this manuscript includes the following:

Movie S1