S C I E N C E A D VA N C E S | R E S E A R C H A R T I C L E

S Y N T H E T I C B I O LO G Y Copyright © 2023
The Authors, some
Multiplex enzymatic synthesis of DNA with single-base rights reserved;
exclusive licensee
resolution American Association
for the Advancement
of Science. No claim to
Damiano Verardo†, Beatrice Adelizzi†, Daniel A. Rodriguez-Pinzon†, Nicolas Moghaddam†, original U.S. Government
Emma Thomée†, Tessa Loman, Xavier Godron†, Adrian Horgan*† Works. Distributed
under a Creative
Enzymatic DNA synthesis (EDS) is a promising benchtop and user-friendly method of nucleic acid synthesis that, Commons Attribution
instead of solvents and phosphoramidites, uses mild aqueous conditions and enzymes. For applications such as License 4.0 (CC BY).
protein engineering and spatial transcriptomics that require either oligo pools or arrays with high sequence
diversity, the EDS method needs to be adapted and certain steps in the synthesis process spatially decoupled.
Here, we have used a synthesis cycle comprising a first step of site-specific silicon microelectromechanical
system inkjet dispensing of terminal deoxynucleotidyl transferase enzyme and 30 blocked nucleotide, and a
second step of bulk slide washing to remove the 30 blocking group. By repeating the cycle on a substrate
with an immobilized DNA primer, we show that microscale spatial control of nucleic acid sequence and
length is possible, which, here, are assayed by hybridization and gel electrophoresis. This work is distinctive
for enzymatically synthesizing DNA in a highly parallel manner with single base control.

Downloaded from [Link] on January 15, 2026

INTRODUCTION others to suggest a hybrid strategy where the enzyme is attached
Enzymatic DNA synthesis (EDS) has the potential to be faster and through the terminal phosphate of the nucleotide, and polyaddition
give higher-purity DNA with none of the drawbacks (1, 2) of the is prevented by a reversible terminator at the 30 position (14).
conventional phosphoramidite DNA synthesis method first pro- Reversible terminators are 30 blocking moieties that can be con-
posed by Beaucage and Caruthers (3). Instead of harmful chemicals verted when needed to extendable 30 -OH (15–18). They can be used
and flammable solvents, it relies on mild aqueous conditions and an to exert single-base resolution, but the choice is quite restricted once
enzyme called terminal deoxynucleotidyl transferase (TdT). Un- stringent limits are placed on deblocking speed and efficiency. Fur-
usually for a polymerase, TdT does not need a template and is thermore, they need to be handled carefully as they may be good
able to complex a broad range of divalent cations (e.g., Co2+) to nucleophiles (19), and the wrong conditions may provoke side re-
lower the pKa (where Ka is the acid dissociation constant) of the actions such as deamination and the formation of adducts (20, 21).
OH moiety at the 30 end of a DNA primer (4). This facilitates nu- Moreover, a necessary development step before their implementa-
cleophilic attack on the 50 -phosphate of an incoming nucleotide to tion is the screening of thousands of TdT variants to find one that
generate an extended primer and one molecule of pyrophosphate as accepts the 30 blocking group, which can be a laborious iterative
a by-product. Although addition is usually spontaneous, the com- process (22).
position of the DNA primer; the presence of secondary structure; A third way to wield control over the sequence has been called
and the position, shape, and size of any modifications on the nucle- “free running enzymatic DNA synthesis” (23). This de novo DNA
otide may all influence the rate at which they are added (5–8). To synthesis method modulates conditions to limit polycondensation
ensure that addition occurs unfettered, a denaturing agent may be to a minimum and is most suited to DNA data storage and related
used to lower the melting temperature of the primer and the enzyme applications (24). One of the strategies proposed uses apyrase to
modified to enable its operation at elevated temperature (9). compete for nucleotides and convert them to less TdT-active mo-
Under optimum conditions, EDS coupling times are faster than nophosphate and diphosphate analogs (25). Another uses reversible
in phosphoramidite synthesis (10), and left to itself, TdT will add caging of Co2+ to regulate TdT activity and provides the only dem-
nucleotides to the DNA primer unchecked (11, 12). Different strat- onstration to date of multiplexed EDS (26). While notable, the
egies have been proposed to prevent polycondensation to enable number of synthesis cycles performed (8) and the number of indi-
custom oligonucleotide synthesis (see fig. S1). One approach uses vidual sites of DNA synthesis (12) were relatively modest, and the
the enzyme to block further primer extension until a labile linker, approach lacked single-base control.
positioned between the enzyme and the nucleotide, is cleaved in a In this work, we report the first multiplex EDS demonstration
subsequent step to detach the enzyme from the primer (13). An ad- with single-base control. We also show a longer length than non-
vantage of this route is fast kinetics without the need for enzyme multiplexing studies that showed single-nucleotide addition. With
engineering [because the nucleotide has a 30 -unblocked OH identi- optimization, the method could provide custom DNA for applica-
cal to a native deoxynucleotide triphosphate (dNTP) and conjuga- tions such as protein engineering. Microarray DNA is the cheapest
tion ensures a high local concentration of the substrate]. A potential available source of synthetic DNA but is currently made using con-
drawback is that it leaves a scar on the nucleobase. This has led ventional phosphoramidite chemistry (27). Here, we have used
silicon microelectromechanical system (MEMS) piezoelectric print-
heads to deposit enzyme and nucleotides with 30 reversible termi-
DNA Script, 67 Avenue de Fontainebleau, 94270 Le Kremlin-Bicêtre, France. nator groups site-specifically onto a substrate. This is radically
*Corresponding author. Email: [Link]@[Link] different from the industrial-scale inkjet DNA synthesis using
†These authors contributed equally to this work.

Verardo et al., Sci. Adv. 9, eadi0263 (2023) 7 July 2023 1 of 12

S C I E N C E A D VA N C E S | R E S E A R C H A R T I C L E

phosphoramidites commercialized by Agilent and Twist Bioscienc- the liquid to be swept off the slide in a single motion. We also note
es (28, 29). Inspired by the POSaM platform (30–32), we have used that some intended improvements might have drawbacks. For in-
low-cost, off-the-shelf components and built a benchtop instru- stance, a thermally conductive gold or silicon substrate might facil-
ment for de novo synthesis of DNA microarrays in research labora- itate warming during incubation but might complicate imaging.
tories. The benchtop advantage EDS affords is that it removes any
requirement for anhydrous solvents, dry atmosphere, airtightness, Surface chemistry
and inert gas, i.e., glove box conditions (33, 34). DNA may be purchased with various 50 modifications (e.g., amino
and thiol) and covalently immobilized to silanized or polymer-
coated substrates for use as a primer for EDS. In the present
RESULTS work, a 50 -DBCO (dibenzocyclooctyne) primer with a photocleav-
The inkjet platform able group (see fig. S2) was incubated with an azide-modified glass
The layout of the inkjet synthesis platform and the relationship with substrate at room temperature in aqueous conditions (see Materials
the EDS synthesis cycle are depicted in Fig. 1. To spatially control and Methods). The copper-free Click SPAAC (strain-promoted
the 30 extension of DNA primers on a substrate, four thin-film azide–alkyne cycloaddition) reaction (46) was observed to be very
silicon MEMS printheads with 600 dots per inch (dpi) native reso- fast (<10 min) in agreement with the literature (47) and resulted in a
lution were used to print inks containing enzymes and nucleotides uniform layer of DNA as judged by signal intensity (see fig. S3).
(A, C, G, and T). In the ink were also additives to prevent evapora- While incubation with different DBCO-modified small molecules
tion, reduce secondary structure, and scavenge adventitious agents and oligomers modified the contact angle, suggesting that there
that might cap the 30 -aminoxy blocking group (e.g., formaldehyde) were residual azide groups remaining on the surface, quenching
(35). Reproducible and reliable printing without satellite spots was was not used in the present work.

Downloaded from [Link] on January 15, 2026

assured by active control of the meniscus pressure at the nozzle plate To determine the relationship between ink and surface and to
and by adjusting the ink’s rheology into a range identified by a plot examine the influence of synthesis parameters more easily, a new
of Ohnesorge number against Reynolds number (36). A lower vis- “move-stop-print” line-by-line printing process controlled by the
cosity ink was chosen to avoid the possibility of negatively affecting IPA was developed (Fig. 2A). It comprised a series of rapid (<500
enzyme activity (37), and tests were done to ascertain whether TdT ms) stepwise movements in the Y direction in between which
activity was altered by the choice of detergent and by jetting, given nozzles in the perpendicular (nozzle row) direction (X ) were actu-
that ejection occurs at several meters per second and involves high ated while the substrate was stationary. Each nozzle actuation de-
shear (38–45). Latency, a negative behavior of ink after a period posited a ~3-pl droplet on the substrate below the nozzle location.
sitting idle in the printhead, was reduced by nozzle tickling, period- Images resembling one-dimensional (1D) barcodes were used to de-
ic dummy printing, and increasing the humidity in the enclosure, termine which nozzles were fired and how many times the ad-
which also helped to ensure that the picoliter droplets deposited on dressed nozzles were fired at each stop (see fig. S4). Among its
the substrate did not evaporate during incubation postprinting. advantages, the method allows the deposition on the substrate of
Automation of the synthesis cycle involved moving a substrate (a any drop volume in precise locations without repeated back and
standard microscope slide coated uniformly with a cleavable DNA forth movements and printing on the fly based on encoder position.
primer) under the printheads to an imaging station, then to a wash It is also more time efficient for large droplet (i.e., wet)/spot (i.e.,
station, and then back to the starting (home) position (Fig. 1A). dry) sizes, which are easier to image and afford DNA spots with
Slide washing was initiated after incubating the enzyme and nucle- higher mass for faster, simpler sample analysis without a need for
otide for 5 to 10 min during which time the slide was imaged, and polymerase chain reaction (PCR) amplification and next-genera-
the positions and sizes of the printed spots were recorded. Com- tion sequencing (NGS) library preparation.
pressed air was used to convey liquids to a wash head equipped Printing tests were done with two jettable inks, one containing a
with four electronic valves (Fig. 1B), each fluidically connected to 50 -DBCO primer (here called “DNA ink”) and the other having an
a row of four nozzles and under the control of an Arduino board enzyme and a 30 blocked nucleotide (here called “elongation ink”).
interfaced with an in-house software called the Inkjet Printer App The first tests aimed to show DNA coupling and DNA elongation
(IPA), written in C#. To remove wash liquids, the substrate was for different spot sizes. Both inks were printed at 75 dpi (see fig. S5)
mounted on a holder with trough and liquids were emptied on the appropriate surface (an azide-modified microscope slide for
under IPA control via a peristaltic pump (Fig. 1C). Slide drying DNA ink and an azide slide modified with a uniform layer of a 50 -
was performed after slide washing using a serpentine motion. DBCO primer for elongation ink) using a different number of drop-
The setup, as described, proved to be robust and effective. lets in each section of the slide (Fig. 2B). Imaging was done with an
However, some changes would be needed to lower instrument epifluorescence microscope after automated slide washing and
cost further while making synthesis faster and reagent use more ef- drying. An additional step of TdT-mediated end-labeling with 6-
ficient. For instance, a much simpler and cheaper means of control- Carboxyfluorescein (6-FAM)–labeled dideoxynucleotide was per-
ling humidity would pull moisture from a benchtop humidifier into formed before imaging in the case of the elongation ink (see Mate-
the enclosure using a fan activated using a proportional–integral– rials and Methods). In both cases, the dry spots were bright, regular
derivative (PID) controller connected to a humidity probe in the in shape, and uniformly distributed (see fig. S6). Both inks were also
enclosure. No effort was made in this work to control temperature printed on both surfaces and imaged wet without any further treat-
at the slide surface or within the Perspex enclosure, but one option ment (see fig. S7). The droplet sizes are reported in Fig. 2C. Using
would be to use a heated slide cover and apply the cover when in- the same drive waveform (48), the DNA ink gave larger droplets
cubating. From a washing point of view, more nozzles would give a than the elongation ink, and both inks spread more on the hydro-
more uniform dispense of the liquid, and an air knife would allow philic DNA surface than on the hydrophobic azide surface.

Verardo et al., Sci. Adv. 9, eadi0263 (2023) 7 July 2023 2 of 12

S C I E N C E A D VA N C E S | R E S E A R C H A R T I C L E

Downloaded from [Link] on January 15, 2026

Fig. 1. The EDS inkjet platform. (A) Schematic of the printer and the EDS synthesis cycle. The four printheads (600-dpi resolution), custom slide holder, and custom
washing/drying station are enclosed in a transparent box that protects against dust. The humidity inside the enclosure is controlled at ~65% RH using a Cellkraft PD-10
(not shown) to prevent ink drying at the printhead or on the substrate. Each printhead (1) deposits in a spatially addressable manner an ink containing a 30 -ONH2–blocked
nucleotide (red) and TdT enzyme onto a glass microscope slide on which is a 50 -end-immobilized DNA primer (50 -surface-DNA-30 -OH). After bright-field imaging with a
microscope (2) to determine droplet properties (presence, size, position, and shape), the substrate is moved in its holder to a wash/dry station (3). One of the wash fluids is
used to neutralize the enzyme reaction. Another converts the blocking group (red) on the 30 end of the extended DNA primer (50 -surface-DNAn+130 -ONH2) to a freely
extendable (blue) 30 -OH (50 -surface-DNAn+130 -OH) so another cycle of elongation can be performed. (B) Side view of the washing and drying station. The wash head has
three rows of four nozzles for deprotection buffer (DB), proteinase K wash buffer, and Milli-Q water. A fourth row is dedicated to slide drying and is connected to a
compressed air source. (C) View from above of the slide holder. The slide holder is mounted on two XY direct drive stages. It is made from polytetrafluoroethylene
and uses two stainless steel springs to push the glass slide against three metal retainer posts. The slide is suspended above a trough that has a waste drain attached
to a peristaltic pump. The pump is switched on during slide washing.

Having shown control over the droplet size, trials were done to changed to cover the range from 150 to 50 dpi (in the X direction).
see whether the separation between droplets (pitch) could be varied Collectively, the results are useful to determine an experimentally
at the same time as size by activating subsets of nozzles in the X di- convenient spot size and density for subsequent EDS tests (60
rection (nozzle row) (Fig. 2D). Printing was done on the azide pulses and 75 dpi). They also allow estimation of the maximum se-
surface using DNA ink, and the printed droplets were imaged quence diversity that might be achievable using the formulated inks
(see fig. S8). The number of pulses (droplets ejected) was varied and a hexagonal close-packed arrangement (~1 M per slide) versus
from 10 to 600, while the number of nozzles activated was the square pattern used here (see fig. S5).

Verardo et al., Sci. Adv. 9, eadi0263 (2023) 7 July 2023 3 of 12

S C I E N C E A D VA N C E S | R E S E A R C H A R T I C L E

Downloaded from [Link] on January 15, 2026

Fig. 2. Printing of inks. (A) The “move-stop-print” principle. (B) Change in spot size. Left: Epifluorescence microscope image of FAM-labeled DNA spots after printing of
DNA ink on an azide surface, coupling, and automated slide washing. The nozzles were pulsed 2, 20, 40, 60, 100, 150, 200, and 300 times from top to bottom. Center:
Image of an array printed with an ink formulated to allow elongation of a uniform DNA surface. Nozzles were actuated 2, 10, 15, 20, 30, 40, 60, and 80 times from top to
bottom. After incubating, nonprinted areas were manually capped with TdT + dideoxyadenosine triphosphate (ddATP). Automated slide washing was then performed to
remove the 30 -ONH2 protection; then, end-labeling was done manually with TdT + FAM-labeled ddATP. Right: Close-up of the right (center) array. (C) Wetting versus
surface type. A different number of droplets was printed in eight sections on the substrate and then imaged wet. To allow imaging, the elongation ink contained 50 μM
fluorescein. The reported mean droplet size is for two printed slides. (D) Control over pitch and droplet size. Images and normalized profiles obtained for DNA ink on an
azide slide using (top to bottom) 50 dpi: 600 pulses, 60 dpi: 300 pulses, 75 dpi: 200 pulses, 100 dpi: 50 pulses, 150 dpi: and 10 pulses. (E) Protein adsorption. Images of an
azide slide (left) and a DNA slide (right) after (top to bottom) first printing (60 pulses) with the TdT–green fluorescent protein (GFP) elongation ink; then automated
deprotection and water washing; then automated proteinase K treatment, deprotection, and water washing; and then printing elongation ink again (60 pulses). Slide
washing leads to nonuniform adsorption of TdT that can be removed by a proteinase K wash. (F) Effect of proteinase K treatment on droplet size. The effect on droplet size
is negligible for the DNA surface. Scale bars, 500 μm.

As a final printing test, an elongation ink at pH 6.0 containing the droplet arrays on the surface after proteinase K washing
green fluorescent protein (GFP)–labeled TdT was printed on the were regular.
azide and DNA surfaces to gauge the extent of protein adsorption
(Fig. 2E). TdT [isoelectric point (pI) ~ 6.5, melting transition tem- Ink formulation
perature (Tm) ~ 42°C] proved to be a more difficult protein to print Inks intended for use in inkjet printers should be degassed to avoid
than model proteins bovine serum albumin (pI ~ 4.5, Tm ~ 63°C) or the formation of air occlusions, should be filtered to avoid nozzle
lysozyme (pI ~ 11.4, Tm ~ 72°C), and this led us to suspect that blocking by debris, and should be inherently stable. To circumvent
protein adsorption might be an issue. It was observed that the drop- nozzle “drop out” caused by blockages, nozzles can be held in
lets of the GFP-TdT elongation ink are smaller than in Fig. 2C. We reserve, but if clogging is extensive, then the only viable solution
attribute this to the additional presence of GFP (pI ~ 6.2, Tm ~ 78°C) is to perform repeated priming and purging of the printhead
and its coupling to an earlier TdT mutant. A second observation during which time the printer is inactive. In the case of inkjet-
was that protein adsorption was nonuniform because of the enabled multiplex EDS, an additional requirement is that the ink
method of slide washing and was more severe on the azide remains active during synthesis (>24 hours if a cycle takes 10 min
surface than on the DNA surface. To counteract this problem, a pro- and a run has 150 cycles). Furthermore, it would be beneficial on
teinase K treatment was tried and found to be an effective solution cost and convenience grounds if the ink remained active for multi-
for renewing the surface in both cases (see fig. S9). Little evidence ple runs given that the inner volume of a printhead will exceed the
was found of proteinase K (or Tween 20) adsorption on the DNA volume printed in a typical synthesis run by one or two orders of
surface, as judged by a second print and by comparison of the size magnitude.
and shape of the droplets with the first (Fig. 2F). Most importantly,

Verardo et al., Sci. Adv. 9, eadi0263 (2023) 7 July 2023 4 of 12

S C I E N C E A D VA N C E S | R E S E A R C H A R T I C L E

Initial results for EDS ink stability were not promising. A precip- extended the lifetime of the ink further (Fig. 3D) and could help in
itate formed in less than 12 hours, and its pink color suggested the other ways. The operating temperature of most piezo-silicon MEMS
formation of a cobalt complex. Components were therefore printheads is between 5° and 40°C, so printing of cooled ink is fea-
removed one by one from the elongation ink (“Ei-1”) to understand sible if the substrate is warmed. The added benefit of cooling is that,
their contribution (Fig. 3A). This confirmed the involvement of at lower temperatures, ink viscosity will be higher, meaning a
cobalt and also the importance of the nucleotide. With an under- reduced need for viscosity modifier, with the former advantageous
standing of which components were most detrimental, we em- for jetting and the latter for ink activity (37). We also observed that
barked on a study of pH and cobalt concentration. The pH of the the presence of glycerol and dimethyl sulfoxide (DMSO; added to
ink was lowered to exclude the possible binding of the his-tag on the modify viscosity and hinder secondary structure formation, respec-
protein (TdT), which is most efficient at a slightly basic pH (~7.5 to tively) allowed the ink to be stored below −18°C in liquid form. This
8). The presence or absence of his-tag, however, had no bearing on raises the possibility of ink transport without lyophilization and sep-
ink stability, but the pH and concentration of cobalt had a pro- aration of components into hermetic compartments of an expensive
nounced effect (Fig. 3B). Lowering both markedly improved microfluidic cartridge.
stability. With the advantages of cooling established, we wanted to see
Having found a more physically stable formulation (“Ei-2”), tests whether it was possible to develop an ink that is room temperature
were done to determine if there was a concomitant improvement in stable. The starting point was to understand the behavior of the
activity. According to gel electrophoresis, activity was much im- most labile species, which was hypothesized to be the nucleotide.
proved and unchanged over 48 to 72 hours (Fig. 3C), a period suf- The terminal (γ) phosphate of a nucleotide is susceptible to hydro-
ficient for at least two or three long synthesis runs. Cooling the ink lysis, and at room temperature, there may be decomposition to the

Downloaded from [Link] on January 15, 2026

Fig. 3. Ink stability and activity. (A) Stability of initial elongation ink (Ei-1). Venn diagrams showing stable (green) formulations and unstable (red) formulations, giving
precipitate when aged 48 hours at 20°C. (B) Stability of elongation inks as a function of pH, [CoCl2], and dimethyl sulfoxide (DMSO). Green denotes stable ink; red denotes
inks that precipitate in ≤48 hours at 20°C. Ei-2 was selected as ink for this work on the basis of its stability and activity. It has 500 μM dTTP-30 -ONH2, 20 μM TdT, 0.25 mM
CoCl2, 15% (v/v) DMSO, 50 mM O-benzylhydroxylamine·HCl, 10% (v/v) glycerol, 0.05% (v/v) Tween 20, 2.5 mM tris-HCl, 40 mM NaCl, 0.5 mM Hepes, 0.5 M cacodylic acid,
and a final pH of 6.0 (measured in the absence of DMSO). (C) Ink activity of select inks. Polyacrylamide gel electrophoresis (PAGE)–urea gel of manually synthesized
sequence TTT-TTT-T. Synthesis was performed with Ei-1 and Ei-2 inks freshly prepared and aged 48 hours at 20°C. (D) Effect of ink cooling. PAGE-urea gel of manually
synthesized TTT-TTT-T with Ei-2 aged for 1 week at 20°, 4°, −20°, and −80°C. (E) Degradation of dTTP-30 -ONH2 according to ion exchange high-pressure liquid chroma-
tography with ultraviolet detection (IE-HPLC UV). Results for ink Ei-1 and Ei-2 as a function of the temperature (4° and 20°C) and aging time. Control ink does not contain
CoCl2 or TdT. (F) Effect of pyrophosphatase. IE-HPLC UV measurement of [dTTP-30 -ONH2] for Ei-1 and Ei-2 inks in the presence and absence of inorganic pyrophosphatase
(PPase). Control inks with and without PPase do not contain CoCl2 or TdT.

Verardo et al., Sci. Adv. 9, eadi0263 (2023) 7 July 2023 5 of 12

S C I E N C E A D VA N C E S | R E S E A R C H A R T I C L E

corresponding diphosphates and monophosphates at a rate of ~1% has only two reaction steps: nucleotide coupling under mild condi-
over 6 weeks (49, 50). High-performance liquid chromatography tions and 30 -ONH2 deprotection with a short exposure (<3 min) to
(HPLC) confirmed the benefit of lowering the pH, cobalt concen- nitrous acid. By comparison, DNA synthesis using phosphorami-
tration, and temperature (Fig. 3E) but still revealed a clear difference dites has five reaction steps: deblocking (detritylation), coupling,
with ink lacking cobalt or TdT (“Control’). The degradation was not oxidation, capping, and removal of base protecting groups; most
therefore simply due to hydrolysis. of which use an organic solvent.
During these studies, two grades of nucleotide became available, To demonstrate inkjet EDS and single-base control, a single-
and precipitation was observed to be faster for the lowest grade con- enzyme ink containing 30 -deoxythymidine 50 -triphosphate
taining more pyrophosphate (see fig. S10), which is known to form (dTTP)–30 -ONH2 was printed multiple times onto a glass slide
insoluble precipitates with cobalt (51). Inorganic pyrophosphatase with a uniform covering of DNA using the move-stop-print
was therefore added to the inks, which led to visual stability even for method. DNA was not printed because there were insufficient print-
ink Ei-1 (Fig. 3F). Unfortunately, it sped up the degradation of the heads to print both the DNA ink and the four enzyme-containing
nucleotide according to HPLC. To explain these results, it was hy- inks (A, G, C, and T) needed for a subsequent real-sequence dem-
pothesized that the pyrophosphatase was encouraging a forward re- onstration. In addition, enzyme adsorption was previously shown to
action involving the α-phosphate of the nucleotide and was be worse for an azide surface. A medium spot size (15 pulses) and
generating pyrophosphate as a by-product with the reaction medi- spot density (75 dpi; ~16,000 synthesis sites/slide) were chosen,
ated by TdT and cobalt. Liquid chromatography–mass spectrome- taking into consideration prior results and a desire to make moni-
try (LC-MS) was therefore conducted on the unmodified ink, with toring of the printing process and analysis of the results straightfor-
and without cobalt, to analyze the mass change of the nucleotide. A ward. First-cycle elongation was followed by capping of DNA
rapid degradation in the presence of cobalt was apparent but not in primer in nonprinted areas through incubation with an elongation

Downloaded from [Link] on January 15, 2026

the metal ion’s absence, and a new peak emerged with a mass cor- buffer containing TdT and dideoxyadenosine triphosphate
responding to a species at 410 mass/charge ratio that is probably an (ddATP). After this step, and after each subsequent round of print-
adduct of glycerol (see fig. S11). Tests using the purer nucleotide ing (60 pulses to allow for any imprecision and ensure coverage of
(performed with and without TdT), cobalt, and glycerol provided the sites of synthesis), the slide was washed with proteinase K, water,
additional support for the proposed nonhydrolysis pathway (see DB, and water again before drying. Figure 4C shows the change in
fig. S12). As noted by Schaudy et al. (5) and Baiga et al. (52), TdT the length of the DNA with each cycle of printing (coupling) and
can prime off non-nucleic acid primers such as hexaethylene glycol. washing (deprotection). The number of sections printed on was
Hence, to preserve activity and stability, the ink should lack endog- reduced by one each cycle to afford eight printed sections, where
enous pyrophosphate or species that could lead to its generation by the DNA length differs by one base. The DNA was subsequently
acting as a pseudo-DNA primer. If the ink does contain pyrophos- imaged on the slide (see fig. S13) and then released from the solid
phate or a potential precursor, then the ink should not contain in- support by photocleaving at λ365 nm for length separation by poly-
organic pyrophosphatase. Other solutions are to screen enzymes for acrylamide gel electrophoresis (PAGE)–urea gel electrophoresis.
activity against any non-nucleic acid primers and to split ink com- This yielded a ladder with single bands with one nucleotide differ-
ponents (e.g., TdT and cobalt) by compatibility and then mix them ence in length.
just before printing or print them separately on the slide surface. Given these promising results, three real sequences of 21 bases
with a similar CG content and melting temperature (see fig. S2)
Spatial EDS with single-base control were next synthesized in duplicate in eight sections on a DNA-
To carry out EDS and add dNTP-30 -ONH2 nucleotides one by one coated glass slide. One sequence was synthesized per section
to a DNA primer as shown in Fig. 4A, we used the deprotection except for sections 1 and 5, where the three sequences were synthe-
buffer (DB) of Hutter et al. (19), a pH-adjusted solution of sized side by side in rows for a first, very simple demonstration of
sodium nitrite and sodium acetate. The mechanism of 30 -ONH2 multiplexing. Competitive hybridization of complementary targets,
(aminoxy) deblocking has not been elucidated to our knowledge, each labeled with a different dye, was then used to reveal the loca-
but the reaction likely proceeds by nucleophilic attack by the 30 - tions of the sites of synthesis and the positions of the three sequenc-
aminoxy on the nitrosonium cation (53) followed by cyclic elimina- es (Fig. 4D). Excellent sequence discrimination, spot uniformity,
tion of nitrous oxide (N2O). The reaction speed and specificity are signal to noise, and signal homogeneity are evident (see fig. S14),
due to the enhanced nucleophilicity of the 30 -aminoxy group result- which are strong indicators of well-controlled single-base extension
ing from an alpha effect and its lower pKa (~6) compared to primary and reproducible printing. Sequence purity was confirmed by gel
amines. Unwanted side reactions are a possibility, but as in the case electrophoresis, with each lane showing a single intense band
of phosphoramidite synthesis, their incidence depends on the from which purity and cycle efficiency (~99%) were estimated
degree of optimization (29). Depurination can be suppressed by ad- (see fig. S15).
justment of pH and addition of salts and other additives (21), while
deamination resulting from nonspecific nitrosation of the base takes Multiplex EDS
place on a time scale significantly longer (4300 times) than the time Having used the move-stop-print method to demonstrate control
required to cleave 98% of the 30 -ONH2 groups (19). Furthermore, over spot size and pitch (density) and then shown microscale
both issues can be avoided entirely by switching to a non-nitrosa- EDS of real sequences, we sought to do a true demonstration of mul-
tion method such as reductive cleavage (54) or through base protec- tiplexing. Instead of printing the three sequences horizontally (in
tion, which is a standard practice in conventional DNA synthesis. the X direction) parallel to each other, they were juxtaposed in
We chose to exemplify high-multiplex EDS with single-base control both the X and Y directions (i.e., a 2D configuration) (Fig. 5A).
using the simplest synthesis cycle and choice of reagents. Our cycle They were also printed in the Z direction with one sequence on

Verardo et al., Sci. Adv. 9, eadi0263 (2023) 7 July 2023 6 of 12

S C I E N C E A D VA N C E S | R E S E A R C H A R T I C L E

Downloaded from [Link] on January 15, 2026

Fig. 4. Spatial EDS using inkjet dispensing. (A) EDS principle and deprotection mechanism. Steps: (i) TdT-mediated elongation of the 30 -OH-DNA primer (iDNA) with
the reversible terminator dNTP-30 -ONH2 to afford 30 -ONH2-DNAn+1 and (ii) nitrosonium-mediated deprotection to 30 -OH-DNAn+1. (B) Inkjet EDS method showing single-
base control. Elongation ink containing dTTP-30 -ONH2 (Ei-2) was printed onto a DNA-covered surface and left to incubate 10 min at 20°C; then, nonprinted areas were
capped by incubation for 5 min with a ddATP elongation ink. The slide was then washed consecutively for 1 min with a buffered solution of proteinase K, 1 min with water,
3 min with a DB containing 0.7 M NaOAc and 1.0 M NaNO2 at pH 5.2, and 1 min with water. All liquids were at room temperature and dispensed under IPA control as was
slide drying with compressed air. (C) PAGE-urea gel displaying single-base length control via inkjet EDS. The image shows the increase in length of a poly(T) sequence
synthesized via 1 to 8 cycles of inkjet EDS. The DNA was enzymatically end-labeled with ddATP-FAM for 10 min, photocleaved from the substrate in 0.1× phosphate-
buffered saline (PBS) at λ365 nm (15 min). (D) Sequence and array-wide spatial control. Epifluorescence microscope image of a whole slide after competitive hybridization
of three complementary fluorescent target oligonucleotides (t-e13, t-q4, and t-q41) to three EDS-printed probe sequences (e13, q4, and q41). (E) EDS synthesized probes
and their complementary targets. EDS starts with a DNA primer bound via its 50 end. The primer contains a UV-cleavable group for release postsynthesis and analysis via
gel electrophoresis. pDNA, printed DNA; tDNA, target DNA; iDNA, initiator(/primer) DNA. (F) Close-up of a section where the three sequences were multiplexed side by
side in the Y direction (perpendicular to the direction of slide movement) and the related fluorescence profiles (channel 1, enhanced GFP; channel 2, DsRed; channel
3, Cy5).

Verardo et al., Sci. Adv. 9, eadi0263 (2023) 7 July 2023 7 of 12

S C I E N C E A D VA N C E S | R E S E A R C H A R T I C L E

Downloaded from [Link] on January 15, 2026

Fig. 5. EDS multiplexing. (A) 2D multiplexing. The three sequences of Fig. 4 were arrayed in an XY pattern, requiring control of selected nozzles in the nozzle row (X
direction). Their locations were revealed by competitive hybridization. (B) 3D multiplexing. The following three 50-mer sequences [e13-poly(T)n+8-q4, q4-poly(T)n+8-q41,
and q41-poly(T)n+8-e13] were synthesized separately and in an XY pattern, and their locations were revealed by competitive hybridization with two target DNAs per
synthesized strand. (C) Determination of synthesis errors (deletions) by hybridization. Fluorescence intensity after hybridization to e13 probes deliberately synthesized
with double deletions. (D) Determination of synthesis errors (substitutions). Fluorescence intensity after hybridization to e13 probes deliberately synthesized with sub-
stitutions. (E) The EDS advantage in spatial transcriptomics. EDS proceeds in the 50 to 30 direction, allowing poly(T) tails to capture mRNA and be reverse transcribed
unhindered. (F) Capture of mRNA and reverse transcription (RT) to cDNA. Poly(T) surface probes were synthesized with different lengths and then incubated with a short
mRNA of sequence 50 -UACACGUUGUCUAUCGCCUU(30A)-30 (see fig. S2). After reverse transcription and dehybridization, the resulting cDNA was end-labeled and cleaved
from the solid support, and the increase in length was confirmed by gel electrophoresis..

top of another to demonstrate that lengths above a typical PCR significant impact on intensity (Fig. 5D), especially when situated
primer (18 to 30 bases) are attainable. Multiplexing was limited to near the interior of the probe (see fig. S19).
just three unique sequences because of the filter set of the micro- With DNA purity confirmed by hybridization and by gel electro-
scope and the desire to keep the analysis as quick and as direct as phoresis, we set out to highlight one of the potential advantages of
possible without PCR. Figure 5B shows the result of competitive hy- spatial EDS. A DNA primer was uniformly immobilized to a sub-
bridization and confirms that the three sequences had the correct strate via its 50 end and on its 30 end was inkjet synthesized a poly-
XYZ coordinates (see also fig. S16). Gel electrophoresis (see fig. thymidine [poly(T)] tail. As EDS proceeds in the 50 to 30 direction,
S17) corroborates the difference in length of the XY and XYZ the synthesized DNA, which could include a barcode, unique mo-
probe sequences of fig. S2. lecular identifier, and handle for amplification, was able to freely
To see whether hybridization could be used to identify major capture the poly(A) (polyadenylate) tail of a short mRNA molecule
and minor synthesis errors caused by printing imprecision, for (Fig. 5E). Reverse transcription was then performed using the
example, a droplet not being printed in an assigned location at a poly(T) tail as a primer to generate cDNA, which was end-
given cycle, the three probes were deliberately synthesized with mis- labeled, cleaved, and analyzed by gel electrophoresis. Figure 5F
matches and deletions and hybridized. Figure 5C shows the example shows the corresponding increase in DNA length (+20 bases),
of double deletions and compares signal with theoretical calcula- which can only happen if there has been mRNA capture (see fig.
tions of the hybridization free energy (ΔG) according to IDT’s Oli- S20). By synthesizing in the 50 to 30 direction, spatial transcriptom-
goAnalyzer (“Hetero-dimer analysis”). As anticipated (29), the ics, which seeks to relate protein expression to 2D location in histo-
hybridization signal varied inversely with the number of deletions logical sections, is made easier. In the method of Stahl et al. (55), the
on the 50 -terminus while translation of deletions to the middle of transcriptome of only 1007 locations (occupying an area of ~41
the probe exerted a greater effect than deletions at the extremities mm2) was elucidated because the barcoded mRNA capture probes
because of the reduced degree of complementarity (see fig. S18). first had to be synthesized individually in the 30 to 50 direction using
Substitutions, which are less likely to be caused by printing errors the phosphoramidite method by an external supplier. Then, they
but might occur because of poor slide washing, also had a had to be arrayed on a surface in the opposite direction before
tissue application. A much higher barcoding density (~106 K/

Verardo et al., Sci. Adv. 9, eadi0263 (2023) 7 July 2023 8 of 12

S C I E N C E A D VA N C E S | R E S E A R C H A R T I C L E

mm2) for subcell resolution was achieved using a bead array (56), MATERIALS AND METHODS
but the oligos required had to be synthesized first on a phosphora- Immobilization of DNA initiators on glass substrates and
midite synthesizer, then immobilized to beads then extended via DNA cleavage
split-and-pool ligation, and then decoded through several rounds The 2D azide glass slides (PolyAn, Berlin, Germany) were mounted
of hybridization and fluorescence imaging after transfer of the bar- in a ProPlate one-well slide gasket from Grace Bio-Labs Inc. (Bend,
coded beads into the wells of a microfabricated substrate (57). OR, USA) and incubated with 5 μM 50 DBCO-triethyleneglycol–
Spatial EDS is potentially a much simpler way to diverse, and modified oligonucleotide (Eurogentec, Belgium) in 0.1 M NaOAc
dense arrays of ordered barcoded probes with their 30 poly(T) and 0.5 M NaCl buffer (pH 4.5) at 20°C in a chamber at 70% relative
ends correctly orientated unhindered away from the surface for humidity (RH). After 1 hour, the coupling solution was removed
the capture and transcription of mRNA. Access to a printer in re- and the slide was rinsed with 1500 μl of 0.1 M NaOAc and 0.5 M
search laboratories would also allow researchers to modify the re- NaCl buffer (pH 4.5), 5 × 1500 μl of phosphate-buffered saline
sulting cDNA, for example, by adding a common 30 primer for (PBS), 50 ml of DB (1 hour), and 50 ml of Milli-Q (5 min), followed
sequencing. by immediate drying with compressed air. For DNA printing, the
DNA primer ink (pH 4.5; measured without DMSO) was formulat-
ed in Milli-Q:DMSO [65:35 (v/v)] with 5 μM initiator/primer DNA
DISCUSSION (iDNA), 325 mM NaCl, and 65 mM NaOAc.
We report an advance in de novo DNA synthesis: spatial microscale
EDS with single-base control. A printer equipped with four silicon Move-stop printing and automation
MEMS piezo inkjet printheads has been shown to be able to deposit The 1D barcode images (see fig. S4) were generated using an in-
picoliter volumes of ink necessary for EDS with micron precision in house software called the Inkjet Printer App and printed according

Downloaded from [Link] on January 15, 2026

preselected locations on a glass slide in which a DNA primer has to the cycle and position on the array. The positions of the bars in
been immobilized. Within each ink droplet on the surface and in the image determine the nozzles actuated in the nozzle row, and the
every synthesis cycle, a template-free polymerase (TdT) adds a lengths of the bars determine how many actuations (pulses) each
single nucleotide to the 30 end of the DNA primer, with the risk nozzle makes. A 5-V trigger signal was used to initiate printing
of multiple additions precluded by a 30 blocking group that is after each move of the translation stage.
removed by washing of the slide. If required, the DNA can be
cleaved enzymatically at the 50 end postsynthesis to give the Ink studies
desired DNA without primer (58). We have shown that hybridiza- The ink [pH 6.6 (Ei-1) or pH 6.0 (Ei-2)] contained 10% (v/v) glyc-
tion and gel electrophoresis can be used to confirm length and erol, 0.025% (v/v) Tween 20, 2.5 mM tri-·HCl, 40 mM NaCl, 0.5
gauge sequence accuracy for an EDS synthesized microarray. Key mM Hepes, 0.5 M cacodylic acid, 50 mM O-benzohydroxylamine
issues such as ink stability have been identified and addressed to hydrochloride, 0.25 (Ei-2) or 1 (Ei-1) mM CoCl2, 500 μM 30 -
ensure reproducible results and ease of use. Similar to the benchtop ONH2-dNTP (N = A, C, G, T), 20 μM TdT, and 15% (v/v)
PoSAM platform (30), our printer is simple and inexpensive, being DMSO. Inks for stability tests were stored in plastic Eppendorf
assembled from off-the-shelf components. Unlike PoSAM, our tubes in the dark at either −80°, −20°, 4°, or 20°C. Visual stability
printer permits control over spot size (mass) and offers far more was evaluated by taking pictures with a white light screen as back-
than 9600 sites of synthesis, while it also eliminates the requirement ground. Activity tests were performed via manual EDS. Ink degra-
of existing inkjet printers for glove box conditions. The cost of a dation was evaluated via ion exchange HPLC: Agilent 1260 Infinity
phosphoramidite-synthesized oligo pool is estimated to be $1 per II [Thermo Fisher Scientific DNAPac PA200 column (4·250 mm);
103 to 105 base pairs (59). As we have shown, an EDS reaction solvent A, 25 mM tris (pH 8.0); solvent B, 1 M LiCl in Milli-Q; flow
cycle needs only micromolar quantities of reagent and can have rate: 1 ml/min, column temperature: 25°C] equipped with an ultra-
fewer steps, which means fewer washes with water, which violet (UV)–visible detector. Before injection, inks were passed
happens to be significantly less expensive than acetonitrile. Al- through an Amicon membrane filter (3 kDa, 14,000 rcf, 20 min).
though an inkjet DNA printer should consume very little reagent Nucleotide quantification was carried out using the automated
once running and should generate relatively little waste, the peak integration of the software. The percentual decrease is based
storage and disposal of aqueous reagents is less onerous than on absolute peak integration (mAU·s), setting the peak area of the
organic reagents. Thus, we anticipate that the cost of an EDS nucleotide measured at 0 hours as 100%. The mass of the degrada-
oligo pool will be inferior. To achieve the longest lengths, highest tion products was confirmed via HPLC-MS (column HSST2 2.1
density, and lowest error rates, further optimization of machine pa- mm by 150 mm by 1.8 μm) (Waters, MA, USA). Elution was per-
rameters and EDS reagents and protocols will be necessary. Looking formed with the following: solvent A, water solution containing 120
forward, spatial EDS could, one day, become routine and common- mM hexafluoroisopropanol (HFIP) and 5 mM triethylamine (pH 7;
place in research laboratories and accelerate biological research by adjusted with formic acid), and solvent B, methanol solution con-
enabling researchers to iterate quickly with much greater flexibility taining 120 mM HFIP and 5 mM triethylamine at 40°C. Optical de-
than at present. tection was carried out with a photodiode array; mass detection was
performed via electrospray ionization (capillary voltage of 2.2 kV,
cone voltage of 45 V, desolvation temperature of 400°C, desolvation
flow of 500 liters/hour, and cone flow 30 liters/hour).

Verardo et al., Sci. Adv. 9, eadi0263 (2023) 7 July 2023 9 of 12

S C I E N C E A D VA N C E S | R E S E A R C H A R T I C L E

Enzymatic DNA synthesis 5 min. cDNA products were end-labeled and cleaved as previously
Reagents described.
Elongation steps in the synthesis cycle used ink Ei-2 (see the “Ink Analysis
studies” section). Deprotection steps used the buffer (DB) of Hutter The arrays were imaged with a Zeiss Axiolmager Z2 epifluorescence
et al. (19), i.e., 0.7 M NaOAc and 1.0 M NaNO2 (pH 5.2). The pro- microscope (×5 magnification, channels, GFP λex 470, DsRed λex
teinase K wash buffer was proteinase K (0.05 mg ml−1) in a 10 mM 555, and Cy5 λex 625) equipped with an automatic stage. Light-
tris (pH 8.0) solution containing 0.25 mM EDTA and 100 mM emitting diode intensities and exposure times were adjusted to
NaCl. End-labeling of DNA and capping of primer was performed the fluorophore used and the nature of the experiment.
with a pH 6.6 solution containing 0.5 M cacodylic acid, 1 mM Image analysis was performed with ImageJ (bio-format plugin)
CoCl2, 20 μM TdT, 500 μM ddATP (for capping), or 50 μM after defining boundary conditions. Diameters in Fig. 2 (C and F)
ddATP-FAM or ddATP-Cy5 (for end-labeling). All solutions were were calculated by including all spots of the section (i.e., ≈1200).
prepared with Milli-Q water. If not otherwise specified, then re- The values are the mean diameters from analyzing two slides, i.e.,
agents were purchased from Sigma-Aldrich, Jena Bioscience, AK ≈2400 spots. The error bars are ±SD. The diameters reported in
Scientific, Fisher Chemicals, and New England Biolabs and used Fig. 5 (C and D) are the mean values ± SD from the measurement
without further purification. TdT and dNTP-30 -ONH2 were pro- of 1200 spots in each section. Measurement was performed on a
duced internally by DNA Script. single slide.
Protocols Cleaved oligos were mixed (1:1) with Brilliant Blue and gel
Inkjet EDS was performed without any optimization of cycle times loading buffer and loaded onto a polyacrylamide gel [30 ml of poly-
or processes. The cycle consisted of move-stop printing (see above), acrylamide, 80 μl of ammonium persulfate, and 28 μl of TEMED
incubation (elongation) at 20°C for 10 min, and then slide washing (N,N,N0 ,N0 -tetramethylethylenediamine)] and electrophoresed at

Downloaded from [Link] on January 15, 2026

(1 min with proteinase K, 1 min with water, 3 min with DB, and 1 1500 V (EV3330 Consort). The gel was then imaged with the Amer-
min with water). sham Typhoon Biomolecular Imager, and the images were analyzed
End-labeling (10 min at 20°C) was performed after synthesis with Bio-Rad Image Lab software with gamma, high, and low pa-
with 1500 μl of end-labeling solution while incubating the slide at rameters set to 1.8, 800, and 150, respectively. The same Typhoon
70% RH. Slides were washed with 1500 μl of DB (2 × 2 min), 1500 μl instrument was used to image the whole slide in fig. S3.
of Milli-Q water (3 × 1 min), followed by drying with com-
pressed air.
Photocleaving was done by applying an eight-well gasket and Supplementary Materials
adding 150 μl of 0.1× PBS to each well. The substrate was then sub- This PDF file includes:
Figs. S1 to S20
jected to UV illumination at λ365 nm (Analytik Jena; 8 W) for 15
min at 20°C. The cleaved oligonucleotides were recovered and then
View/request a protocol for this paper from Bio-protocol.
vacuum concentrated (SpeedVac, Eppendorf, Hamburg, Germany)
for 45 min at 45°C.
Competitive hybridization used three labeled targets (t-e13-488,
REFERENCES AND NOTES
t-q4-555, and t-41-647) purchased from Eurogentec (Seraing, 1. M. A. Jensen, R. W. Davis, Template-independent enzymatic oligonucleotide synthesis
Belgium) and diluted to a final concentration of 1 μM each in 5× (TiEOS): Its history, prospects, and challenges. Biochemistry 57, 1821–1832 (2018).
saline sodium citrate buffer (SSC) containing 0.1% Tween 20. A 2. K. Hoff, M. Halpain, G. Garbagnati, J. S. Edwards, W. Zhou, Enzymatic synthesis of designer
one-well gasket was applied to the slide and the surface incubated DNA using cyclic reversible termination and a universal template. ACS Synth. Biol. 9,
283–293 (2020).
with 1300 μl of hybridization solution for 15 min at 60°C followed
3. L. Beaucage, M. H. Caruthers, Deoxynucleoside phosphoramidites—A new class of key
by 2 hours at 20°C (300 rpm) in a humid chamber at 70% RH. This intermediates for deoxypolynucleotide synthesis. Tetrahedron Lett. 22, 1859–1862 (1981).
was followed by slide washing at 20°C with 1300 μl of 5× SSC, 1300 4. E. A. Moteaa, A. J. Berdis, Terminal deoxynucleotidyl transferase: The story of a misguided
μl of 4× SSC, 1300 μl of 2.5× SSC, 1300 μl of 1× SSC, and 5 × 1300 μl DNA polymerase. Biochim. Biophys. Acta 1804, 1151–1166 (2010).
in Milli-Q water and then immediate drying under compressed air. 5. E. Schaudy, J. Lietard, M. M. Somoza, Sequence preference and initiator promiscuity for de
Synthesized 50-mer XYZ probes were hybridized in a stepwise novo DNA synthesis by terminal deoxynucleotidyl transferase. ACS Synth. Biol. 10,
1750–1760 (2021).
manner because of the occurrence of t-e13-488 and t-q4-555
6. S. Barthel, S. Palluk, N. J. Hillson, J. D. Keasling, D. H. Arlow, Enhancing terminal deoxy-
quenching (through fluorescence resonance energy transfer) by t- nucleotidyl transferase activity on substrates with 30 terminal structures for enzymatic de
q41-647 (see fig. S16). novo DNA synthesis. Genes 11, 102 (2020).
Reverse transcription was performed by adapting the protocol 7. D. Tauraitė, J. Jakubovska, J. Dabužinskaitė, M. Bratchikov, R. Meškys, Modified nucleotides
from New England Biolabs. Hybridization of mRNA onto synthe- as substrates of terminal deoxynucleotidyl transferase. Molecules 22, 672 (2017).
sized poly(T) oligomers (see fig. S2) in ribonuclease-free water 8. S. Deshpandea, Y. Yangb, A. Chilkotia, S. Zauschera, Enzymatic synthesis and modification
of high molecular weight DNA using terminal deoxynucleotidyl transferase. Methods
([mRNA] = 0.6 μM, dNTP = 1 mM, 65°C for 5 min followed by
Enzymol. 627, 163–188 (2019).
rapid cooling for 3 min at 0°C) was followed by reverse transcription 9. J. P. S. Chua, M. K. Go, T. Osothprarop, S. Mcdonald, A. G. Karabadzhak, W. S. Yew,
(1 hour, 42°C) in 1× ProtoScript II buffer with ProtoScript II reverse S. Peisajovich, S. Nirantar, Evolving a thermostable terminal deoxynucleotidyl transferase.
transcription (10 U/μl) and 10 mM dithiothreitol. Deactivation of ACS Synth. Biol. 9, 1725–1735 (2020).
the enzyme was carried out following the proteinase K treatment 10. L.-F. Song, Z.-H. Deng, Z.-Y. Gong, L.-L. Li, B.-Z. Li, Large-scale de novo oligonucleotide
described above. Dehybridization of mRNA was done by consecu- synthesis for whole-genome synthesis and data storage: Challenges and opportunities.
Front. Bioeng. Biotechnol. 22, 689797 (2021).
tive washes at 60°C with 3 M NaCl for 30 min, 5× SSC for 5 min, 4×
SSC for 1 min, 2.5× SSC for 1 min, 1× SSC for 1 min, and Milli-Q for

Verardo et al., Sci. Adv. 9, eadi0263 (2023) 7 July 2023 10 of 12

S C I E N C E A D VA N C E S | R E S E A R C H A R T I C L E

11. V. Tjong, H. Yu, A. Hucknall, S. Rangarajan, A. Chilkoti, Amplified on-chip fluorescence 36. S. Krainer, C. Smit, U. Hirn, The effect of viscosity and surface tension on inkjet printed
detection of DNA hybridization by surface-initiated enzymatic polymerization. Anal. Chem. picoliter dots. RSC Adv. 9, 31708–31719 (2019).
83, 5153–5159 (2011). 37. S. Uribe, J. G. Sampedro, Measuring solution viscosity and its effect on enzyme activity.
12. L. Tang, L. A. Navarro Jr., A. Chilkoti, S. Zauscher, High-molecular-weight polynucleotides Biol. Proced. Online 5, 108–115 (2003).
by transferase-catalyzed living chain-growth polycondensation. Angew. Chem. Int. Ed. Engl. 38. C. C. Cook, T. Wang, B. Derby, Inkjet delivery of glucose oxidase. Chem. Commun. 46,
56, 6778–6782 (2017). 5452–5454 (2010).
13. S. Palluk, D. H. Arlow, T. de Rond, S. Barthel, J. S. Kang, R. Bector, H. M. Baghdassarian, 39. N. Komuro, S. Takaki, K. Suzuki, D. Citterio, Inkjet printed (bio)chemical sensing devices.
A. N. Truong, P. W. Kim, A. K. Singh, N. J. Hillson, J. D. Keasling, De novo DNA synthesis using Anal. Bioanal. Chem. 405, 5785–5805 (2013).
polymerase-nucleotide conjugates. Nat. Biotechnol. 36, 645–650 (2018). 40. E. Gdor, S. Shemesh, S. Magdassi, D. Mandler, Multienzyme inkjet printed 2D arrays. ACS
14. J. Lackey, D. Dodd, Methods of synthesizing oligonucleotides using tethered nucleotides. Appl. Mater. Interfaces 7, 17985–17992 (2015).
WO2022086866A1. 41. G. M. Nishioka, A. A. Markey, C. K. Holloway, Protein damage in drop-on-demand printers.
15. A. S. Mathews, H. Yang, C. Montemagno, Photo-cleavable nucleotides for primer free J. Am. Chem. Soc. 126, 16320–16321 (2004).
enzyme mediated DNA synthesis. Org. Biomol. Chem. 14, 8278–8288 (2016). 42. J. T. Delaney Jr., P. J. Smith, U. S. Schubert, Inkjet printing of proteins. Soft Matter 5,
16. F. Chen, E. A. Gaucher, N. A. Leal, D. Hutter, S. A. Havemann, S. Govindarajan, E. A. Ortlund, 4866–4877 (2009).
S. A. Benner, Reconstructed evolutionary adaptive paths give polymerases accepting re- 43. Y. Bai, D. Zhang, Q. Guo, J. Xiao, M. Zheng, J. Yang, Study of the enzyme activity change due
versible terminators for sequencing and SNP detection. Proc. Natl. Acad. Sci. U.S.A. 107, to inkjet printing for biosensor fabrication. ACS Biomater Sci. Eng. 7, 787–793 (2021).
1948–1953 (2010). 44. S. Di Risio, N. Yan, Piezoelectric ink-jet printing of horseradish peroxidase: Effect of ink
17. H. Saneyoshi, K. Nakamura, K. Terasawa, A. Ono, Development of bioreduction labile viscosity modifiers on activity. Macromol. Rapid Commun. 28, 1934–1940 (2007).
protecting groups for the 20 -hydroxyl group of RNA. Org. Lett. 22, 6006–6009 (2020). 45. T. T. Biswas, J. Yu, V. A. Nierstrasz, Effects of ink characteristics and piezo-electric inkjetting
18. M. Flamme, S. Hanlon, I. Marzuoli, K. Püntener, F. Sladojevich, M. Hollenstein, Evaluation of parameters on lysozyme activity. Sci. Rep. 9, 18252 (2019).
30 -phosphate as a transient protecting group for controlled enzymatic synthesis of DNA 46. N. Z. Fantoni, A. H. El-Sagheer, T. Brown, A hitchhiker’s guide to click-chemistry with
and XNA oligonucleotides. Commun. Chem. 5, 68 (2022). nucleic acids. Chem. Rev. 121, 7122–7154 (2021).
19. D. Hutter, M.-J. Kim, N. Karalkar, N. A. Leal, F. Chen, E. Guggenheim, V. Visalakshi, J. Olejnik, 47. M. Shelbourne, X. Chen, T. Brown, A. H. El-Sagheer, Fast copper-free click DNA ligation by
S. Gordon, S. A. Benner, Labeled nucleoside triphosphates with reversibly terminating

Downloaded from [Link] on January 15, 2026

the ring-strain promoted alkyne-azide cycloaddition reaction. Chem. Commun. 47,
aminoalkoxyl groups. Nucleos. Nucleot. Nucl. 29, 879–895 (2010). 6257–6259 (2011).
20. K. Haghbeen, S. Rezaeian, M. Bakavoli, A synthetic approach for evaluation of cytosine- 48. S. Kim, M. Cho, S. Jung, The design of an inkjet drive waveform using machine learning. Sci.
diazonium susceptibility. J. Appl. Chem. Res. 11, 84–96 (2017). Rep. 12, 4841 (2022).
21. R. An, Y. Jia, B. Wan, Y. Zhang, P. Dong, J. Li, X. Liang, Non-enzymatic depurination of 49. M. G. Mohsen, D. Ji, E. T. Kool, Polymerase synthesis of four-base DNA from two stable
nucleic acids: Factors and mechanisms. PLOS ONE 9, e115950 (2014). dimeric nucleotides. Nucleic Acids Res. 47, 9495–9501 (2019).
22. X. Lu, J. Li, C. Li, Q. Lou, K. Peng, B. Cai, Y. Liu, Y. Yao, L. Lu, Z. Tian, H. Ma, W. Wang, J. Cheng, 50. H.-G. Ihlenfeldt, A. Schmidt, K. Muhlegger, V. Leitenberger, Stabilized aqueous nucleotide
X. Guo, H. Jiang, Y. Ma, Enzymatic DNA synthesis by engineering terminal deoxynucleo- triphosphate solution, Patent U.S. 2002/0004230 A1 (2001).
tidyl transferase. ACS Catal. 12, 2988–2997 (2022).
51. L. B. Rogers, C. A. Reynolds, Interaction of pyrophosphate ion with certain multivalent
23. M. A. Jensen, P. B. Griffin, R. W. Davis, Free-running enzymatic oligonucleotide synthesis for cations in aqueous solutions. J. Am. Chem. Soc. 71, 2081–2085 (1949).
data storage applications. bioRxiv 355719 [Preprint]. 30 June 2018. [Link]
52. T. Baiga, M. A. Burley, A. Smith, Modified DNA polymerases. US2019 0078065 and
1101/355719.
US2019 0078126.
24. N. Bhan, A. Callisto, J. Strutz, J. Glaser, R. Kalhor, E. S. Boyden, G. Church, K. Kording,
53. D. L. H. Williams, Nitrosation Reactions and the Chemistry of Nitric Oxide (Elsevier, 2004).
K. E. J. Tyo, Recording temporal signals with minutes resolution using enzymatic DNA
54. S. Park, N. P. Westcott, W. Luo, D. Dutta, M. N. Yousaf, General chemoselective and redox-
synthesis. J. Am. Chem. Soc. 143, 16630–16640 (2021).
responsive ligation and release strategy. Bioconjug. Chem. 25, 543–551 (2004).
25. H. H. Lee, R. Kalhor, N. Goela, J. Bolot, G. M. Church, Terminator-free template-independent
55. P. L. Stahl, F. Salmen, S. Vickovic, A. Lundmark, J. F. Navarro, J. Magnusson, S. Giacomello,
enzymatic DNA synthesis for digital information storage. Nat. Comm. 10, 2383 (2019).
M. Asp, J. O. Westholm, M. Huss, A. Mollbrink, S. Linnarsson, S. Codeluppi, A. Borg, F. Ponten,
26. H. Lee, D. J. Wiegand, K. Griswold, S. Punthambaker, H. Chun, R. E. Kohman, G. M. Church,
P. I. Costea, P. Sahlen, J. Mulder, O. Bergmann, J. Lundeberg, J. Frisen, Visualization and
Photon-directed multiplexed enzymatic DNA synthesis for molecular digital data storage.
analysis of gene expression in tissue sections by spatial transcriptomics. Science 353,
Nat. Comm. 11, 5246 (2020).
78–82 (2016).
27. B. P. Kuiper, R. C. Prins, S. Billerbeck, Oligo pools as an affordable source of synthetic DNA
56. S. Vickovic, G. Eraslan, F. Salmen, J. Klughammer, L. Stenbeck, D. Schapiro, T. Aijo,
for cost-effective library construction in protein- and metabolic pathway engineering.
R. Bonneau, L. Bergenstrahle, J. Fernandez Navarro, J. Gould, G. K. Griffin, A. Borg,
Chembiochem 23, e202100507 (2022).
M. Ronaghi, J. Frisen, J. Lundeberg, A. Regev, P. L. Stahl, High-definition spatial tran-
28. T. R. Hughes, M. Mao, A. R. Jones, J. Burchard, M. J. Marton, K. W. Shannon, S. M. Lefkowitz, scriptomics for in situ tissue profiling. Nat. Methods 16, 987–990 (2019).
M. Ziman, J. M. Schelter, M. R. Meyer, S. Kobayashi, C. Davis, H. Dai, Y. D. He,
57. K. L. Gunderson, S. Kruglyak, M. S. Graige, F. Garcia, B. G. Kermani, C. Zhao, D. Che,
S. B. Stephaniants, G. Cavet, W. L. Walker, A. West, E. Coffey, D. D. Shoemaker, R. Stoughton,
T. Dickinson, E. Wickham, J. Bierle, D. Doucet, M. Milewski, R. Yang, C. Siegmund, J. Haas,
A. P. Blanchard, S. H. Friend, P. S. Linsley, Expression profiling using microarrays fabricated
L. Zhou, A. Oliphant, J.-B. Fan, S. Barnard, M. S. Chee, Decoding randomly ordered DNA
by an ink-jet oligonucleotide synthesizer. Nat. Biotechnol. 19, 342–347 (2001).
arrays. Genome Res. 14, 870–877 (2004).
29. E. M. LeProust, B. J. Peck, K. Spirin, H. B. McCuen, B. Moore, E. Namsaraev, M. H. Caruthers,
58. S. Creton, Efficient product cleavage in template-free enzymatic synthesis of polynucleo-
Synthesis of high-quality libraries of long (150mer) oligonucleotides by a novel depuri-
tides, Patent U.S. WO2020165137A1 (2020).
nation controlled process. Nucleic Acids Res. 38, 2522–2540 (2010).
59. S. Kosuri, G. M. Church, Large-scale de novo DNA synthesis: Technologies and applications.
30. C. Lausted, T. Dahl, C. Warren, K. King, K. Smith, M. Johnson, R. Saleem, J. Aitchison, L. Hood,
Nat. Methods 11, 499–507 (2014).
S. R. Lasky, POSaM: A fast, flexible, open-source, inkjet oligonucleotide synthesizer and
microarrayer. Genome Biol. 5, R58 (2004).
Acknowledgments: We would like to thank the following people at DNA Script: M. Artigue for
31. I. Saaem, K.-S. Ma, A. N. Marchi, T. H. LaBean, J. Tian, In situ synthesis of DNA microarray on
help with the manufacture of the microscope slide holder and washing station; E. Atse for gel
functionalized cyclic olefin copolymer substrate. ACS Appl. Mat. Inter. 2, 491–497 (2010).
electrophoresis; K. Khodaverdi and E. Champion for useful discussions regarding the stability
32. J. Quan, I. Saaem, N. Tang, S. Ma, N. Negre, H. Gong, K. P. White, J. Tian, Parallel on-chip and activity of the enzyme; J. Loc’h and N. Tavernier for supply of enzyme without glycerol;
gene synthesis and application to optimization of protein expression. Nat. Biotechnol. 29, S. Macewicz and J. Jedeloh for useful discussions regarding formats and applications; W. Wu
449–452 (2011). and A. Blond for supply and purification of nucleotides; I. Aparin for HPLC support; and E. Dizin,
33. H. Li, Y. Huang, Z. Wei, W. Wang, Z. Yang, Z. Liang, Z. Li, An oligonucleotide synthesizer D. Canet, and T. Ybert for critical reading of the manuscript. We also thank M. Smith, D. Morris,
based on a microreactor chip and an inkjet printer. Sci. Rep. 9, 5058 (2019). and S. Morris at NIH/NHGRI for support. Funding: This research is based on work supported in
34. J. S. Hargreaves, R. Kaiser, P. K. Wolber, The degradation of dG phosphoramidites in sol- part by the National Human Genome Research Institute (NHGRI) of the National Institutes of
ution. Nucleos. Nucleot. Nucl. 34, 691–707 (2015). Health (NIH) under award number R01HG011736. The views and conclusions contained here
35. S. Laulhe, “Aminooxy reagents for synthesis and analysis: Expanding the role of oximation,” are those of the authors and should not be interpreted as necessarily representing, either
thesis, University of Louisville (2013). expressed or implied, those of NIH. Author contributions: A.H. and X.G. conceived the project.
A.H., D.V., and B.A. designed the experiments. E.T. prepared the glass substrates for synthesis,

Verardo et al., Sci. Adv. 9, eadi0263 (2023) 7 July 2023 11 of 12

S C I E N C E A D VA N C E S | R E S E A R C H A R T I C L E

hybridized and imaged the slides after synthesis, and cleaved the DNA for gel analysis. N.M., T.L., patent applications relevant to the work in this publication have been filed, e.g., EP21737100.4.
and B.A. formulated and prepared the inks used for synthesis. A.H., D.V., and D.A.R.-P. developed Data and materials availability: All data needed to evaluate the conclusions in the paper are
the instrumentation. D.A.R.-P. and D.V. developed the control software. E.T., T.L., and B.A. present in the paper and/or the Supplementary Materials.
analyzed the data. A.H., D.V., and B.A. developed the EDS protocol. A.H., D.V., and B.A.
supervised the project. D.V. and B.A. prepared the figures. A.H. wrote the manuscript, and all Submitted 30 March 2023
authors read and discussed it. Competing interests: The authors are employed by DNA Script. Accepted 5 June 2023
T.L. was employed at DNA Script but is now a Ph.D. student at the Eindhoven University of Published 7 July 2023
Technology. All other authors declare that they have no other competing interests. One or more 10.1126/sciadv.adi0263

Downloaded from [Link] on January 15, 2026

Verardo et al., Sci. Adv. 9, eadi0263 (2023) 7 July 2023 12 of 12