Precise Alkoxyamine Design to Enable Automated Tandem Mass Spectrometry Sequencing of Digital Poly(phosphodiester)s

Article abstract of DOI:10.1002/anie.202010171

A major step towards reliable reading of information coded in the sequence of long poly(phosphodiester)s was previously achieved by introducing an alkoxyamine spacer between information sub-segments. However, MS/MS decoding had to be performed manually to

Full text of DOI:10.1002/anie.202010171

A Journal of the Gesellschaft Deutscher Chemiker
Angewandte
International Edition Chemie
www.angewandte.org
Accepted Article
Title: Precise alkoxyamine-design enables automated tandem mass
spectrometry sequencing of digital poly(phosphodiester)s
Authors: Kévin Launay, Jean-Arthur Amalian, Eline Laurent, Laurence
Oswald, Abdelaziz Al Ouahabi, Alexandre Burel, Florent
Dufour, Christine Carapito, Jean-Louis Clément, Jean-
François Lutz, Laurence Charles, and Didier Gigmes
This manuscript has been accepted after peer review and appears as an
Accepted Article online prior to editing, proofing, and formal publication
of the final Version of Record (VoR). This work is currently citable by
using the Digital Object Identifier (DOI) given below. The VoR will be
published online in Early View as soon as possible and may be different
to this Accepted Article as a result of editing. Readers should obtain
the VoR from the journal website shown below when it is published
to ensure accuracy of information. The authors are responsible for the
content of this Accepted Article.
To be cited as: Angew. Chem. Int. Ed. 10.1002/anie.202010171
Link to VoR: https://doi.org/10.1002/anie.202010171

Angewandte Chemie International Edition
10.1002/anie.202010171
Precise alkoxyamine-design enables automated tandem mass spectrometry sequencing
of digital poly(phosphodiester)s
Kévin Launay, Jean-Arthur Amalian, Eline Laurent, Laurence Oswald,[b] Abdelaziz Al
[
a]
[a]
[b]
Ouahabi,[b] Alexandre Burel,[c] Florent Dufour,[b,c] Christine Carapito,[c] Jean-Louis
Clément, Jean-François Lutz, Laurence Charles,[a]* Didier Gigmes[a]*
[
a]
[b]*
[
a]
Aix Marseille Université, CNRS, Institute for Radical Chemistry, UMR 7273, 23 Av
Escadrille Nomandie-Niemen, 13397 Marseille Cedex 20, France
[
b]
Université de Strasbourg, CNRS, Institut Charles Sadron UPR22, 23 rue du Loess, 67034
Strasbourg Cedex 2, France
[
c]
Université de Strasbourg, CNRS, Laboratoire de Spectrométrie de Masse BioOrganique
LSMBO), IPHC, 23 rue du Loess, 67034 Strasbourg Cedex 2, France
(
Abstract
A major step towards reliable reading of information coded in the sequence of long
poly(phosphodiester)s was previously achieved by introducing an alkoxyamine spacer
between information sub-segments. However, MS/MS decoding had to be performed
manually to safely identify useful fragments of low abundance compared to side-products
from the amide-based alkoxyamine used. Here, alternative alkoxyamines were designed to
prevent side-reactions and enable automated MS/MS sequencing. Different styryl-TEMPO
spacers were prepared to increase radical delocalization and stiffness of the structure. Their
dissociation behavior was investigated by EPR and best results were obtained with spacers
containing in-chain benzyl ring, with no side-reaction during synthesis or sequencing.
Automated decoding of these polymers was performed using the MS-DECODER software,
which interprets fragmentation data recorded for each sub-segment and re-align them in their
original order based on location tags.
Key words: alkoxyamine synthesis; digital information; mass spectrometry; radical ions;
sequence-defined synthetic polymers; structural design
1
This article is protected by copyright. All rights reserved.

Angewandte Chemie International Edition
10.1002/anie.202010171
Introduction
As inspired by the information-storage capacity of DNA,[1] digital polymers are functional
macromolecules that store binary information at the molecular level in form of a controlled
comonomer sequence.[2-9] Indeed, a string of information can be implemented in a polymer
using two or more comonomers intentionally defined as single bits or groups of bits, allowing
encryption of any desired binary sequence in the chains. These polymers are typically
prepared by automated solid-phase synthesis, which consists of attaching monomers one-by-
one on a solid-support and hence permits to insert coding motifs at precise positions in chains
containing up to 100-150 repeating units. Longer chains are not attainable by stepwise
synthesis and an efficient strategy to circumvent this limitation is to produce large libraries of
medium-sized macromolecules, as demonstrated to store about a megabyte of data in DNA
chains.[10-12] Writing information in the structure of non-naturals polymers is also very
appealing as they offer many advantages over DNA. Because it can proceed by simpler
protecting-group-free methods,[13] solid-phase synthesis of abiotic polymers is often easier
than for DNA. Moreover, the macromolecular structure of non-natural species can be
optimized to control their fragmentation for best sequencing by tandem mass spectrometry
(MS/MS). Such structural designs always aim at reducing the number of dissociation routes,
in order to avoid signal dilution when dealing with long chains, while still enabling full
sequence coverage. For this purpose, different strategies have been reported: removing
reactive functions,[14] controlling the oxidation degree of specific atoms,[15] or introducing
linkages of low dissociation energy to make any other bonds MS/MS silent.[16] The latter
approach was successfully employed to decipher large amounts of information stored in
sequence-defined poly(phosphodiester)s using a routine mass spectrometer.[17] Indeed, the
structure of long chains produced via automated phosphoramidite chemistry[18-21] was re-
engineered to enable their de novo sequencing, in spite of the numerous fragmentation routes
experienced by their repeating units when subjected to collision-induced dissociation (CID).
To do so, reagents including alkoxyamine linkages were used to introduce weak bonds
between short coded sub-segments (for example, between two consecutive bytes). As a result,
the reading step is simply based on MS/MS sequencing of sub-segments released upon
homolysis of all C–ON bonds in these alkoxyamine-modified poly(phosphodiester)s. Indeed,
using a two-step protocol (Supporting Information, Figure S1), intact synthetic chains ionized
as polyanions are first subjected to in-source collisional activation that cleaves each inter-byte
alkoxyamine linkage, and so-released sub-segments are then individually submitted to a
second activation stage for MS/MS sequencing based on polyphosphate dissociation rules.[22]
2
This article is protected by copyright. All rights reserved.

Angewandte Chemie International Edition
10.1002/anie.202010171
Due to the lack of precursor ion selection before the first activation stage, this sequencing
3
protocol based on two CID steps is designated as a pseudo-MS experiment. The use of
specific tags to label each sub-segment for its initial location in the chains permits to readily
reconstruct the information originally encoded in the backbone of these digital
poly(phosphodiester)s. Very recently, this simple strategy was used in conjunction with an
expanded monomer alphabet (e.g. using 4 or 8 coded monomers instead of two) to afford
high-capacity digital poly(phosphodiester)s.[23] Indeed, 144 bits of data were stored on and
extracted from a polymer single chain, which is to our knowledge the longest non-natural
decodable sequence ever reported.
Although this analytical strategy allows full sequence coverage of all coded sub-segments,
and hence reliable recovery of messages stored in poly(phosphodiester)s, automation of the
reading step with the MS-DECODER software[24] is currently impossible. Indeed, automated
3
identification of fragments in pseudo-MS spectra requires ions to be produced with
sufficiently high abundance, which is hardly achieved with the cleavable spacer used so far.
This spacer is composed of an amide-containing segment, (CH
2
)
3
NH(CO)C(CH
3
2
) , connected
to a 2,2,6,6-tetramethylpiperidine-1-oxyl (TEMPO) nitroxide: upon C–ON bond homolysis in
2
MS experiments, the former moiety becomes the  termination of all sub-segments but the
last one, whereas TEMPO becomes the α termination of all sub-segments but the first one
(Supporting Information, Scheme S1). While the TEMPO nitroxide is highly stable, the
carbon-centered radical constituting the -termination of sub-segments fragments is too
3
reactive during the second activation stage (pseudo MS experiments). The observed radical-
induced dissociation routes efficiently competes with phosphate bond cleavages, which gives
rise to an overall decrease of useful fragment abundance and hence jeopardizes their
automated identification.
In this study, phosphoramidite reagents including new TEMPO-based alkoxyamines were
developed and employed to prepare automatically-decodable poly(phosphodiester)s. To
prevent the above-mentioned side reactions which occurred with the amide moiety (Scheme
1
a), herein, we developed original alkoxyamine spacers bearing a highly delocalized rigid
linker, namely the 1-ethyl-phenyl moiety. The effect of the linker position, i.e. as pendant
group (Ring OutSide Chain, namely ROSC, Scheme 1b) or in the main chain (Ring InSide
Chain, namely RISC, Scheme 1c), on the alkoxyamine dissociation selectivity was studied by
electron paramagnetic resonance (EPR) of the monomeric spacer.
3
This article is protected by copyright. All rights reserved.

Angewandte Chemie International Edition
10.1002/anie.202010171
Scheme 1. Comparison between the radicals formed upon C–ON bond homolysis with (a) the
former and new alkoxyamine spacers following the (b) ROSC or (c) RISC strategy.
The new structural designs were then scrutinized for their efficiency at limiting undesired
radical-induced reactions during pseudo-MS3 of the first sub-segment using
poly(phosphodiester)s composed of two bytes, prior to studying longer polymers.
Results and discussion
Reactivity of the terminal carbon-centered radical
In order to rationalize the development of new alkoxyamines, dissociative reactions
experienced by sub-segments released upon C–ON bond homolysis of TEMPO-based digital
poly(phosphodiester)s were studied in details. Prior to using the expanded phosphoramidite
alphabets which have been recently described,[23] the model digital polymers employed to
evaluate alkoxyamine spacers were encrypted using two monomers only. Rules for coding
and decoding this binary monomer alphabet are described in details in previous
publications.[17, 20, 21] As previously mentioned, these polymers are prepared with an
•
alkylamide-TEMPO spacer (Scheme 1a) which leaves the (CH
2
)
3
NH(CO)C (CH
3
2
) moiety as
the  termination of the first sub-segment during the first activation stage. This is exemplified
with the two-byte containing polymer P0 (Supporting Information, Table S1), which CID
leads to two triply charged sub-segments (Supporting Information, Figure S2) when activated
as its preferential 6– form (i.e., 3 deprotonated phosphate per sub-segment). Activation of the
m/z 433.7 sub-segment containing the first byte conducts to two main types of dissociation
pathways (Figure 1). The first one comprises cleavages of backbone phosphate bonds that
produce ions useful for sequencing. These are α-containing (i.e., a z-
containing (i.e., w
z-, x z-, y z-, z
annotated with colored square labels in Figure 1).[22] However, sequence recovery is
achieved with α-containing fragments only because most product ions containing the original
termination are missing, as illustrated by many empty cells in the w/x/y/z raws of the
coverage table (inset of Figure 1).
, b
z-, c
z-, d
z-
) and -
i
i
i
i
z-) product ions typically employed to reconstruct the sequence
i
i
i
i
(

4
This article is protected by copyright. All rights reserved.

Angewandte Chemie International Edition
10.1002/anie.202010171
Figure 1. CID spectrum of the m/z 433.7 sub-segment containing the first byte of P0. Besides
phosphate bond cleavages leading to fragments annotated with colored square labels, the
triply deprotonated precursor (black structure) experiences three main radical-induced side
reactions (depicted in Scheme 2) to generate by-products I (m/z 400.4, in red), II (m/z 532.1,
in pink), and III (m/z 396.1, in brown). Further dissociation of these byproducts leads to signal
annotated with pink triangles. The purple star at m/z 650.6 designates an oxidation product of
the precursor ion. Inset: coverage table of the 01001000 byte with each cell reporting the
charge states of fragments produced after phosphate bond cleavages (Supporting Information,
Table S3). Due to α = HO, some fragments are not expected (n.e. in grey cells). Secondary
–
–
product ions are in grey, including [0 – H] at m/z 137.0 and [1 – H] at m/z 165.0 (annotated
by #).
These missing fragments are due to the occurrence of the second type of reactions induced by
the reactive carbon-centered radical termination of the dissociating precursor ion. As depicted
•
in Scheme 2a, H abstraction by this radical carbon would occur at the first methylene moiety
of the  end-group to generate the triply charged by-product I at m/z 400.4 after elimination
•
of a 100.1 Da radical (in red, Figure 1). Alternatively, H could be transferred from the first
methylene moiety of the last coding unit (Scheme 2b) to yield the doubly charged by-product
II at m/z 532.1 together with the m/z 237.1 anion (both in pink in Figure 1). Finally, the
radical-induced elimination of dimethylcyclobutanamide (113.1 Da) depicted in Scheme 2c
would generate the triply charged by-product III at m/z 396.1 (in brown, Figure 1). Of note,
5
This article is protected by copyright. All rights reserved.

Angewandte Chemie International Edition
10.1002/anie.202010171
collisional activation is also observed to promote oxidation of the triply charged m/z 433.7
precursor, as revealed by the doubly charged fragment at m/z 650.6 (designated by a purple
star in Figure 1). Such charge stripping reactions were first reported in high energy (keV)
collision regime[25] but were also observed at lower energy as in the case of
oligonucleotides.[26] It should also be noted that the charge state of the dissociating sub-
segment has no influence on the type of observed fragmentation reactions. This last result
suggests that long range stabilizing effects reported for some distonic anions[27, 28] are not
significant in the studied species.
Scheme 2. Mechanisms proposed to account for radical-induced pathways experienced by
sub-segments containing the first byte of poly(phosphodiester)s built with the alkylamide-
TEMPO spacer.
Overall, mechanisms proposed in Scheme 2 for the three main radical-induced side-reactions
•
all rely on a high flexibility of the linear (CH
2
)
3
NH(CO)C (CH
3
)2 terminal moiety.
Accordingly, these reactions should no longer proceed with an alternative termination of
increased steric hindrance or increased stiffness. In addition, any moiety improving radical
stabilization via delocalization of the unpaired electron would help reducing its reactivity. To
do so, the ROSC and RISC strategies depicted in Scheme 1 were explored.
Alkoxyamine spacer with increased electron delocalization (ROSC strategy)
6
This article is protected by copyright. All rights reserved.

Angewandte Chemie International Edition
10.1002/anie.202010171
Increasing the radical delocalization is correlated to an increase of the C–ON homolytic
dissociation rate constant k
and therefore a diminution of its activation energy, Ea,d.[29] This
may lead to a poorly stable polymer and the activation energy of the possible ROSC structures
was therefore estimated from the k calculated with the multiparameter equation (1)
established by Bertin et al. in 2005 for TEMPO-based alkoxyamines:[30]
log kd,393 = –14.8 (±0.7) + 13.9 (±0.9) · σRS + 13.6 (±3.2) · σ + 6.6 (±0.7) · ν
This equation is an empirical evaluation of k at 120°C from the radical stabilization constant
RS, the universal electrical Hammett constant σ , and the steric constant ν (detailed
d
d
I
(1)
d
σ
I
calculations are given in Supporting Information). Earlier MS/MS experiments have shown
that, while CID of poly(phosphodiester)s built with a di-methylated spacer (Table 1, a) only
leads to C–ON bond homolysis, phosphate bond cleavages become competitive for polymers
constructed with a mono-methylated spacer (Table 1, b).[17] The dissociation rate constant of
the latter alkoxyamine was therefore estimated with equation (1) and considered as upper
limit of k
d
. From calculations summarized in Table 1, the 1-ethyl-phenyl radical structure
without amide moiety (Table 1, f) appears as the structure of choice to combine a low
activation energy (high k
d
) and a high stability (low k ) of the final polymer.
d
alkyl
radical
σ
σ
ν
RS
0.20
0.07
1.42
0.18
0.09
1.01
0.41
0.10
1.45
0.39
0.12
1.20
0.36
0.07|0.06
1.28
0.34
0.09|0.07
0.86
I
5
.3.10–4
3.7.10–4
1.8.10–5
143.9
76.39
94.0
1.56
0.41|0.30
–
1
0.02
K
d,393 (s )
|
–
1
120.7
106.7
111.1|112.2 132.8|134.0
E (kJ.mol )
a
Table 1. Calculated molecular descriptors and activation energy corresponding to the
dissociation rate constant calculated with equation (1). An averaged frequency factor A
d
=
1
4
–1
2
.4.10 s was used to estimate E
a
from kd,393. The σ values of structures e) and f) are
I
directly connected to a phosphate in the final polymer (left values) or to a DMT-protecting
group in the initial monomer tested in EPR (right values).
7
This article is protected by copyright. All rights reserved.

Angewandte Chemie International Edition
10.1002/anie.202010171
Among the various possibilities to prepare alkoxyamine compounds, one of the most
robust and convenient consists of performing a coupling reaction between an alkyl radical and
a nitroxide under atom transfer radical addition (ATRA) conditions.[31] The first alkoxyamine
candidate includes a pendant phenyl group (hence named Ring OutSide Chain, ROSC) and
was prepared according to the 4-step chemical pathway described in Scheme 3. Typically, 2-
chloro-2-phenylethan-1-ol (a2) obtained by chlorination of phenylethane-1,2-diol (a1) with
HClaq was protected with dimethoxytrityl chloride (DMT-Cl) in 85% yield. The resulting
halide a3 was reacted with 4-hydroxy-TEMPO in ATRA conditions (CuBr, Cu0,
pentamethyldiethylenetriamine (PMDETA) in degassed THF) to produce the alkoxyamine a4
in
74%
yield.
Final
functionalization
with
O-2-cyanoethyl-N,N-
diisopropylchlorophosphoramidite led to the ROSC spacer a5 in 75% yield, as confirmed by
NMR (Supporting Information, Figure S3) and electrospray ionization high resolution mass
spectrometry (ESI-HRMS) performed in both MS and MS/MS modes (Supporting
Information, Figure S4). The C–ON bond homolytic dissociation rate constant (k
d
) determined
–4
–1
by EPR for ROSC was 5.4 x 10 s at 128°C, which corresponds to an activation energy of
–1
1
35.5 kJ mol , consistent with literature results associated to 2,2,6,6,-tetramethyl-1-(1-
phenyl-ethoxy)piperidine.[32]
Scheme 3. Synthesis of ROSC, a5. (i) HClaq 37%, CHCl
3
, r.t, 1h; (ii) DMT-Cl, anhydrous
0
pyridine, anhydrous THF, r.t, overnight; (iii) 4-hydroxyTEMPO, CuBr, Cu , PMDETA, THF,
r.t., overnight; (iv) O-2-cyanoethyl-N,N-diisopropylchlorophosphoramidite, DIPEA,
anhydrous DCM, r.t, 1h. See details in supporting information.
The ROSC phosphoramidite reagent was then employed to prepare the P1 polymer
composed of two bytes, HO-00000000-ROSC-11111111-T (Supporting Information, Table
S1). The negative mode ESI mass spectrum shows a monodisperse sample, with P1
distributed over charge states ranging from –3 to –9 (Supporting Information, Figure S5).
8
This article is protected by copyright. All rights reserved.

Angewandte Chemie International Edition
10.1002/anie.202010171
During the first activation stage, homolysis of the ROSC spacer cleaves the [P1 – 6H]6–
precursor ion in two complementary triply charged sub-segments (Supporting Information,
Figure S6). Activation of the m/z 407.4 ion containing the first byte leads to the pseudo-MS3
spectrum shown in Figure 2. The major product ion in this spectrum is a doubly charged
species at m/z 504.0 (in red), which formation would require a backbiting process (Supporting
Information, Scheme S2) which is obviously not hindered by the bulky benzyl group. The
alternative elimination of styrene from the terminal styryl radical yields the triply charged m/z
3
72.7 species (in pink), which quite low abundance can be explained by a consecutive fast
radical-induced depolymerization process (Supporting Information, Scheme S3). Oxidation of
the precursor ion is also observed, yielding the doubly charged fragment at m/z 611.1
(
designated by a purple star in Figure 2). As a result of these side reactions, many w/x/y/z
product ions containing the original  termination of the m/z 407.4 precursor ion are missing
see coverage table in inset of Figure 2). Yet, full sequence coverage of byte 1 can be
achieved with α-containing fragments, all detected but some with very minor abundance
Supporting Information, Table S4). This is in contrast with the sub-segment containing the
second byte of P1, which CID yields all expected α- and -containing product ions
Supporting Information, Figure S7 and Table S5).
(
(
(
Figure 2. CID spectrum of the m/z 407.4 sub-segment containing the first byte of P1. The
triply deprotonated precursor ion (black structure) experiences radical-induced side reactions
leading to i) the m/z 504.0 ion (which further dissociates into products designated with red
9
This article is protected by copyright. All rights reserved.

Angewandte Chemie International Edition
10.1002/anie.202010171
triangles) and ii) the m/z 327.7 ion (in pink) after elimination of styrene. The purple star at m/z
6
11.1 designates an oxydation product of the precursor ion. Inset: coverage table of the
0
0000000 byte with each cell reporting the charge states of fragments produced after
phosphate bond cleavages (Supporting Information, Table S4). Due to α = HO, some
fragments are not expected (n.e. in grey cells). Secondary product ions are in grey, including
–
[
0 – H] at m/z 137.0 (#).
Alkoxyamine spacer with increased delocalization and stiffness (RISC strategy)
In order to prevent CID side-reactions still observed for the ROSC-based P1, a second
alkoxyamine was designed with a phenyl ring inside the main chain to also increase the
stiffness of the  moiety of the first sub-segment. The so-called RISC (Ring InSide Chain)
species was prepared in 5 steps (Scheme 4), starting with Wohl-Ziegler bromination of 4-
benzaldehyde b1 which afforded b2 in 20% yield, followed by formation of the alkoxyamine
b3 in ATRA conditions (82% yield). The benzaldehyde was then reduced with NaBH to
4
afford diol b4 in 71% yield, which primary alcohol was selectively protected with DMT-Cl to
produce b5 in 64% yield. Final functionalization with O-2-cyanoethyl-N,N-
diisopropylchlorophosphoramidite afforded the targeted RISC b6 in 42% yield, as confirmed
by NMR (Supporting Information, Figure S8) and ESI-HRMS and MS/MS (Supporting
Information, Figure S9). EPR study of RISC permitted to determine a dissociation rate
–
4
–1
–1
constant of 8.7·10 s at 127°C, corresponding to an activation energy of 133.6 kJ mol
consistent with values estimated in Table 1 and literature data.[32]
Scheme 4. Synthesis of RISC, b6. (i) BMS, anhydrous THF, rt, overnight; (ii) DMT-Cl,
anhydrous pyridine, anhydrous THF, r.t, overnight; (iii) 4-hydroxyTEMPO, CuBr, Cu0,
PMDETA,
degassed
THF,
r.t.,
overnight;
(iv)
O-2-cyanoethyl-N,N-
diisopropylchlorophosphoramidite, DIPEA, anhydrous DCM, r.t, 1 h. See details in
supporting information.
1
0
This article is protected by copyright. All rights reserved.

Angewandte Chemie International Edition
10.1002/anie.202010171
The byte-truncated polymer P2 (Supporting Information, Table S1) prepared with the
RISC phosphoramidite reagent was observed with charge states ranging from –4 to –10 in
negative mode ESI-MS (Supporting Information, Figure S10). Activated homolysis of RISC
yields the expected complementary sub-segments (Supporting Information, Figure S11) and
further CID of the m/z 412.0 ion containing the first byte of P2 gives rise to the pseudo-MS3
data reported in Figure 3. All product ions observed in this spectrum result from cleavages of
phosphate bonds, including those holding the original  termination of the dissociating sub-
segment. No side-product resulting from radical-induced reactions could be detected and,
interestingly, charge stripping was not observed to occur for the m/z 412.0 precursor. These
results clearly indicate a very low MS/MS reactivity of the RISC segment constituting the 
termination of the dissociating sub-segment.
Figure 3. CID spectrum of the m/z 412.0 sub-segment containing the first byte of P2, showing
that dissociation only occurs via phosphate bond cleavages. Inset: coverage table of the
0
0000000 byte with each cell reporting the charge states of fragments (Supporting
Information, Table S6). Due to α = HO, some fragments are not expected (n.e. in grey cells).
z–
Most x
i
ions cannot be determined (hence annotated with an asterisk) because they interfere
13
z–
–
with C isotope of a
37.0 (#).
i
ions. Secondary product ions are in grey, including [0 – H] at m/z
1
1
1
This article is protected by copyright. All rights reserved.

Angewandte Chemie International Edition
10.1002/anie.202010171
These CID data allow full coverage of the byte sequence but, as illustrated by the inset table
of Figure 3, some fragment series are incomplete. On the one hand, most x z–
ions cannot be
validated due to an isotopic interference with a z–
ions. On the other hand, some product ions
containing the α termination are missing (e.g., the largest c z–
ions). This is actually a
i
i
i
6–
sensitivity issue, due to the quite low abundance of the [P2 – 6H] primary precursor yielding
small amounts of the m/z 412.0 secondary precursor. Indeed, in contrast to the monodisperse
ROSC-based P1 polymer, ESI(–)-MS of the P2 sample shows four main impurities (named A-
D) of lower mass but higher intensity than the targeted P2 polymer (Supporting Information,
Figure S10). When subjected to CID, all impurities exhibit the same dissociation pattern, with
C–ON bond homolysis yielding a single product ion at m/z 600.5 corresponding to the sub-
segment containing byte 2 at the right-hand side of the alkoxyamine linkage (Supporting
Information, Figure S12). For each species, the defective part at the left-hand side of the C–
ON bond has a very low mass (134.1 Da for A, 148.2 Da for B, 196.1 Da for C and 247.1 Da
for D) compared to the perfect one (1225.2 Da) in P2. Moreover, this defective part is never
detected as a charged fragment, which indicates that it does not contain any phosphate groups.
Based on these features and using accurate mass measurements as a support, structural
assumptions could be proposed for impurities A (Supporting Information, Figure S12a) and B
(
Supporting Information, Figure S12b). In both cases, proposed structures suggest a
nucleophilic substitution in benzylic position with ammonia (for impurity A) or methylamine
for impurity B), the two reagents used to cleave chains from their solid support at the end of
(
the synthesis. Because such impurities were not formed during the cleavage step of the
ROSC-based P1, where the last phosphate group of byte 1 was in a homobenzylic position,
the RISC structure was modified accordingly and the RISC2 spacer was investigated.
RISC2 was hence synthesized in 6 steps starting from commercially available 4-
acetylphenylacetic acid (Scheme 5) starting with NaBH reduction of 2-(4-acetylphenyl)acetic
4
acid (c1) yielding alcohol c2 in 67% yield followed with chlorination in THF with
concentrated HCl in 72% yield. The resulting 2-(4-(1-chloroethyl)phenyl)acetic acid (c3) was
then selectively reduced with BMS to produce alcohol c4 in 73% yield, which was then
involved in the same synthesis route than ROSC in Scheme 3, with DMT-protection (c5 in
8
8% yield), ATRA (c6 in 86% yield) and phosphoramidite functionalization (c7 in 48%
yield).
1
2
This article is protected by copyright. All rights reserved.

Angewandte Chemie International Edition
10.1002/anie.202010171
Scheme 5. Synthesis of RISC2, c7. (i) NaBH
4
, EtOH, rt, overnight; (ii) HCl 37%, THF, 0°C,
4
h; (iii) BMS, anhydrous THF, rt, overnight; (iv) DMT-Cl, anhydrous pyridine, anhydrous
0
THF, r.t, overnight; (v) 4-hydroxy-TEMPO, CuBr, Cu , PMDETA, degassed THF, r.t.,
overnight; (vi) O-2-cyanoethyl-N,N-diisopropylchlorophosphoramidite, DIPEA, anhydrous
DCM, r.t, 1h. See details in Supporting Information.
RISC2 was fully characterized by NMR (Supporting Information, Figure S13) and mass
spectrometry (Supporting Information, Figure S14). This synthesis took advantage of the
selectivity of the borane reduction of a benzoic acid in presence of a benzyl chloride.[33] It
should be noted that the molecule is prone to dimerization before DMT protection since the
primary alcohol resulting from the borane reduction is a good Lewis base. The c4
intermediate was hence stored at –18°C before step (ii) and dissolved at low temperature
before slow addition to a well stirred solution of DMT-Cl with 1.5 eq of anhydrous pyridine.
The dissociation rate constant determined by EPR for RISC2 was 1.1·10–3 s–1 at 133°C,
–1
corresponding to an activation energy of 134.7 kJ mol consistent with values estimated in
Table 1 and literature data.[32]
The byte-truncated poly(phosphodiester) P3 prepared with RISC2 has two identical bytes
composed of eight 0 units, which corresponds to a C77
H
156
N
3
74
O P17 elemental composition
and a monoisotopic mass of 2833.4076 Da (Supporting Information, Table S1). Negative
mode ESI-MS shows a monodisperse sample, with P3 detected with charge state ranging from
3
– to 7– (Supporting Information, Figure S15). This result shows that the structure of RISC2
has kept intact during the cleavage step, and hence confirms our assumptions regarding the
origin of impurities A and B found in the P2 sample. Collisional activation of [P3 – 6H]6–
conducts to homolysis of the C–ON bond in RISC2 and subsequent release of the two
expected complementary products (Supporting Information, Figure S16). As shown in Figure
4
, the CID spectrum of the m/z 416.7 sub-segment containing the first byte of P3 only
displays fragments of quite high abundance resulting from phosphate bond cleavages. The
1
3
This article is protected by copyright. All rights reserved.

Angewandte Chemie International Edition
10.1002/anie.202010171
reason why all -containing fragments are detected except x
i
ions remains unclear. Overall,
both MS and MS/MS data clearly show that RISC2 has an optimal structure to prevent (i)
degradation issues while releasing chains from their solid support and (ii) undesired radical-
3
induced during pseudo-MS sequencing experiments.
Figure 4. CID spectrum of the m/z 416.7 sub-segment containing the first byte of P3, showing
that dissociation only occurs via phosphate bond cleavages. Inset: coverage table of the
0
0000000 byte with each cell reporting the charge states of fragments (Supporting
Information, Table S7). Due to α = HO, some ions are not expected (n.e. in grey cells). Peaks
annotated in grey are secondary fragments including deprotonated 0 units at m/z 137.0 (#).
Sequencing of RISC2-terminated central bytes
The performance of RISC2 to prevent side reactions during MS/MS of first bytes was also
tested for central bytes that experience additional undesired dissociations due to the presence
of nucleotide tag (used to retrieve initial bye location, Supporting Information, Figure S1) in
the vicinity of the terminal radical moiety (Supporting Information, Scheme S1c). When using
the amide-containing spacer, these alternative reactions are identical regardless of the
•
nucleotide tag, also rely on a high flexibility of the linear (CH
2
)
3
NH(CO)C (CH
3
)2 terminal
moiety (Supporting Information, Scheme S4) and efficiently compete with phosphate bond
cleavages useful for sequencing (Supporting Information, Figure S17). The RISC2 spacer was
hence used to prepare the four-byte poly(phosphodiester) P4 (Supporting Information, Table
1
4
This article is protected by copyright. All rights reserved.

Angewandte Chemie International Edition
10.1002/anie.202010171
S1) as a monodisperse sample (Figure S18). The [P4 – 12H]12– precursor ion was selected to
favor production of sub-segments as triply charged ions during the first activation stage
(
Supporting Information, Figure S19). CID spectra of the second byte labeled with adenine
Figure 5a) and of the third byte labeled with cytosine (Figure 5b) do not exhibit any side-
(
reaction products but only display ions resulting from phosphate bond cleavages. As
illustrated by inset tables of Figure 5, full sequence coverage is readily achieved for both sub-
segments. Amongst all expected fragments, only one ion (c 2–) remains not assignable due to
7
coincidence of its m/z 599.1 value with that of the precursor ion. Mass characteristics of the
RISC2 spacer were then implemented in the MS-DECODER algorithm developed for
alkoxyamine-containing poly(phosphodiester)s, allowing automated sequence and localization
of each byte to be successfully performed in 50 to 100 ms (Supporting Information, Figure
S20).
Figure 5. CID spectrum of a) the m/z 599.1 sub-segment labeled with adenine and containing
the second byte of P4, and b) the m/z 591.1 sub-segment labeled with cytosine and containing
the third byte of P4. Inset: coverage table of the 00000000 byte with each cell reporting the
charge states of fragments (Supporting Information, Tables S8-S9). Peaks annotated in grey
are secondary fragments including deprotonated 0 units at m/z 137.0 (#).
1
5
This article is protected by copyright. All rights reserved.

Angewandte Chemie International Edition
10.1002/anie.202010171
Automated sequencing of high-capacity digital poly(phosphodiester)s
In order to demonstrate that a complete data file stored in a single molecule can also be
read in a fully automated manner, the RISC2 alkoxyamine was used in conjunction with an
expanded phosphoramidite alphabet to store a black and white image of 80 pixels in the P5
polymer (Supporting Information, Table S1). Recently developed to enable a storage density
of 2 bits per monomer,[23] this alphabet is based on a library of four binary dyads 00, 01, 10
and 11 (Supporting Information, Scheme S5). The image was created in a grid of 10 columns
and 8 raws, converted into a binary sequence using the convention “white = 0, black = 1”, and
the binary string was then converted into a coded monomer sequence according to the dyad-
language. The so-obtained five-block P5 poly(phosphodiester) was produced as a
monodisperse sample, as supported by ESI-MS data (Supporting Information, Figure S21a).
The observed charge state distribution ranges from 8– to 19–, with one of the preferential
charge states corresponding to 3 negative charges per sub-segment. Activated homolysis of
RISC2 in this [P5 – 15H]15– precursor yields the expected complementary sub-segments
(Supporting Information, Figure S21b). The m/z value measured for each triply charged sub-
segment permits to retrieve their initial location thanks to the tag system but is associated to a
set of possible dyad combinations (Supporting Information, Table S10), in contrast to the
unique 0/1 composition obtained when using the alphabet where one unit codes for one bit.
This is a consequence of the mass of dyads used here, with 138.0 Da for 00, 152.0 Da for 01,
1
66.0 for 10 and 180.1 for 11, leading to mass coincidences when combined within a given
sub-segment. For example, the first sub-segment of P5 detected at m/z 435.4 as a triply
charged ion has a mass of 1309.2 Da which includes the 1160 Da contribution of the coding
string that be obtained with 6 “00” + 1 “01” + 1 “11”, 6 “00” + 2 “10”, 5 “00” + 2 “01” + 1
“10”, or 4 “00” + 4 “01”. As a result, it takes more time for the MS-DECODER to decode
each sub-segment written with dyads (1 second, on average) compared to the 100 ms
requested to retrieve the binary information encoded with monomers “0” and “1”. Yet, all
3
pseudo-MS spectra recorded for P5 (Supporting Information, Figure S21c) were successfully
processed. Based on the location tag system, the MS-DECODER software enables proper
alignment of all sub-segments according to their initial sequence in the chain (Figure 6, left
part) and, providing that the user defines the image definition (10 columns x 8 raws), readily
reconstruct the black and white image (Figure 6, right part).
1
6
This article is protected by copyright. All rights reserved.

Angewandte Chemie International Edition
10.1002/anie.202010171
Figure 6. Output of the MS-DECODER software for automated decoding of the 80 pixel
image of a space invider stored in P5.
Conclusion
In summary, it was shown herein that precise alkoxyamine-design allows preparation of
automatically-decodable digital poly(phosphodiester)s. Among all alkoxyamines studied in
this work, RISC2 turned out to be, by far, the most efficient. Indeed, this particular design
efficiently suppresses side-peaks that usually prevent MS/MS automated decryption with the
software MS-DECODER. As a result, model digital polymers could be decrypted in a fully
3
automated manner: after loading all pseudo-MS files recorded for a given polymer, the
algorithm was developed to decode each sub-segment and to align all sub-segments according
to their initial order in the chain thanks to the location tag system. These results constitute an
important step-forward in the field of non-biological digital polymers. Poly(phosphodiester)s
being currently the only family of synthetic polymers allowing high capacity storage,[23] the
automatic decryption of these polymers opens up very interesting perspectives for fast and
parallel sequencing. Using RISC2 in conjunction with expanded phosphoramidite alphabets
was shown to allow fast (5 seconds) interpretation of a black and white image stored in a
synthetic poly(phosphodiester) and, when combined with appropriate compression
algorithms, this may also allow efficient processing of larger amounts of information as
requested to encode texts, videos or secure data. These results further indicate that precise
macromolecular-engineering is a key strategy for improving the storage properties of digital
macromolecules. Consequently, the performances of non-natural informational polymers may
soon equal or even surpass those of biological polymers.[34]
Experimental section
See Supporting Information
1
7
This article is protected by copyright. All rights reserved.

Angewandte Chemie International Edition
10.1002/anie.202010171
Acknowledgments
These results were obtained with the support of the French National Research Agency (ANR
project 00111001, grant numbers ANR-16-CE29-0004-01 and ANR-16-CE29-0004-02). L.C.
acknowledges support from Spectropole, the Analytical Facility of Aix-Marseille University,
by allowing a special access to the instruments purchased with European Funding (FEDER
OBJ2142-3341).
References
[
1] V. Zhirnov, R. M. Zadegan, G. S. Sandhu, G. M. Church, W. L. Hughes, Nature Mater.
016, 15, 366.
2
[
[
[
2] J.-F. Lutz, M. Ouchi, D. R. Liu, M. Sawamoto, Science 2013, 341, 1238149.
3] H. Colquhoun, J.-F. Lutz, Nat. Chem. 2014, 6, 455.
4] R. K. Roy, A. Meszynska, C. Laure, L. Charles, C. Verchin, J.-F. Lutz, Nat. Commun.
2
015, 6, 7237.
[
[
5] M. Rutten, F. W. Vaandrager, J. Elemans, R. J. M. Nolte, Nat. Rev. Chem. 2018, 2, 365.
6] S. Martens, A. Landuyt, P. Espeel, B. Devreese, P. Dawyndt, F. Du Prez, Nat. Commun.
2
018, 9, 4451.
[
[
7] A. C. Boukis, M. A. R. Meier, Eur. Polym. J. 2018, 104, 32.
8] Z. H. Huang, Q. N. Shi, J. Guo, F. Y. Meng, Y. J. Zhang, Y. T. Lu, Z. F. Qian, X. P. Li,
N. C. Zhou, Z. B. Zhang, X. L. Zhu, Nat. Commun. 2019, 10, 1918.
9] J. M. Lee, M. B. Koo, S. W. Lee, H. Lee, J. Kwon, Y. H. Shim, S. Y. Kim, K. T. Kim,
Nat. Commun. 2020, 11, 56.
[
[
[
10] G. M. Church, Y. Gao, S. Kosuri, Science 2012, 337, 1628.
11] N. Goldman, P. Bertone, S. Y. Chen, C. Dessimoz, E. M. LeProust, B. Sipos, E. Birney,
Nature 2013, 494, 77.
[
[
[
12] Y. Erlich, D. Zielinski, Science 2017, 355, 950.
13] T. T. Trinh, C. Laure, J.-F. Lutz, Macromol. Chem. Phys. 2015, 216, 1498.
14] L. Charles, G. Cavallo, V. Monnier, L. Oswald, R. Szweda, J.-F. Lutz, J. Am. Soc. Mass
Spectrom. 2017, 28, 1149.
[
[
15] B. Liu, Q. Shi, L. Hu, Z. Huang, X. Zhu, Z. Zhang, Polym. Chem. 2020, 11, 1702.
16] J.-A. Amalian, A. Al Ouahabi, G. Cavallo, N. F. Konig, S. Poyer, J.-F. Lutz, L. Charles,
J. Mass Spectrom. 2017, 52, 788.
[
[
17] A. Al Ouahabi, J.-A. Amalian, L. Charles, J.-F. Lutz, Nat. Commun. 2017, 8, 967.
18] N. Appukutti, C. J. Serpell, Polym. Chem. 2018, 9, 2210.
1
8
This article is protected by copyright. All rights reserved.

Angewandte Chemie International Edition
10.1002/anie.202010171
[
[
[
[
[
19] M. Vybornyi, Y. Vyborna, R. Haner, Chem. Soc. Rev. 2019, 48, 4347.
20] A. Al Ouahabi, M. Kotera, L. Charles, J.-F. Lutz, ACS Macro Letters 2015, 4, 1077.
21] A. Al Ouahabi, L. Charles, J.-F. Lutz, J. Am. Chem. Soc. 2015, 137, 5629.
22] S. A. McLuckey, G. J. Vanberkel, G. L. Glish, J. Am. Soc. Mass Spectrom. 1992, 3, 60.
23] E. Laurent, J.-A. Amalian, M. Parmentier, L. Oswald, A. Al Ouahabi, F. Dufour, K.
Launay, J.-L. Clement, D. Gigmes, M. A. Delsuc, L. Charles, J.-F. Lutz,
Macromolecules 2020, 53, 4022.
[
[
[
24] A. Burel, C. Carapito, J.-F. Lutz, L. Charles, Macromolecules 2017, 50, 8290.
25] R. G. Cooks, J. H. Beynon, T. Ast, J. Am. Chem. Soc. 1972, 94, 1004.
26] D. P. Little, R. A. Chorush, J. P. Speir, M. W. Senko, N. L. Kelleher, F. W. McLafferty,
J. Am. Chem. Soc. 1994, 116, 4893.
[
[
[
[
[
27] G. Gryn'ova, D. L. Marshall, S. J. Blanksby, M. L. Coote, Nat. Chem. 2013, 5, 474.
28] G. Gryn'ova, M. L. Coote, J. Am. Chem. Soc. 2013, 135, 15392.
29] S. Marque, H. Fischer, E. Baier, A. Studer, J. Org. Chem. 2001, 66, 1146.
30] D. Bertin, D. Gigmes, S. R. A. Marque, P. Tordo, Macromolecules 2005, 38, 2638.
31] K. Matyjaszewski, B. E. Woodworth, X. Zhang, S. G. Gaynor, Z. Metzner,
Macromolecules 1998, 31, 5955.
[
[
[
32] S. Marque, C. Le Mercier, P. Tordo, H. Fischer, Macromolecules 2000, 33, 4403.
33] E. R. Burkhardt, K. Matos, Chem. Rev. 2006, 106, 2617.
34] J. F. Lutz, ACS Macro Letters 2020, 9, 185.
1
9
This article is protected by copyright. All rights reserved.

Angewandte Chemie International Edition
10.1002/anie.202010171
Entry for the Table of Contents
Reliable reading of information coded in the sequence of long poly(phosphodiester)s was
previously achieved by introducing an alkoxyamine spacer between information sub-
segments. In the present work, we show that the structure of this spacer can be designed in a
rational manner to prevent side-reactions during MS/MS sequencing and hence enable
automated reading of each byte of information in about 100 ms.
2
0
This article is protected by copyright. All rights reserved.