Gas-phase sugar formation using hydroxymethylene as the reactive formaldehyde isomer

Article abstract of DOI:10.1038/s41557-018-0128-2

Carbohydrates (CH₂O)_n are the formal adducts of carbon (atoms) to water with a repeating unit that structurally resembles H–C?–OH (hydroxymethylene). Although hydroxymethylene has been suggested as a building block for sugar formation, it is a reactive species that had escaped detection until recently. Here we demonstrate that formaldehyde reacts with its isomer hydroxymethylene to give glycolaldehyde in a nearly barrierless reaction. This carbonyl–ene-type transformation operates in the absence of base and solvent at cryogenic temperatures similar to those found in extraterrestrial environments or interstellar clouds. Hydroxymethylene acts as a building block for an iterative sugar synthesis, as we demonstrate through the formation of the triose glyceraldehyde. The thermodynamically preferred ketose dihydroxyacetone does not form, and the formation of further branched sugars in the iterative synthesis presented here is unlikely. The results therefore provide a link between the well-known formose (Butlerow) reaction and sugar formation under non-aqueous conditions.

Full text of DOI:10.1038/s41557-018-0128-2

Articles
https://doi.org/10.1038/s41557-018-0128-2
Gas-phase sugar formation using hydroxymethylene
as the reactive formaldehyde isomer
1
,2
1,2
1
1
André K. Eckhardt , Michael M. Linden , Raffael C. Wendeꢀ ꢀ , Bastian Bernhardtꢀ ꢀ and
1
Peter R. Schreinerꢀ ꢀ *
Carbohydrates (CH O) are the formal adducts of carbon (atoms) to water with a repeating unit that structurally resembles
2
n
̈
H–C–OH (hydroxymethylene). Although hydroxymethylene has been suggested as a building block for sugar formation, it is
a reactive species that had escaped detection until recently. Here we demonstrate that formaldehyde reacts with its isomer
hydroxymethylene to give glycolaldehyde in a nearly barrierless reaction. This carbonyl–ene-type transformation operates in
the absence of base and solvent at cryogenic temperatures similar to those found in extraterrestrial environments or interstel-
lar clouds. Hydroxymethylene acts as a building block for an iterative sugar synthesis, as we demonstrate through the forma-
tion of the triose glyceraldehyde. The thermodynamically preferred ketose dihydroxyacetone does not form, and the formation
of further branched sugars in the iterative synthesis presented here is unlikely. The results therefore provide a link between the
well-known formose (Butlerow) reaction and sugar formation under non-aqueous conditions.
he simplest sugar, glycolaldehyde (1a, 2-hydroxyacetalde- and aqueous solution, and at very low temperatures. Such con-
hyde), is considered to be the first entry in the famous for- ditions would render 1a and subsequently formed higher sugars
1
2
T
mose (Butlerow ) sugar-forming reaction in basic aqueous much more long-lived, as required by hypotheses that deem sug-
14,15
formaldehyde (H CO, 2a) solutions, but the detailed mechanism ars essential to the origin of life . As 2a is ubiquitously found in
2
1
6
for the initial dimerization of two molecules of 2a—formally elec- space it has been suggested to play a major role in the abiological
1
7
trophiles—to 1a and further on to higher sugars remains a puzzle. carbon cycle .
While 1a has long been assumed simply to be the slowly forming The photoreaction of 2a apparently leads to 1a in cryogenic
formaldehyde dimerization product, such a direct and uncatalysed argon matrices , as does the irradiation of methanol–carbon mon-
3,18
19
1
20,21
dimerization has been deemed chemically impossible , as only dis- oxide ices in the laboratory . Many such reactions suggest car-
22
̈
proportionation of 2a into methanol and formic acid (Cannizzaro benes , especially hydroxymethylene (H–C–OH, 3a), to be the key
3
reaction) takes place in aqueous solutions of 2a in the absence of 1a. intermediate; this was, however, only characterized spectroscopi-
4
23
24
The notion of ‘glycolaldehyde autocatalysis’ in the formose reac- cally in 2008 , although the mass spectrum was published in 1987 .
25,26
tion does not reveal the details of the underlying chemical reaction This parent molecule of the new hydroxycarbene family was also
5
mechanism , and typically requires the presence of liquid water, a suggested to form in the reactions of excited-state atomic carbon
27,28
strong base, high reactant concentrations and ambient tempera- atoms with water . We have shown recently that hydroxycarbenes
6
t
tures . Moreover, the sugars formed in the formose reaction show (3) (R=H, CH , CF , Ph, Bu, OH and cyclopropyl) can readily be
3
3
7
8
very limited long-term stability and any stereochemical purity
prepared through the thermal decomposition of the corresponding
25,26
(
not addressed in the formose reaction per se) would be lost because α-ketocarboxylic acids (4) . For instance, the formaldehyde iso-
̈
of the multiple base-catalysed propagation and equilibration steps. mer H–C–OH (3a, R=H) results from pyrolysis of glyoxylic acid
23
̈
In other words, the aqueous formose reaction is fascinating in its (4a) , while H C–C–OH (3b, R=CH ), that is, isomeric acetalde-
3
3
29
own right but it suffers from very severe limitations that make it hyde (2b), forms from pyruvic acid (4b) . We now demonstrate
probably less relevant for questions concerning sugar formation in a that 3a serves as an iterative building block for the formation of
prebiotic and extraterrestrial context.
simple sugars (Fig. 1). This has been suggested from computational
30
A variety of sugar-related compounds, including the ketose dihy- studies and on the basis of the product distributions in the reac-
31
droxyacetone (5b), were found in the Murchison and Murray mete- tions of excited carbon atoms with water . Of course, 3a would
9
,10
orites and it was concluded ‘that polyols were present on the early have to be formed in other ways in space than in the laboratory, but
Earth and therefore at least available for incorporation into the first there are plenty of options within the realm of C, H, O chemistry,
9
32–34
forms of life’ . This implies that aqueous and ambient conditions including the photoreaction of CO or CO with water ; this issue
2
11
are not necessarily needed for carbohydrate/polyol production . is not addressed here. Note that this makes a remote connection
12,13
Although 1a was only shown to exist in outer space in 2000 , this to Fischer–Tropsch chemistry, which utilizes the same constituents
3
5,36
finding was an important indication that the processes leading to and 3a as the key intermediate in the production of hydrocarbon
biologically relevant molecules are also taking place in non-Earth- fuels from CO and H on catalyst surfaces.
2
like environments. This was interpreted as evidence for a gas-phase
Proposed intermediates 3 had escaped detection because they
13
formose reaction by a hitherto unknown mechanism .
undergo tunnelling reactions to their aldehyde isomers with half-
The key question therefore is how the first molecules of 1a lives of typically several hours under cryogenic matrix isolation
could form from 2a as the starting material in the absence of a base conditions2 ; these tunnelling reactions are also detectable at
5,26
1
2
Institute of Organic Chemistry, Justus Liebig University, Giessen, Germany. These authors contributed equally: André K. Eckhardt and Michael M. Linden.
*
e-mail: prs@uni-giessen.de
NAꢁꢂRE CꢃEMꢄSꢁRꢅ | www.nature.com/naturechemistry

Articles
NATure CHemisTry
Formose
O
Chemically
not possible
uncatalysed
O
O
O
H
H
2
a
+
OH
Aqueous
solution
H
OH
Autocatalytic
H
H
H
H
H
H
OH
H
H
Formaldehyde (2a)
Glycolaldehyde (1a)
Glyceraldehyde (5a)
Light
O
ΔT
Very fast
Very fast
C
H
H
C
H
O
O
H
CO2H
–CO2
H
Gas-phase
surfaces
4a
Hydroxymethylene (3a)
Light
Working hypothesis
Autocatalysis, water, base and high
concentrations are not required
CO + H2
Fig. 1 | Mechanistic hypotheses related to sugar formation from formaldehyde. Simplified depiction of the traditional autocatalytic (with glycolaldehyde
1a)) formose reaction (top half) in aqueous media in the presence of base at ambient temperatures, and the proposed new reaction path via
(
hydroxymethylene (3a) (bottom half) in the gas phase or on surfaces. Note that the initiation step in the formose reaction for the formation of the first
molecule of 1a is chemically not possible in an uncatalysed fashion.
O
H
O
H
O
O
+
H
H
O
H
O
HVFP, 1,000 °C
– CO2
O
OH
H
2a
H
H
H
H
H
H
H
O
4a
3a
Carbonyl-ene reaction
1a
Fig. 2 | Mechanistic hypothesis for the uncatalysed ‘carbonyl–ene’ reaction. Formaldehyde (2a) reacts with hydroxymethylene (3a), generated from
glyoxylic acid (4a) via CO extrusion, to give glycolaldehyde (1a). Note the exchange of atoms from starting materials to products. The C=O and C–OH
2
oxygen atoms have changed places. HVFP, high-vacuum flash pyrolysis.
1
13
much higher temperatures in the gas phase (320–350K) where they
H as well as C NMR and gas chromatography–mass spectrom-
37
are, as expected, much faster . This reactivity, however, does not etry (GC–MS) analysis) of various sugars depending on the choice
play a role in the reactions presented here as they are extremely fac- of carbene precursor and reaction partners. The matrix isolation
ile owing to very low activation barriers. In the context of initiation experiments revealed that the mechanism proposed for the reac-
of formose-type reactions, the key feature of 3 is that it formally tions of hydroxycarbenes with aldehydes is productive even at cryo-
turns electrophile 2 into a nucleophile, so that reaction of 2 with 1 genic temperatures, without water and a base, and in high dilution.
may occur smoothly. The adducts of 3 with the thiazolin-2-ylidene
We investigated the reaction of 3a with 2a produced in situ dur-
38
23
moiety of vitamin B are the famous ‘Breslow intermediates’ in the ing the pyrolysis of 4a . After pyrolysis we obtained a colourless
1
39,40
41
catalysed Umpolung of aldehydes to give acyloins .
polymer in our liquid N cooling trap. We analysed the pyrolysis
2
products via NMR spectroscopy in D O (Fig. 3 and Supplementary
2
Results and discussion
Figs. 34–44), which efficiently suppresses dimerization of the
Here we provide evidence that the simplest carbohydrates form sugars because of the preferential formation of the correspond-
1
through a long discussed yet hitherto unobserved very facile, unca- ing hydrates. The H NMR spectrum in DMSO-d showed a large
6
42
1
talysed reaction of ‘active aldehydes’ with aldehydes (or ketones). number of multiplets that made the unequivocal analysis extremely
39,43
Although this has not been shown in a prebiotic context , we will challenging (Supplementary Figs. 3–8). Additionally, there are mul-
demonstrate here that this is indeed viable. As initiation and homol- tiple equilibria between the monomer of 1a, five-membered and
ogation follow the same mechanistic pattern (Figs. 1 and 2), this is six-membered cyclic dimer structures, and the hydrated forms,
4
5
an alternative to the often proposed aldol condensation reactions which make the spectral identification practically impossible .
that require aqueous conditions and are thermodynamically driven We also employed matrix isolation techniques to unequivocally
4
4
(
cf. Lobry de Bruyn-Alberda-van Ekenstein rearrangement ). Aldol identify the products of these reactions via infrared spectroscopy
condensation steps are suppressed under our reaction conditions.
(Supplementary Figs. 9–14).
We analysed the colourless polymer by heating it under vacuum
Flow pyrolyses. Two types of experiment were performed, namely and mixing the resulting gas with argon (1:200). The correspond-
quantitative flow pyrolysis (FP; for the experimental set-up see ing matrix isolation spectrum (Supplementary Fig. 9) was domi-
Supplementary Fig. 1) and high-vacuum flash pyrolysis (HVFP) nated by 2a resulting from the polymer polyoxymethylene (POM),
experiments coupled with cryogenic matrix isolation infrared (IR) with visible traces of methanol, formic acid, acetic acid (6) and 1a.
spectroscopy. The FP experiments allowed the identification (via Methanol and formic acid are possibly the reaction products of a
NAꢁꢂRE CꢃEMꢄSꢁRꢅ | www.nature.com/naturechemistry

NATure CHemisTry
Articles
a
H2O
HO OH
H
H
O
HO OH
H
H2O
O
OH
e
OH
H
a
a
c
H
H
CH3
H
–H2O
H
H
CH OH
3
b
d
1a
O
OH
HO OH OH
H2O
H
H
j
H
9.69
9.67
9.65
9.63
9.61
H
H
g
H
h
e
–H2O
δ (ppm)
H
OH
H
OH
HO OCH3
f
5a
i
H
H
NMR spiking experiment:
d
c
h
FP spectrum
+1 mg 5a
c
h
HO OCH3
+5 mg 5a
j
H
H
i
5.07 5.05 5.03 5.01 4.99 4.97 4.95 4.93
δ (ppm)
*
*
b
g
f
5.8
5.6
5.4
5.2
5.0
4.8
4.6
4.4
4.2
4.0
3.8
3.6
3.4
δ (ppm)
b
HO OCH3
CH3OH
H
H
*
HO OH
H
H
i
h
d
c
j
δ (ppm)
δ (ppm)
OH
H
δ (ppm)
δ (ppm)
O
HO OH
d
O
HO OH OH
NMR spiking experiment:
FP spectrum
H2O
f
H2O
b OH
OH
i
H
H
H a
H c
H e
g
H h
j
H
H
–H2O
H
–H2O
H
H
H
OH
H
OH
+1 mg 5a
5a
1a
+5 mg 5a
105
100
95
90
85
80
75
δ (ppm)
70
65
60
55
50
45
1
Fig. 3 | NMR spectra of products resulting from the pyrolysis of glyoxylic acid. a, H NMR (600ꢀMHz, D O) spectrum taken immediately after pyrolysis
2
of glyoxylic acid (4a) and subsequent spiking with 5a. The main FP products are various polymers of polyoxymethylene (marked with an asterisk) and
the hydrated monomer methanediol (4.90ꢀppm) from unimolecular decarboxylation and rearrangement of 4a, formic acid from decarbonylation (see
Supplementary Figs. 38 and 39 for complete spectra), methanol and glycolaldehyde (1a) resulting from the bimolecular reaction of hydroxycarbene (3a)
with formaldehyde 2a. We assigned the remaining small signals to the triose glyceraldehyde 5a; these were confirmed in a spiking experiment with the
13
pure compound. b, Corresponding C APT (151ꢀMHz, D O) spectrum. C–H and CH carbon atoms are pointing upwards, whereas CH and quaternary
2
3
2
carbon atoms are pointing downwards. The assigned signals for hydrated 5a increased significantly by adding 1ꢀmg and 5ꢀmg of pure 5a dimer to the FP
pyrolysate in an NMR spiking experiment. Signals of unhydrated 1a and 5a are not visible due to their very low signal intensities.
NAꢁꢂRE CꢃEMꢄSꢁRꢅ | www.nature.com/naturechemistry

Articles
NATure CHemisTry
0
.08
0.07
.06
0.05
.04
0.03
.02
0.01
.00
gas-phase or surface Cannizzaro reaction, which is typical for for-
After pyrolysis
After annealing
Hydroxyacetone reference
46,47
O
mose-type reactions , while the origin of 6 is currently unclear.
Hence, this kind of product identification complements the NMR
0
H
H
spectra in D O and provides firm evidence that the simplest sugar,
2
2a
1
a, can indeed form through the gas-phase reaction of 3a with its
H
isomer 2a without the need for water or a strong base as suggested
in Fig. 2. Higher sugars cannot be identified with this experimental
method because they are not volatile enough for the preparation of
an argon gas mixture.
0
O
O
3b
OH
H3C
H3C
H
H
0
1
c
As the NMR spectra in D O were dominated by POM, we adapted
2
our experimental set-up to remove POM and installed a second cool-
ing trap. The first cooling trap was held at –18°C so that most of the
0
1
13
2
a only condensed in the colder second cooling trap (for H and C
1
,290
1,285
1,280
1,275
1,270
1,265
1,260
1,255
NMR spectra see Fig. 3 and Supplementary Figs. 34–44). The main
products after 90min FP in these remarkably clean pyrolysis spectra
are the same as above, as is evident from a comparison with refer-
ence spectra and gas chromatograms (Supplementary Figs. 15–33).
~
–1
Wavenumber ν (cm
)
Fig. 4 | Reaction of methylhydroxycarbene (3b) with formaldehyde (2a)
to hydroxyacetone (1c). The black spectrum was obtained directly after
pyrolysis and trapping in the matrix. After annealing this matrix for 5ꢀmin at
8ꢀK the red spectrum was recorded after cooling back to 3ꢀK. Bands around
,260ꢀcm correspond to the C–O stretching vibration of 3b, whereas the
48
We also observed 2b, but its origin is currently unclear . The forma-
tion of 1a was confirmed by attached proton test (APT) and hetero-
nuclear single quantum coherence (HSQC) NMR spectral analysis
2
1
–
1
(
Supplementary Figs. 35–37) as well. The characteristic signals of
–1
large signal around 1,250ꢀcm on the right originates from the H–C–H
1
1
a in H NMR (Fig. 3a) are the singlet at 9.63ppm for the aldehyde
–1
rocking vibration of 2a. After annealing, the signal at 1,286ꢀcm increased
proton (a in Fig. 3a) and the singlet at 4.44ppm for the CH protons
2
significantly, belonging to the C–O–H deformation and CH wagging
2
(
b), which, however, show no coupling to each other. The character-
vibration of the newly formed 1c.
istic signals for the hydrated form of 1a are the doublet at 3.51ppm
for the CH protons (d) together with the triplet at 5.05ppm for the
2
C–H proton (c) of the hydrated carbonyl group. Note that there is continued iterative sugar synthesis from 1a and 3a should not lead
a signal overlap for proton c, distorting the expected characteristic to branched sugars (cf. Supplementary Fig. 65; see also the
13
triplet (see Supplementary Figs. 34–44 for further details). The C mechanistic discussion below).
APT signals (Fig. 3b) for hydrated 1a are at 66.4ppm (d in Fig. 3b,
We also performed analogous FP experiments with 4b to investi-
29
pointing downwards) for the CH OH carbon atom and at 91.6ppm gate the reaction of 3b with 2b and to demonstrated generality; for
2
(
c, pointing upwards) for the hydrated carbonyl carbon atom (the the crude NMR spectra of the FP see Supplementary Figs. 49 and 50.
hydrate dominates the equilibrium).
The main pyrolysis products of this reaction are acetoin (1b), 6 and
We also found weak remaining signals in the spectra that we 2b. GC–MS data (Supplementary Figs. 51 and 52) confirm the forma-
assigned to triose glyceraldehyde (5a). With longer pyrolysis times tion of 1b, which in all likelihood is the product of the bimolecular
(
3h) these signals increased (Fig. 3). To solidify evidence for the reaction of 3b and 2b (for computations see Supplementary Fig. 66).
formation of 5a we spiked our pyrolysis sample with 1mg and 5mg
We also experimented with a second reaction partner in the FP
of pure glyceraldehyde dimer, respectively (Fig. 3). The assigned cross-reaction experiments by investigating the reaction of 3b with
signals increased while all other signals remained unchanged. The acetone (7) as a model ketone. When 4b was pyrolysed in the pres-
1
characteristic signals of 5a in H NMR (Fig. 3a) are the singlet at ence of a large excess of 7 (for spectra, chromatograms and com-
9
.69ppm for the aldehyde proton (e) that overlaps with the quartet putations see Supplementary Figs. 53–57 and 67), the rather clean
of acetaldehyde (9.68ppm), the weak triplet around 4.43ppm for the formation of 3-methylacetoin (1d) indicates that these types of reac-
C–H proton (f) and a doublet around 3.98ppm for the CH protons tion indeed are general and very probably follow a common mecha-
2
(
g) that overlap with various other dimer structures of 1a and 5a. nism. Naturally, this reaction would constitute an entry point into
The preferred hydrated form of 5a shows two characteristic multi- the series of branched sugars, which are not discussed here.
plets around 3.70ppm for the i and j protons as well as the doublet
at 4.95ppm for the C–H proton (h) of the hydrated carbonyl group. Matrix isolation studies. We utilized cryogenic matrix isolation
For hydrated 5a we observed weak signals at 76.0ppm (Fig. 3b, i, spectroscopy to complement our FP–NMR findings because it
pointing upwards) for the CHOH carbon atom, at 63.9ppm (j, point- employs much more sensitive infrared spectroscopy of molecules
ing downwards) for the CH OH carbon atom and at 91.7 ppm (h, trapped in solid noble gases in the 3–28K range (mimicking inter-
2
pointing upwards) for the hydrated carbonyl carbon atom in the stellar conditions), thereby providing upper bounds for the acti-
1
3
C APT spectrum (Fig. 3b). These signals increased significantly in vation energy of our mechanistic proposal (Fig. 2). As several
the conducted NMR spiking experiment. Moreover, GC–MS data infrared signals overlap in the critical regimes for the products of
SupplementaryFigs.45–47)furtherconfirmedtheformationof5a, the parent reaction to 1a, we chose to investigate the cross-reaction
which in all likelihood is the subsequent iteration reaction prod- of 3b with 2a first.
(
uct of the bimolecular reaction of 3a and 1a (Fig. 1).
As before, 3b was generated by HVFP of 4b; the infrared spec-
29
Shevlin and colleagues identified a ratio of 11.5:1 for 5a to the tral data are in full accord with previous studies . In contrast to the
thermodynamically favoured 5b in the reaction of arc-generated FP experiments, no traces of acetoin 1b were found in the spectra
3
1
excited-state atomic carbon atoms with an excess of water , but the because the identification of 1b is hampered by several signal over-
observed reaction products must come from a higher energy regime laps as in the parent reaction. When the HVFP products are trapped
than the conditions reported in the present work where 5b did not in argon matrices doped with 1% 2a, new infrared signals belong-
form. The absence of 5b in our experiments was also confirmed by ing to hydroxyacetone (1c) appeared. After annealing the matrix
GC–MS (Supplementary Figs. 45–47) and in an analogous NMR for 5min at 28K to allow diffusion and thereby enable bimolecular
spiking experiment with pure 5b (Supplementary Figs. 42–44). reactions, the signals of 1c increased significantly (Fig. 4). That is,
As a consequence of the kinetic gas-phase conditions employed even at such a low temperature, the bimolecular reaction does occur;
–
1
here and the absence of water, isomerization is suppressed and the this puts an upper limit on the activation barrier of E ~1kcalmol .
a
NAꢁꢂRE CꢃEMꢄSꢁRꢅ | www.nature.com/naturechemistry

NATure CHemisTry
Articles
a
b
TSins
0.4
Alder (ene) reaction
1
H
H
O
+
O
OH
2
3
a
a
1
.76
CH2
0
2.99
TSene
–7.3
Cs complex
8.2
.77
–
Carbonyl-ene reaction for the formation of 1a
1
H
H
O
O
O
O
O
OH
3.29
1a
+
Glycolaldehyde 1a
CH2
H
–
73.3
H
3
a
2a
1a
1
/2
CCSD(T)/cc-pVQZ//CCSD(T)/cc-pVTZ
ξ (reaction coordinate) (a.m.u. bohr)
Fig. 5 | Reaction mechanism of the carbonyl–ene-type reaction. a, Computed potential energy hypersurface for the formation of glycolaldehyde (1a) in
the reaction of hydroxycarbene (3a) with formaldehyde (2a). Ab initio computations at the CCSD(T)/cc-pVQZ//CCSD(T)/cc-pVTZ+ZPVE level of theory
depict the proposed carbonyl–ene-type reaction via the corresponding transition structure (ꢁS ) that is associated with a vanishingly small barrier, as
ene
confirmed by our cryogenic matrix isolation experiments. b, Comparison of established and new reaction patterns. C indicates the point group.
s
The formation of 1c was also confirmed by experiments using Instead, 1c is the carbonyl–ene reaction product, in perfect agree-
1
3
formaldehyde isotopologues (H CO, H CO or D CO) reacting ment with computations (Supplementary Fig. 68).
2
2
2
1
3
with 3b and 1- C labelled 3bʹ. This led to a set of cross-exper-
In conclusion, we have presented the formation of the simplest
iments that can readily be analysed through their isotopic shifts sugars, glycolaldehyde and glyceraldehyde, via hydroxymethylene
in the infrared spectra, which can also be computed with high as Breslow’s ‘active formaldehyde ’. This was demonstrated through
confidence (see Supplementary Tables 3–6 for details). When a combination of quantitative flow pyrolysis and cryogenic matrix
the matrix was irradiated with a wavelength of λ=436 nm for at isolation experiments, followed by NMR and infrared spectroscopic
least 5 min after pyrolysis, 3b rearranged to 2b and vinyl alco- characterization, respectively. The experimental results were verified
2
9
hol (8), respectively . After annealing of the irradiated matrix by high-level ab initio CCSD(T)/cc-pVQZ coupled cluster compu-
no new bands were observed, confirming our proposal that 3b is tations. In the flow pyrolysis experiments with glyoxylic and pyru-
the reactive species in the formation of 1c. The observed infra- vic acid as the corresponding hydroxycarbene precursors, we show
red bands belonging to 1c (using pure reference material for the that dimerization of two aldehyde isomers (hydroxycarbene+alde-
assignments) are summarized in Supplementary Tables 2–6. The hyde) is possible and general. Furthermore, we show that formation
spectral changes during the bimolecular reaction triggered by the of the triose glyceraldehyde is kinetically favoured, and isomeriza-
annealing process can readily be followed through the intensity tion to the thermodynamically more stable dihydroxyacetone does
change of the C–O–H deformation and CH wagging vibration not occur in the gas phase. These findings are in stark contrast to
2
–
1
of 1c at 1,286 cm (Fig. 4). Additional bands and their changes the thermodynamic isomerizations found in the aqueous formose
upon reaction are depicted and described in the Supplementary reaction and should hamper the formation of branched sugars.
Information (Supplementary Figs. 59–64).
With the formation of hydroxyacetone from the cross-reaction
between methylhydroxycarbene and formaldehyde under matrix
Reaction mechanism. For the formation of 1a in the parent reac- isolation conditions we have demonstrated that hydroxycarbenes
tion of 3a with 2a, two concerted mechanisms have been consid- react with aldehydes in a nearly barrierless carbonyl–ene-type reac-
3
0
ered . First, a bond insertion of the hydroxycarbene centre into tion mechanism to the corresponding sugars. An alternative C–H
4
9
the C–H bond of 2a and, second, a carbonyl–ene-like allowed bond insertion of the hydroxycarbene into the C–H bond of the
27,30,31
five-centre–six-electron reaction (Fig. 5)
. The latter is aldehyde was not observed. The carbonyl–ene mechanism is also
apparently a variant of the established hetero-Alder (ene) reac- applicable to the reaction between hydroxycarbenes with ketones as
tion where a carbene lone pair (of 3) takes the place of the olefinic shown in flow pyrolysis experiments of pyruvic acid with an excess
double bond (Fig. 5b). This reduces the six-membered transition of acetone. With these experimental results in hand we provide an
structure to a five-membered one. Our ab initio computations at alternative concept for interstellar glycolaldehyde and subsequent
the CCSD(T)/cc-pVQZ//CCSD(T)/cc-pVTZ+ZPVE (zero-point sugar formation.
vibrational energy) level of theory indicate that the parent reaction
of 3a with 2a is extremely facile and associated with a vanishingly
Methods
‡
–1
small barrier (∆H =0.9kcalmol ), in perfect agreement with our
0
Matrix apparatus design. For all matrix isolation studies, a Sumitomo cryostat
experimental observations. C–H bond insertions common for sin- system consisting of a F-70 compressor unit and an RDK 408D2 closed-cycle
50
glet carbenes typically have much higher barriers . The reaction refrigerator was used. ꢀe vacuum shroud surrounding the coldhead was ꢁtted
with polished KBr windows, and the coldhead was equipped with a CsI window
barrier for the iterative sugar synthesis does not change significantly
for infrared measurements. Spectra were recorded with a Bruker Vertex 70 Fourier
(
Supplementary Fig. 65). The reaction barrier for the gas-phase
–1
transform infrared spectrometer with a spectral range of 4,000–350cm and a
–1
isomerization of 5a to 5b is 42.7kcalmol at CCSD(T)/cc-pVQZ//
MP2/aug-cc-pVTZ.
–1
resolution of 0.7cm , a SiC globar MIR light source, a KBr beamsplitter and a
DLaTGS detector. For a single measurement, a total of 50 scans were accumulated.
All measurements were performed at 3K. ꢀe temperature of the cold window was
measured with a Si diode connected to a Lakeshore 336 temperature controller.
For the combination of high-vacuum ꢂash pyrolysis and matrix isolation we
employed a small, home-built, water-cooled oven, which was directly connected
to the vacuum shroud of the cryostat. ꢀe pyrolysis zone consisted of an empty
The carbonyl–ene mechanism was confirmed by our FP and
matrix isolation cross-reaction experiments. In the FP experiments
with 3b and 7, no C–H insertion product was observed. Equally, not
even traces of the C–H insertion product lactaldehyde were detected
in the reaction of 3b with 2a under matrix isolation conditions. quartz tube with an inner diameter of 8mm, which was resistively heated over a
NAꢁꢂRE CꢃEMꢄSꢁRꢅ | www.nature.com/naturechemistry

Articles
NATure CHemisTry
length of 50mm by a coaxial wire. ꢀe temperature was monitored with a NiCr–Ni
thermocouple. ꢀe host gas ꢂow was regulated by a Pfeiꢃer EVN 116 gas dosing
valve with separate shut-oꢃ. For all experiments we used Ar of 99.999% purity.
Pyruvic acid (received from Sigma Aldrich, further purified by distillation, and
degassed by repeated freeze–pump–thaw cycles) was evaporated at –20°C from a
storage bulb into the hot quartz pyrolysis tube (900°C). At a distance of ~70mm,
all pyrolysis products were co-condensed with a large excess of Ar containing
9. Cooper, G. et al. Carbonaceous meteorites as a source of sugar-related
organic compounds for the early Earth. Nature 414, 879–883 (2001).
10. Kebukawa, Y., Kilcoyne, A. L. D. & Cody, G. D. Exploring the potential
formation of organic solids in chondrites and comets through polymerization
of interstellar formaldehyde. Astrophys. J. 771, 19 (2013).
11. Meinert, C. et al. Ribose and related sugars from ultraviolet irradiation of
interstellar ice analogs. Science 352, 208–212 (2016).
1% formaldehyde (typically 500–600mbar from a 2,000ml storage bulb) on the
12. Hollis, J. M., Lovas, F. J. & Jewell, P. R. Interstellar glycolaldehyde: the ꢁrst
sugar. Astrophys. J. 540, L107–L110 (2000).
13. Halfen, D. T., Apponi, A. J., Woolf, N., Polt, R. & Ziurys, L. M. A systematic
study of glycolaldehyde in Sagittarius B2(N) at 2 and 3 mm: criteria for
detecting large interstellar molecules. Astrophys. J. 639, 237–245 (2006).
14. Sutherland, J. D. & Whitꢁeld, J. N. Prebiotic chemistry: a bioorganic
perspective. Tetrahedron 53, 11493–11527 (1997).
15. Jortner, J. Conditions for the emergence of life on the early Earth: summary
and reꢂections. Phil. Trans. Royal Soc. Lond. B 361, 1877–1891 (2006).
16. Snyder, L. E., Buhl, D., Zuckerma., B. & Palmer, P. Microwave detection of
interstellar formaldehyde. Phys. Rev. Lett. 22, 679–681 (1969).
surface of the matrix window at 12K. Gaseous formaldehyde was prepared by
thermal decomposition of paraformaldehyde (received from Sigma Aldrich). After
deposition the matrix window was annealed at 28K for 5min to allow diffusion. A
high-pressure mercury lamp (HBO 200, Osram) with a monochromator (Bausch &
Lomb) was used for irradiation.
1
3
Analogous experiments were also performed with 1,2- C-pyruvic acid (99at%
1
3
13
13
C received from Sigma Aldrich), paraformaldehyde- C (99at% C received from
Cambridge Isotope Laboratories) and paraformaldehyde-d (98at% D, received
2
from Sigma Aldrich).
Flow pyrolysis apparatus design. The home-built flow pyrolysis set-up consisted
of four main compartments: sample reservoir, pyrolysis zone, collection area
17. Kebukawa, Y. & Cody, G. D. A kinetic study of the formation of organic
solids from formaldehyde: implications for the origin of extraterrestrial
organic solids in primitive Solar System objects. Icarus 248, 412–423 (2015).
18. Shigemasa, Y., Matsuda, Y., Sakazawa, C. & Matsuura, T. Formose reactions.
II. Photochemical formose reaction. Bull. Chem. Soc. Jpn 50, 222–226 (1977).
19. Sodeau, J. R. & Lee, E. K. C. Intermediacy of hydroxymethylene (HCOH) in
low-temperature matrix photochemistry of formaldehyde. Chem. Phys. Lett.
57, 71–74 (1978).
(
liquid-nitrogen-cooled cooling trap(s)) and vacuum system, as depicted in
Supplementary Fig. 1. The sample reservoir consisted of two storage flasks,
one orthogonal to the flow direction equipped with a needle valve to control
volatile reactants. The following pyrolysis zone consisted of a quartz tube with
a total length of 60.0cm and a diameter of 3.0cm with a cylindrical furnace
(
42.0cm). A second access supplied the system with deuterated solvents, Ar and
the possibility to provide reactants behind the pyrolysis zone. The whole system
ran on a Vaccubrand RZ6 rotary pump with a maximum level of pressure of
20. Maity, S., Kaiser, R. I. & Jones, B. M. Infrared and reꢂectron time-of-ꢂight
mass spectroscopic study on the synthesis of glycolaldehyde in methanol
−
4
about 4.0×10 mbar. The average pressure during the experiments was typically
(CH OH) and methanol-carbon monoxide (CH OH-CO) ices exposed to
3 3
−
2
2
.0×10 mbar.
ionization radiation. Faraday Discuss. 168, 485–516 (2014).
All chemicals were purchased from Sigma Aldrich and were purified through
21. Bennett, C. J. & Kaiser, R. I. On the formation of glycolaldehyde
distillation or sublimation. Degassed chemicals and Ar-flushed reaction set-ups
avoided oxidative conditions during the pyrolysis experiments. The flow pyrolysis
apparatus was evacuated over at least 60min with oven temperatures applied
during the experiments. The resulting pyrolysis reaction products were trapped
in one of the cooling traps. Pyrolysis experiments were conducted for all starting
materials and products to verify their stability under experimental conditions.
NMR spectra were recorded on Bruker AV400 and AV600 spectrometers at
2 3
(HCOCH OH) and methyl formate (HCOOCH ) in interstellar ice analogs.
Astrophys. J. 661, 899–909 (2007).
22. Eckhardt, A. K. & Schreiner, P. R. Spectroscopic evidence for
aminomethylene (H−C−NH )—the simplest amino carbene. Angew. Chem.
Int. Ed. 57, 5248–5252 (2018).
̈
2
23. Schreiner, P. R. et al. Capture of hydroxymethylene and its fast disappearance
through tunnelling. Nature 453, 906–909 (2008).
2
98K. Chemical shifts (δ) are given in ppm relative to tetramethylsilane (TMS,
δ=0.00ppm) as the internal standard or to the respective solvent residual
peaks (CDCl : δ=7.26 and 77.16ppm; DMSO-d : δ=2.50 and 39.52ppm). All
depicted C NMR spectra in D O were calibrated to trimethylsilylpropionate
24. Feng, R., Wesdemiotis, C. & McLaꢃerty, F. W. Gaseous negative ions from
neutral molecules and positive ions: new information for neutralization–
reionization mass spectrometry. J. Am. Chem. Soc. 109, 6521–6522 (1987).
25. Ley, D., Gerbig, D. & Schreiner, P. R. Tunnelling control of chemical
reactions—the organic chemist’s perspective. Org. Biomol. Chem. 10,
3781–3790 (2012).
3
6
1
3
2
(
TSP, δ=1.7ppm) as external standard. GC–MS was carried out on Hewlett
Packard 5890 or Agilent Technologies 7820A gas chromatographs with Hewlett
Packard 5971 or Agilent Technologies 5977B mass selective detectors (EI, 70eV),
respectively, equipped with J&W Scientific fused-silica DB-5MS or HP-5MS
columns (30m×0.25mm).
26. Schreiner, P. R. Tunneling control of chemical reactions: the third reactivity
paradigm. J. Am. Chem. Soc. 139, 15276–15283 (2017).
27. Flanagan, G., Ahmed, S. N. & Shevlin, P. B. Formation of carbohydrates
in the reaction of atomic carbon with water. J. Am. Chem. Soc. 114,
3892–3896 (1992).
Computations. Details of computations are provided in the Supplementary
Information.
28. Schreiner, P. R. & Reisenauer, H. P. ꢀe ‘non-reaction’ of ground-state triplet
carbon atoms with water revisited. ChemPhysChem 7, 880–885 (2006).
29. Schreiner, P. R. et al. Methylhydroxycarbene: tunneling control of a chemical
reaction. Science 332, 1300–1303 (2011).
Data availability. All relevant data generated and analysed during this study,
including infrared, NMR and GC–MS spectra and chromatograms and optimized
coordinates for all discussed compounds, are included in this Article and
its Supplementary Information, and are also available from the authors upon
reasonable request.
30. Kemper, M. J. H., Hoeks, C. H. & Buck, H. M. A theoretical study on the
2
reactivity and spectra of H CO and HCOH. A dimeric model for non-zero
pressure formaldehyde photochemistry. J. Chem. Phys. 74, 5744–5757 (1981).
1. Ahmed, S. N., McKee, M. L. & Shevlin, P. B. ꢀe unusual reactivity of
hydroxymethylene. J. Am. Chem. Soc. 107, 1320–1324 (1985).
3
3
3
Received: 22 October 2017; Accepted: 26 July 2018;
Published: xx xx xxxx
2. Cleaves, H. J. ꢀe prebiotic geochemistry of formaldehyde. Precamb. Res. 164,
1
11–118 (2008).
3. Pinto, J. P., Gladstone, G. R. & Yung, Y. L. Photochemical production of
References
formaldehyde in earths primitive atmosphere. Science 210, 183–184 (1980).
1
2
3
4
5
6
7
.
.
.
.
.
.
.
Breslow, R. On the mechanism of the formose reaction. Tetrahedron Lett. 1,
34. Lucchese, R. R. & Schaefer, H. F. Metal–carbene complexes and possible role
of hydroxy-carbene in formaldehyde laser photochemistry. J. Am. Chem. Soc.
100, 298–299 (1978).
22–26 (1959).
Butlerow, A. Bildung einer zuckerartigen Substanz durch Synthese. Justus
Liebigs Ann. Chem. 120, 295–298 (1861).
35. Hibbitts, D., Dybeck, E., Lawlor, T., Neurock, M. & Iglesia, E. Preferential
activation of CO near hydrocarbon chains during Fischer–Tropsch synthesis
on Ru. J. Catal. 337, 91–101 (2016).
Schwartz, A. W. & Degraaf, R. M. ꢀe prebiotic synthesis of carbohydrates—a
reassessment. J. Mol. Evol. 36, 101–106 (1993).
Socha, R. F., Weiss, A. H. & Sakharov, M. M. Auto-catalysis in the formose
reaction. React. Kinet. Catal. Lett. 14, 119–128 (1980).
36. Deng, J. et al. Linked strategy for the production of fuels via formose
reaction. Sci. Rep. 3, 1244 (2013).
Huskey, W. P. & Epstein, I. R. Auto-catalysis and apparent bistability in the
formose reaction. J. Am. Chem. Soc. 111, 3157–3163 (1989).
Shapiro, R. Prebiotic ribose synthesis—a critical analysis. Orig. Life Evol.
Biosph. 18, 71–85 (1988).
37. Schäfer, M. et al. Hydrogen tunneling above room temperature evidenced by
infrared ion spectroscopy. J. Am. Chem. Soc. 139, 5779–5786 (2017).
38. Breslow, R. On the mechanism of thiamine action. IV. Evidence from studies
on model systems. J. Am. Chem. Soc. 80, 3719–3726 (1958).
39. Ritson, D. & Sutherland, J. D. Prebiotic synthesis of simple sugars by
photoredox systems chemistry. Nat. Chem. 4, 895–899 (2012).
40. Ritson, D. J. & Sutherland, J. D. Synthesis of aldehydic ribonucleotide and
amino acid precursors by photoredox chemistry. Angew. Chem. Int. Ed. 52,
5845–5847 (2013).
Larralde, R., Robertson, M. P. & Miller, S. L. Rates of decomposition of ribose
and other sugars—implications for chemical evolution. Proc. Natl Acad. Sci.
USA 92, 8158–8160 (1995).
Toxvaerd, S. ꢀe role of carbohydrates at the origin of homochirality in
biosystems. Orig. Life Evol. Biosph. 43, 391–409 (2013).
8.
NAꢁꢂRE CꢃEMꢄSꢁRꢅ | www.nature.com/naturechemistry

NATure CHemisTry
Articles
4
1. Paul, M. et al. Breslow intermediates from aromatic N-heterocyclic carbenes
benzimidazolin-2-ylidenes, thiazolin-2-ylidenes). Angew. Chem. Int. Ed. 57,
310–8315 (2018).
Acknowledgements
(
8
This work was supported by the Volkswagen Foundation (‘What is Life’ grant 92 748),
the Fonds der Chemischen Industrie (doctoral fellowship to A.K.E.) and the Justus Liebig
University (graduate fellowship to M.M.L.). The authors thank H. Hausmann for support
with the NMR spectroscopic measurements.
4
4
4
2. Poust, S. et al. Mechanistic analysis of an engineered enzyme that catalyzes
the formose reaction. ChemBioChem 16, 1950–1954 (2015).
3. Woods, P. M. et al. On the formation of glycolaldehyde in dense molecular
cores. Astrophys. J. 750, 19 (2012).
Author contributions
4. Angyal, S. J. in Glycoscience: Epimerisation, Isomerisation and rearrangement
reactions of carbohydrates Vol. 215 (ed. Stütz, A. E.) 1–14 (Springer,
Berlin, 2001).
A.K.E. conducted all matrix isolation experiments and carried out all computations.
A.K.E., M.M.L., and B.B. carried out all FP experiments. A.K.E. and R.C.W. conducted all
data analysis. P.R.S. conceived the original working hypothesis. A.K.E. and P.R.S. wrote
the manuscript.
4
5. Kua, J., Galloway, M. M., Millage, K. D., Avila, J. E. & De Haan, D. O.
Glycolaldehyde monomer and oligomer equilibria in aqueous solution:
comparing computational chemistry and NMR data. J. Phys. Chem. A 117,
2997–3008 (2013).
Competing interests
4
4
6. Sutherland, J. D. ꢀe origin of life—out of the blue. Angew. Chem. Int. Ed. 55,
The authors declare no competing interests.
104–121 (2016).
7. Delidovich, I. V., Simonov, A. N., Taran, O. P. & Parmon, V. N. Catalytic
formation of monosaccharides: from the formose reaction towards selective
synthesis. ChemSusChem 7, 1833–1846 (2014).
Additional information
Supplementary information is available for this paper at https://doi.org/10.1038/
s41557-018-0128-2.
4
4
5
8. Schwartz, A. W. & Degraaf, R. M. Photoreductive formation of acetaldehyde
from aqueous formaldehyde. Tetrahedron Lett. 34, 2201–2202 (1993).
9. Clarke, M. L. & France, M. B. ꢀe carbonyl ene reaction. Tetrahedron 64,
Reprints and permissions information is available at www.nature.com/reprints.
Correspondence and requests for materials should be addressed to P.R.S.
9003–9031 (2008).
0. Bourissou, D., Guerret, O., Gabbaï, F. P. & Bertrand, G. Stable carbenes.
Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in
Chem. Rev. 100, 39–92 (2000).
published maps and institutional affiliations.
NAꢁꢂRE CꢃEMꢄSꢁRꢅ | www.nature.com/naturechemistry