An Unusual Iron-Dependent Oxidative Deformylation Reaction Providing Insight into Hydrocarbon Biosynthesis in Nature

Article abstract of DOI:10.1021/acscatal.6b00592

The development of new pathways for next-generation biofuel production has spurred recent investigations into the mechanisms of enzymes that biosynthesize hydrocarbons. One widely distributed group of enzymes, aldehyde decarbonylases, catalyze unusual deformylation reactions in which long-chain fatty aldehydes are converted to alkanes. These enzymes are all iron-dependent and in insects are represented by a cytochrome P450 enzyme that releases the aldehyde carbon as CO₂. Here we describe a novel nonenzymatic reaction of an α-cyclopropyl-substituted aldehyde that mimics the enzyme reaction. This aldehyde is oxidatively deformylated in the presence of aqueous iron(II) salts and oxygen to yield an alkyl-substituted cyclopropane and CO₂, in a reaction that competes with the more conventional oxidation of the aldehyde to the carboxylic acid. Like the enzymatic reaction, the reaction occurs with retention of the aldehyde proton in the alkane product and probably proceeds through an iron-peroxo species. Computational reaction discovery tools were used to search for potential reaction pathways and investigate their energetic feasibility. These identified a plausible reaction pathway leading to the experimentally observed products and reproduced the transfer of the aldehyde proton to the cyclopropane product. These studies provide further insight into how enzymes may control reactive iron-oxo species to catalyze the diverse range of iron-dependent oxidative transformations observed in biology.

Full text of DOI:10.1021/acscatal.6b00592

Research Article
pubs.acs.org/acscatalysis
An Unusual Iron-Dependent Oxidative Deformylation Reaction
Providing Insight into Hydrocarbon Biosynthesis in Nature
Benjamin R. Ellington,† Bishwajit Paul,‡,§ Debasis Das,‡ Andrew K. Vitek,‡ Paul M. Zimmerman,‡
and E. Neil G. Marsh*,†,‡
‡
†Department of Biological Chemistry and Department of Chemistry, University of Michigan, Ann Arbor, Michigan 48109, United
States
S
* Supporting Information
ABSTRACT: The development of new pathways for next-
generation biofuel production has spurred recent investiga-
tions into the mechanisms of enzymes that biosynthesize
hydrocarbons. One widely distributed group of enzymes,
aldehyde decarbonylases, catalyze unusual deformylation
reactions in which long-chain fatty aldehydes are converted
to alkanes. These enzymes are all iron-dependent and in
insects are represented by a cytochrome P450 enzyme that
releases the aldehyde carbon as CO2. Here we describe a novel
nonenzymatic reaction of an α-cyclopropyl-substituted alde-
hyde that mimics the enzyme reaction. This aldehyde is
oxidatively deformylated in the presence of aqueous iron(II)
salts and oxygen to yield an alkyl-substituted cyclopropane and
CO2, in a reaction that competes with the more conventional oxidation of the aldehyde to the carboxylic acid. Like the enzymatic
reaction, the reaction occurs with retention of the aldehyde proton in the alkane product and probably proceeds through an
iron−peroxo species. Computational reaction discovery tools were used to search for potential reaction pathways and investigate
their energetic feasibility. These identified a plausible reaction pathway leading to the experimentally observed products and
reproduced the transfer of the aldehyde proton to the cyclopropane product. These studies provide further insight into how
enzymes may control reactive iron−oxo species to catalyze the diverse range of iron-dependent oxidative transformations
observed in biology.
KEYWORDS: aldehyde decarbonylase, enzyme models, isotope effects, computational reaction discovery, nonheme iron oxygenases,
cytochrome p450
complex, integral membrane proteins that have proved
recalcitrant to expression in active form and remain poorly
understood.2
INTRODUCTION
■
Growing interest in renewable energy sources has resulted in
the identification and characterization of various biosynthetic
pathways involved in hydrocarbon synthesis. One such pathway
identified in cyanobacteria,1,2 plants,2,3 and insects4−6 generates
aliphatic hydrocarbons from fatty acids, which serve as
waterproofing agents in plants2,3 and contact pheromones in
insects.5 The final enzymes in these pathways catalyze the
unusual deformylation of Cn aldehydes to generate Cn−1 alkanes
and carbon monoxide (plants),3 carbon dioxide (insects),5 or
formate (cyanobacteria).2,7,8 (Here we use the term deformy-
lation to describe these reactions; the term decarbonylation is
also commonly used in the enzyme literature.) Interestingly
these are all iron-dependent enzymes: the insect decarbonylase
is a cytochrome P450 enzyme, CYP4G1,5 whereas the plant
and cyanobacterial decarbonylases are nonheme iron en-
zymes.2,7,8 The cyanobacterial decarbonylase, also called
aldehyde deformylating oxygenase (cADO), is a soluble protein
for which several crystal structures have been solved and
various mechanistic investigations have been undertaken in
recent years.2,7−17 The plant and insect enzymes are more
The mechanisms of both P45018 and nonheme iron
oxygenases19 have been extensively studied. A hallmark of
these enzymes is the wide variety of oxidative transformations
that they catalyze, including hydroxylations,20 decarbon-
ylations,1 demethylations,21 and epoxidations,22,23 among
others. How enzymes discriminate among the different reaction
pathways available to the reactive high-valent iron−oxo species
that are initially generated at the active site continues to
challenge our understanding of their mechanisms. For the
aldehyde decarbonylase enzymes, this discrimination involves
directing the reactive iron−oxygen species toward C−C bond
cleavage to effect removal of the aldehyde carbon rather than
simply effecting C−H bond cleavage that would result in
oxidation of the aldehyde to a carboxylic acid.2,7−17
Received: February 26, 2016
Revised: April 5, 2016
© XXXX American Chemical Society
3293
DOI: 10.1021/acscatal.6b00592
ACS Catal. 2016, 6, 3293−3300

ACS Catalysis
Research Article
Figure 1. Reaction of 1 with Fe2+ and O2 resulting in deformylation: (A) scheme depicting reaction of 1 with Fe2+ and O2; (B) GCMS trace
demonstrating conversion of 1 to 1,2-cyclopropylundecane (2); (C) dependence of the formation of 2 on [Fe2+]; (D) dependence of the formation
of 3 on [Fe2+].
Studies of metal complexes that mimic the active sites of
enzymes have played an important role in elucidating the
mechanisms of both nonheme12,24−27 and P45028 iron
oxygenases: for example, in probing the ability of different
iron−oxo intermediates12,25,26,29 and different spin states24,27 to
catalyze nucleophilic deformylations and electrophilic oxida-
tions of aldehydes (and other substrates). Here we describe an
unanticipated reaction involving the oxidative deformylation of
an α-cyclopropyl aldehyde to produce CO2 and a cyclopropyl
alkane that is catalyzed by aqueous Fe2+ ions in the presence of
oxygen or by Fe3+ ions in the presence of H2O2. The reaction
mimics the oxidative deformylation catalyzed by the insect
P450 aldehyde decarbonylase.5,6 The presence of the cyclo-
propyl group is essential for the deformylation, as simple n-
aldehydes only undergo oxidation to carboxylic acids under
these conditions. An atomistic reaction mechanism determined
by computational reaction discovery tools shows that C−C
bond breaking is favored in α-cyclopropyl in comparison to
alkyl aldehydes, providing a plausible explanation for the
observed reactivity.
Initial studies on the reaction of 1 with cADO suggested that
the compound might undergo deformylation, as indicated by
the appearance of a peak at 5.9 min that coeluted with an
authentic standard of 1,2-cyclopropylundecane (2) (Figure 1).
However, subsequent control experiments established that the
deformylation reaction did not depend on cADO and, in fact,
required only aqueous Fe2+ and O2 to occur. (Typically
Fe(NH3)6SO4 was used in these experiments, although other
ferrous salts, e.g. FeCl2, were equally effective.) The auxiliary
reducing system, which is necessary for cADO activity, was
found to inhibit the nonenzymatic reaction.
The deformylation of 1 occurred in competition with the
more conventional oxidation of the aldehyde group to the
corresponding 2,3-cyclopropyldodecanoic acid (3). In H2O, the
ratio of alkane 2 to carboxylic acid 3 formed was typically ∼1:4.
Further control experiments using dodecanal, α-dodecenal, and
α-oxiranyldodecanal failed to detect any n − 1 alkane products
under the same reaction conditions. Instead these aldehydes, as
expected, were exclusively oxidized to dodecanoic acid, α-
dodecenoic acid, and α-oxiranyldodecanoic acid, respectively
(see the Supporting Information). These results pointed to the
unusual role of the cyclopropyl group in directing the oxidative
chemistry of 1 toward deformylation.
We further investigated the stoichiometry with which 2 and 3
were formed with respect to the amount of Fe2+ in the reaction
(Figure 1C,D). Reactions containing 500 μM 1 and 20−500
μM FeSO4(NH3)6 were prepared microaerobically in 10 mM
sodium acetate buffer (pH 5.0) and then exposed to air before
capping and shaking at 37 °C and 200 rpm for 15 min. The
organic compounds were then extracted and analyzed as
described in Experimental and Computational Methods. These
experiments revealed that both 2 and 3 were formed in linear
proportion to [Fe2+] until the concentration exceeded 100−200
μM, after which the yields leveled off with the balance of
material recovered as unreacted 1. The leveling off is likely due
to depletion of oxygen in the reaction. Interestingly, two Fe2+
RESULTS
■
Recent studies in our laboratory have used α-oxiranyl- and β-
cyclopropyl-substituted long-chain aldehydes to probe the
mechanism of cADO.10,11 These sterically undemanding
substrate modifications provided valuable insight into the
mechanism of deformylation, pointing to a radical mechanism
for C−C bond cleavage.11 Therefore, we were motivated to
synthesize the α-cyclopropyl-substituted dodecanal derivative 1
to further probe the enzyme mechanism. We reasoned that
introducing a cyclopropyl group at the site of C−C bond
scission would significantly destabilize any radical intermediate
potentially generated on C2, so that 1 might function as either
an inhibitor or a very slow substrate.
Reaction of α-Cyclopropyl-Aldehyde with Iron and
Oxygen. Compound 1 was synthesized by standard methods,
as described in Experimental and Computational Methods.
3294
DOI: 10.1021/acscatal.6b00592
ACS Catal. 2016, 6, 3293−3300

ACS Catalysis
Research Article
Figure 2. pH dependence for formation of compounds 2 (A; fitted to a single pKa) and 3 (B) from the reaction of 1 with Fe2+ and O2.
ions were required per molecule of 2 formed, whereas about 2
equiv of carboxylic acid 3 was formed per Fe2+ ion.
situ by 13C NMR. The only 13C-labeled material evident from
the spectrum was the carboxylic acid arising by oxidation of the
aldehyde; therefore, it seemed likely that the aldehyde carbon
was converted to CO2 in the reaction. To examine this
possibility, the reaction was conducted in sealed flasks and the
headspace analyzed by HRMS. The reaction of unlabeled 1
resulted in a ∼2.5-fold increase in CO2 above background,
whereas reaction of 13C-labeled 1 resulted in the appearance of
the expected 13CO2 peak at m/z 45 (Figure 3). Interestingly,
the production of CO2 as a byproduct mimics the poorly
understood insect P450 enzyme (CYP4G1) that catalyzes the
oxidative decarbonylation of fatty aldehydes.4,5
The stoichiometry with which 3 was formed is in accord with
the well-documented ability of transition-metal ions, including
iron, to catalyze the oxidation of aldehydes to carboxylic
acids.30,31 However, the observation that formation of 2
required 2 equiv of Fe2+ suggested that Fe2+ might be acting
to reduce O2 to hydrogen peroxide and that the deformylation
of 1 may occur through a metal peroxide intermediate. To test
this, assays were conducted under anaerobic conditions in
which 300 μM of 1 was reacted with 100 μM of FeCl3 and 1−
10 mM hydrogen peroxide. Under these conditions 1 was
converted to 2 and 3 in approximately the same ratios as
observed with Fe2+ and O2. This observation indicates that an
iron-bound peroxide species is likely the oxidant in the reaction.
Although Fe3+ and hydrogen peroxide were shown to function
in the reaction, experimentally it proved easier to conduct
further studies on the reaction using Fe2+ and O2 as the oxidant.
pH Dependence. To obtain further information about the
conditions leading to the oxidation of 1 to form 2 and 3, we
studied the pH dependence of the reaction over the pH range
4−10. The formation of both products was favored by slightly
acidic conditions, although the pH dependences of the
reactions were not identical. The formation of 2 was described
by a sigmoidal pH curve that was well fitted by a single
apparent pKa of 7.2 (Figure 2A). However, the relationship of
this pH dependence to the mechanism proposed below, which
is based on the isotopic labeling experiment and computer
simulations described later, is currently unclear. We speculate
that this apparent pKa may reflect the pKa of a water ligand to
iron, as water bound to Fe3+ would be expected to have a pKa in
this range. Deprotonation of the aqueous iron species would
lead to the formation of iron(III) hydroxide, which would likely
be inactive in the reaction.
In contrast, the formation of 3 exhibited more complex pH-
dependent behavior (Figure 2B). The yield of 3 gradually
declined between pH 5 and 8, with the reaction barely
proceeding above pH 9. It is likely, as discussed below, that
there may be more than one mechanism for the oxidation of 1
to 3, which have different pH dependences. This would be in
accord with the ability of iron and other transition-metal
complexes to catalyze the oxidation of aldehydes to carboxylic
acids in both their 2+ and 3+ oxidation states.30,31
Fate of Aldehyde Carbon. To determine the fate of the
carbonyl carbon, 1 was synthesized with the carbonyl carbon
13C-labeled and reacted with Fe2+ and O2 in water containing
10% D2O. The products of the reaction were then analyzed in
Figure 3. Normalized total ion counts of m/z 44 and m/z 45 peaks,
representing CO2 and 13CO2, recovered from headspace above
reactions: (A) 0 μM 1, 1000 μM Fe2+; (B) 500 μM unlabeled 1 and
1000 μM Fe2+; (C) 500 μM 13C-labeled 1 and 1000 μM Fe2+. Peaks
were normalized to atmospheric N2 ion counts.
Fate of Aldehyde Proton. Experiments on the insect P450
enzyme indicated that the aldehyde hydrogen was retained in
the n − 1 alkane.4,5 To determine the fate of the aldehyde
hydrogen in the reaction of 1, we initially synthesized 1 with
the aldehyde proton deuterium-labeled. However, reaction of
deuterium-labeled 1 with Fe2+ and O2 resulted in the
compound being converted almost entirely to the carboxylic
acid 3, with negligible formation of 2. This indicated that a
significant kinetic isotope effect was operating against the
formation of 2, thereby favoring the alternate reaction pathway
leading to the formation of 3. We note that although the
formations of both 2 and 3 require cleavage of the aldehyde C−
H bond and would therefore be expected to exhibit kinetic
isotope effects, the sizes of these effects could be intrinsically
different. Presumably, the formation of 2 is either associated
with a larger kinetic isotope effect than the formation of 3 or
the isotope effect on the formation of 3 is suppressed by
another isotope-insensitive step. The unfavorable isotopic
3295
DOI: 10.1021/acscatal.6b00592
ACS Catal. 2016, 6, 3293−3300

ACS Catalysis
Research Article
branching made it unfeasible to examine the deuterium content
of 2 by NMR or MS. We therefore adapted our approach by
synthesizing triply deuterated 1 (referred to as d3-1), in which
the α-carbon and the bridging cyclopropyl methylene were
deuterated (Figure 4A). This allowed us to test, instead,
whether protium was transferred from the aldehyde carbon.
d3-1 was reacted with Fe2+ and O2 in either H2O or D2O and
the product, 2, recovered and analyzed by H NMR. The
1
spectra, shown in Figure 4, were very similar whether the
reaction was carried out in H2O or D2O. Although small
differences in the integration of the peaks are evident, we do
not consider these to be meaningful, as there are also
discrepancies in the integration of the peaks in the reference
compound. The reason for this is unclear. The spectra indicate
that the proton in the product was transferred from the
aldehyde rather than the solvent. Moreover, protonation
appeared to be equally favored on either the trans or cis face
of the cyclopropyl ring. This suggests a mechanism in which
C−C bond scission precedes hydrogen transfer.
Computational Search of Potential Reaction Path-
ways. The experimental observations described above
presented a challenge to rationalize mechanistically. Therefore,
to gain further insight into the atomistic details of this unusual
reaction, we turned to computational simulations. Reaction
discovery methods25,26 were used to search for potential
pathways leading to the observed reaction products for 1.
These methods allow for exploration and evaluation of
chemically reasonable elementary steps and produce active
reaction paths by identifying the most kinetically and
thermodynamically feasible elementary steps (see the Support-
ing Information for additional details of these methods). On
the basis of the experimental observations above, we assumed
the first step in the reaction to be the formation an iron-bound
peroxo species, for which there is evidence in the mechanism of
cADO.16 This species could be formed by the reaction of O2
with Fe2+, followed by reduction by a second Fe2+ to generate
an aqueous Fe3+−OO− species, as discussed above. The iron−
peroxo species subsequently undergoes nucleophilic addition to
the aldehyde to form the metal peroxyhemiacetal species
represented by structure A1 in Figure 5. We therefore started
with a model comprising the (H2O)5Fe3+−OOR complex A1
Figure 4. Stereorandom transfer of the aldehyde proton of 1 to the
cyclopropyl alkane product 2: (A) general reaction scheme showing
deuterium labeling employed; (B) dideuterated 1-nonylcyclopropane
standard; (C) d3-labeled 1 reacted in H2O; (D) d3-labeled 1 reacted in
D2O. The protium content at HT and HC does not depend on the
isotopic composition of the solvent.
Figure 5. Deformylation and oxidation mechanisms for reaction of 1 with aqueous iron−hydroperoxide complex predicted by computational
reaction pathway discovery: (A) mechanistic scheme for the oxidative deformylation of 1; (B) mechanistic scheme for a branching pathway leading
to aldehyde oxidation. Note that D2 can also form directly through Fe−OO−, as previously described in the literature.30,31
3296
DOI: 10.1021/acscatal.6b00592
ACS Catal. 2016, 6, 3293−3300

ACS Catalysis
Research Article
Figure 6. Potential energy surface for deformylation of α-cyclopropyl substrate in the hextet spin state.
(in which the alkyl chain of 1 is abbreviated as a methyl group)
to represent the first key intermediate in the reaction.
Despite little intuition about how the mechanism might
proceed, the computational reaction search identified a
plausible mechanism for deformylation, shown schematically
in Figure 5. This mechanistic pathway represents the most
kinetically and thermodynamically feasible pathway that was
located for deformylation, identified from among over 3000
elementary reaction steps that were screened (see the
Supporting Information for further details). The potential
energy surface for deformylation of A1 and the computed
structures of the intermediates and transition states in the
mechanism are shown in Figure 6. The reaction search also
identified a pathway from A1 leading to carboxylic acid
formation, as discussed below. However, we note that there are
other pathways to carboxylic acid formation, which have been
described in the literature30,31 and which involve direct
abstraction of the aldehyde hydrogen by Fe3+−OO−. We
expect that these pathways also contribute to carboxylic acid
formation in the present case.
Interestingly, the reaction pathway search identified the
configuration about the C−O bond as playing an important
role in the partitioning of the reaction between deformylation
and oxidation. Starting from the metal−peroxyhemiacetal
structure A1 (Figure 5), rotation about the C−O bond aligns
the acetal’s hydroxyl group close to the Fe−O peroxide bond to
yield configuration B1. Next, homolytic fission of the peroxide
O−O bond generates the high-valent iron−oxo species and the
acetal radical (structure C1). From C1, deformylation begins by
abstraction of the acidic proton by the high-valent iron−oxo
species. The resulting species then fragments to generate a
formyl radical and a reactive cyclopropyl radical (Figure 6,
TSCD1), which immediately abstracts the formyl hydrogen to
generate CO2 and 2.
The minimum energy paths for the deformylation process
show TSCD1 to be concerted and asynchronous (see Figure S24
in the Supporting Information). Importantly, the concerted
nature of TSCD1 means that the aldehyde hydrogen is retained
in the cyclopropyl ring. We note, however, that experimentally
hydrogen transfer was found to be stereorandom, suggesting
that a very short lived cyclopropyl radical may be an
intermediate, rather than the concerted transfer pathway
identified by computation. The potential energy surface for
the deformylation reaction mechanism (Figure 6) indicates that
the rate-limiting step for deformylation of 1 is computed to be
the aldehyde hydrogen abstraction (TSCD1 at 14.4 kcal/mol).
This agrees with the experimental finding that C−H scission is
rate-limiting, as indicated by a large deuterium isotope effect.
Whereas the steps leading to the peroxo species A1 and the
subsequent formation of C2 and D2 are well prece-
dented,10,16,32 the reaction of C1 to D1 is very unusual.
These steps, however, are supported by the deuterium-labeling
results described above and, significantly, were identified
independently of the experimental results using reaction
discovery simulations. For example, the reaction search was
not constrained by the requirement that aldehyde hydrogen be
retained in the product alkane or even that the reaction would
proceed to deformylation.
The reaction pathway searches found an exactly analogous
deformylation pathway that appeared to be feasible for n-
propanal (the model for dodecanal). The n-propanal
deformylation pathway was found to have a rate-limiting
barrier 3.0 kcal/mol higher in energy than the cyclopropyl-
forming pathway (see the Supporting Information for full
details). The higher barriers indicate that deformylation is
slower for the noncyclopropyl substrate, where carboxylic acid
formation will be more competitive. Further studies were also
undertaken to calculate the potential energy surface for the
individual steps in the hypothesized mechanism using various
density functionals (see the Supporting Information). These
studies supported the feasibility of the pathways identified in
the reaction search, and all density functionals used in this
study agree that the high-spin hextet Fe(III) pathway of Figure
6 is plausible and consistent with the experimental results.
Alternatively, to effect carboxylic acid formation, rotation
about the C−O bond in A1 can also bring the aldehyde
hydrogen in close proximity to the Fe−O peroxide bond
(configuration B2). From the B2 configuration, homolysis of
the peroxide O−O bond leads to C−H abstraction and
consequent oxidation of the aldehyde to the carboxylic acid 3
(structures C2 and D2).
3297
DOI: 10.1021/acscatal.6b00592
ACS Catal. 2016, 6, 3293−3300

ACS Catalysis
Research Article
(Figure 5) is likely insufficiently reactive to undergo C−C bond
cleavage to release formate, as occurs in the reaction of
cyclohexane carbaldehyde.34 Instead, hydrogen abstraction
from the acetal oxygen by the high-valent iron−oxo species is
presumed to provide the energetic driving force for the
reaction. The acetal diradical species thus formed would be
expected to break down extremely rapidly to give formyl and
cyclopropyl radicals, at which point the cyclopropyl radical
would be perfectly positioned to abstract the formyl hydrogen.
This would result in the energetically favorable formation of
CO2 and transfer of the aldehyde hydrogen to the cyclopropyl
group.
DISCUSSION
■
The deformylation reaction we identified for 1 represents a
highly unusual mode of oxidation for an aldehyde, because in
this case the aldehyde carbon is completely oxidized to CO2
while the α-carbon is reduced by transfer of the aldehyde
hydrogen. The reaction mimics that catalyzed by insect P450
aldehyde decarbonylase, CYP4G1, and it is somewhat
surprising that a very simple oxidizing systemeither aqueous
Fe2+ and O2 or Fe3+ and H2O2is able to recapitulate the
reaction of a rather complex enzyme.4,5 To our knowledge this
is the first nonenzymatic example of this type of oxidative
deformylation of an aldehyde to generate CO2 and an alkane.
An obvious question is as follows: how does appending a
simple cyclopropyl ring adjacent to the aldehyde carbon alter
the oxidation pathway? The oxidation of aldehydes to
carboxylic acids in the presence of oxygen and transition-
metal complexes30,31 is by far the more common pathway and,
indeed, predominates in the reaction of 1. These reactions have
been extensively studied and shown to proceed by various
mechanisms involving reactive oxygen species.31,33 The key
step in the reaction is abstraction of the aldehyde hydrogen
atom by a reactive oxygen species, e.g. metal−peroxo or metal−
oxo, to generate an acyl radical.30,31,33 For the observed
deformylation of 1 to occur, this reaction pathway must
somehow be suppressed.
A less commonly encountered mode of aldehyde reaction is
oxidative decarbonylation to release of the aldehyde carbon as
formate. A well-studied example is the deformylation of
cyclohexane carbaldehyde by metal peroxides that has been
shown to proceed by an eliminative mechanism that results in
the formation of cyclohexene and formate.34 Enzymatic
reactions that liberate formate include the P450-dependent
deformylation of cyclohexane carbaldehyde catalyzed by
CYP2B4 and the aromatization reaction of androst-4-ene-
3,17-dione to estrone catalyzed by aromatase.35,36 Under
reductive conditions this deformylation pathway can lead to
the formation of alkanes, rather than alkenes. This was recently
demonstrated for an iron(III)−peroxo complex that under
reducing conditions effectively deformylated phenylpropional-
dehyde to yield formate and ethylbenzene.12 This reaction
mimics that of cyanobacterial aldehyde deformylating oxygen-
ase (cADO), in which the intermediate alkyl radical, formed
after release of the aldehyde carbon as formate, is reduced to
form the alkane product.1,2,10,11
We propose that the cyclopropyl group may influence the
reaction pathway in two ways. First, the well-established
electron-withdrawing nature of the cyclopropyl group37 would
make the aldehyde more electrophilic, thereby facilitating the
formation of the initial metal−peroxyhemiacetal complex A1 in
Figure 5. DFT calculations support this notion and indicate that
the transition state for metal-peroxyacetal formation is about
2.5 kcal mol−1 lower for 1 than for propanal (see the
Supporting Information). In contrast, metal−peroxyhemiacetal
formation is not required for carboxylic acid formation, as this
can proceed through direct abstraction of the aldehyde
hydrogen by the metal−peroxide complex.30,38
Second, the positioning of the cyclopropyl group does not
allow a radical β-elimination mechanism to operate that would
result in the formation of formate and an alkene, as is seen for
the oxidative deformylation of cyclohexane carbaldehyde.10
Furthermore, formation of the postulated cyclopropyl radical is
energetically very unfavorable; therefore, the formyl radical C1
CONCLUSIONS
■
Our studies on an α-cyclopropyl aldehyde originally designed
as a mechanistic probe of the reaction catalyzed by cADO have
uncovered a novel mode of a nonenzymatic deformylation that
occurs in the presence of Fe2+ and O2. Computational reaction
pathway searches identified a plausible mechanistic pathway
that is in accord with isotope-labeling studies. The discussion
above suggests how the cyclopropyl group may fine-tune the
energetics of the reaction to allow deformylation instead of only
oxidation. It may be possible to incorporate more sophisticated
stereoelectronic control into the design of either the substrate
aldehyde or iron complex to bias the reaction further in favor of
deformylation.
In the context of understanding enzymatic deformylation
reactions, we note that an enzyme could exert more precise
stereoelectronic control of the reaction pathway. First, the
electrophilicity of the aldehyde could be increased by various
mechanisms such as general acid or Lewis acid catalysis.
Second, by controlling the orientation of the substrate with
respect to the iron center, hydrogen atom abstraction by the
reactive iron−oxo species could be directed toward the acetal
O−H bond and away from the aldehyde C−H bond and C−H
bonds on the β-carbon. Thus, these studies provide further
insight into how enzymes may control reactive iron−oxo
species to catalyze the diverse range of iron-dependent
oxidative transformations observed in biology.
EXPERIMENTAL AND COMPUTATIONAL
METHODS
■
Materials. Ferrous ammonium sulfate was purchased from
Sigma-Aldrich. NADH and ferric chloride were from Acros
Organics. D2O (99.9%) and d6-DMSO (99.9%) were obtained
from Cambridge Isotope Laboratories, Inc. Hydrogen peroxide
(30% in water) and ferrous chloride were from Fischer
Scientific.
Synthesis of Cyclopropyl Aldehydes. The synthesis of
trans-2-nonylcyclopropane-1-carbaldehyde (α-cyclopropyl do-
decanal, 1) was accomplished by standard methods starting
from commercially available (E)-dodec-2-en-1-ol, in which the
cyclopropyl group is positioned α to the carbonyl group.
Briefly, the alcohol group on (E)-dodec-2-en-1-ol was protected
as a TBDMS silyl ether followed by the conversion of the
double bond to a cyclopropyl group using diethylzinc and
diiodomethane to yield the trans stereoisomer. Finally, the
deprotection of TBDMS group by methanolic hydrochloric
acid, followed by the TEMPO oxidation of the alcohol, yielded
the desired trans-2-nonylcyclopropane-1-carbaldehyde (1) (see
the Supporting Information for details).
3298
DOI: 10.1021/acscatal.6b00592
ACS Catal. 2016, 6, 3293−3300

ACS Catalysis
Research Article
Assays. Enzymatic assays were performed as described
previously.11 For nonenzymatic assays, concentrated stocks of
aldehydes (1, dodecanal, α-dodecenal, α-oxiranyldodecanal) in
DMSO (10 mM) and Fe(NH3)6SO4 in H2O (10 mM) were
prepared in a glovebox. Standard assays consisted of 100 μM
Fe(NH3)6SO4 and 300−500 μM 1 in 10 mM sodium acetate
buffer (pH 5.0) at a total volume of 500 μL. Vials were
removed from the glovebox and exposed to air briefly before
capping and incubating at 37 °C and 200 rpm for 10 min.
Assays were quenched with the addition of 50 mM EDTA and
organics extracted in ethyl acetate. To study the pH
dependence of the reaction, assays were conducted in aqueous
buffer over the pH range 4−10 (10 mM Na acetate pH 4−6, 10
mM HEPES pH 6−8, 1 mM CHES pH 8−10) that contained
100 μM Fe2+ and 500 μM 1 at 37 °C with shaking at 200 rpm
for 15 min. The products of the reaction were then extracted,
derivatized, and analyzed by GC-MS. To study the reaction of 1
with ferric iron and hydrogen peroxide, 100 μM FeCl3 and 1−
10 mM H2O2 were substituted and the reactions performed
anaerobically before quenching, extraction, and analysis by GC-
MS.
hypothesizes potential intermediates that are plausibly one
elementary step away from starting reactants,43,44 (2) the
growing string method determines the exact transition state and
minimum energy reaction path for each proposed elementary
step,45−47 and (3) steps 1 and 2 are repeated for
thermodynamically and kinetically feasible intermediates until
complete, low-barrier pathways for product formation are
found. These methods allow for efficient exploration of the
relevant chemical space and discovery of plausible reaction
paths without relying on prior knowledge of the elementary
steps involved.48 For full computational details, see the
Supporting Information.
ASSOCIATED CONTENT
* Supporting Information
The Supporting Information is available free of charge on the
ACS Publications website at DOI: 10.1021/acscatal.6b00592.
■
S
Details of the synthesis and characterization of the
compounds used in these studies and details of
computational simulations including calculated free
energies and coordinates for calculated molecular
structures (PDF)
Carboxylic acid quantification was accomplished using the
standard diazomethane derivatization protocol.39 The organic
compounds extracted from the assay were treated with a few
drops of diazomethane solution and the esterified product
analyzed by GCMS.
AUTHOR INFORMATION
Corresponding Author
*E.N.G.M.: tel, 734 763 6096; e-mail, nmarsh@umich.edu.
■
Solvent Proton Incorporation. Assays were prepared as
described above in either H2O or D2O (99%). Aldehyde
solutions were made up as stock solutions in d6-DMSO to
facilitate subsequent 1H NMR analyses. A typical assay
contained 1 mM ferrous ammonium sulfate and 500 μM d3-1
in a total volume of 4 mL. For assays performed in D2O, the
final H2O concentration was <∼1% after adding all the assay
components. Five identical 4 mL reaction mixtures were set up
under each set of assay conditions and shaken at 37 °C at 200
rpm for 15 min. The reaction mixtures were sequentially
extracted with a total volume of 1.2 mL of CDCl3 (99.9%). The
CDCl3 layers were dried over sodium sulfate and filtered before
Present Address
§(B.P.) Department of Studies in Chemistry, Bangalore
University, Bangalore-560001, India
Notes
The authors declare no competing financial interest.
ACKNOWLEDGMENTS
■
This work was supported in part by the grant CAREER
1551994 from the National Science Foundation to P.M.Z. and
by grants CHE 1152055 and CBET 1336636 from the National
Science Foundation and GM 093088 from the National
Institutes of Health to E.N.G.M.
1
analysis by H NMR. Spectra were acquired at 700 MHz
(Varian) at 22 °C.
HRMS Analysis of C1 Products. Compound 1 (13C-
labeled at the carbonyl carbon or unlabeled) and Fe(NH3)6SO4
were dissolved in air-saturated H2O to 500 μM and 1 mM,
respectively, in a final volume of 2 mL. Samples were placed in
sealed 5 mL round-bottom flasks equipped with HRMS
adapters and incubated at 37 °C and 200 rpm for 30 min.
The headspace was analyzed by a VG 70-250-S magnetic sector
mass spectrometer, and total ion counts were collected.
Computational Methods. All quantum chemical calcu-
lations were performed by density functional theory (DFT)
using the Q-Chem 4.3 quantum chemistry package.41
Unrestricted PBE40 with hextet spin and the triple-ζ, polarized
cc-pVTZ basis set was used to calculate energies with the SMD
solvation model42 using H2O as the implicit solvent (in
addition to the five explicit inner-sphere H2O molecules).
Methylcyclopropyl carbaldehyde was used to approximate 1,
where the substrate’s alkyl chain was truncated to reduce
computational cost. Thermodynamic corrections were applied
to the solvated energies at a temperature of 298 K.
REFERENCES
■
(1) Schirmer, A.; Rude, M. A.; Li, X.; Popova, E.; Cardayre, S. B.
Science 2010, 329, 559−562.
(2) Marsh, E. N.G.; Waugh, M. W. ACS Catal. 2013, 3, 2515−2521.
(3) Schneider-Belhaddad, F.; Kolattukudy, P. Arch. Biochem. Biophys.
2000, 377, 341−349.
(4) Reed, J. R.; Quilici, D. R.; Blomquist, G. J.; Reitz, R. C.
Biochemistry 1995, 34, 16221−16227.
(5) Reed, J. R.; Vanderwel, D.; Choi, S.; Pomonis, J. G.; Reitz, R. C.;
Blomquist, G. J. Proc. Natl. Acad. Sci. U. S. A. 1994, 91, 10000−10004.
(6) Qiu, Y.; Tittiger, C.; Wicker-Thomas, C.; Le Goff, G.; Young, S.;
Wajnberg, E.; Fricaux, T.; Taquet, N.; Blomquist, G. J.; Feyereisen, R.
Proc. Natl. Acad. Sci. U. S. A. 2012, 109, 14858−14863.
(7) Das, D.; Eser, B. E.; Han, J.; Sciore, A.; Marsh, E. N. G. Angew.
Chem., Int. Ed. 2011, 50, 7148−7152.
(8) Li, N.; Nørgaard, H.; Warui, D. M.; Booker, S. J.; Krebs, C.;
Bollinger, J. M. J. Am. Chem. Soc. 2011, 133, 6158−61661.
(9) Andre, C.; Kim, S. W.; Yu, X.-H.; Shanklin, J. Proc. Natl. Acad. Sci.
U. S. A. 2013, 110, 3191−3196.
(10) Das, D.; Ellington, B.; Paul, B.; Marsh, E. N. G. ACS Chem. Biol.
2014, 9, 570−577.
Research discovery tools developed by the Zimmerman
group were used to determine the deformylation mechanism
that results in CO2 and n − 1 alkane formation. This strategy
proceeds as follows: (1) a graph-based structure generator
(11) Paul, B.; Das, D.; Ellington, B.; Marsh, E. N. G. J. Am. Chem. Soc.
2013, 135, 5234−5237.
(12) Shokri, A.; Que, L., Jr. J. Am. Chem. Soc. 2015, 137, 7686−7691.
3299
DOI: 10.1021/acscatal.6b00592
ACS Catal. 2016, 6, 3293−3300

ACS Catalysis
Research Article
(13) Waugh, M. W.; Marsh, E. N. G. Biochemistry 2014, 53, 5537−
5543.
(14) Buer, B. C.; Paul, B.; Das, D.; Stuckey, J. A.; Marsh, E. N. G.
(48) Nett, A. J.; Zhao, W.; Zimmerman, P. M.; Montgomery, J. J. Am.
Chem. Soc. 2015, 137, 7636−7639.
ACS Chem. Biol. 2014, 9, 2584−2593.
(15) Pandelia, M. E.; Li, N.; Nørgaard, H.; Warui, D. M.; Rajakovich,
L. J.; Chang, W.-c.; Booker, S. J.; Krebs, C.; Bollinger, J. M. J. Am.
Chem. Soc. 2013, 135, 15801−15812.
(16) Rajakovich, L. J.; Nørgaard, H.; Warui, D.; Chang, W.-c.; Li, N.;
Booker, S. J.; Krebs, C.; Bollinger, J. M.; Pandelia, M.-E. J. Am. Chem.
Soc. 2015, 137, 11695−11709.
(17) Aukema, K. G.; Makris, T. M.; Stoian, S. A.; Richman, J. E.;
Munck, E.; Lipscomb, J. D.; Wackett, L. P. ACS Catal. 2013, 3, 2228−
̈
2238.
(18) Modi, A. R..; Dawson, J. H. In Monooxygenase, Peroxidase and
Peroxygenase Properties and Mechanisms of Cytochrome P450; Hrycay, E.
G., Bandiera, S. M., Eds.; Springer International: Geneva, Switzerland,
2015; pp 63−82.
(19) Kovaleva, E. G.; Lipscomb, J. D. Nat. Chem. Biol. 2008, 4, 186−
193.
(20) Merkx, M.; Kopp, D. A.; Sazinsky, M. H.; Blazyk, J. L.; Muller,
J.; Lippard, S. J. Angew. Chem., Int. Ed. 2001, 40, 2782−2807.
(21) Cascella, B.; Mirica, L. M. Biochemistry 2012, 51, 8699−8701.
(22) Liu, P.; Murakami, K.; Seki, T.; He, X.; Yeung, S.-M.; Kuzuyama,
T.; Seto, H.; Liu, H. J. Am. Chem. Soc. 2001, 123, 4619−4620.
(23) Liu, P.; Liu, A.; Yan, F.; Wolfe, M. D.; Lipscomb, J. D.; Liu, H.
Biochemistry 2003, 42, 11577−11586.
(24) Jensen, K. B.; McKenzie, C. J.; Nielsen, L. P.; Pedersen, J. Z.;
Svendsen, H. M. Chem. Commun. 1999, 1313−1314.
(25) Annaraj, J.; Suh, Y.; Seo, M. S.; Kim, S. O.; Nam, W. Chem.
Commun. 2005, 4529−4531.
(26) Cho, J.; Jeon, S.; Wilson, S. A.; Liu, L. V.; Kang, E. A.; Braymer,
J. J.; Lim, M. H.; Hedman, B.; Hodgson, K. O.; Valentine, J. S.;
Solomon, E. I.; Nam, W. Nature 2011, 478, 502−505.
(27) Shan, X.; Rohde, J.; Koehntop, K. D.; Zhou, Y.; Bukowski, M.
R.; Costas, M.; Fujisawa, K.; Que, L., Jr. Inorg. Chem. 2007, 46, 8410−
8417.
(28) Goh, Y. M.; Nam, W. Inorg. Chem. 1999, 38, 914−920.
(29) Pirovano, P.; Magherusan, A. M.; McGlynn, C.; Ure, A.; Lynes,
A.; McDonald, A. R. Angew. Chem., Int. Ed. 2014, 53, 5946.
(30) McNesby, J. R.; Heller, C. A. Chem. Rev. 1954, 54, 325−346.
(31) Chudasama, V.; Fitzmaurice, R. J.; Caddick, S. Nat. Chem. 2010,
2, 592−596.
(32) Rat, S.; Menage, S.; Thomas, F.; Niviere, V. Chem. Commun.
2014, 50, 14213−14216.
(33) Walling, C. In Active Oxygen in Chemistry; Foote, C., Valentine,
J., Greenberg, A., Liebman, J., Eds.; Chapman and Hall: London, 1995;
pp 105−140.
(34) Vaz, A. D.; Pernecky, S. J.; Raner, G. M.; Coon, M. J. Proc. Natl.
Acad. Sci. U. S. A. 1996, 93, 4644−4648.
(35) Meunier, B.; de Visser, S. P.; Shaik, S. Chem. Rev. 2004, 104,
3947−3980.
(36) Newcomb, M.; Hollenberg, P. F.; Coon, M. J. Arch. Biochem.
Biophys. 2003, 409, 72−79.
(37) Parker, W.; Halton, B. Alicyclic Chemistry; Royal Society of
Chemistry: London, 2006; Vol. 5, pp 1−99.
(38) van der Donk, W. A.; Krebs, C.; Bollinger, J. M. Curr. Opin.
Struct. Biol. 2010, 20, 673−683.
(39) Hudlicky, M. J. Org. Chem. 1980, 45, 5377−5378.
(40) Perdew, J. P.; Burke, K.; Ernzerhof, M. Phys. Rev. Lett. 1996, 77,
3865−3868.
(41) Shao, Y.; et al. Mol. Phys. 2015, 113, 184−215.
(42) Marenich, A. V.; Cramer, C. J.; Truhlar, D. G. J. Phys. Chem. B
2009, 113, 6378−6396.
(43) Zimmerman, P. M. J. Comput. Chem. 2013, 34, 1385−1392.
(44) Zimmerman, P. M. Mol. Simul. 2015, 41, 43−54.
(45) Zimmerman, P. M. J. Chem. Phys. 2013, 138, 184102.
(46) Zimmerman, P. M. J. Chem. Theory Comput. 2013, 9, 3043−
3050.
(47) Zimmerman, P. M. J. Comput. Chem. 2015, 36, 601−611.
3300
DOI: 10.1021/acscatal.6b00592
ACS Catal. 2016, 6, 3293−3300