Báo cáo khoa học: Characterization of heme-binding properties of Paracoccus denitrificans Surf1 proteins doc

Characterization of heme-binding properties of
Paracoccus denitrificans Surf1 proteins
Achim Hannappel
1
, Freya A. Bundschuh
1
and Bernd Ludwig
1,2
1 Molecular Genetics Group, Institute of Biochemistry, Goethe University Frankfurt, Frankfurt am Main, Germany
2 Cluster of Excellence Macromolecular Complexes, Frankfurt am Main, Germany
Introduction
Cytochrome c oxidase (COX; also kown as cyto-
chrome aa
3
or complex IV) is a key player in cellular
respiration: as the terminal redox complex of the elec-
tron transport chain, it reduces molecular oxygen to
water and couples the free energy of this reaction to
the generation of a transmembrane proton gradient.
The mitochondrial enzyme consists of up to 13 subun-
its, of which only the key subunits I–III are encoded in
the organelle genome. These very subunits are also
found in the corresponding bacterial oxidases and
share both structural and functional homologies. Sub-
unit I, as the core of the enzyme, houses two heme a
moieties (heme a and a
3
) and a copper ion [1,2].
The biosynthetic pathway of heme a cofactor syn-
thesis starts with the conversion of heme b to heme o
by the addition of a farnesyl side chain. This reaction

is catalysed by the enzyme heme o synthase, named
Cox10 in eukaryotes and CtaB in bacteria [3]. In a
subsequent step, the C-8 methyl group of heme o is
oxidized to a formyl group catalysed by heme a
synthase, termed Cox15 in eukaryotes and CtaA in
bacteria [3]. The Bacillus subtilis CtaA homolog is the
best-studied variant to date, but its exact cofactor-
binding stoichiometries and enzymatic mechanism is
still unresolved [4–6]. With its eight transmembrane
helices, heme a synthase is supposed to provide two
different heme-binding sites, one for the heme b redox
cofactor and one where the substrate heme o may bind
and be oxidized to heme a [6].
The assembly factor Surf1 has long been known for
its involvement in oxidase biogenesis. Its functional
loss leads to Leigh syndrome in humans, a severe
Keywords
CtaA; cytochrome c oxidase; heme a
synthase; Leigh syndrome; oxidase
assembly
Correspondence
B. Ludwig, Institute of Biochemistry,
Goethe University Frankfurt, Max-von-Laue-
Strasse 9, D-60438 Frankfurt am Main,
Germany
Fax: +49 69 798 29244
Tel: +49 69 798 29237
E-mail:
Website: chem.
uni-frankfurt.de/

(Received 8 February 2011, revised 4 March
2011, accepted 14 March 2011)
doi:10.1111/j.1742-4658.2011.08101.x
Biogenesis of cytochrome c oxidase (COX) is a highly complex process
involving >30 chaperones in eukaryotes; those required for the incorpora-
tion of the copper and heme cofactors are also conserved in bacteria.
Surf1, associated with heme a insertion and with Leigh syndrome if defec-
tive in humans, is present as two homologs in the soil bacterium Paracoc-
cus denitrificans, Surf1c and Surf1q. In an in vitro interaction assay, the
heme a transfer from purified heme a synthase, CtaA, to Surf1c was
followed, and both Surf proteins were tested for their heme a binding pro-
perties. Mutation of four strictly conserved amino acid residues within the
transmembrane part of each Surf1 protein confirmed their requirement for
heme binding. Interestingly the mutation of a tryptophan residue in trans-
membrane helix II (W200 in Surf1c and W209 in Surf1q) led to a drastic
switch in the heme composition, with Surf1 now being populated mostly
by heme o, the intermediate in the heme a biosynthetic pathway. This tryp-
tophan residue discriminates between the two heme moieties, apparently
coordinates the formyl group of heme a, and most likely presents the
cofactor in a spatial orientation suitable for optimal transfer to its target
site within subunit I of cytochrome c oxidase.
Abbreviation
COX, cytochrome c oxidase.
FEBS Journal 278 (2011) 1769–1778 ª 2011 The Authors Journal compilation ª 2011 FEBS 1769
neurodegenerative disorder characterized by lesions in
the central nervous system [7–9]. Patients suffer a 80–
90% loss of COX activity in all tissues and the assem-
bly process is stalled at intermediate levels [10]. The
yeast homolog Shy1p is one of the best-characterized
Surf1 homologs and studies point at a stabilizing role

in early assembly intermediates containing both the
core subunits I and II [11–13].
Analysis of purified COX from surf1 deletion
mutants in a bacterial background showed that the
loss of oxidase activity is attributed to a disturbed
heme a incorporation into COX subunit I, particularly
affecting the level of heme a
3
[14,15]. The soil bacte-
rium Paracoccus denitrificans encodes two different
Surf1 homologs, with Surf1c specifically acting on the
aa
3
-type cytochrome c oxidase, and Surf1q exclusively
serving the ba
3
-type quinol oxidase [15]. In addition,
we showed that both Paracoccus Surf proteins indeed
bind heme a, both in vivo and in vitro, with a 1 : 1
stoichiometry and affinities in the submicromolar
range. A conserved periplasmically located histidine
residue near the second transmembrane helix is the
presumed ligand for the central iron atom, because a
histidine-to-alanine mutant showed strongly diminished
heme a affinities [16]. A corresponding mutation in the
yeast homolog Shy1p resulted only in modest growth
impairment on nonfermentable carbon sources, in line
with diminished levels, but no complete loss of heme
affinities [17].
In early oxidase assembly events, a triple contribu-

tion of Surf1 may be imagined: (a) modulation of
heme a synthase activity by abstracting the enzymatic
end product, (b) supply of an readily available protein-
bound heme a pool, and (c) incorporation of heme a
into oxidase subunit I.
In this study, we show that Surf1 specifically receives
heme a from heme a synthase under in vitro condi-
tions. We further investigated highly conserved resi-
dues within the transmembrane part of the Surf1
protein and show that they all contribute to heme
binding in either bacterial version of the Surf1 protein,
with a strategic tryptophan residue possibly interacting
with the formyl group of heme a.
Results
Lacking the machinery for the final step in heme a bio-
synthesis, Escherichia coli is well suited for heterologous
expression experiments of biogenesis factors involved in
heme a delivery to COX: E. coli cells neither encode a
Surf1 homolog nor provide the heme a-synthesizing
machinery and therefore offer a genetic background
that allows well-defined analysis of interactions between
these components. In an earlier study, we showed that
the Surf1 proteins from Paracoccus denitrificans can
bind heme a under in vivo conditions when coexpressed
in E. coli together with the heme a-biosynthesizing
enzymes CtaB and CtaA [16]. To further investigate
whether CtaA is the interaction partner for Surf1 and
thereby directly supplies the cofactor, we characterized
the CtaA protein on a purified level.
The P. denitrificans CtaA protein was heterologously

expressed in E. coli to low levels under the control of
the constitutive tet promoter. As judged by SDS ⁄
PAGE analysis, the protein is isolated to > 95% pur-
ity via a TEV-cleavable C-terminal hexahistidine tag
(Fig. S1). Depending on the presence or absence of
CtaB, as well as on the growth conditions (high aera-
tion in baffled flasks vs. limiting aeration in standard
flasks), the protein is obtained in one of three spec-
trally distinct forms: b-form, bo-form and ba-form
(Figs S2 and S3). The b-form is acquired when the
protein is expressed in the absence of CtaB; the
bo-form in the presence of CtaB under limiting oxygen
conditions; the ba-form in the presence of CtaB and
high aeration (Table S1).
Heme a is specifically transferred from CtaA to
Surf1 in vitro
The ba-form of CtaA, carrying predominatly hemes b
and a, but also traces of heme o, was further used in
an in vitro heme transfer assay. For this, purified CtaA
and apo-Surf1c were mixed in a 1 : 2 molar ratio and
after 1 h of incubation at room temperature reseparat-
ed via immobilized metal-affinity chromatography.
Both proteins are recovered efficiently with only minor
cross-contaminations of CtaA in the Surf1c fraction
(Fig. 1A). Redox spectra (dithionite minus ferricya-
nide) taken under native conditions clearly show that
CtaA specifically loses heme a during incubation with
apo-Surf1c (Fig. 1B). The heme a peak maximum at
592.5 nm is decreased by  66%, whereas the heme b
peak at 560 nm remains unchanged. However, Surf1c

specifically takes up heme a exhibiting an absorption
maximum at 595 nm not present in its apo-form
before incubation with CtaA (Fig. 1C). The small
shoulder at 560 nm in the Surf1c fraction after incuba-
tion is caused by the minor cross-contamination with
CtaA visible in the polyacrylamide gel.
This transfer assay clearly shows that both proteins
transiently interact in vitro and that during this interac-
tion heme a
is specifically transferred from its site of
synthesis in CtaA to Surf1c.
A conserved Trp specifies heme a binding in Surf1 A. Hannappel et al.
1770 FEBS Journal 278 (2011) 1769–1778 ª 2011 The Authors Journal compilation ª 2011 FEBS
Heme binding of Surf1 mutants
Earlier expression studies of wild-type Surf1 proteins
in E. coli, in the presence of CtaB and CtaA, con-
tained next to heme a also small amounts of heme o in
Surf1, a fact interpreted as unspecific binding under
the expression conditions chosen [16]. Indeed the for-
mation of heme a in the recombinant E. coli cells
is highly sensitive to aeration levels as CtaA can be
purified with different heme compositions (Figs S2
and S3).
In order to circumvent spectral impurities caused by
nonoptimal expression conditions in E. coli, we chan-
ged to an autoinductive medium [18]. Expression in
autoinductive medium offers some advantages favour-
ing heme a biosynthesis in the heterologous E. coli sys-
tem: expression of Surf1 is induced at higher cell
densities and expression times are prolonged, allowing

the heme a biosynthesis enzymes CtaB and CtaA to
match the higher expression levels observed for Surf1.
Indeed, the change to autoinductive medium avoids
such spectral impurities for the Surf1 wild-type pro-
teins and yields homogenous preparations containing
heme a as sole heme species (see below). In native
redox spectra, absorption maxima of 595 nm for wild-
type Surf1c and 600 nm for wild-type Surf1q are
obtained, as described previously [16].
To determine the heme-binding properties of the
Surf1 proteins, several conserved amino acid residues
were mutated that locate within the transmembrane
helices near the periplasmic side of the membrane. The
chosen residues (Table 1) are the only ones fully con-
served in a comparison of 61 different Surf1 sequences
found in the KEGG database ( />kegg/). Their side chains may serve as ligands for func-
tional groups such as propionates or the formyl side
chain of heme a. In the design of the mutations, care
Fig. 1. In vitro heme a transfer from CtaA to Surf1c. Purfied CtaA
and apo-Surf1c were mixed in a 1 : 2 molar ratio and separated
after 1 h of incubation at room temperature via immobilized metal-
affinity chromatography. A Coomassie Brilliant Blue stained
SDS ⁄ polyacrylamide gel of 1 lg protein per lane (A) shows the pro-
teins before (b.t.) and after (a.t.) the transfer. Native redox differ-
ence spectra of 20 l
M CtaA (B) and Surf1c (C) are shown.
A comparison of the spectra taken before (b.t., dotted line) and
after (a.t., solid line) the transfer shows a diminished heme a con-
tent for CtaA, whereas heme b remains unaffected. The transferred
heme a is found in the Surf1c fraction (a.t., solid line) that did not

contain any heme before transfer (b.t., dotted line).
Fig. 2. Gel electrophoresis of purified Paracoccus denitrificans
Surf1 mutant proteins. Coomassie Brilliant Blue-stained 12%
SDS ⁄ polyacrylamide gel of Surf1 mutant proteins after purification
via immobilized metal affinity chromatography. Approximately
2.5 lg of protein was loaded per lane.
A. Hannappel et al. A conserved Trp specifies heme a binding in Surf1
FEBS Journal 278 (2011) 1769–1778 ª 2011 The Authors Journal compilation ª 2011 FEBS 1771
was taken to introduce conservative mutations to
avoid disturbing the overall structural determinants
and to merely tackle their potential electron-donor
functions involved in hydrogen bonding.
The mutated proteins were heterologously expressed
in autoinductive medium in the presence of CtaA and
CtaB and purified via their N-terminal decahistidine
tag, as described previously [16]. All proteins were
obtained without major impurities (Fig. 2). In the sam-
ples of Surf1c Q25A and Surf1q Q26A, a potential
degradation product can be seen at low molecular
mass that does not stain in a Western blot against the
decahistidine tag (not shown).
Native redox spectra reveal that all mutants are
impaired in their heme a content compared to wild-
type. The mutants W24F, Q25A, H193A and Y196F
(Surf1c nomenclature) completely lack any heme a, the
same being observed for the corresponding Surf1q
mutants. By contrast, mutation of a tryptophan in the
second transmembrane helix (W200F for Surf1c,
W209F for Surf1q) resulted in a peak maximum differ-
ing from the wild-type proteins (558 nm for Surf1c

W200F with a small shoulder at 595 nm; 565 nm for
Surf1q with a shoulder at 600 nm).
A conserved tryptophan in the second
transmembrane helix of Surf1 may be
responsible for heme a formyl group
coordination
In order to clarify the nature of the heme species in
Surf1c W200F and Surf1q W209F, pyridine hemo-
chrome spectra were taken under denaturing condi-
tions (Fig. 4). Both mutant proteins show a peak at
552 nm and a small shoulder at 587 nm, pointing to
heme types o and a.
To unequivocally establish the nature of the heme
species, HPLC analysis was performed. For this, puri-
fied Surf1c W200F and Surf1q W209F protein was
extracted with acidified acetone and hemes were analy-
sed on a reversed-phase column using an acetonitrile
gradient. A comparison with samples of known heme
content clearly shows that both mutant proteins con-
tain heme o and heme a. Assuming similar extinction
coefficients in acetonitrile for the Soret for both
heme o and heme a, we estimate  25% heme a and
75% heme o for Surf1c W200F, and  40% heme a
and 60% heme o for Surf1q respectively.
The cytochrome c oxidase defect in a D surf
strain can not be rescued by Surf1c W200F
Upon deletion of the surf1c gene in P. denitrificans the
resulting COX is compromised in its heme content to
 50% compared with wild-type levels, and COX
activity is reduced to 28%. This phenotype can be res-

cued by the expression of wild-type Surf1c in trans
[15]. In order to assess the effect of the Surf1c W200F
mutant on oxidase, COX was purified from a P. deni-
trificans Dsurf1c strain expressing Surf1c W200F (con-
firmed by immunescreen; not shown), and analysed
enzymatically as well as spectroscopically. Surf1c
W200F cannot recover oxidase activity, and the
Table 1. Mutations introduced in Surf1c and Surf1q and their
resulting heme composition after heterologous expression in
Escherichia coli in the presence of Paracoccus denitrificans CtaB
and CtaA.
Heme type
Surf1c Wild-type a
W24F –
Q25A –
H193A
a
–
Y196F –
W200F o ⁄ a
Surf1q Wild-type a
W25F –
Q26A –
H202A
a
–
Y205F –
W209F o ⁄ a
a
Mutant first described in Bundschuh et al. [16].

Fig. 3. Native redox difference spectra of
purified Surf1 mutant proteins. Native redox
spectra of 50 l
M purified Surf1 mutant pro-
teins in the a-region for the Surf1c mutants
(A) and the corresponding Surf1q mutants
(B).
A conserved Trp specifies heme a binding in Surf1 A. Hannappel et al.
1772 FEBS Journal 278 (2011) 1769–1778 ª 2011 The Authors Journal compilation ª 2011 FEBS
enzyme is still impared in its heme content as is the
deletion strain FA3 (Table 2), clearly confirming that
this Surf1 mutant version is nonfunctional during
oxidase assembly under in vivo conditions.
Discussion
Failures in cytochrome c oxidase biogenesis are a com-
mon cause of respiratory deficiencies that may lead to
severe diseases such as Leigh syndrome. In eukaryotic
systems, an increasing number of protein factors has
been associated with COX biogenesis, because power-
ful genetic screening systems available for the yeast
model. In bacteria, known biogenesis factors appear to
be limited to those possibly involved in metal cofactor
delivery to oxidase subunits [19]. Recently we estab-
lished that the P. denitrificans Surf1 proteins stoichio-
metrically bind heme a with submicromolar affinities
[16]. In this study, we further analyse their heme-
binding properties using site-directed mutagenesis and
provide clear evidence for a specific interaction with
heme a synthase.
To establish heme a biosynthesis in E. coli, endoge-

nously modifying heme b only to the level of heme o,
two additional genes (ctaB and ctaA) were introduced
and expressed under high aeration, as shown
previously [20,21]. Here, we demonstrate that this
observation is also true for the Paracoccus heme a-syn-
thesizing machinery. The Paracoccus heme a synthase,
CtaA, can be purified after heterologous expression in
E. coli with three distinct heme cofactor compositions:
b-form, bo-form and ba-form (Figs S1–S3). Only with
CtaB present and under high aeration levels can CtaA
be enzymatically fully active when expressed in E. coli.
To our knowledge, this is the first study working with
purified Paracoccus CtaA enzyme, confirmimg results
obtained for the corresponding Bacillus subtilis and
Rhodobacter sphaeroides enzymes [6,20].
Because Surf1 binds heme a when coexpressed in
E. coli together with the heme a-synthesizing enzymes,
we postulated an in vivo interaction between CtaA and
Surf1 [16]. To check whether CtaA is the actual
heme a source for Surf1, we followed the heme trans-
fer in vitro by mixing purified CtaA and apo-Surf1c.
Indeed, heme a is being taken up by Surf1 from its site
of synthesis, CtaA, requiring a specific interaction
between both proteins to mediate heme transfer. Stud-
ies in yeast demonstrated that heme a synthase expres-
sion and activity is strictly regulated [22]. In this
context, Surf1 is an attractive candidate for a regula-
tion of CtaA activity because it abstracts the enzymatic
end product, thereby initiating a new round of heme a
synthesis. In this function, Surf1 would prevent any

undesirable release of heme a from CtaA, thus avoid-
ing the accumulation of free heme a that would be det-
rimental to the cell. As we show experimentally here,
both proteins transiently interact and readily separate
once the transfer process is completed (Fig. 1). Because
our experimental approach does not offer any kinetic
resolution we can only speculate on the nature of the
transfer process. It may be that the transfer process, as
such, leads to structural rearrangements that cause a
decrease in the reciprocal affinity, either triggered by
Fig. 4. Heme analysis of Surf1c W200F and
Surf1q W209F. Pyridine hemochromogen
redox spectra of 20 l
M Surf1 mutant pro-
teins (A) and HPLC analysis of heme
extracts derived from Surf1c W200F and
Surf1q W209F in an 50–100% acetonitrile
gradient shown as dotted line in (B). Both
mutant proteins show an increased ratio of
heme o over heme a.
Table 2. Activity and heme a content of COX samples purified
from a Paracoccus surf1c deletion strain (FA3), Surf1c complemen-
tation (FA3.61) and a complementation with Surf1c W200F
(FA3.61-W200F).
Turnover
[s
)1
]
%wt
activity

% Heme
a content
a
Wild-type 344 100 100
FA3
b
95 28 41
FA3.61
b
292 85 96
FA3.61–W200F 123 36 43
a
As determined by pyridine hemochromogen redox spectra using
an extinction coefficient of De
587–620
= 21.7 mM
)1
Æcm
)1
; heme a
content relative to wild-type oxidase (16.1 nmolÆ mg
)1
).
b
first
described in Bundschuh et al. [15].
A. Hannappel et al. A conserved Trp specifies heme a binding in Surf1
FEBS Journal 278 (2011) 1769–1778 ª 2011 The Authors Journal compilation ª 2011 FEBS 1773
the loss of heme a in CtaA or the gain of the cofactor
by Surf1. After separation from CtaA, Surf1 might

instantly donate the heme moiety to COX subunit I,
potentially in a cotranslational fashion.
An interaction of heme a synthase and Surf1 was
also detected in yeast, where heme a synthase, Cox15,
copurifes with the yeast Surf1 homolog Shy1p, but the
authors of the study did not discuss this observation in
any further detail [23]. Additional biochemical and
biophysical studies are needed to investigate the inter-
action between both proteins and clarify the nature of
the heme transfer process.
Because the formation of heme a in E. coli cells is
highly dependent on growth parameters, as discussed
above, reproducible expression conditions for Surf1
are needed, avoiding variations from one preparation
to another. The switch to an autoinductive medium
fulfils these requirements and compensates the rather
slow production of heme a in the heterologous host
system by induction of Surf1 expression at high cell
densities and extended induction times. Such experi-
mental conditions allow the access to spectroscopically
pure Surf1 samples containing heme a as the sole heme
species, enabling us to test the heme-binding properties
of the two Paracoccus homologs using site-directed
mutagenesis. All mutations introduced into conserved
residues within the transmembrane part of the proteins
led to impaired heme a binding compared with wild-
type, further emphasizing that heme binding must be
an important physiological role of the protein, at least
in the bacterial system (Fig. 3).
Missense mutations of human Surf1 tend to be

responsible for mild clinical phenotypes of Leigh syn-
drome, allowing prolonged survival of affected patients
[24]. Recently, a mutagenesis study on Shy1p tried to
mimic clinically relevant missense mutations in yeast,
also investigating the role of the conserved tyrosine resi-
due in transmembrane helix II (Fig. 5). The mutation to
aspartate did not have major effects in the yeast back-
ground [17]. Here we show that both bacterial versions
lose their heme-binding properties once the tyrosine is
mutated. The same is true also for W24F and Q25A
(Surf1c nomenclature), as well as the corresponding
Surf1q mutations. All mutations tested here compro-
mise any hydrogen-bonding properties of the side
chains, presumably without affecting the overall struc-
ture. Those residues might, therefore, be involved in
binding the propionate side chains of the heme moiety.
Interestingly, in the tryptophan to phenyalanine
mutant in transmembrane helix II (W200F in Surf1c,
W2009F in Surf1q), heme o is the predominant
heme cofactor, indicating that this residue is the criti-
cal determinant to bind the formyl group of heme a,
thus discriminating against the heme o moiety (Fig. 5).
This tryptophan residue would orient the cofactor in a
way that facilitates subsequent correct incorporation
of both hemes into COX subunit I. Because the trypto-
phan to phenylalanine mutant binds an unphysiologi-
cal heme type we expected a negative effect on COX
biogenesis in vivo. Indeed, the Surf1c W200F mutant is
not able to compensate the surf1 deletion phenotype in
P. denitrificans (Table 2).

Nevertheless, an alternative heme incorporation
pathway into COX subunit I must be operative inde-
pendent of Surf1, because residual COX activity is
reported for surf1 deletions in any organism studied so
far. A direct interaction between heme a synthase and
subunit I could be envisaged. As we now show a direct
transfer of heme from heme a synthase to Surf1, it is
tempting to speculate whether a ternary complex
exists, comprised of heme a synthase, Surf1 and sub-
unit I. Surf1 could at least stabilize such a complex
and facilitate the flow of heme a from its site of syn-
thesis to its target site(s) within subunit I. But as the
Surf protein itself stably binds heme a, the mechanism
Fig. 5. Schematic model of heme a binding to Surf1. Based on
sequence alignments highly conserved amino acid residues can be
identified that all locate in the predicted transmembrane helices
near the periplasmic side of the membrane. The histidine residue is
most likely involved in iron (blue sphere) coordination [16], whereas
the tryptophan in the second helix may specify the binding of the
formyl group of heme a (red sphere) thus correctly orienting the co-
factor for later integration into COX subunit I. Secondary structure
elements are based on predictions by the program Sable [31].
A conserved Trp specifies heme a binding in Surf1 A. Hannappel et al.
1774 FEBS Journal 278 (2011) 1769–1778 ª 2011 The Authors Journal compilation ª 2011 FEBS
of incorporation into subunit I may proceed in a
sequential mode as well.
Although it is still not clear what exact physiological
function Surf1 exhibits in mitochondrial oxidase bio-
genesis and whether eukaryotic Surf1 proteins also
bind heme a, it becomes obvious from studying the

bacterial homologs that Surf1 is directly involved in
binding and incorporation of heme a into oxidase. In
future, bacterial model systems may be particularly
helpful because they easily offer access to sufficient
amounts of protein needed for protein biochemical
and biophysical studies.
Materials and methods
Cloning of CtaA
The ctaA gene was amplified via PCR using P. denitrificans
genomic DNA (strain Pd1222) as template with the forward
primer (5¢-ATATACATATGGCTAGCATGACTGGTGG
ACAGCAAATGGGTCGCGGATCCATGTCGCGCCCG
ATCGAGAA-3¢) introducing an N-terminal t7 tag and a
NdeI site, and a reverse primer (5¢ -TATATCTCGAGT
CA(ATGGTG)
3
GCCCTGAAAATAAA GATTC TCA CCC
GGACCGGGACCTCGGACAGTTCCCCGGAC-3¢) add-
ing a TEV site, a C-terminal hexahistidine tag and an XhoI
site. The product was digested with NdeI ⁄ XhoI and cloned
into the plasmid pET24a, generating pHA01. For constitu-
tive expression under control of the tet-promoter, ctaA–
His
6
was subcloned into the plasmid pGR52. This plasmid
encodes the P. denitrificans heme a synthesizing genes ctaB
and ctaA in their wild-type versions [16]. To this end,
ctaA–His
6
was excised from pHA01 using an endogenous

AatII site and an XhoI site and subcloned into AatII ⁄ XhoI-
digested pGR52, resulting in plasmid pHA02. For constitu-
tive expression of CtaA–His
6
without CtaB, the ctaA gene
was amplified via PCR using pHA02 as template with the
forward primer (5¢-TCAAGGTGTACAAAGGAGATACT
CATGTCGCGCCCGATCGAGAAG-3¢) introducing a
Acc65I site, a ribosomal-binding site and the reverse primer
mentioned above. The PCR product was digested with
Acc65I ⁄ XhoI and cloned into the vector pGR50 [16]
restricted by BsrGI ⁄ XhoI, resulting in pHA23. The gener-
ated plasmids were verified by sequencing and can be used
for constitutive expression of CtaA with a TEV-cleavable
C-terminal hexahistidine tag in the presence (pHA02) or
the absence (pHA23) of CtaB.
Mutagenesis of Surf1
Templates for site-directed mutagenesis of His
10
–surf1 genes
were pET22b derivative plasmids housing surf1c (pFA48)
or surf1q (pFA49) as described previously [16]. Mutagenesis
reactions were either performed using a QuikChange muta-
genesis protocol (Stratagene, La Jolla, CA, USA) or via
inverse PCR. Mutants derived from a QuikChange protocol
(Surf1c: Q25A, Y196F, W200F; Surf1q: Q26A) were
obtained using a single primer harbouring the desired muta-
tion and introducing a restriction site for screening pur-
poses. Mutants derived from inverse PCR reactions
(Surf1c: W24F; Surf1q: W25F, Y205F, W209F) were

obtained using a mutagenic forward primer also introduc-
ing a restriction site for screening purposes and a reverse
primer allowing the amplification of the expression plasmid
by PCR. All introduced mutations were confirmed by
sequencing and mutant plasmids were subsequently trans-
formed into the E. coli expression strain C41(DE3) that
already contained the plasmid pGR52.
For cloning of a P. denitrificans complementation plas-
mid, the gene encoding Surf1c W200F was subcloned from
pFA48–W200F into the broad host range vector pFA61
that expresses wild-type Surf1c under the control of the cta-
promoter [15]. For this, both plasmids were digested with
NdeI ⁄ SacI and the resulting fragments were ligated, result-
ing in plasmid pFA61–W200F. After sequence confirmation
the plasmid was finally conjugated via triparental mating
into the Paracoccus surf1c-deletion strain FA3 [15],
resulting in FA3.61–W200F.
Expression and purification of CtaA
E. coli DH5-a strains constitutively expressing CtaA were
grown in 15–30 L of Luria–Bertani medium supplemented
with 10 lm iron (III) chloride. For high aeration, baffled
flasks were used, whereas oxygen-limited growth was per-
formed in normal flasks. The main culture was inoculated
to 1% with an overnight culture and grown for 18 h at
32 °C. Cells were harvested at 5000 g for 15 min, resus-
pended in 20 mm sodium phosphate, pH 8.0, 10 mm
sodium chloride, 1 mm EDTA and membranes were
prepared by established methods.
Membranes were solubilized in the presence of 5% (w ⁄ v)
Triton X-100 in buffer A (50 mm sodium phosphate,

pH 8.0, 300 mm sodium chloride, 10 mm imidazole) at a
final protein concentration of 10 mgÆmL
)1
. The solubilized
material was diluted with buffer A to reduce the Triton con-
centration to 2.5% prior to loading on a nickel–nitriloacetic
acid column (column volume CV = 20 mL; Qiagen, Hilden,
Germany). Bound material was washed with 7.5 CV of
buffer B (50 mm sodium phosphate, pH 8.0, 300 mm
sodium chloride, 0.02% n-dodecyl-b -d-maltoside) contain-
ing 20 mm imidazole. After washing with 40 mm imidazole
in buffer B (7.5 CV), His-tagged CtaA was eluted from the
column by 100 mm imidazole (7.5 CV). This fraction was
concentrated to 1 mL and imidazole was removed by a
30-mL Sephadex G-25 desalting column (Pharmacia, Frei-
burg, Germany). The His-tag was cleaved off by TEV prote-
ase (Tobacco Etch Virus protease, final concentration of
0.02 mgÆmL
)1
) under gentle shaking for 18 h at 4 °C.
A. Hannappel et al. A conserved Trp specifies heme a binding in Surf1
FEBS Journal 278 (2011) 1769–1778 ª 2011 The Authors Journal compilation ª 2011 FEBS 1775
The TEV-digested material was again loaded on a nickel–
nitriloacetic acid column (CV = 5 mL) and washed with
buffer B. CtaA is found in the flow-through, whereas con-
taminating proteins, TEV protease and uncleaved CtaA
bind to the column material. The flow-through is concen-
trated and the protein concentration is determined by a
modified Lowry protocol [25,26].
Expression and purification of Surf1

Heterologous expression of apo-Surf1 in E. coli was per-
formed in the absence of the Paracoccus heme a-synthesiz-
ing machinery as described previously [16]. Expression of
Surf1 proteins in the presence of CtaB and CtaA (encoded
on pGR52) was performed in autoinductive ZYM-5052
medium supplemented with 10 lm iron (III) chloride,
according to Studier [18]. Nine litres of culture were inocu-
lated to 1% in baffled flasks and grown for 24 h at 32 °C.
Cells were harvested at 5000 g for 15 min and resuspended in
20 mm sodium phosphate, pH 8.0, 10 mm sodium chloride,
1mm EDTA. Membranes were prepared using established
methods and the protein concentration was determined using
the Lowry assay. The Surf1 proteins (apo-form, heme a
containing and site-directed mutants) were solubilized
from membranes and purified by a nickel–nitriloacetic acid
column as described previously [16].
Heme a transfer from CtaA to Surf1
CtaA (ba-form) and apo-Surf1c were mixed in buffer 1
(20 mm sodium phosphate, pH 8.0, 150 mm sodium chloride,
0.02% n-dodecyl-b-d-maltoside) to a final concentration of
0.75 mgÆmL
)1
CtaA and 0.895 mgÆmL
)1
apo-Surf1c, corre-
sponding to a molar ratio of CtaA ⁄ apo-Surf1c of 1 : 2. The
mixture was incubated for 1 h at 25 °C and the two proteins
were separated again on a nickel–nitriloacetic acid column
(CV = 5 mL; Qiagen). After loading the sample, the column
was washed with 5 CV of buffer 2 (50 mm sodium phos-

phate, pH 8.0, 300 mm sodium chloride, 0.02% n-dodecyl-b-
d-maltoside) and the CtaA containing flow through was col-
lected. A washing step with 5 CV of buffer 2 containing
50 mm imidazole followed, before Surf1c was eluted from the
column with 5 CV of buffer 2 containing 250 mm imidazole.
All fractions were collected, concentrated and washed with
buffer 1 to remove imidazole. After final concentration of
the samples, the protein concentration was determined by
the Lowry assay and samples were further analysed by
SDS ⁄ PAGE and UV ⁄ Vis spectroscopy.
Purification of cytochrome c oxidase and COX
activity measurement
COX purification from Paracoccus membranes was carried
out as described previously [27]. COX activity measurements
were performed at room temperature in buffer (20 mm
potassium phosphate, pH 8, 20 mm potassium chloride,
0.05% n-dodecyl-b-d-maltoside) with 20 lm reduced horse
heart cytochrome c as substrate on a Hitachi U-3000 spec-
trophotometer (De
550
(cyt c) = 19.4 mm
)1
Æcm
)1
).
PAGE
For SDS ⁄ PAGE, samples were denatured in SDS-containing
buffer for 20 min at 37 °C. Electrophoresis was performed
on 12% polyacrylamide gels according to Laemmli [28].
Spectral analysis

Redox difference spectra were recorded in the visible range
using potassium ferricyanide for oxidation and sodium
dithionite for reduction. For denaturing redox difference
spectra samples were taken up in 20% (v ⁄ v) pyridine, 0.1 m
NaOH, and heme a concentration was determined using
the extinction coefficient De
587–620
= 21.7 mm
)1
Æcm
)1
[29].
Heme extraction and HPLC analysis
For analysis of the Surf1c W200F and Surf1q W209F
mutant proteins, 10 mg were subjected to acidic acetone ⁄
ether extraction [30]. After evaporation of the ether, the
precipitate was resolved in 50 lL dimethylsulfoxide. The
resulting heme extract was diluted to 50% acetonitrile ⁄ TFA
and filtered through a 0.2 lm nanosep MF filter (Pall,
Dreieich, Germany). The hemes were separated on a lRPC-
C2 ⁄ C18 column (GE Healthcare, Mu
¨
nchen, Germany) using
a linear 50–100% acetonitrile gradient over 5 CV at 4 °C.
Reference samples of different heme types were prepared
from E. coli (hemes b and o) and from purified P. denitrif-
icans COX (heme a).
Acknowledgements
We thank Andrea Herrmann for excellent technical
assistance and Thuy Van-Tran and Sina Weidenweber

for performing initial experiments in the mutagenesis
project. We acknowledge financial support from DFG
(SFB 472 ‘Molecular Bioenergetics’, and Cluster of
Excellence ‘Macromolecular Complexes’ EXC 115).
References
1 Hosler JP, Ferguson-Miller S & Mills DA (2006)
Energy transduction: proton transfer through the respi-
ratory complexes. Annu Rev Biochem 75, 165–187.
2 Kaila VR, Verkhovsky MI & Wikstrom M (2010) Pro-
ton-coupled electron transfer in cytochrome oxidase.
Chem Rev 110, 7062–7081.
A conserved Trp specifies heme a binding in Surf1 A. Hannappel et al.
1776 FEBS Journal 278 (2011) 1769–1778 ª 2011 The Authors Journal compilation ª 2011 FEBS
3 Mogi T, Saiki K & Anraku Y (1994) Biosynthesis and
functional role of haem O and haem A. Mol Microbiol
14, 391–398.
4 Hederstedt L, Lewin A & Throne-Holst M (2005)
Heme A synthase enzyme functions dissected by muta-
genesis of Bacillus subtilis CtaA. J Bacteriol 187, 8361–
8369.
5 Mogi T (2009) Probing structure of heme A synthase
from Bacillus subtilis by site-directed mutagenesis.
J Biochem 145, 625–633.
6 Svensson B, Andersson KK & Hederstedt L (1996)
Low-spin heme A in the heme A biosynthetic protein
CtaA from Bacillus subtilis. Eur J Biochem 238,
287–295.
7 Leigh D (1951) Subacute necrotizing encephalomyelopa-
thy in an infant. J Neurol Neurosurg Psychiatry 14,
216–221.

8 Zhu Z, Yao J, Johns T, Fu K, De Bie I, Macmillan C,
Cuthbert AP, Newbold RF, Wang J, Chevrette M et al.
(1998) SURF1, encoding a factor involved in the bio-
genesis of cytochrome c oxidase, is mutated in Leigh
syndrome. Nat Genet 20, 337–343.
9 Tiranti V, Hoertnagel K, Carrozzo R, Galimberti C,
Munaro M, Granatiero M, Zelante L, Gasparini P,
Marzella R, Rocchi M et al. (1998) Mutations of
SURF-1 in Leigh disease associated with cytochrome c
oxidase deficiency. Am J Hum Genet 63, 1609–1621.
10 Van Coster R, Lombres A, De Vivo DC, Chi TL,
Dodson WE, Rothman S, Orrechio EJ, Grover W,
Berry GT, Schwartz JF et al. (1991) Cytochrome c
oxidase-associated Leigh syndrome: phenotypic features
and pathogenetic speculations. J Neurol Sci 104, 97–
111.
11 Nijtmans LG, Taanman JW, Muijsers AO, Speijer D &
Van den Bogert C (1998) Assembly of cytochrome-c
oxidase in cultured human cells. Eur J Biochem 254,
389–394.
12 Nijtmans LG, Artal Sanz M, Bucko M, Farhoud MH,
Feenstra M, Hakkaart GA, Zeviani M & Grivell LA
(2001) Shy1p occurs in a high molecular weight com-
plex and is required for efficient assembly of cyto-
chrome c oxidase in yeast. FEBS Lett 498, 46–51.
13 Barrientos A, Korr D & Tzagoloff A (2002) Shy1p is
necessary for full expression of mitochondrial COX1 in
the yeast model of Leigh’s syndrome. EMBO J 21, 43–
52.
14 Smith D, Gray J, Mitchell L, Antholine WE &

Hosler JP (2005) Assembly of cytochrome-c oxidase
in the absence of assembly protein Surf1p leads to loss
of the active site heme. J Biol Chem 280, 17652–
17656.
15 Bundschuh FA, Hoffmeier K & Ludwig B (2008) Two
variants of the assembly factor Surf1 target specific ter-
minal oxidases in Paracoccus denitrificans
. Biochim Bio-
phys Acta 1777, 1336–1343.
16 Bundschuh FA, Hannappel A, Anderka O & Ludwig B
(2009) Surf1, associated with Leigh syndrome in
humans, is a heme-binding protein in bacterial oxidase
biogenesis. J Biol Chem 284, 25735–25741.
17 Bestwick M, Jeong MY, Khalimonchuk O, Kim H &
Winge DR (2010) Analysis of Leigh syndrome muta-
tions in the yeast SURF1 homolog reveals a new mem-
ber of the cytochrome oxidase assembly factor family.
Mol Cell Biol 30, 4480–4491.
18 Studier FW (2005) Protein production by auto-induc-
tion in high density shaking cultures. Protein Expr Purif
41, 207–234.
19 Greiner P, Hannappel A, Werner C & Ludwig B (2008)
Biogenesis of cytochrome c oxidase – in vitro
approaches to study cofactor insertion into a bacterial
subunit I. Biochim Biophys Acta 1777, 904–911.
20 Brown BM, Wang Z, Brown KR, Cricco JA & Hegg
EL (2004) Heme O synthase and heme A synthase
from Bacillus subtilis and Rhodobacter sphaeroides
interact in Escherichia coli. Biochemistry 43, 13541–
13548.

21 Brown KR, Brown BM, Hoagland E, Mayne CL &
Hegg EL (2004) Heme A synthase does not incorporate
molecular oxygen into the formyl group of heme A.
Biochemistry 43, 8616–8624.
22 Wang Z, Wang Y & Hegg EL (2009) Regulation of the
heme A biosynthetic pathway: differential regulation of
heme A synthase and heme O synthase in Saccharomy-
ces cerevisiae. J Biol Chem 284, 839–847.
23 Mick DU, Wagner K, van der Laan M, Frazier AE,
Perschil I, Pawlas M, Meyer HE, Warscheid B & Reh-
ling P (2007) Shy1 couples Cox1 translational regulation
to cytochrome c oxidase assembly. EMBO J 26, 4347–
4358.
24 Piekutowska-Abramczuk D, Magner M, Popowska E,
Pronicki M, Karczmarewicz E, Sykut-Cegielska J,
Kmiec T, Jurkiewicz E, Szymanska-Debinska T,
Bielecka L et al. (2009) SURF1 missense mutations
promote a mild Leigh phenotype. Clin Genet 76, 195–
204.
25 Lowry OH, Rosebrough NJ, Farr AL & Randall RJ
(1951) Protein measurement with the Folin phenol
reagent. J Biol Chem 193, 265–275.
26 Markwell MA, Haas SM, Bieber LL & Tolbert NE
(1978) A modification of the Lowry procedure to sim-
plify protein determination in membrane and lipopro-
tein samples. Anal Biochem 87, 206–210.
27 Hendler RW, Pardhasaradhi K, Reynafarje B & Lud-
wig B (1991) Comparison of energy-transducing capa-
bilities of the two- and three-subunit cytochromes aa3
from Paracoccus denitrificans and the 13-subunit beef

heart enzyme. Biophys J 60, 415–423.
28 Laemmli UK (1970) Cleavage of structural proteins
during the assembly of the head of bacteriophage T4.
Nature 227, 680–685.
A. Hannappel et al. A conserved Trp specifies heme a binding in Surf1
FEBS Journal 278 (2011) 1769–1778 ª 2011 The Authors Journal compilation ª 2011 FEBS 1777
29 Williams JN Jr (1964) A Method for the simultaneous
quantitative estimation of cytochromes a, b, c
1
, and c in
mitochondria. Arch Biochem Biophys 107, 537–543.
30 Weinstein JD & Beale SI (1983) Separate physiological
roles and subcellular compartments for two tetrapyrrole
biosynthetic pathways in Euglena gracilis. J Biol Chem
258, 6799–6807.
31 Adamczak R, Porollo A & Meller J (2005) Combining
prediction of secondary structure and solvent accessibil-
ity in proteins. Proteins 59, 467–475.
Supporting information
The following supplementary material is available:
Fig. S1. Gel electrophoresis of purified Paracoccus den-
itrificans CtaA proteins.
Fig. S2. Native redox difference spectra of purified
CtaA proteins.
Fig. S3. Pyridine hemochromogen redox spectra of
purified CtaA proteins.
Table S1. Strains used for the heterologous expression
of CtaA from Escherichia coli.
This supplementary material can be found in the
online version of this article.

Please note: As a service to our authors and readers,
this journal provides supporting information supplied
by the authors. Such materials are peer-reviewed and
may be re-organized for online delivery, but are not
copy-edited or typeset. Technical support issues arising
from supporting information (other than missing files)
should be addressed to the authors.
A conserved Trp specifies heme a binding in Surf1 A. Hannappel et al.
1778 FEBS Journal 278 (2011) 1769–1778 ª 2011 The Authors Journal compilation ª 2011 FEBS