acute regulatory lipid-transfer START domains of StAR STARD1 and MLN64 STARD3 Julian Reitz1, Katja Gehrig-Burger1, Jerome F.. Among the START proteins are the StAR protein itself STARD1
Trang 1acute regulatory lipid-transfer (START) domains of StAR (STARD1) and MLN64 (STARD3)
Julian Reitz1, Katja Gehrig-Burger1, Jerome F Strauss III2and Gerald Gimpl1
1 Institute of Biochemistry, Gutenberg-University Mainz, Germany
2 Department of Obstetrics & Gynecology, Virginia Commonwealth University, Richmond, VA, USA
Cholesterol is an essential multifunctional lipid in most
eukaryotic cells It exerts a strong influence on the
physical state of the plasma membrane, forms
choles-terol–sphingolipid-rich microdomains such as caveolae
and lipid rafts, is necessary for the activity of several
membrane proteins, and serves as the precursor for
steroid hormones [1–5] Despite many efforts, the
path-ways and mechanisms of cellular cholesterol trafficking
are currently not well understood Misfunctions of
cholesterol transport are linked to a variety of diseases
[6,7]
The biosynthesis of steroid hormones requires the
transfer of cholesterol from multiple sources to the
inner mitochondrial membrane, where steroidogenesis
begins with the conversion of cholesterol to
pregneno-lone The translocation of cholesterol to the inner
mitochondrial membrane, the rate-limiting step in steroidogenesis, is mediated by steroidogenic acute regulatory protein (StAR, STARD1) [8–12] The mechanism by which STARD1 moves cholesterol to the inner mitochondrial membrane is currently unclear [13] Mutations that inactivate STARD1 in humans lead to an impaired ability of the adrenal gland to pro-duce steroid hormones, a potentially lethal disease known as congenital lipoid adrenal hyperplasia [14] Ablation of the StarD1 gene in mice also causes impaired steroidogenesis and adrenal lipid accumula-tion [15] STARD1 is synthesized as a 37 kDa phos-phoprotein with an N-terminal mitochondrial targeting sequence that is cleaved during mitochondrial entry (Fig 1A) Deletion of 62 N-terminal residues (N-62 STARD1), including the leader peptide, resulted in a
Keywords
cholesterol; MLN64; STARD1; STARD3;
START proteins
Correspondence
G Gimpl, Institute of Biochemistry,
Gutenberg-University Mainz, Becherweg 30,
55128 Mainz, Germany
Fax: +49 6131 3925348
Tel: +49 6131 3923829
E-mail: gimpl@uni-mainz.de
(Received 14 January 2008, revised 5
Febru-ary 2008, accepted 14 FebruFebru-ary 2008)
doi:10.1111/j.1742-4658.2008.06337.x
The steroidogenic acute regulatory (StAR)-related lipid transfer (START) domains are found in a wide range of proteins involved in intracellular trafficking of cholesterol and other lipids Among the START proteins are the StAR protein itself (STARD1) and the closely related MLN64 protein (STARD3), which both function in cholesterol movement We compared the cholesterol-binding properties of these two START domain proteins Cholesterol stabilized STARD3-START against trypsin-catalyzed degrada-tion, whereas cholesterol had no protective effect on STARD1-START [3H]Azocholestanol predominantly labeled a 6.2 kDa fragment of STARD1-START comprising amino acids 83–140, which contains residues proposed to interact with cholesterol in a hydrophobic cavity Photoaffinity labeling studies suggest that cholesterol preferentially interacts with one side wall of this cavity In contrast, [3H]azocholestanol was distributed more or less equally among the polypeptides of STARD3-START Overall, our results provide evidence for differential cholesterol binding of the two most closely related START domain proteins STARD1 and STARD3
Abbreviations
MLN64 (= STARD3), metastatic lymph node 64; MbCD, methyl-b-cyclodextrin; NBD-cholesterol, 22-[N-(7-nitrobenz-2-oxa-1,3-diazol-4-yl)amino]-23,24-bisnor-5-cholen-3-ol; SELDI, surface-enhanced laser desorption/ionization; StAR (= STARD1), steroidogenic acute regulatory protein; START, steroidogenic acute regulatory protein lipid-transfer domain.
Trang 2cytosolic protein with full activity, as shown in intact
cells and in isolated mitochondria [16–18] The
func-tionally active C-terminal domain of STARD1
con-tains the StAR-related lipid-transfer (START) domain
START domains consist of 200–210 amino acids and are found in a wide range of proteins involved in several cellular functions, including lipid transport, signal transduction, and transcriptional regulation [19] Among the START proteins are the StAR protein itself (STARD1) and the closely related metastatic lymph node 64 (MLN64) protein (STARD3) Both proteins function as cholesterol-binding proteins [20,21] Their START domains share 37% sequence identity
STARD3 is overexpressed in certain breast cancers [22] The protein contains four transmembrane helices that target it to the membrane of late endosomes [23] (Fig 1A) However, the physiological function of STARD3 is currently unclear It may be involved in steroidogenesis in the human placenta, which lacks STARD1 [24,25] The START domain at the C-termi-nal half of STARD3 is believed to be exposed to the cytosol In its isolated form, STARD3-START is able
to promote steroidogenesis even more efficiently than intact STARD3 [26] The crystal structure of the unli-ganded START domain of human STARD3 has been resolved [20] This structure shows a hydrophobic tun-nel that expands throughout the length of the START domain and is perfectly sized to accommodate a single cholesterol molecule [20] A similar structure has been reported for the cholesterol-regulated START pro-tein 4 (STARD4) [27] For another START propro-tein, the phosphatidylcholine transfer protein (STARD2), it has been directly shown that the tunnel represents the binding site of the lipid, in this case phosphatidylcho-line [28]
To understand the molecular mechanism how cho-lesterol is transferred by STARD1 and STARD3, the cholesterol-binding sites of these proteins have to be identified As a crystal structure of a cholesterol– START complex is not yet available, other methods are required to explore the cholesterol–protein interac-tion One approach is molecular modeling based on the knowledge of the unliganded STARD3 structure Two such modeling studies have been recently per-formed for the START domains of STARD1 and STARD3 [29,30] This led to the proposal that STARD1-START shuttles cholesterol carried in its hydrophobic cavity between the outer and inner mito-chondrial membranes [20] However, spectral and bio-chemical data supported the view that STARD1 partially unfolds and forms molten globules in the low-pH environment of the outer mitochondrial membrane These intermediates were hypothesized to facilitate the cholesterol transfer of STARD1 to the mitochondrial inner membrane through a mechanism that does not involve sterol shuttling [31,32]
A
START
START
97
66
45
31
21
14
– – –
–
– –
C
m/z
m/z
0
5
10
15
20
0 10 20 30 40
50
29162.8+H 26167.8+H
Fig 1 Expression of the START domains of STARD1 and STARD3.
(A) Domain organization of the START proteins STARD1 (285 amino
acids) and STARD3 (445 amino acids) Both proteins possess a
ste-rol-binding START domain ( 200 amino acids) in their C-terminal
regions The N-terminal targeting sequence of STARD1 is cleaved
upon entry into the mitochondria, and is nonessential for the activity
of STARD1 [16–18] The N-terminal part of STARD3 possesses four
transmembrane segments that target the protein to late endosomes.
The START domain in STARD3 is exposed to the cytosol and is
func-tionally active in its isolated form [26] (B) Purification of the START
domains of STARD1 and STARD3 expressed in Escherichia coli The
proteins were purified from E coli, resolved by SDS ⁄ PAGE, and
identified by Coomassie blue staining Lane 1: marker Lane 2:
STARD1-START (2 lg of protein) Lane 3: STARD3-START (6 lg of
protein) (C) SELDI-TOF of STARD1-START and STARD3-START.
Trang 3Here, we analyzed the cholesterol-binding
character-istics of the two most related START proteins,
STARD1 and STARD3 Photoaffinity labeling with
radiolabeled 6-azocholestanol as the photoreactive
cho-lesterol probe was employed to characterize and
com-pare the cholesterol binding of the START domains
This cholesterol analog (previously often termed
photocholesterol) has already been successfully applied
for various proteins [23,33–36] Overall, this study
addresses the question of whether or not the related
START domains of StARD1 and StARD3 interact
with cholesterol in a similar manner
Results
Expression of the START domains
The recombinant START proteins each contain a His6
-tag at their C-terminus The proteins were expressed in
BL21 Escherichia coli and purified by affinity
chroma-tography using an Ni2+–nitrilotriacetic acid agarose
matrix Figure 1B shows the Coomassie stains of the
purified proteins The apparent molecular masses of the
His-tag START proteins in the SDS⁄ PAGE system were
slightly greater than the calculated molecular masses of
25 769 Da (pI 6.42) and 26 847 Da (pI 8.43) for
STARD1-START and STARD3-START, respectively
(Fig 1B) This discrepancy has also been observed by
Arakane et al [17] in the case of STARD1-START To
explore this issue, we also determined the molecular
masses of both START proteins by surface-enhanced
laser desorption/ionization (SELDI)-TOF MS
Molecu-lar masses of 26 167 and 29 162 Da were found for
STARD1-START and STARD3-START, respectively
(Fig 1C) Whereas the molecular mass of
STARD1-START is relatively close (+398 Da) to the calculated
value of 25.7 kDa, the mass of STARD3-START is
about 2.3 kDa higher than that calculated for the
unmodified polypeptide This could reflect
post-transla-tional protein modification The expression levels of
STARD1-START and STARD3-START were similar
Cholesterol binding of the START proteins
In order to verify the cholesterol binding of the
START proteins, we used the fluorescent cholesterol
reporter
22-[N-(7-nitrobenz-2-oxa-1,3-diazol-4-yl)amino]-23,24-bisnor-5-cholen-3-ol (NBD-cholesterol) This
cho-lesterol analog has successfully been employed to
analyze the cholesterol binding of STARD1-START
[21,31] A strong increase in the fluorescence intensity
of NBD-cholesterol occurs when the ligand binds to
the hydrophobic environment of the START proteins
This has recently been studied in detail by Petrescu et al [21] in the case of STARD1-START The binding of NBD-cholesterol to each of the START proteins shows
a saturating profile (supplementary Fig S1A,B) The curves were fitted using a nonlinear regression algorithm according to one-site models, and yielded KD values of 161 ± 45 nm (n = 3) for STARD1-START and 58 ± 16 nm (n = 3) for STARD3-START Thus, STARD3-START bound NBD-cholesterol with a slightly higher affinity than did STARD1-START Two-site models did not result in significantly better fittings of the binding data
According to one model of START domain action,
a pH-dependent molten globule transition of STARD1
is required for sterol transfer activity at the level of the mitochondrial outer membrane [31,32] Therefore, we also measured the fluorescence of NBD-cholesterol (500 nm) bound to STARD1-START (10 nm) at an acidic pH At pH 3, the sterol binding of STARD1-START was about three-fold lower than the sterol binding measured at pH 7.4 (data not shown)
Analysis of the stabilizing effect of cholesterol
on START proteins Cholesterol and its analogs are able to stabilize pro-teins against proteolysis or thermal degradation [37]
To test whether this occurs in the case of the START proteins, we analyzed the migration behavior of these proteins in SDS gels under various conditions
First, the START proteins were incubated (for
20 min at 25C) in the presence of cholesterol, photo-cholesterol, or buffer control The proteins were irradi-ated with UV light for 10 min prior to separation by SDS⁄ PAGE, western blotting, and immunodetection with antibody to His (supplementary Fig S2A) It is important to note that the His-tag is localized at the C-terminus of both proteins, so that only molecular species with an intact C-terminus are visible on the immunoblots The immunoblot revealed no significant differences among treated and untreated START pro-teins Faint staining was observed for the putative dimer forms of the proteins in addition to the predom-inant monomer ( 30 kDa) bands We did not find a slight increase in the molecular size of the START pro-teins in the photoactivated samples of the photocholes-terol-containing samples Most probably, the labeled species is below the detection limit, due to the low photoaffinity yield (< 9%)
We next analyzed the resistance of the START pro-teins to degradation in the presence and absence of cholesterol The proteins were pretreated either with buffer solution or cholesterol–methyl-b-cyclodextrin
Trang 4(MbCD) (0.1 mm) for 20 min at 25C Then, the
sam-ples were incubated for increasing times (6 h, 24 h,
80 h) at 40C prior to separation by SDS ⁄ PAGE,
western blotting, and immunodetection with antibody
to His (supplementary Fig S2B) For
STARD1-START, we did not observe any evidence of
degrada-tion during the time course of this experiment In
contrast, in the case of STARD3-START, an
addi-tional band with a slightly decreased apparent
molecu-lar mass (by 3–4 kDa) appeared after an incubation
period of 24 h or longer The presence of cholesterol
did not influence the appearance of this additional
band (supplementary Fig S2B)
When the samples were treated with trypsin (10 min
or 40 min at 37C), additional bands were observed
on the immunoblots for both START proteins
(Fig 2) Two additional molecular species with slightly
higher electrophoretic mobilities appeared for STARD1-START The presence of cholesterol did not inhibit the appearance of these additional bands, nor did it affect the protein patterns of the immunoblots STARD3-START was more sensitive to trypsinolysis (Fig 2) When trypsin was incubated for 40 min, most
of the STARD3-START was either totally degraded
or, more probably, had its C-terminus bearing the His-tag cleaved Incubations with trypsin for more than
60 min resulted in immunoblots with no detectable START proteins (not shown) However, cholesterol was clearly able to inhibit the trypsinolysis of STARD3-START (Fig 2)
Cholesterol labeling of STARD1-START
To determine the cholesterol docking site within the START domains of STARD1 and STARD3, we per-formed photoaffinity labeling with [3
H]photocholester-ol and subsequent chemical or enzymatic cleavage of the photoactivated samples Highly reproducible frag-mentation patterns were obtained when the protein was subjected to chemical cleavage by cyanogen bromide (CNBr), which hydrolyzes peptide bonds C-terminal to Met residues The predicted cleavage products are listed in Table 1 for STARD1-START
In the case of STARD1-START, the [3 H]photocholes-terol radiolabel was incorporated nearly quantitatively into a single band at about 6.2 kDa (Fig 3) Even when we increased the protein amounts from 20 lg (Fig 3, filled symbols) to 60 lg (Fig 3, open symbols), the label was predominantly incorporated in a
6.2 kDa fragment A control labeling of STARD1-START with [3H]photocholesterol but without UV irradiation did not reveal any bands (Fig 3, dia-monds) Similarly, when cholesterol was added to the samples at a ‡ 50-fold molar excess over [3 H]photo-cholesterol, the appearance of the 6.2 kDa fragment
+ + + + – + + + +
–
Try
+ – + – – + – + –
–
Cho
40´
10´
40´
10´
STARD3-START STARD1-START
31–
Fig 2 Stability of the START domains of human STARD1 and
STARD3 in the presence or absence of cholesterol The START
pro-teins (1 lgÆlL)1) were preincubated with buffer solution or
choles-terol-MbCD (Cho) (0.1 m M ) for 20 min at 25 C Then, the samples
were incubated in the presence of trypsin (Try) for 10 min or
40 min at 37 C The proteins were precipitated with acetone,
dis-solved in water, separated by SDS ⁄ PAGE, and subjected to
wes-tern blotting, using antibody to His and Amersham ECL Plus for
detection.
Table 1 Cleavage and fragmentation of STARD1-START by CNBr The molecular mass data are calculated average masses [M + H] + according to the program PEPTIDE MASS (Expasy).
Trang 5was suppressed (not shown) A predicted fragment of
this size (6236 Da) corresponds to STARD1-START
residues 83–140, as listed in Table 1 Owing to partial
cleavage, CNBr fragments with sizes similar to the
6236 Da species are possible, such as the combined
fragments with molecular masses of 5185 Da
(= 2300 + 2885 Da), 5179 Da (= 2885 + 2294 Da),
and 6598 Da (= 2885 + 2294 + 1419 Da) To
deter-mine whether partially cleaved fragments are present
within this molecular range, we performed MS (see
inset in Fig 3) The sample for SELDI-TOF MS was
prepared as described for STARD1-START, except
that unlabeled photocholesterol was used instead of
[3H]photocholesterol In the mass spectrum, two major
peaks are observed within the molecular range 4000–
7000 m⁄ z, a 5194 Da species and a 6263 Da species
The 5194 Da species could represent either the
com-bined 5185 Da fragment or the (possibly oxidized)
par-tially uncleaved 5179 Da fragment The 6263 Da peak
should represent the 6236 Da fragment, perhaps
modi-fied by formylation (+26 Da) Covalent coupling of one molecule of photocholesterol should add a mass of about 386 Da to the 6236 Da fragment, resulting in a
6.6 kDa species A small shoulder area to the right
to the 6263 Da peak (Fig 3, inset) might include such
a species However, a partial uncleaved 6598 Da frag-ment (see above) would overlap with this species and does not allow us to reach a definite conclusion on this point STARD1-START protein labeled with photo-cholesterol and cleaved by CNBr did not reveal sub-stantial differences in the mass spectra in comparison with samples untreated with photocholesterol prior to cleavage with CNBr, probably because of the low photoaffinity yield (< 9%), which results in the labeled species being below the detection limit
Affinity labeling with [3H]photocholesterol and subsequent CNBR cleavage were carried out for STARD1-START at neutral and acidic pH Typical fragmentation profiles are demonstrated in Fig 4A (at neutral pH) and Fig 4B (at acidic pH) Quantitation
of the results is shown in Table 2 Cholesterol labeling
of the 6.2 kDa fragment was lower at pH 3.0 than at
pH 7.4 Moreover, in gel slices at and close to the gel front, a markedly higher incorporation of radioactivity was found at acidic pH than at neutral pH These gel slices contain oligopeptide fragments with molecular masses < 2 kDa, including unbound [3 H]photocholes-terol According to the fragmentation pattern (Table 1), these could represent peptides with molecular masses of
1705, 751, and 302 Da Obviously, at pH 3, the choles-terol labeling of STARD1-START is less specific than the labeling at pH 7.4
Cholesterol labeling of STARD3-START
In case of STARD3-START, photoaffinity labeling with [3H]photocholesterol and subsequent CNBr cleav-age revealed several peaks, which were numbered from
1 to 5 (Fig 5, circles) The predicted cleavage products for STARD3-START are listed in Table 3 Peak 1 cor-responds to molecular mass > 26.6 kDa, and should represent uncleaved STARD3-START Peaks 2 and 3 can be assigned to the predicted fragments of
13 262 Da (residues 93–212) and 10 556 Da (resi-dues 1–92), respectively (Table 3) Peak 4 corresponds
to the fragment of size 2972 Da (residues 213–236) Peak 5 represents unbound [3H]photocholesterol (Fig 5, dotted line) SELDI-TOF of CNBr-cleaved STARD3-START revealed major peaks oat 3187,
11 575, 14 332, and 25 918 Da, and a minor peak at
29 152 Da (not shown) The 25 918 Da species ( 11 575 + 14 332 Da) should be partially cleaved polypeptide Thus, each of the masses of the three
Gel slice number
0 10 20 30 40 50 60 70 80 90 100
0
5000
10 000
15 000
20 000
25 000
30 000
35 000
26.6 17.0 14.4 6.5 3.5 1.4
m/z
0
2
4
6
8
6263.1+H
5194.9+H
Fig 3 Cholesterol labeling and chemical cleavage of
STARD1-START STARD1-START (20 lg of protein, filled circles and
dia-monds, and 60 lg of protein, open circles) was incubated with
[ 3 H]photocholesterol (50 l M ) for 20 min at 25 C Then, the
sam-ples were either UV-irradiated (circles) or not UV-irradiated (control,
diamonds) for 10 min at 4 C The protein was precipitated with
acetone, dissolved in water, and subjected to chemical cleavage
by CNBr for 24 h at 37 C The proteins were separated by
SDS ⁄ PAGE The gel was cut into 1 mm slices and incubated
over-night at room temperature with a scintillation cocktail The
radioac-tivity of each slice was counted The molecular mass (in kDa) was
estimated from a control lane loaded with molecular size markers,
and is given at the top of each panel The reference line (dotted)
corresponds to unbound [ 3 H]photocholesterol The inset shows a
SELDI-TOF mass spectrum of STARD1-START cleaved by CNBr in
(and calibrated for) the mass range 4000–7000 m ⁄ z The sample
for MS was prepared as described, except that unlabeled
photo-cholesterol was used instead of [ 3 H]photocholesterol.
Trang 6fragments is higher (215–1070 Da) than calculated for
the corresponding unmodified polypeptide This
sug-gests that unknown post-translational protein
modifi-cations are more or less equally distributed along the
length of the protein In control experiments in the
presence of an excess of unlabeled cholesterol, low
amounts of radioactivity were detected in the gel slices
over the whole length of the gel (except at peak 5,
cor-responding to unbound photocholesterol) (Fig 5,
dia-monds) Similar low amounts of radioactivity were
observed when the START protein was denaturated
by heat (5 min at 95C) (not shown)
Discussion
We have explored the cholesterol binding of the
START domains of the two most related START
pro-teins, STARD1 and STARD3 Both proteins bound
the fluorescent cholesterol reporter NBD-cholesterol
with high affinity With respect to the sterol binding of
STARD1-START, our results were within the range previously reported [21] Cholesterol is able to stabilize proteins, e.g by protecting them from thermal dena-turation or proteolytic degradation, as shown for the oxytocin receptor [37], the Torpedo californica acetyl-choline receptor [38], and rhodopsin [39] When STARD3-START was incubated for many hours (24–
80 h) at 40C, an additional band (truncated by
3 kD in apparent molecular mass) appeared in immunoblots This additional molecular species could represent either a denaturated form of the protein with higher electrophoretic mobility or an N-terminal trun-cated fragment of STARD3-START resulting from cleavage by a protease still present in our preparation
In each case, the presence of cholesterol was not able
to suppress the appearance of this additional molecular species However, cholesterol had a protective effect against the trypsinolysis of STARD3-START, whereas the cleavage of STARD1-START was not affected Both START proteins possess several cleavage sites
Table 2 Efficiency of labeling of the 6.2 kDa fragment with [ 3 H]photocholesterol in STARD1-START Labeling was performed with [3H]photocholesterol (50 l M ) and STARD1-START (5 l M ) The samples were UV-irradiated for 10 min at 4 C at the indicated pH in a volume
of 100 lL The protein was precipitated with acetone, dissolved in water, and subjected to chemical cleavage by CNBr for 24 h at 37 C The proteins were separated by SDS⁄ PAGE The gel was cut into 1 mm slices The slices were incubated with scintillation cocktail, and the radioactivity of each slice was counted To calculate the labeling efficiency, the radioactivity in the peak area ( 15 slices) corresponding to a molecular mass of 6.2 kDa was integrated Control samples were treated under the same conditions except for the UV crosslinking step These control values (integrated radioactivity of 15 slices corresponding to a molecular mass of 6.2 kDa) were subtracted from the sample data Labeling efficiency is the amount of [3H]photocholesterol incorporated into the 6.2 kDa fragment of STARD1-START (0.5 nmol), with 100% being equal to 0.5 nmol of the photolabel The data are means ± SD (n = 3) To obtain the relative labeling efficiencies, the data were normalized to 100%.
Gel slice number
Radioactivity (dpm) Radioactivity (dpm)
0
15 000
A
Gel slice number
0 10 20 30 40 50 60 70 80 90 100 0 10 20 30 40 50 60 70 80 90 100
0
15 000
B
*
*
Fig 4 Cholesterol labeling and CNBr cleavage of STARD1-START at different pH values The START proteins (each 20 lg of protein) were incubated with [ 3 H]photocholesterol (50 l M ) for 20 min at 25 C at pH 7.4 (A) or pH 3.0 (B) Then, the samples were UV-irradiated for
10 min at 4 C The protein was cleaved by CNBr and further processed as described in the legend for Fig 3 The asterisks mark the position of the 6.2 kDa band The reference lines (dotted) correspond to the gel front line containing unbound [ 3 H]photocholesterol and fragments of less than 1 kDa.
Trang 7(Arg and Lys residues) for trypsin within their
N-ter-minal sequence, which could lead to the observed
frag-mentation pattern One simple explanation of the data
is that the N-terminal region of STARD3-START
directly interacts with cholesterol, thus impeding the
access of trypsin Alternatively, cholesterol could
stabi-lize a conformation of the protein that is more
resis-tant to trypsinolysis
What is known about the cholesterol-binding site of
the START domains of STARD1 and STARD3? The
crystal structure of human STARD3-START revealed
an a⁄ b-fold consisting of a nine-stranded twisted
b-sheet and four a-helices [20] The START domains
of STARD3 [20], STARD4 [27], phosphatidylcholine
transfer protein [28,40], and related bacterial proteins
share this basic structure [41,42] A STARD1-START model based on the structure of STARD3-START is shown in Fig 6A,B in two views The view in Fig 6B
is related to that in Fig 6A by a 90 rotation about the y-axis The b-strands in the order b1–b2–b3–b9–b8–
b7–b6–b5–b4 form a U-shaped unclosed b-barrel with a predominant hydrophobic cavity that is optimally sized
to bind a single cholesterol molecule (Fig 6A) The roof of the cavity is mainly formed by the C-terminal
a4-helix The access of cholesterol to this cavity may
be enabled by conformational changes of the a4-helix and the adjacent loops In the case of STARD1-START, we have identified a 6.2 kDa fragment comprising amino acids 83–140 as a major cholesterol-binding site (Fig 7, residues 83–140, highlighted in gray) The corresponding structures, colored yellow in Fig 6A,B, are the b-strands b7–b6–b5–b4 including the W3-loop (connecting b5 and b6) and part of the
a3-helix This suggests that cholesterol bound in the cavity is preferentially in contact with one side wall of this cavity The geometry of the cavity in STARD1-START is well suited for a ligand with the size and shape of cholesterol [29,30] Critical residues proposed
to interact with cholesterol are localized within the fragment containing amino acids 83–140 These resi-dues are in magenta in Fig 6B For example, the acidic side chain of Glu107 in STARD1-START (Glu169 in STARD1) (corresponding to Asp117 in STARD3-START) was proposed to be involved in specific cholesterol binding, most likely with the 3b-hydroxyl group of cholesterol [20] Cholesterol might also interact with the conserved and buried Arg residue at position 126 in STARD1-START (Arg136
in STARD3-START) [20] The charged residues Glu107 and Arg126 in human STARD1-START, which are equivalent to Glu168 and Arg187 in the hamster STARD1 model, were found to form a salt bridge at the bottom of the hydrophobic pocket of the START domain [29,30] In STARD3-START, these residues may interact with the 3b-hydroxyl group of cholesterol via hydrogen bonding to an included water molecule [30], as was concluded from molecular
Gel slice number
0 10 20 30 40 50 60 70
0
1000
2000
3000
4000
5000
6000
7000
26.6 17.0 14.4 6.5 3.5 1.4
1
2 3
4 5
Fig 5 Cholesterol labeling and chemical cleavage of
STARD3-START The protein (20 lg) was incubated with [ 3
H]photocholes-terol (50 l M ) for 20 min at 25 C As a control, STARD3-START
(20 lg) was incubated with [3H]photocholesterol (50 l M ) in the
presence of a 50-fold molar excess of cholesterol (diamonds).
Then, the samples were UV-irradiated, cleaved by CNBr, and
further processed as described in the legend for Fig 3 The
molecular mass (in kDa) was estimated from a control lane
loaded with molecular size markers, and is given at the top of
panel The reference line (dotted) corresponds to unbound
[ 3 H]photocholesterol.
Table 3 Cleavage and fragmentation of STARD3-START by CNBr The molecular mass data are calculated average masses [M + H]+ according to the program PEPTIDE MASS (Expasy).
Molecular mass
TVYTIEVPFHGKTFILKTFLPCPAELVYQEVILQPERM
LSSGIATSHSAKPPTHKYVRGENGPGGFIVLKSASNPRVCTFVWILNTDLKGRLPRYLIHQSLAATM
Trang 8modeling and structure-based thermodynamics [29,30].
Water molecules were in fact discovered inside the
STARD3 crystal [20] The replacement of the two
charged residues Glu107 and Arg126 in
STARD1-START by hydrophobic residues of similar volume
resulted in the total loss of STARD1 activity [30]
According to molecular modeling, another residue
located within the 6.2 kDa fragment could be involved
in cholesterol interaction: Leu137 (Leu199) in STARD1-START (STARD1), and the corresponding Ser147 (Ser362) in STARD3-START (STARD3) [29,30] In STARD1-START, cholesterol might contact Leu137 indirectly, mediated by at least one water mol-ecule, whereas in STARD3-START cholesterol was suggested to form a direct hydrogen bond with Ser147 [29,30] Nevertheless, the major contributions to the
C
N
β4
α1
α4 Ω3
Ω2
Ω1
β5 β6 β7
β1 β2 β3 α2 α3
β8 β9
N
C
E L R
Fig 6 Model of STARD1-START The model was build after sequence alignment of STARD1-START with STARD3-START, for which a crys-tal structure is known [20] For a better depiction of the elongated hydrophobic pocket, the same ribbon diagram is displayed from two different views [(A) and (B)] using the program CHIMERA [51] The view in (B) is related to that in (A) by a 90 rotation about the y-axis The photocholesterol docking region is shown in yellow, and comprises half of the a3-helix and the strands b3–b7, including their connecting loops The residues Glu107 (E), Arg126 (R) and Leu137 (L) (all marked in magenta) are located within this region and have been proposed to interact with cholesterol (see Discussion) Otherwise, the model is colored according to the secondary structure, with helices in red, b-strands in green, and loops in gray.
Fig 7 Alignment of the START domains of human STARD1 and STARD3 Sequence identities are marked by a star, and residues contribut-ing to the tunnel in STARD3 are marked in bold STARD1 missense mutations causcontribut-ing congenital adrenal hyperplasia are underlined The numbering of residues within the whole sequences of STARD3 and STARD1, respectively, is in parentheses STARD1-START and STARD3-START share 37% sequence identity and 60% amino acid similarity Residues 83–140, corresponding to the photocholesterol-interacting fragment in STARD1-START, are marked in bold and highlighted in gray.
Trang 9energy of cholesterol binding are most likely provided
by nonpolar contacts with side chains lining the
hydro-phobic cavity of STARD1-START [29]
In contrast to STARD1-START, STARD3-START
did not show preferential incorporation of
photocho-lesterol into a single polypeptide If one assumes the
same cholesterol-binding site as in STARD1-START,
one should expect that photocholesterol is primarily
incorporated into the CNBr fragment 93–212
How-ever, this was clearly not the case Instead, cholesterol
labeling of STARD3-START was distributed more or
less equally among the three fragments This could
indicate that the cholesterol molecule localized within
the binding pocket of STARD1-START possesses a
lower degree of freedom than the cholesterol molecule
inside the tunnel of STARD3-START Although both
START domains show high structural similarity, a
recent modeling approach provided evidence for slight
differences in the orientation of the cholesterol ring
within their cavities that may result in distinct contact
sites for photocholesterol [29]
How is the nearly solvent-inaccessible cavity opened
or closed in response to cholesterol loading and
release? Access into the cavity is mainly occluded by
the C-terminal a4-helix and the adjacent loops
(Fig 6A) Conformational changes of the amphipathic
a4-helix allow opening of the cavity This scenario is
supported by spectroscopic measurements
demonstrat-ing a loss of helical structure in STARD1 after binddemonstrat-ing
of the cholesterol reporter NBD-cholesterol [21] The
a4-helix is believed to contact the phospholipid bilayer
of the outer mitochondrial membrane [43] According
to one hypothesis, STARD1 thereby undergoes an
acid-inducible structural change to a molten globule
state [44] Biophysical data provided evidence for a
stronger association of STARD1 with the
mitochon-drial outer membrane (e.g with the protonated
phos-pholipid head groups) at an acidic pH ( 3.5) [45]
We show here that under acidic pH conditions, the
efficiency in photocholesterol labeling of
STARD1-START was significantly but not dramatically
decreased Thus, a putative molten globule state of
STARD1-START might be slightly more capable
of releasing its bound cholesterol However, the
STARD1-mediated translocation of cholesterol into
the mitochondria is not well understood Probably,
STARD1 acts in concert with other proteins, such as
STARD4 and the peripheral benzodiazepine receptor,
to transfer cholesterol from the outer to the inner
membrane of the mitochondrion [43,46]
Taken together, our observations provide evidence
for differential cholesterol interactions with the two
most closely related START proteins The importance
of the cholesterol-binding site in STARD1-START is underlined by the fact that several disease-related mutations or truncations in human STARD1 appear
to correspond to residues lining the interior of the hydrophobic cavity, or in the C-terminal a-helix, when mapped onto the STARD3-START structure [14,18,47]
However, it is important to mention that any con-clusions drawn from studies employing cholesterol analogs such as NBD-cholesterol or photocholesterol have to be judged with caution [35] For example, photocholesterol is structurally different from choles-terol, having, associated with the B-ring, an additional ring structure consisting of two nitrogen atoms, and could be involved in significantly different interactions (e.g hydrogen bonding) with certain amino acid side chains Thus, it cannot be excluded that the difference
in photocholesterol binding does not truly reflect a dif-ference in binding of native cholesterol An ultimate understanding of the interaction of cholesterol with START proteins requires the structure(s) of choles-terol-occupied START proteins
Experimental procedures Expression of the START domains
The recombinant START proteins were produced in BL21
E coli expressing human STARD3-START (amino acids 216–445) [26], or N-62-STARD1 (STARD1-START),
as previously described [17] Each of the expressed proteins contained a His6-tag at the C-terminus The bacteria were cultivated in LB medium containing 25 lgÆmL)1kanamycin for STARD1-START or 25 lgÆmL)1 ampicillin for STARD3-START For expression of the proteins, 400 mL
of medium (with antibiotic) was inoculated with 1 mL of overnight culture The medium was shaken at 37C until an attenuance of 0.5–1.0 at 600 nm was achieved Expression was induced by the addition of 0.5 m isopropyl-b-d-thio-galactopyranoside After 4.5 h, the bacteria were pelleted The pellet was resuspended on ice in 10 mL of the fol-lowing buffer: 300 mm NaCl, 50 mm NaH2PO4, 20 mm Tris⁄ HCl (pH 7.4), and 10 mm b-mercaptoethanol The bacteria were sonicated on ice (3· 15 pulses of 1 s, output level 7), using a Branson Sonifier 250 (Branson, Danbury,
CT, USA) The suspension was centrifuged at 4C for
30 min at 20 000 g (J2-21-centrifuge; Beckman, Munich, Germany) The supernatant was incubated with 500 lL of
Ni2+–nitrilotriacetic acid–agarose matrix (Qiagen, Hilden, Germany) The mixture was rotated at 4C overnight The matrix was placed in a column and washed with 20 mL
of the following buffer: 300 mm NaCl, 50 mm NaH2PO4
(pH 8.0), and 20 mm imidazole STARD1-START was eluted with 2 mL of the following buffer: 300 mm NaCl,
Trang 1050 mm NaH2PO4 (pH 8.0), and 250 mm imidazole To
avoid aggregation of STARD3-START, the STARD3
elu-tion buffer contained 40% (w⁄ v) glycerol The eluted
pro-teins were dialyzed (molecular mass cutoff 12 kDa; Sigma,
Schnelldorf, Germany) against the following buffer: 50 mm
KCl, 50 mm Hepes (pH 7.4), and 1 mm dithiothreitol For
dialysis of STARD3-START, the following buffer was
used: 150 mm NaCl, 50 mm KCl, 50 mm Tris (pH 7.4),
10 mm dithiothreitol, and 40% (w⁄ v) glycerol
Immunoblotting
Proteins were separated by SDS⁄ PAGE and were
trans-fered onto a nitrocellulose membrane using a tank blot
sys-tem Immunodetection was performed with appropriate
antibodies: mouse anti-His serum (1 : 2000) and mouse
anti-peroxidase Ig (1 : 1000) The proteins were detected
with Amersham ECL Plus (GE Healthcare Life Sciences,
Munich, Germany) The results were displayed and
docu-mented using a VersaDoc 3000 imaging system (Bio-Rad,
Munich, Germany)
Photoaffinity labeling
Photoaffinity labeling of the START proteins was performed
using the photoreactive cholesterol analog [3
H]6,6-azocho-lestanol (termed [3H]photocholesterol) [3H]Photocholesterol
was synthesized according to an established protocol
[48] Twenty micrograms of protein in a final volume of
200 lL were incubated with [3H]photocholesterol (50 lm,
30–185 GBqÆmmol)1) for 20 min at room temperature The
sterol was complexed with MbCD (0.6 mgÆmL)1) For UV
irradiation, either a 200 W Hg-lamp (k 330 nm; Leitz,
Wetzlar, Germany) or a Transilluminator 4000 (Stratagene,
Heidelberg, Germany) was used The distance between the
lamp of the Transilluminator and the samples was about
5 cm During the irradiation, the samples were incubated on
ice in 1.5 mL reaction tubes The samples were irradiated for
10 min When the 200 W Hg-lamp was used, the samples
were irradiated in a cooled quartz cuvette with a magnetic
stir-bar The crosslinking efficiency obtained with the
Trans-illuminator was found to be similar to that obtained with the
200 W Hg-lamp The proteins were precipitated with 1 mL
of cold acetone ()20 C) The sample was stored at )20 C
for at least 1 h The proteins were pelleted by centrifugation
at 20 000 g for 10 min at 4C The supernatant was
removed The pellet was dried with gaseous N2 The protein
pellets were subjected to SDS⁄ PAGE or to chemical or
enzymatic cleavage
Cleavage of proteins
For chemical cleavage, CNBr (Fluka, Germany) was used
The pellet (20 lg of protein) was resuspended in 30 lL of
H2O Seventy microliters of formic acid containing 100 lg
of CNBr were added The sample was incubated for 24 h
at 37C in the dark The solvent was evaporated with gaseous N2 For enzymatic cleavage, the protease LysC (Roche, Germany) was used The pellet (20 lg of protein) was resuspended in 20 lL of the following buffer: 100 mm
NH4HCO3 (pH 8.5) One microgram of LysC in 1 lL of the same buffer was added, and the sample was incubated
at 37C for 24 h in the dark in a gaseous N2atmosphere
SDS⁄ PAGE
To determine the molecular masses of the proteins, the Laemmli protocol was employed For the separation of small protein fragments, the method described by Schaegger and von Jagow [49] was used
Scintillation counting
The fragments of the labeled and cleaved proteins were sep-arated by tube gels (100 mm in length, 4 mm in diameter)
or slab gels (50 mm in length, 1.5 mm in thickness) The gels were cut into 1 mm slices Each slice was incubated overnight at room temperature in a scintillation vial (Canb-erra Packard, Dreieich, Germany) with 4 mL of the follow-ing scintillation cocktail: 90% (v⁄ v) Lipoluma; 9% (v ⁄ v) Lumasolve; and 1% (v⁄ v) H2O (Lumac-LSC; Perkin-Elmer, Groningen, the Netherlands) For scintillation counting, a Tri-Carb 2100 TR-counter (Packard, Dreieich) was used
Fluorescence spectroscopy
The fluorescent cholesterol reporter NBD-cholesterol was used to verify the cholesterol binding of STARD1-START and STARD3-START The measurements were performed with a Photon Technologies International (Birmingham, NJ, USA) spectrofluorometer (Quantamaster) The proteins were diluted with 25 mm potassium phosphate buffer (pH 7.4) including 0.0002% Tween-20 to a final concentration of
10 nm The sample was transferred in a quartz cuvette that was placed in a cuvette holder equipped with a magnetic stir-bar The sterol was added from ethanolic stock solutions The samples were incubated for 10 min at 37C before the fluorescence was recorded at constant temperature (37C) NBD-cholesterol was excited at 473 nm Fluorescence emis-sion was monitored at 530 nm Excitation and emisemis-sion bandpasses were set to 4 nm To reduce light scatter, a cutoff filter (495 nm) was placed in the emission path The binding data were calculated using sigmaplot (version 8.0)
MS
A SELDI-TOF mass spectrometer (Ciphergen Biosystems, Go¨ttingen, Germany) was used to measure the molecular