(Received for publication, September 17, 1996, and in revised form, December 26, 1996)
From the Department of Biochemistry, Stockholm
University, S-106 91 Stockholm, Sweden, the § Department of
Molecular Biology, Graduate School of Medical Science, Kyushu
University, Fukuoka 812, Japan, and the ¶ Department of
Molecular Biology, Karolinska Institute Center for Biotechnology,
NOVUM, S-141 57 Huddinge, Sweden
Insertion into the endoplasmic reticulum membrane of model proteins with one, two, and four transmembrane segments and different distributions of positively charged residues in the N-terminal tail and the polar loops has been studied both in vitro and in vivo. Membrane insertion of these same constructs has previously been analyzed in Escherichia coli, thus making possible a detailed comparison between the topological rules for membrane protein assembly in prokaryotic and eukaryotic cells. In general, we find that positively charged residues have similar effects on the membrane topology in both systems when they are placed in the N-terminal tail but that the effects of charged residues in internal loops clearly differ. Our results rule out a sequential start-stop transfer model where successive hydrophobic segments insert with alternating orientations starting from the most N-terminal one as the only mechanism for membrane protein insertion in eukaryotic cells.
What features of the amino acid sequence of integral membrane proteins control their insertion and orientation in the membrane? A number of recent studies in Escherichia coli have suggested that hydrophobic segments both target bacterial inner membrane proteins to the membrane and drive their insertion into the lipid bilayer, whereas flanking positively charged amino acids determine the final orientation of the hydrophobic transmembrane segments (1).
Statistical analyses suggest that a similar "positive inside" rule is applicable also to eukaryotic membrane proteins (2-7), although the tendency for positively charged residues to be absent from extracellularly exposed parts is weaker in eukaryotic than in prokaryotic membrane proteins. Nevertheless, a handful of eukaryotic membrane proteins have been successfully expressed in E. coli with what appears to be the correct topology (8-10), and at least one example of an E. coli inner membrane protein that inserts with the correct topology into mammalian microsomes in vitro has been described (11, 12).
Mutagenesis experiments on a number of bitopic eukaryotic plasma membrane proteins with a single N-terminal transmembrane segment have revealed a clear effect of arginine and lysine residues on membrane orientation, although other factors such as the presence of negatively charged residues, the length and folding properties of the polar N-terminal tail preceding the transmembrane segment, and the length of the transmembrane segment itself have also been shown to be important (13-23). However, the effects of charged residues on the topology of polytopic eukaryotic membrane proteins have not been systematically studied so far, and it is not clear whether the topological rules relating amino acid sequence to membrane topology are the same in prokaryotic and eukaryotic cells.
Building on our earlier work on membrane protein assembly in E. coli, we now report a systematic comparative study of the insertion into the ER1 membrane, both in vitro and in vivo, of a range of model proteins with one to four transmembrane segments and different distributions of positively charged residues in the flanking regions. Most of these proteins have previously been characterized in E. coli, allowing a detailed comparison between membrane protein insertion in prokaryotic and eukaryotic systems.
Unless otherwise stated, all enzymes
were from Promega. T7 DNA polymerase was from Pharmacia Biotech Inc.
Endo H was from Boehringer Mannheim. [35S]Met was from
Amersham Corp. Ribonucleotides, deoxyribonucleotides, dideoxyribonucleotides, protein A-Sepharose, and the cap analog m7G(5)ppp(5
)G were from Pharmacia. Plasmid pGEM1, transcription buffer, and rabbit reticulocyte lysate were from Promega. BHK 21 cells
were grown in medium from Life Technologies, Inc. The glycosylation
acceptor peptide
N-benzoyl-Asn-Leu-Thr-N-methylamide and the
nonacceptor peptide
N-benzoyl-Asn-Leu-(allo)Thr-N-methylamide were synthesized according to Ref. 24. Oligonucleotides were from Kebo
Lab (Stockholm, Sweden).
Site-specific mutagenesis was performed
according to the method of Kunkel (25) as modified by Geisselsoder
et al. (26). All mutants were confirmed by DNA sequencing of
single-stranded M13 DNA using T7 DNA polymerase. For cloning into and
expression from the pGEM1 plasmid, the 5 end of the lep
gene was modified first by the introduction of an XbaI site
and second by changing the context 5
to the initiator ATG codon to a
"Kozak consensus" sequence (27). Thus, the 5
region of the gene
was modified to: ATAACCC
GCCACC
GCGAAT
(XbaI site and initiator codon underlined). Mutants with
four transmembrane segments were made by fusing the H1-P1-H2 region
(from the upstream XbaI site to a KpnI site
introduced at the codon corresponding to Asp99 in the wild
type sequence) of one Lep construct to a KpnI site introduced just upstream of the start codon of a second Lep construct (see Fig. 1). The KpnI site introduced three new amino acid
residues (Glu-Val-Pro) at the fusion site. The
XbaI-SmaI fragment carrying the constructs was
cloned into pGEM1 behind the SP6 promoter. For cloning into the pSFV1
in vivo expression vector (28), constructs were polymerase
chain reaction-amplified from pGEM1 using polymerase chain reaction
primers introducing BamHI and SmaI sites flanking the relevant gene. The BamHI-SmaI fragment
carrying the constructs was then cloned into the pSFV1 polylinker
region.
Assay of Membrane Topology in E. coli
E. coli strain MC1061 (29) transformed with the pING1 vector (30) carrying the leader peptidase mutant WT-3K under control of the arabinose promoter was grown at 37 °C in M9 minimal medium supplemented with 100 µg/ml ampicillin, 0.5% fructose, and all amino acids (50 µg/ml each) except methionine. An overnight culture was diluted 1:25 in fresh medium, shaken for 3.5 h at 37 °C, induced with arabinose (0.2%) for 5 min, and labeled with [35S]methionine (150 µCi/ml). After 1 min, nonradioactive methionine was added (500 µg/ml) and incubation was stopped by chilling on ice. Cells were spun at 14,000 rpm for 2 min, resuspended in ice-cold buffer (40% w/v sucrose, 33 mM Tris pH 8.0), and incubated with lysozyme (5 µg/ml) and 1 mM EDTA for 15 min on ice. Aliquots of the cell suspension were incubated for 1 h on ice, either with 0.75 mg/ml TPCK-treated trypsin or with (0.75 mg/ml TPCK-treated trypsin + 1.0 mg/ml trypsin inhibitor + 0.64 mg/ml phenylmethylsulfonyl fluoride). After the addition of trypsin inhibitor and phenylmethylsulfonyl fluoride, samples were acid-precipitated (trichloroacetic acid, 10% final concentration), resuspended in 10 mM Tris/2% SDS, immunoprecipitated with antisera to Lep and OmpA (an outer membrane control (31)), washed, and analyzed by SDS-polyacrylamide gel electrophoresis.
In Vitro Transcription and Translation in Reticulocyte LysateThe pGEM1 plasmids carrying the relevant constructs were linearized at a SmaI site immediately downstream of the coding region prior to in vitro transcription. Synthesis of RNA by SP6 RNA polymerase and translation in reticulocyte lysate in the presence of dog pancreas microsomes was performed as described (32) or according to the protocol provided by Promega. Translocation of polypeptides to the lumenal side of the microsomes was assayed by prevention of N-linked glycosylation through competitive inhibition by the addition of a glycosylation acceptor tripeptide but not by a nonacceptor tripeptide and by proteinase K treatment of the microsomes (12). Alkaline extraction of microsomes was carried out as described (33). SDS-polyacrylamide gel electrophoresis gels were scanned in a FUJIX Bas 1000 PhosphorImaging Plate Scanner and analyzed using the MacBAS software (version 2.1).
In Vivo Expression in BHK CellsThe pSFV1 vector containing the relevant constructs were linearized at a unique SphI site downstream of the cloned gene. The linearized plasmid was used as template for in vitro transcription, and mRNA was transfected into cells by electroporation as described earlier (34). 7 h post-transfection, cells were pulse-labeled with [35S]Met for 15 min, chased for 30 min, and solubilized in Nonidet P-40 (1%) lysate buffer (35). The lysate was centrifuged at 4000 × g for 5 min. The supernatant was used for protein A-mediated immunoprecipitation using a Lep polyclonal antiserum. The precipitate was solubilized in Endo H buffer (1% SDS, 50 mM sodium-citrate, pH 6.0) and incubated for 5 min at 70 °C. The sample was then divided in two and either mock treated or Endo H-treated (5 milliunits) for 6 h at 37 °C and analyzed by SDS-polyacrylamide gel electrophoresis under nonreducing conditions.
In order to be able to make direct comparisons between the topological effects of positively charged amino acids in membrane proteins expressed in prokaryotic and eukaryotic systems, we chose the well characterized E. coli inner membrane protein leader peptidase (Lep) as a model protein. Lep has been used extensively in studies of membrane protein assembly in E. coli (1) and has also been shown to insert efficiently into ER-derived dog pancreas microsomes (11, 12).
Lep constructs with one and two transmembrane segments and various
point mutations, additions, or deletions (Fig. 1) were made from the wild type gene by site-directed mutagenesis. Molecules with four transmembrane segments were constructed by fusing the H1-L1-H2 region from one construct with the entire coding region from
another (36). To map the cytoplasmic or extra-cytoplasmic location of
the N-terminal tail (Ncyt and Next) and the
C-terminal P2 domain (Ccyt and Cext), acceptor
sites for N-linked glycosylation that can only be modified
if translocated to the lumen of the ER were introduced into these
regions by site-directed mutagenesis (note that all constructs with an
N-terminal glycosylation acceptor site have an extra N-terminal
extension of 14 residues, except in wild type-derived (WT*) construct
and construct H2 (Fig. 2), where the extension is 16 residues). All genes were cloned behind the SP6 promoter in the pGEM1
vector for expression in vitro and into the Semliki Forest
virus vector pSFV1 (28) for expression in vivo.
A Single N-terminal Transmembrane Segment Is Sufficient for Targeting and Insertion of a Bacterial Protein into the ER Membrane
In E. coli, Lep adopts a somewhat unusual
membrane topology with both the N and C termini facing the periplasm.
When expressed in vitro in the presence of microsomes, both
termini are again translocated, whereas the P1 loop remains on the
cytoplasmic side of the membrane (12). To check whether both the H1 and
H2 transmembrane segments are necessary for targeting and insertion
into the ER membrane, transmembrane segment H2 was deleted from WT*
with an N-terminal extension including a glycosylation acceptor site
and a mutation (Asn214 Gln) that removes the only
acceptor site in the P2 domain, and the protein was expressed in the
absence and the presence of microsomes. As seen in Fig. 2, the
H2
mutant was glycosylated on the N-terminal acceptor site to the same
extent as WT* (lanes 2 and 5). When the
microsomes were treated with proteinase K, a protease-resistant
fragment representing the H2-P2 domain (12) was produced from WT* (Fig.
2, lane 3), whereas mutant
H2 was completely degraded
(Fig. 2, lane 6). After disruption of the microsomes with
the detergent Triton X-100, no protease-resistant material remained for
either construct (data not shown). Together, these results demonstrate
that H1 has an intrinsic ability to target to the ER membrane and to
insert in the same Next-Ccyt orientation as
adopted in E. coli.
Lep has a highly charged
cytoplasmic loop (P1) that cannot be translocated across the inner
E. coli membrane, presumably because the positively charged
amino acids immediately downstream of the first hydrophobic domain (H1)
prevent translocation (37, 38). Thus, even though the addition of one
or more lysines to the N-tail preceding H1 can cause certain Lep
constructs lacking most of the charged residues in the P1 loop to
insert into the inner membrane in an inverted
Ncyt-Ccyt orientation with P1 exposed to the
periplasm (31, 39), no translocation of the P1 loop was observed even when as many as three lysines were added to the N-tail in wild type Lep
(mutant WT-3K). This mutant was expressed in E. coli, and
its topology was probed by trypsin treatment of spheroplasts followed
by immunoprecipitation with a Lep polyclonal antiserum (Fig.
3A). Very little protein remained after
protease treatment (Fig. 3A, lane 2), and there
was no protease-protected fragment diagnostic for translocation of the
P1 domain (31). Parallel analysis of the protease-sensitivity of the
mature and precursor forms of the outer membrane protein OmpA was
performed to ensure that the spheroplasts were intact. The protease
sensitivity of the periplasmically exposed mature OmpA demonstrates
that no intact cells remained, whereas the protease-resistance of
cytoplasmic, nontranslocated precursor OmpA demonstrates that the
spheroplasts were intact (31). Thus, the P2 domain in WT-3K is exposed
to the periplasmic space, whereas the P1 domain is cytoplasmic.
It has been shown that translocation of the P2 domain across the microsomal membrane is much less sensitive to the presence of positively charged amino acids immediately downstream of H2 than when expression is carried out in E. coli (11), suggesting that the addition of one or more lysines to the normally uncharged N-tail of Lep might result in an inverted orientation in microsomes with the highly charged P1 loop in the lumen, in contrast to what is seen in E. coli. Wild type Lep expressed in vitro in the presence of dog pancreas microsomes was efficiently glycosylated on its only potential acceptor site for N-linked glycosylation (Asn214) (Fig. 3B, top panel, lanes 1 and 2), demonstrating the lumenal disposition of the P2 domain. The addition of a single N-terminal lysine between Met4 and Phe5 resulted in a marked reduction in glycosylation efficiency (Fig. 3B, top panel, lanes 3 and 4), and with two and four N-terminal lysines the glycosylation efficiency was only ~20%, consistent with a predominantly Ncyt-Ccyt topology. The decrease in glycosylation was paralleled by a concomitant decrease in the amount of protease-resistant H2-P2 fragment (data not shown). In all cases, both the glycosylated and nonglycosylated forms were properly integrated into the microsomal membrane as shown by their retention in the pellet after alkaline extraction (Ref. 33 and data shown for the 0K and 2K mutants in Fig. 3B, bottom panel). We conclude that positively charged residues in the N-terminal tail have a strong influence on the topology of Lep expressed both in the E. coli and microsomal systems, whereas positively charged residues in the P1 loop between the two transmembrane segments block translocation only in E. coli.
The Length of the Loop between Two Transmembrane Segments Does Not Affect TopologyAnother variable that has been shown to influence
the topology of Lep-derived constructs in E. coli is the
length of the P1 loop; although this 39-residue-long loop cannot be
translocated (Fig. 3A), translocation is possible when its
length is increased to more than ~60 residues (40). It has been
speculated that this may be related to the participation of the Sec
translocase in the translocation of long loops across the inner
membrane; only a small number of positively charged residues would be
allowed in short loops (60 residues) that cannot use the Sec
translocase for translocation, whereas longer loops that use the more
permissive Sec pathway would have less stringent sequence requirements
(41).
To test if the length of the P1 loop affects topology also in the ER
membrane, a series of constructs with zero, one, two, and four
N-terminal lysines and P1 lengths ranging from 19 to 88 residues was
analyzed in the microsomal in vitro system. As shown in Fig.
4, the length of the P1 loop has at best a minor and not
very consistent effect on the topology in the 1K series of mutants and
has no effect for the 0K, 2K, and 4K series, in contrast to the
situation in E. coli.
The construct with 88 residues in the P1 loop was glycosylated somewhat more efficiently than the other constructs when two and four lysines were present in the N-tail (Fig. 4B); however, in this particular case there is an additional glycosylation site in the P1 loop, ~15 residues downstream of H1. The higher glycosylation efficiency may thus result from partial glycosylation of this site rather than from translocation of the P2 domain. To ascertain that this was indeed the case, a second glycosylation site was inserted into the middle of the P1 loop, and the site in P2 was simultaneously removed. The resulting construct 2K(88*) with two N-terminal lysines and with two glycosylation sites in P1 and none in P2 was efficiently glycosylated (Fig. 4A, lanes 11 and 12); 20% of the chains were glycosylated on both sites, and 70% were glycosylated only on one, demonstrating that no more than ~10% of the molecules have the wild type orientation (or fail altogether to integrate into the membrane). N-terminal positively charged residues thus efficiently promote an inverted, Ncyt-Ccyt topology in the ER membrane irrespective of the length of the P1 loop.
The Topology of "Nonfrustrated" Proteins Follows the Positive Inside RulePrevious topological analyses in E. coli of Lep-derived constructs with four transmembrane segments and different distributions of positively charged residues in the N-tail and the loops connecting the transmembrane segments have demonstrated that "frustrated" constructs, i.e. molecules where the distribution of positively charged residues is such that neither the Next-Cext nor the Ncyt-Ccyt topology will have all the highly charged loops in the cytoplasm, often adopt a "leave one out" topology with only three of the hydrophobic segments inserted across the membrane (36).
To test if the influence of positively charged residues is equally
strong in the ER, a subset of the constructs previously characterized
in E. coli was expressed in the microsome system. First, two
nonfrustrated constructs that adopt, respectively, the
Next-Cext (i.e. N and C terminus in
the periplasm) and the Ncyt-Ccyt topology in
E. coli were analyzed. As shown in Fig. 5A, construct 0K/3K/0K/3K with three lysines
in each of the loops between the first and second and between the third
and fourth transmembrane segments was efficiently glycosylated both in
the C-terminal P2 domain (Fig. 5A, middle panel,
lane 2) and on an Asn-Ser-Thr acceptor site introduced into
the N-tail (Fig. 5A, top panel, lane
2) and thus adopts the Next-Cext topology
just as in E. coli. The nonglycosylated full-length product
of construct *0K/3K/0K/3K (Fig. 5A, top panel,
lane 7) has the same size as the corresponding 0K/3K/0K/3K
construct expressed in E. coli (36) (Fig. 5A,
top panel, lane 8).
Proteinase K treatment of the microsomes (Fig. 5A, top
panel, lane 5) produced a protected fragment that runs
at the approximate location of the weak band in lane 1 of
the top panel of Fig. 5A marked by a black
square in the middle and bottom panels and
considerably below this band in the top panel. This suggests
that the protected fragment is glycosylated when the P2 domain carries
a glycosylation site (Fig. 5A, middle and
bottom panels) but not when the glycosylation site is in the
N-tail (fig. 5A, top panel). Indeed, the
protected fragment in the middle and bottom
panels of Fig. 5A has the same size as a glycosylated,
protease-protected fragment derived from microsome-integrated wild type
Lep that was previously shown to correspond to the glycosylated H2-P2
domain (Ref. 12 and data not shown), and we conclude that the protected
fragment in Fig. 5A results from cleavage in the L1
loop.
With glycosylation sites simultaneously present in both the N-tail and
the P2 domain, a majority of the molecules were glycosylated on both
sites (Fig. 5A, bottom panel, lane 2),
suggesting that all membrane-inserted molecules are glycosylated in the
P2 domain and that most of these are also glycosylated on the N-tail.
To rule out that the rather small number of residues between the N-terminal glycosylation site and H1 cause the somewhat inefficient glycosylation of the N-tail (12), we also made constructs where this
distance was increased to 19 and 25 residues. Both were glycosylated with the same efficiency as the original construct (data not shown), suggesting that the N-terminal glycosylation site is either
intrinsically less easily modified than the site in the P2 domain or
that the N-tail is somewhat less efficiently translocated than P2. In
any case, these data show that the majority of the molecules have the
expected Next-Cext orientation with the L1
loop (and most likely also the L1 loop) exposed on the cytoplasmic side
of the microsomal membrane.
In contrast, construct 3K/0K/3K/0K with three lysines in the N-tail and in the loop between transmembrane segments 2 and 3 was found to be essentially unglycosylated on both the N- and C-terminal acceptor sites (Fig. 5B, top panel), suggesting the same Ncyt-Ccyt topology as observed in E. coli. The trivial explanation that this particular construct is not inserted into the membrane was ruled out by alkaline extraction of the microsomes (33); as seen in Fig. 5B (bottom panel), a large fraction of the molecules partitioned with the wash supernatant when translation was carried out in the absence of microsomes (or when microsomes were added post-translationally; data not shown), whereas essentially no molecules were found in this fraction when translation was carried out in the presence of microsomes, demonstrating proper membrane integration. Further, the long L1 loop in the related construct 3K/88/3K/0K (see below) is efficiently glycosylated, again consistent with proper membrane assembly and an Ncyt-Ccyt topology.
Both nonfrustrated constructs with a glycosylation acceptor site in the P2 domain were also expressed in vivo in BHK cells. As in vitro, construct 3K/0K/3K/0K was not glycosylated, whereas construct 0K/3K/0K/3K was fully glycosylated (Fig. 5C), demonstrating efficient targeting and insertion into the ER membrane (we cannot completely rule out that some molecules may fail to integrate and are rapidly degraded). The two lower molecular weight bands marked with black and white squares are also faintly seen in most of the in vitro experiments. These bands run at roughly the same position as nonglycosylated and glycosylated wild type Lep, respectively (data not shown), and may result either from endogenous proteolysis or from internal initiation at one or other of the four methionine codons present in the L2 loop (see Fig. 1).
We conclude that nonfrustrated constructs with four transmembrane segments based on a bacterial inner membrane protein can be targeted and inserted into the ER membrane both in vitro and in vivo and that they adopt the same topology as in E. coli.
Membrane Insertion of Frustrated ProteinsTwo frustrated
constructs, previously shown to adopt leave one out topologies in
E. coli, were also tested in the ER system. Construct
0K/3K/3K/0K was glycosylated on the N terminus and in the P2 domain
both in vitro and in vivo and was doubly
glycosylated when acceptor sites were simultaneously present in these
two regions (Fig. 6A, bottom),
suggesting that the topology is predominantly Next-Cext with four transmembrane segments
(Fig. 6B, bottom panels). This is in contrast to
E. coli, where this construct adopts a leave one out
topology with only three transmembrane segments (36).
On the other hand, the N-tail in construct 3K/0K/0K/3K was not glycosylated, whereas the P2 domain was efficiently glycosylated (Fig. 6A, left top panels). When expressed in vivo, this construct was again fully glycosylated in the P2 domain (Fig. 6A, right top panel). As in E. coli, the dominating topology is thus Ncyt-Cext, with only three transmembrane segments (Fig. 6B). It is formally possible that the second hydrophobic segment is translocated into the lumen with the first, third, and fourth segments spanning the membrane; however, given that there is no precedence for the translocation of hydrophobic segments across the ER membrane, we consider this unlikely. Because this construct is slightly less efficiently glycosylated in the P2 domain in vitro than the nonfrustrated construct 0K/3K/0K/3K (65% versus 75%) and because most of the nonglycosylated molecules are found in the membrane fraction after alkaline extraction in both cases (data not shown), a minor proportion of the molecules (~10%) may have the P2 domain in the cytoplasm. This is not the case in vivo, however. Which of the two uncharged loops that is exposed to the lumenal side cannot be determined with the glycosylation assay. Protease treatment of the microsomes did not give interpretable results for this mutant, because no unique protease-resistant fragment was produced.
In order to address this point further, two constructs where the L1
loop between the first and second transmembrane segments is 88 rather
than 25 residues long were also tested. A glycosylation acceptor site
placed in this loop is far enough from transmembrane segment H1 (~15
residues) to be accessible to the lumenally disposed glycosyltransferase, and the location of the loop can thus be mapped.
In the nonfrustrated construct 3K/88/3K/0K, the long loop was quite
efficiently glycosylated (Fig. 7A, lane
3), and no doubly glycosylated molecules were apparent when the
acceptor site in the P2 domain was also present (Fig. 7A,
lane 2), demonstrating that the dominating topology is
Ncyt-Ccyt, as expected. The slight increase in
glycosylation efficiency (from 45 to 55%; compare lanes 2 and 3, Fig. 7A) when the acceptor site in the P2
domain was added suggests that a small fraction of the molecules may have the long loop in the cytoplasm and the P2 domain in the lumen.
In the frustrated construct 3K/88/0K/3K, both the long loop and the P2 domain could be glycosylated. With one acceptor site in the long loop and one in the P2 domain, molecules with both one and two glycosyl moieties were apparent (Fig. 7A, lane 4); with an additional acceptor site present in the P2 domain, the molecules were glycosylated either once or on all three sites (Fig. 7A, lane 5). The absence of twice glycosylated molecules modified only in the P2 domain in the latter construct demonstrates that no molecules have a topology with the P2 domain and the short L2 loop between the second and third transmembrane segments in the lumen. Thus, the majority of the molecules have both the long loop and the P2 domain in the lumen (Ncyt-Cext, leave one out topology) (Fig. 7B), although the singly glycosylated molecules in lane 5 of Fig. 7A suggest the additional presence of molecules with four transmembrane segments and Ncyt-Ccyt orientation. Similar proportions of nonglycosylated and glycosylated molecules were found in the membrane pellet after alkaline extraction (data not shown).
We have constructed a number of model proteins with either one, two, or four hydrophobic transmembrane segments and have analyzed their membrane assembly both in vitro in dog pancreas microsomes and in vivo in BHK cells. Most of these constructs have previously been analyzed in E. coli (36, 39, 40), and we can thus make a detailed comparison between the sequence determinants of membrane protein topology in prokaryotic and eukaryotic cells.
In general, we find that positively charged residues placed in the N-tail have similar effects on the topology of model proteins in both the prokaryotic and eukaryotic systems, i.e. that such residues prevent translocation of the N-tail across the membrane. However, there are also some obvious differences.
In constructs with two transmembrane segments, the length of the connecting loop does not seem to be important for its ability to be translocated across the ER membrane (Fig. 4), whereas in E. coli short and long loops apparently use different mechanisms for translocation and react differently to the presence of positively charged residues (40-43). Thus, when two lysines are added to the N terminus of wild type Lep, most of the molecules insert into the ER membrane in an inverted Ncyt-Ccyt orientation (Fig. 3), whereas addition of three lysines to the N-tail has no effect on the orientation when the protein is expressed in E. coli. Our data thus suggest that positively charged residues placed in the N-tail are critical topological determinants in both E. coli and mammalian ER, whereas charged residues in the following P1 loop have a major effect on the topology only in E. coli.
Three of the four constructs with four transmembrane segments also behave similarly in the two systems. The nonfrustrated constructs 3K/0K/3K/0K and 0K/3K/0K/3K insert according to the positive inside rule both in E. coli (36) and in the ER. However, of the two frustrated constructs, one (3K/0K/0K/3K) adopts a leave one out Ncyt-Cext topology in both E. coli and the ER, whereas the other (0K/3K/3K/0K) behaves differently in the two systems. Whereas in E. coli this construct has only three transmembrane segments and Next-Ccyt topology, both the N- and C termini are quite efficiently translocated across the ER membrane. This suggests that the L2 loop between the second and third hydrophobic segments is lumenal in these molecules, despite its high charge.
In summary, in all constructs with a highly charged N-tail the first transmembrane segment is oriented Ncyt, and in constructs with no positively charged residue in the N-tail the orientation is Next. This is consistent with earlier statistical (2-6, 44) and experimental (17, 19, 21) studies. There is also a clear tendency for highly charged internal loops to remain on the cytoplasmic side, although this tendency is stronger in E. coli than in the ER system. The apparently greater ease with which highly charged loops are translocated in the eukaryotic system may be related to the absence of a membrane potential in the ER (cf. Ref. 45) and may explain the earlier observation that extra-cytoplasmic loops in eukaryotic plasma membrane proteins have on average a higher content of positively charged residues than found in bacterial inner membrane proteins (2-4).
We have previously suggested (36, 46) that the so-called Sec-independent mechanism of membrane protein assembly in E. coli is based on helical hairpins composed of two neighboring hydrophobic segments connected by a short loop lacking in positively charged residues that insert en bloc. A purely sequential insertion process, on the other hand, where the orientation of all downstream transmembrane segments is dictated by the orientation of the most N-terminal one (47), is inconsistent with the leave one out topologies found in E. coli and with the Ncyt-Cext topology found for constructs 3K/0K/0K/3K an 3K/88/0K/3K in the ER, as well as with the results of a recent study on the topology of P-glycoprotein mutants in the ER (48). Possibly, proteins with widely spaced hydrophobic segments insert in a sequential fashion (49, 50), whereas regions with closely spaced hydrophobic segments insert according to a helical hairpin mechanism.
Finally, we note that the observation that the topological information present in certain of our constructs is interpreted differently in prokaryotic and eukaryotic cells suggests that problems encountered when trying to overexpress eukaryotic membrane proteins in prokaryotic hosts may in some cases be related to incorrect topologies. If so, redesigning the protein to conform better to the stricter prokaryotic version of the positive inside rule may be one way to obtain higher expression levels (51).
Glycosylation acceptor and nonacceptor tripeptides were kindly provided by Dr. Henrik Garoff (Karolinska Institute Center for Biotechnology).