We reasoned that Mgm1p carrying an engineered hydrophobic segment (H-segment) in place of the first TM segment should give rise to the two isoforms in varying amounts, depending on whether the H-segment is integrated into the inner membrane or not. We therefore replaced the first hydrophobic segment of Mgm1p by a set of 19-residue H-segments of varying hydrophobicity (), for which we have already measured membrane insertion in the mammalian and yeast ER (Hessa et al, 2007
). For immunodetection, a triple-haemagglutinin (HA) tag was fused to the C-terminus of Mgm1p in all constructs. C-terminal epitope tagging of Mgm1p with either HA or c-myc
does not compromise protein function (Meeusen et al, 2006
; Zick et al, 2009
) (and our unpublished data), showing that the C-terminal HA tag disrupts neither targeting nor topogenesis. Each Mgm1p construct was expressed in yeast from a low-copy CEN
plasmid, and the relative amounts of the two isoforms were quantified from western blots and used to calculate an apparent free energy of membrane insertion,
is the gas constant and T
is the absolute temperature, T
=298 K, and fl
denote the fractions of l
-Mgm1p and s
-Mgm1p molecules, respectively).
A hydrophobicity scale for mitochondrial inner membrane proteins
We first measured ΔGapp
for a series of 19-residue Leu/Ala-based H-segments, with the composition GGPG-n
)A-GPGG (all constructs are listed in Supplementary Table S1
). The secondary-structure breaking GGPG…GPGG flanks were included to ‘insulate' the H-segments from the surrounding protein sequence (Hessa et al, 2005
). As anticipated, H-segments with a higher number of Leu residues produced a higher fraction of the membrane-anchored l
-Mgm1p isoform relative to s
-Mgm1p (). The measured ΔGapp
is linear with the number of Leu residues (n
), with ΔGapp
=0 kcal/mol (50% membrane insertion) reached for, n
≈5–6 (). This ‘threshold hydrophobicity' is increased by roughly one Ala → Leu replacement compared with the yeast ER (Hessa et al, 2009
) and by two such replacements compared with the mammalian ER (Hessa et al, 2005
). These differences may at least in part depend on charged residues outside the GGPG…GPGG flanks (see below).
Figure 2 Membrane insertion efficiency of Leu/Ala-based H-segments. (A) Whole-cell lysates from yeast transformants expressing HA-tagged Mgm1p constructs with H-segments containing varying number of Leu residues were analysed by SDS–PAGE and western blotting. (more ...)
As a control for proper mitochondrial import, we compared the size of the l-Mgm1p isoforms from whole-cell lysates prepared from yeast transformants carrying either HA-tagged Mgm1p or the HA-tagged Mgm1p(6L/13A) construct with the corresponding in vitro-translated full-length products (including the presequence) (). A clear difference in molecular weight was observed between the in vitro-translated presequence-containing Mgm1p and the Mgm1p isoforms isolated from whole-cell lysates, indicating that the presequence has been cleaved from the Mgm1p constructs and therefore that they are correctly imported into the mitochondria.
To determine the contribution of each of the 20 naturally occurring amino acid to ΔGapp
, we prepared a second set of H-segments (Supplementary Table S1
). Each test amino acid was placed in the middle of the H-segment, so that it would be embedded in the hydrophobic core of the lipid bilayer when integrated into the membrane. As the maximal sensitivity of the assay is obtained when ΔGapp
≈0 kcal/mol (50% membrane insertion), the Ala/Leu composition of the H-segments was adjusted such that, for each test amino acid, the efficiency of insertion was not too far from 50%. Assuming a simple additive model for the contributions of Leu and Ala residues to ΔGapp
for the 19-residue H-segments shown in , we have n
+1.64 kcal/mol, which yields
=−0.20 kcal/mol and
=0.09 kcal/mol. Possible contributions to ΔGapp
from the Gly-Pro flanking sequences should be small (Hessa et al, 2005
), and are ignored here. For the remaining 18 natural amino acids (X), ΔGappX
was calculated from measured ΔGapp
, see Supplementary Table S1
shows the mitochondrial ΔGappX scale. As expected, the hydrophobic amino acids are at the lower end of the scale, whereas the charged amino acids are found at the higher end. Comparison to the scale obtained for membrane insertion into the mammalian ER shows a good correlation (R2=0.81) (), but the mitochondrial scale spans a somewhat broader range compared with the ER scale (−0.2 to +3.0 kcal/mol versus −0.5 to +1.8 kcal/mol) and the zero point is different, as noted above.
Figure 3 ‘Biological' ΔGXapp scales. (A) The mitochondrial ΔGappX scale derived from Leu/Ala-based H-segments with the indicated amino acid placed in the middle of the H-segment (see Supplementary Table S1). Averages from at least three (more ...)
Meier et al (2005)
have reported that the presence of Pro residues in TM segments renders TIM23-mediated membrane insertion less efficient. To better understand the role of Pro in TM helix insertion, we analysed a set of H-segments carrying a single or a symmetrically disposed pair of Pro residues in different positions. As seen in , Pro residues strongly reduce membrane insertion when located in the middle section of the H-segment and to a lesser but significant extent when near the ends. This suggests that formation of an α-helical structure of the H-segment is an important determinant for TIM23-mediated recognition and insertion of the H-segment into the inner membrane.
Figure 4 Position-dependent effects of different amino acids on membrane insertion. (A) Symmetric pair-scan for Pro residues in the mitochondrial inner membrane (circles) and the mammalian ER membrane (squares). (B) Single-Pro scan of H-segment insertion into (more ...)
We next assessed the positional dependence of the effects on membrane insertion elicited by pairs of aromatic residues (Phe, Trp, Tyr; ) and by single-charged residues (Lys, Asp; ). Membrane insertion increased when Trp and Tyr were placed near the ends of the H-segment, while the effect of Phe was largely independent of position. The positively charged Lys residue was found not to affect insertion compared with the parent 6L/13A H-segment when placed in the first or last four positions. These results are consistent with our previous findings for the mammalian ER (Hessa et al, 2007
). In contrast, the negatively charged Asp residue was only tolerated when placed in the four C-terminal positions, whereas in the mammalian ER Asp is tolerated at both the N- and C-terminal ends of the H-segment.
Effects of charged and polar flanking residues
Finally, we investigated the role of charged and polar flanking sequences on H-segment insertion. On average, three positively charged Arg or Lys flanking residues (RRPR…, …RPRR, KKPK…, …KPKK) reduce ΔGapp by 1.0–1.3 kcal/mol from both the matrix and IMS sides compared with the GGPG… and …GPGG flanks (). This corresponds to approximately four Ala → Leu replacements in the hydrophobic part of the H-segment.
Figure 5 Effects of charged flanking residues on membrane insertion. Flanking sequences have the general design XXPX…XPXX (denoted 3X-3X) where X=G, R, K, or D. The results are expressed as ΔΔGapp values relative to the corresponding construct (more ...)
In contrast, a negatively charged DDPD… flanking sequence on the matrix side increases ΔGapp
by 1.1 kcal/mol, while a …DPDD sequence on the IMS side has little effect compared with the …GPGG flanking sequence, in agreement with the results for a single-Asp residue in . The same qualitative effects are seen with XXPX…,…XPXX flanking sequences where X=Glu, Gln, Asn, and His (see constructs #107–114, Supplementary Table S1
); that is, it is not the charge per se
but the high polarity of the side chain that matters in these cases. Interestingly, His does not behave as a positively charged residue in this context, which is consistent with the high pH of the mitochondrial matrix (around 8.0 in HeLa cells and rat cardiomyocytes (Llopis et al, 1998
)), a value that is well above the pKa for the imidazole side chain.
Positively charged residues thus promote membrane insertion from both sides of the membrane. This is in contrast to the mammalian ER, where positively charged residues only promote insertion if they are present on the cytoplasmic side of the H-segment (Hessa et al, 2007
). Negatively charged and highly polar residues reduce membrane insertion only when placed at the matrix side of the H-segment in the mitochondrial system, and only when placed at the lumenal side of the H-segment in the mammalian ER (Hessa et al, 2007
A second test protein: CoxVa
Given the importance of flanking residues located outside the hydrophobic segment itself and the relatively high threshold hydrophobicity for membrane insertion seen with the Mgm1p constructs, we also determined the threshold hydrophobicity using a second inner membrane protein, CoxVa. CoxVa has a single TM segment that is integrated into the inner membrane via the TIM23 translocon (Glaser et al, 1990
; Miller and Cumsky, 1993
). We replaced the CoxVa TM segment by GGPG-n
)A-GPGG H-segments of varying hydrophobicity and determined the efficiency of membrane insertion by an established protease-accessibility assay (Glaser et al, 1990
) (). As seen in , 50% membrane insertion is observed for n
≈2–3, compared with n
≈5–6 for Mgm1p, corresponding to a difference of ~1 kcal/mol in the threshold hydrophobicity between the CoxVa and Mgm1p constructs. Although minor differences between in vivo
and in vitro
import assays may explain part of this difference (as has been seen for insertion into the ER (Hessa et al, 2005
)), it is likely that sequence context outside the GGPG…GPGG flanks also impact the threshold hydrophobicity.
Figure 6 Analysis of H-segments in the context of the CoxVa protein. (A) CoxVa has an N-terminal mitochondrial-targeting peptide that is cleaved upon import by the matrix-localized mitochondrial processing peptidase (MPP). A segment consisting of residues 96–122 (more ...)
Positively and negatively charged flanking residues have similar effects on membrane insertion in the context of CoxVa as seen for Mgm1p, that is, positively charged flanks tend to increase insertion while negatively charged flanking residues on the matrix side of the hydrophobic segment strongly reduce insertion ().
Impairment of import-motor function increases membrane insertion
Herlan et al (2004)
have shown that mutational impairment of the mitochondrial import-motor components Tim44p and Pam18p/Tim14p reduces the formation of s
-Mgm1p, implying that membrane integration of Mgm1p is increased in the absence of motor activity. We therefore tested the effect of the import motor on the balance between s-
Mgm1p and l
-Mgm1p in constructs with different H-segments. In case of wild-type Mgm1p, we observed a strong reduction of the amount of s
-Mgm1p in the temperature-sensitive mutant pam16-3
of the motor subunit Pam16p (Frazier et al, 2004
) grown at 30°C, the highest temperature at which cells can still grow (). The relative levels of s-
Mgm1p were reduced in pam16-3
cells also for Mgm1p carrying H-segments of varying hydrophobicity, but to a smaller degree than for wild-type Mgm1p (). Similarly, for constructs with H-segments carrying charged or polar flanking residues, we saw a reduction in relative s-
Mgm1p levels (); the effects of the pam16-3
mutation were especially large when the charged or polar residues are at the matrix-facing, N-terminal end of the H-segment. It thus appears that a fully functional import motor increases the threshold for H-segment membrane insertion in the context of Mgm1p, possibly by pulling on the nascent chain.
Figure 7 The mitochondrial import motor affects membrane insertion. (A) Wild-type Mgm1p and two Mgm1p constructs carrying H-segments GGPG-5L/14A-GPGG and DDPD-8L/11A-GPGG were expressed in the temperature-sensitive pam16-3 mutant strain and in the isogenic PAM16 (more ...)