The N-terminal region of Cisd1 and Cisd2 targets them to mitochondria and the endoplasmic reticulum, respectively
The primary targeting of proteins during or shortly after their translation is often controlled by their N-terminal sequence. In both Cisd1 and Cisd2, a hydrophobic segment is present in the N-terminal portion (Fig. 1a). Accordingly, this region of the protein may target CISD1 and CISD2 to the ER or to mitochondria, or to both compartments. In order to verify this hypothesis, we first produced in HEK cells Cisd1 or Cisd2 together with a yellow-shifted green fluorescent protein targeted to the ER (ER-YFP) and a red fluorescent protein targeted to mitochondria (mito-RFP) (Fig. 1b). Importantly, we expressed native Cisd1 and Cisd2 proteins. Indeed, fusion to additional sequences (peptide tags or fluorescent proteins) facilitates detection of proteins, but may alter their intracellular localization. To avoid potential artefacts, we developed recombinant antibodies recognizing specifically either the Cisd1 cytosolic domain (RB251) [11] or the Cisd2 cytosolic domain (RB253) [12, 13]. RB251 detected the presence of endogenous Cisd1 in HEK cells (Additional file 1). On the contrary RB253 did not detect significant amounts of Cisd2 in HEK cells (Additional file 1). The signal generated by RB253 in untransfected HEK cells was so weak that we cannot unambiguously determine if it corresponds to a non-specific binding of the antibody to cellular structures, or to the specific detection of a very small amount of endogenous Cisd2. Each antibody detected a large amount of its target protein in cells transfected with adequate expression plasmids (Additional file 1). As expected, we observed that in transfected cells, Cisd1 was localized in mitochondria, and absent from the ER (Fig. 1b). In addition, as previously reported [8], overexpression of Cisd1 resulted in the aggregation of mitochondria (Fig. 1b, arrowhead). On the contrary, Cisd2 was clearly colocalized with ER-targeted YFP, in particular at the level of the nuclear envelope, and not detected in mitochondria (Fig. 1b). We cannot exclude that a small portion of Cisd2 was localized in mitochondria, but then its concentration would be much weaker than in the ER. The structure of the ER often appeared slightly altered in cells expressing Cisd2 (Fig. 1b): the ER was less finely dispersed in the cytosol, and the nuclear envelope was less regularly organized. However, the limited resolution of immunofluorescence microscopy and the significant cell-to-cell variability prevented us from ascertaining this point unambiguously. We made almost identical observations in HEK cells and in three other cell types (Hela, HCT116 and Huh–7): endogenous Cisd1 is localized to mitochondria and its overexpression induces clustering of mitochondria (Additional file 2). Endogenous Cisd2 is undetectable. In transfected cells it is localized in the ER and perturbs its structure (Additional file 3).
We then produced in HEK cells a chimeric protein (Cisd12) composed of the N-terminal portion of Cisd1, fused to the cytosolic region of Cisd2 (Fig. 1a; Additional file 4). Cisd12 was clearly targeted to mitochondria, and remarkably, like Cisd1 it was capable of causing aggregation of mitochondria (Fig. 1b, arrowhead). We also produced a chimeric protein (Cisd21) composed of the N-terminal portion of Cisd2, fused to the cytosolic domain of Cisd1 (Fig. 1a) and observed that it was localized in the ER. Since there is a significant cell-to-cell variability in the morphology of the ER and of mitochondria, a second set of pictures is presented, essentially confirming the results shown in Fig. 1b (Additional file 5). Together these results demonstrate that the N-terminal region of Cisd1 and Cisd2 ensures their targeting to the mitochondria and to the ER, respectively.
Cisd2 is retained in the ER by its TMD and its cytosolic KKxx motif
Our next aim was to determine how Cisd2, once inserted in the ER, avoids being transported to the cell surface with the bulk of membrane proteins. In order to separate co-translational insertion in the ER from later intracellular transport events, and to provide an invariant method to detect various mutant proteins, we fused Cisd2 at its N-terminal end with a signal sequence and the coding sequence of the luminal domain of CD1b (Fig. 2a; Additional file 6). The resulting CD1b-M1 protein was produced in HEK cells and largely colocalized with ER-targeted YFP, in particular in the nuclear envelope (Fig. 2b). Similar to cells expressing Cisd2, the fine structure of the ER appeared altered in some cells expressing CD1b-M1 (Fig. 2b). In some cells the structure of the nuclear envelope was profoundly modified, with CD1b-M1 unevenly distributed around the nucleus (Fig. 2c; asterisks). For all CD1b chimeric proteins, a duplicate set of images is provided in Additional file 7.
Two elements could in principle ensure efficient ER retention of CD1b-M1: its TMD and its C-terminal KKxx sequence. The TMD of Cisd2 is unusually short (16AA), a feature which can ensure ER retention of some transmembrane proteins [14]. To test this hypothesis, we produced in HEK cells a chimeric protein composed of the CD1b extracellular domain fused to the TMD of Cisd2, and a short cytosolic domain (CD1b-M2). CD1b-M2 was mostly localized in the ER (Fig. 2b), and also to the Golgi apparatus where it was colocalized with giantin, a marker of the Golgi apparatus (Fig. 2c; arrow). We also tried to detect the presence of CD1b fusion proteins by surface immunolabeling. While CD1b-M1 was virtually absent from the cell surface, a significant fraction of CD1b-M2 was detected at the cell surface (Fig. 3). These results indicate that the TMD of Cisd2 does confer ER retention, but not as efficiently as the full Cisd2 protein.
The second element that may ensure ER localization of Cisd2 is the KKEV sequence found at its C-terminus. In order to test this hypothesis, we first produced a CD1b-M3 fusion protein composed of the CD1b extracellular and TMD fused to the cytosolic domain of Cisd2 (Fig. 2a). The CD1b-M3 protein was mostly localized in the ER (Fig. 2b), with no apparent colocalization with a Golgi marker (Fig. 2c). However, a small but significant amount of CD1b-M3 was detected at the cell surface (Fig. 3). In order to ascertain whether ER localization of CD1b-M3 was due to the presence of a C-terminal KKxx motif, we disrupted the putative KKxx motif by mutating its two lysine residues to serine residues (CD1b-M4; Fig. 2a) or by adding 4 additional serine residues at the C-terminal end of Cisd2 (CD1b-M5; Fig. 2a). Both CD1b-M4 and CD1b-M5 were poorly retained in the ER or the Golgi apparatus (Fig. 2b and c), and mostly present at the cell surface (Fig. 3). These conclusions were confirmed by western blot analysis of the CD1b chimeric proteins (Additional file 8): only CD1b-M1, CD1b-M4 and CD1b-M5 showed high molecular weight species typical of proteins bearing mature sugars, indicating that they have escaped massively the ER to reach the plasma membrane.
Overall, these results indicate that the efficient ER localization of CD1b is achieved by the additive effect of its short TMD and its C-terminal cytosolic KKxx motif.
The Cisd2 KKEV binds COPI, Cisd1 KKET does not
Although the presence of a functional KKEV ER retrieval motif in Cisd2 seems relevant, the presence of a very similar sequence (KKET) in mitochondrial Cisd1 is more surprising. To investigate this point, we assessed the ability of the Cisd1 and Cisd2 C-terminal regions to bind cytosolic COPI. For this, peptides corresponding to the 10 C-terminal residues of Cisd1 and Cisd2 were synthesized and coupled to avidin at their N-terminal extremity (Fig. 4a). For comparison we used the well-characterized WBP1 KKxx COPI-binding motif and a mutant (WBP1-SS) unable to bind COPI [9]. As previously observed [9], when cell lysates were incubated with these peptides, COPI bound to WBP1 but not to WBP1-SS (Fig. 4b). The biggest subunits of the COPI complex (α: 170 kDa; β, β’ and γ: 110 kDa) were clearly detected by Coomassie blue staining (Fig. 4b; stars). COPI also bound, though less efficiently, to the Cisd2 peptide: COPI subunits were barely detectable in a Coomassie-stained gel (Fig. 4b), but readily apparent in a silver-stained gel (Fig. 4c). COPI did not visibly bind the Cisd1 peptide. These observations confirm the fact that Cisd2 presents a functional COPI-binding KKxx motif. On the contrary, despite the high degree of sequence similarity, the Cisd1 C-terminus does not have the ability to bind efficiently the COPI complex.
Cisd2 expression alters the structure of the ER
As detailed above, immunofluorescence analysis suggested that expression of Cisd2 alone or fused to the CD1b extracellular domain (CD1b-M1) altered the structure of the ER and of the nuclear envelope. However, the limited resolution of light microscopy and the heterogenous morphology of the ER prevented us from drawing firm conclusions.
In order to assess the effect of Cisd2 on the ER morphology at the ultrastructural level, we expressed in HEK cells the CD1b-M1 chimera, and used a specific antibody to recognize the CD1b extracellular domain by cryo-immuno electron microscopy. In cells where expression of CD1b-M1 was detected, instead of being widely deployed in the cytosol, the ER network often appeared to partially collapse onto the nuclear envelope (arrowheads) with which it connected on multiple points (Fig. 5a). This was also visible in sections of Epon-embedded cells where a better contrast is obtained, although the presence of the CD1b-M1 cannot be assessed (Fig. 5b).
Since Cisd2 has both the ability to interact with COPI and to modify the structure of the ER, we further studied if high levels of Cisd2 altered the intracellular distribution of COPI. For this, we first selected and characterized recombinant antibodies that recognize the fully assembled COPI complex by immunofluorescence [15, 16]. We then used these antibodies to assess the localization of COPI in cells expressing either Cisd2, or a CD1b-tagged Cisd2. We did not observe in these cells a significant change in the localization of COPI: like in mock-transfected cells, COPI was localized mostly in the peri-Golgi region, as well as in dots throughout the cytosol (Fig. 6). COPI did not noticeably accumulate in regions where the ER structure appeared abnormally expanded.