Tic20 and Tic110 display a differential expression pattern
Due to errors in the annotation of AtTic20-I, currently available Affymetrix micro-arrays do not contain specific oligonucleotides for this isoform and therefore cannot be used to investigate the expression levels of AtTic20-I [27]. We designed specific primers for Tic20 and Tic110 in pea and Arabidopsis and performed a quantitative RT-PCR (qRT-PCR) analysis to obtain comprehensive and more reliable quantitative data about the expression of Tic20 than those available from semi-quantitative analysis and the Massively Parallel Signature Sequencing database [19, 26, 27].
For the analysis, RNA was isolated from leaves and roots of two-week-old pea seedlings as well as four-week-old Arabidopsis plants. Arabidopsis was grown hydroponically to provide easy access to root tissue. In all samples, expression of Tic20 was analysed in direct comparison to Tic110 (Figure 1).
In pea, expression of both genes was found to be lower in root tissue as compared to leaves. In roots, PsTic110 RNA is 40% more abundant, while in leaves the expression levels of PsTic20 and PsTic110 seem to be in a similar range. In Arabidopsis, AtTic20-I and AtTic110 are expressed to a lower extent in roots than in leaves, similar to pea (Figure 1B). These results seemingly contradict those of Hirabayashi et al. [26], who concluded a comparable expression level of Tic20-I in shoots and roots. However, they used a non-quantifiable approach in contrast to our quantitative analysis. Furthermore, in our experiments the overall expression of AtTic20-I and AtTic110 differs notably from that in pea, AtTic110 RNA being about 3.5 and 6 times more abundant than AtTic20-I in leaves and roots, respectively.
We also designed specific primers for the second Tic20 homolog in Arabidopsis, AtTic20-IV, and our quantitative method was sufficiently sensitive to precisely define its RNA levels in Arabidopsis leaves and roots, allowing direct comparison with the expression of AtTic20-I and AtTic110 (Figure 1B). Transcription of AtTic20-IV had also been investigated in parallel to AtTic110 by Teng et al. [27], who observed a differential ratio of expression using two different methods, of which one was not even sensitive enough to detect AtTic20-IV. A very recent study [26] also investigated the expression of AtTic20-IV, however, without any quantification of their data.
Our data show that AtTic20-IV is present in leaves and roots with transcript levels similar to AtTic20-I, but less abundant than AtTic110. Interestingly, and in accordance with the data presented by Hirabayashi et al. [26], transcript levels of AtTic20-IV in roots are higher than those of AtTic20-I, while the opposite is true in leaf tissue. It can be speculated that the observed expression pattern reflects tissue-specific differentiation of both genes. AtTic20-IV may still partially complement for the function of AtTic20-I, as becomes evident from the viability of attic20-I knockout plants and the yellowish phenotype of attic20-I mutants overexpressing AtTic20-IV [26, 27]. However, the severe phenotype of attic20-I plants, in conjunction with the observed differential expression pattern, clearly indicates specific functions of the two homologs. Furthermore, a higher AtTic110 expression rate as observed in antisense attic20-I lines might indicate another possible compensatory effect [19].
The expression pattern of the three investigated genes was found to be similar in Arabidopsis growing hydroponically with or without sucrose (Figure 1B) or on soil (data not shown). However, gene expression was generally higher in plants growing without sucrose.
Tic20 protein is much less abundant than Tic110 in envelope membranes
Semi-quantitative analysis of Tic20 and Tic110 on protein level was performed using immunoblots of envelope membranes isolated from two-week-old pea and four-week-old Arabidopsis plants. In parallel, calibration curves were generated using a series of known concentrations of overexpressed and purified proteins (Figure 2A, B, D and 2E). After quantification of immunoblots from envelopes, amounts of PsTic20, PsTic110, AtTic20 and AtTic110 were determined using the corresponding calibration curve. The amount of PsTic110 in IE was found to be almost eight times higher than that of PsTic20 (Figure 2C), which differs strikingly from the similar transcript levels of the two genes detected in leaves (Figure 1A), indicating profound differences in posttranslational processes such as translation rate and protein turnover. In Arabidopsis, the absolute amount of AtTic110 is nearly the same as in pea (Figure 2F), however, Arabidopsis envelopes represent a mixture, containing both outer and IE vesicles. Thus, the relative amount of AtTic110 is possibly higher than in pea. Surprisingly, the amount of AtTic20 is more than 100 times lower than that of AtTic110, showing an even greater difference in comparison to the observed RNA expression levels (Figure 2F). Taking the different molecular size of Tic110 and Tic20 into account (~5:1), we still observe 20 times more AtTic110 than AtTic20 protein. In pea, we found 1.4 times more Tic110 RNA than Tic20, whereas in Arabidopsis the ratio of Tic110 to Tic20 is 20.3. The number of channel forming units must even be more different, since Tic110 was shown to form dimers [11], whereas Tic20 builds very large complexes between 700 kDa (this study) and 1 MDa [20]. Thus, two Tic110 molecules would be necessary to form a channel in contrast to Tic20, which would require many more molecules to form the pore. Though we cannot exclude that Tic20 might be subject to degradation by an unknown protease in vivo, protease treatments with thermolysin of right-side out IE vesicles in vitro clearly shows that Tic20 is very protease resistant, even in the presence of detergent. In contrast, Tic110 is easily degraded already without addition of detergent (Additional file 1). This argues against more rapid degradation of Tic20 compared to Tic110 during preparation of IE. The difference in Tic110 to Tic20 ratios both on the RNA and protein level between pea and Arabidopsis may be due to the different age of the plants or the different needs under the given growth conditions, and suggests that there is no strict stoichiometry between the two proteins. Moreover, the low abundance of Tic20 in comparison to Tic110 in envelopes (see also additional file 2) clearly demonstrates that Tic20 cannot be the main channel of the Tic translocon as previously suggested [21, 24], since it cannot possibly support the required import rates of some highly abundant preproteins that are needed in the chloroplast.
Tic20 forms high molecular weight complexes separately from Tic110
Experimental data suggested a common complex between Tic110 and Tic20 in chloroplast envelope membranes using a cross-linking approach [21]. However, the interaction was not visible in the absence of Toc components, making a stable association unlikely. Furthermore, no evidence for a common complex was found by Kikuchi et al. [20] using solubilized chloroplasts of pea and Arabidopsis for two-dimensional blue native/SDS-PAGE (2D BN/SDS-PAGE) analysis. Likewise, the difference in Tic110 to Tic20 ratios both on the RNA and protein level between pea and Arabidopsis indicates that a common complex, in which both proteins cooperate in translocation channel formation in a reasonable stoichiometry, is improbable.
To clarify this issue, we addressed these partly conflicting results by using IE vesicles, which should minimize the possible influence of the interaction with Toc components on complex formation. Pea IE vesicles were solubilized in 5% digitonin and subjected to 2D BN/SDS-PAGE. Immunoblots revealed that both Tic20 and Tic110 are present in distinct high molecular weight complexes (Figure 3A): Tic110-containing complexes migrate at a size of ~ 200-300 kDa, whereas Tic20 displays a much slower mobility in BN-PAGE and is present in complexes exceeding 700 kDa, in line with the results from Kikuchi et al. [20]. However, at a similar molecular weight of 250 kDa on BN-PAGE not only Tic110 but also Hsp93, Tic62 and Tic55 were described [30]. The molecular weight of a complex containing all of these components would be much higher. Therefore, components of the Tic complex might associate with Tic110 very dynamically resulting in different compositions under different conditions, or alternatively, there are different complexes present at the same molecular weight.
An open question to date is the identity of possible interaction partners of Tic20 in the complex. Tic22, the only Tic component located in the intermembrane space, is a potential candidate, since both proteins were identified together in cross-linking experiments [21]. However, only minor amounts of Tic20 and Tic22 were shown to co-localize after gel filtration of solubilized envelope membranes [21]. A second candidate for common complex formation is PIC1/Tic21: Kikuchi et al. [20] demonstrated that a one-megadalton complex of Tic20 contains PIC1/Tic21 as a minor subunit. PIC1/Tic21 was proposed to form a protein translocation channel in the Tic complex, mainly based on protein import defects of knockout mutants and on structural similarities to amino acid transporters and sugar permeases [27]. An independent study by Duy et al. [31] favours the hypothesis that PIC1/Tic21 forms a metal permease in the IE of chloroplasts, rendering the import-related role questionable. This discrepancy will have to be addressed in the future.
To test the complex formation of Tic20 in vitro without the involvement of other proteins, we used Tic20-proteoliposomes for 2D BN/SDS-PAGE analysis, similarly to IE vesicles (Figure 3B). The migration behaviour of the protein resembles that observed in IE: the majority of the protein localizes in high molecular weight range, however, the signal appears more widespread and a portion is also detected at lower molecular weights, possibly as monomers. This observation reveals that Tic20 has the inherent ability to homo-oligomerize in the presence of a lipid bilayer. The less distinct signal could be due to different solubilization of Tic20 by digitonin in IE vesicles vs. liposomes, or could be an indication that additional subunits stabilize the endogenous Tic20 complexes, which are not present after the reconstitution. However, we interpret these observations as support for the hypothesis that the major component of the one megadalton complex in IE are homo-oligomers composed of Tic20.
The N- and C-termini of Tic20 face the stromal side
In silico analysis of Tic20 predicts the presence of four hydrophobic transmembrane helices positioning both N- and C-termini to one side of the membrane (TMHMM Server [23] and [21, 25]). According to these predictions, three cysteins (Cys) in PsTic20 face the same side, while the fourth would be located in the plane of the membrane. We used pea IE vesicles prepared in a right-side-out orientation [32] to determine the topology of Tic20 employing a Cys-labelling technique. To this end, the IE vesicles were incubated with a membrane-impermeable, Cys-reactive agent (metoxypolyethylenglycol-maleimide, PEG-Mal) that adds a molecular weight of 5,000 Da to the target protein for each reactive Cys residue. In our experiments PEG-Mal did not strongly label any Cys residues of Tic20 under the conditions applied (Figure 4A), indicating the absence of accessible Cys residues on the outside of the membrane. Only one faint additional band of higher molecular weight was detectable (Figure 4A, marked with asterisk), possibly due to a partially accessible Cys located within the membrane. In the presence of 1% SDS, however, all four Cys residues present in PsTic20 are rapidly PEGylated, as demonstrated by the appearance of four intense additional bands after only five minutes of incubation. The observed gain in molecular weight per modification is bigger than the expected 5 kDa for each Cys, but this can be attributed to an aberrant mobility of the modified protein in the Bis-Tris/SDS-PAGE used in the assay.
Our results support a four transmembrane helix topology in which both the C- and N-termini are facing the stromal side of the membrane (Figure 4B), with no Cys residues oriented towards the intermembrane space. Cys108 is most likely located in helix one, Cys227 and Cys230 are oriented to the stromal side of helix four and Cys243 is located in the stroma. This topology is also in line with green fluorescent protein-labelling studies by van Dooren et al. [22] indicating that the N- and C-termini also of the Toxoplasma gondii homolog of Tic20 face the stromal side of the inner apicoplast membrane.
Tic20 is mainly α-helical
Tic20 was identified more than a decade ago but since then no heterologous expression and purification procedure has been reported, which could successfully synthesize folded full-length Tic20. Here, we report two efficient Escherichia coli (E. coli) based systems for Tic20 expression and purification from both pea and Arabidopsis: codon optimized PsTic20 (Additional file 3) was overexpressed in a S12 cell lysate in presence of detergents, and AtTic20 overexpression was successfully accomplished by adaptation of a special induction system [33]. Following these steps, both pea and Arabidopsis proteins could be purified to homogeneity by metal affinity purification (Figure 4C).
Using the purified protein, we performed structural characterization studies of Tic20 by subjecting it to circular dichroism (CD) spectroscopy (Figure 4D). The recorded spectra of PsTic20, displaying two minima at 210 and 222 nm and a large peak of positive ellipticity centered at 193 nm, are highly characteristic of α-helical proteins, and thus demonstrate that the protein exists in a folded state after purification in the presence of detergent. The secondary structure of Tic20 was estimated by fitting spectra to reference data sets (DichroWeb server [34, 35]) resulting in an α-helical content of approximately 78%, confirming in silico predictions [21, 25].
Purified Tic20 protein inserts firmly into liposomes
To better characterize Tic20 in a membrane-mimicking environment, heterologously expressed and purified AtTic20 was reconstituted into liposomes in vitro. Initially, flotation experiments were performed to verify a stable insertion. In the presence or absence of liposomes, Tic20 was placed at the bottom of a gradient ranging from 1.6 M (bottom) to 0.1 M (top) sucrose. In the presence of liposomes, Tic20 migrated to the middle of the gradient, indicating a change in its density caused by interaction with liposomes. In contrast, the protein alone remained at the bottom of the gradient (Figure 5A). Proteoliposomes were also treated with various buffers before flotation (for 30 min at 4°C), to test whether the protein is firmly inserted into the liposomal membrane or just loosely bound to the vesicle surface. None of the applied conditions (control: 10 mM MOPS/Tris, pH 7; high ionic strength: 1 M MOPS/Tris, pH 7; high pH: 10 mM Na2CO3, pH 11; denaturing: 6 M urea in 10 mM MOPS/Tris, pH 7) changed the migration behaviour of Tic20 in the gradient (Figure 5B), indicating that Tic20 was deeply inserted into the liposomal membrane. Thus, proteoliposomes represent a suitable in vitro system for the analysis of Tic20 channel activity.
Tic20 forms a channel in liposomes
Even though Tic20 has long been suggested to form a channel in the IE membrane, this notion was solely based on structural analogy to other four-transmembrane helix proteins [21, 24], and no experimental evidence has been provided so far. To investigate whether Tic20 can indeed form an ion channel, Tic20-proteoliposomes were subjected to swelling assays (Figure 5C). Changes in the size of liposomes in the presence of high salt concentrations, as revealed by changes in the optical density, can be used to detect the presence of a pore-forming protein [36]. After addition of 300 mM KCl to liposomes and Tic20-proteoliposomes, their optical densities dropped initially, due to shrinkage caused by the increased salt concentration [37]. However, the optical density of protein-free liposomes remained at this low level, showing no change in their size; whereas in the case of Tic20-proteoliposomes the optical density increased constantly with time. The increase in optical density (and therefore size) strongly supports the presence of a channel in Tic20-proteoliposomes that is permeable for ions, thereby creating an equilibrium between the inner compartment of the proteoliposomes and the surrounding buffer.
To exclude the possible effects of (i) contaminating channel-forming proteins derived from the bacterial membrane and (ii) a protein inserted into the liposomes (but not forming a channel), a further negative control was set up: Tic110 containing only the first three transmembrane helices (NtTic110) was purified similarly to Tic20 and reconstituted into liposomes. We chose this construct, since NtTic110 inserts into the membrane during in vitro protein import experiments [10]. Furthermore, as the full length and N-terminally truncated Tic110 possess very similar channel activities [11, 12], it is unlikely that the N-terminal part alone forms a channel. The insertion of NtTic110 into liposomes was confirmed by incubation under different buffer conditions (high salt concentration, high pH and 6 M urea) followed by flotation experiments, similarly to Tic20 (data not shown). However, these NtTic110-proteoliposomes behaved similarly to the empty liposomes during swelling assays: after addition of salt, the optical density decreased, and except for a small initial increase, it remained at a constant level (Figure 5C). This makes it unlikely that a contamination from E. coli or simply the insertion of a protein into the liposomes caused the observed effect in the optical density of Tic20-proteoliposomes.
To further characterize the channel activity of Tic20, electrophysiological measurements were performed. After the fusion of Tic20-proteoliposomes with a lipid bilayer, ion channel activity was observed (Figure 6A, B). The total conductance under symmetrical buffer conditions (10 mM MOPS/Tris (pH 7.0), 250 mM KCl) was dependent on the direction of the applied potential: 1260 pS (± 70 pS) and 1010 pS (± 50 pS) under negative and positive voltage values, respectively. The channel was mostly in the completely open state, however, individual single gating events were also frequently observed, varying in a broad range between 25 pS to 600 pS (Figure 6A-D). All detected gating events were depicted in two histograms (Figure 6C, D for negative and positive voltages, respectively). Two conductance classes (I and II) were defined both at negative and positive voltage values with thresholds of 220 pS and 180 pS, respectively (Figure 6A-E). Note that gating events belonging to the smaller conductance classes (I) occurred more frequently. The observed pore seems to be asymmetric, since higher conductance classes notably differ under positive and negative voltages. This is probably due to interactions of the permeating ions with the channel, which presumably exhibits an asymmetric potential profile along the pore. Since small and large opening events were simultaneously observed in all experiments, it is very unlikely that they belong to two different pores.
The selectivity of Tic20 was investigated under asymmetric salt conditions (10 mM MOPS/Tris (pH 7.0), 250/20 mM KCl). Similarly to the conductance values, the channel is intrinsically rectifying (behaving differently under negative and positive voltage values), supporting asymmetric channel properties. The obtained reverse potential is 37.0 ± 1.4 mV (Figure 6D). According to the Goldman-Hodgkin-Katz approach, this corresponds to a selectivity of 6.5:1 for K+:Cl--ions, thus indicating cation selectivity similar to Tic110 [11].
To determine the channel's orientation within the bilayer, two side-specific characteristics were taken into account: the highest total conductance under symmetrical buffer conditions was measured under negative voltage values, and the channel rectifies in the same direction under asymmetrical buffer conditions (see voltage ramp, Figure 6D). Therefore, it seems that the protein is randomly inserted into the bilayer.
The pore size was roughly estimated according to Hille et al. [38]. Considering the highest conductance class (350 pS), a channel length of 1-5 nm and a resistivity of 247.5 Ω cm for a solution containing 250 mM KCl, taking into account that the conductivity of the electrolyte solution within the pore is ~5 times lower than in the bulk solution [39], the pore size was estimated to vary between 7.8-14.1 Å. This is in good agreement with the size of protein translocation channels such as Toc75 (14-26Å, [40]) in the outer envelope membrane and Tic110 (15-31 Å, [12]) in the IE. Thus, the size of the Tic20 pore would be sufficient for the translocation of precursor proteins through the membrane.
NtTic110, as a negative control, did not show any channel activity during electrophysiological measurements, indicating that the measured channel is not the result of a possible bacterial contamination (data not shown).
Considering our data presented here and those published in previous studies, we can conclude that the Tic translocon consists of distinct (at least two) translocation channels: On the one hand, Tic110 forms the main translocation pore and therefore facilitates import of most of the chloroplast-targeted preproteins; on the other hand, Tic20 might facilitate the translocation of a subset of proteins. This scenario would match the one found in the inner mitochondrial membrane, where specific translocases exist for defined groups of precursor proteins: the import pathway of mitochondrial carrier proteins being clearly separated from that of matrix targeted preproteins [41]. The situation in chloroplasts does not seem as clear-cut, but an analogous separation determined by the final destination and/or intrinsic properties of translocated proteins is feasible.
The severe phenotype of attic20-I mutants prompts us to hypothesize that Tic20 might be specifically required for the translocation of some essential proteins. According to cross-linking results [21], Tic20 is connected to Toc translocon components. Therefore, after entering the intermembrane space via the Toc complex, some preproteins might be transported through the IE via Tic20. On the contrary, Kikuchi et al. [20] presented that Tic20 migrates on BN-PAGE at the same molecular weight as the imported precursor of the small subunit of Rubisco (pSSU) and that tic20-I mutants display a reduced rate of the artificial precursor protein RbcS-nt:GFP. The authors interpreted these results in a way that Tic20 might function at an intermediate step between the Toc translocon and the channel of Tic110. However, being a substantial part of the general import pathway seems unlikely due to the very low abundance of Tic20. It is feasible to speculate that such abundant proteins as pSSU, which are imported at a very high rate, may interact incidentally with nearby proteins or indifferently use all available import channels. To clarify this question, substrate proteins and interaction partners of Tic20 should be a matter of further investigation.
Additionally, a very recent study [26] suggested AtTic20-IV as an import channel working side by side with AtTic20-I. However, detailed characterization of the protein (e.g. localization, topology) and experimental evidence for channel activity are still missing.