60S dynamic state of bacterial ribosome is fixed by yeast mitochondrial initiation factor 3

Cloning and standard procedures

Different versions of AIM23 (Saccharomyces cerevisiae) and infC (E. coli) genes were cloned into above-mentioned vectors by standard polymerase chain reaction (PCR)-restriction-ligation approach.

Western-blot was performed by standard protocol using the rabbit antibodies against 6-His-tagged recombinant Aim23p (produced on our order by Almabion).

Construction of mutant E. coli strains

Genomic disruption of infC gene coding for IF3 was carried out in the E. coli strain MG1655. Cassette for infC genomic disruption containing the chloramphenicol resistance gene was prepared by PCR from pKD3 plasmid. Primers contained 5′-parts designed for the homologous recombination into the target genome site. The cassette was then delivered into E. coli cells by electroporation. These cells initially contained pKD46 plasmid encoding for recombinase, as well as pACDH plasmid containing infC gene. Clones where recombination took place were selected on chloramphenicol-containing medium and screened by PCR (Thomason, Costantino & Court, 2007).

For transferring the bacterial genetic material to phage P1, five ml of E. coli culture in logarhythmic growth phase was infected by 100 μl of phage suspension. The mixture was incubated at 37 °C for 3 h with shaking and centrifuged at 9,200×g for 10 min. The phage-containing upper fraction was taken and filtered through 0.45 μm filter.

For generation of the experimental E. coli strains, the MG1655 cells containing pBAD plasmid with cloned infC gene were transformed by the plasmids coding for IF3 or different variants of Aim23p. About two ml of ON cultures were pelleted and resuspended in one ml of 10 mM CaCl₂, five mM MgSO₄. Suspensions were 100 μl aliquoted, then half a volume of P1 lysate was added, and the mixtures were incubated at 37 °C for 30 min without shaking. Then one ml of lysogeny broth (LB) medium and 200 μl of sodium citrate were added followed by the incubation at 37 °C for 1 h with shaking. The cells were plated on the agar dishes with antibiotics, 0.02% arabinose and five mM sodium citrate. Screening of the clones was performed by PCR.

The growth curves of E. coli strains were registered in automatic mode using microplate reader Infinite M200 PRO (Tecan Instruments, Mannedorf, Switzerland).

Ribosome purification and analysis

Ribosomes were isolated from E. coli strain MG1655 according to Rivera, Maguire & Lake (2015) with minor changes. Briefly, bacterial cells from one l culture with OD₆₀₀ ∼0.6 were collected, lysed, ribosomes from clarified lysate were sedimentated through 10% sucrose cushion, and dissolved in minimal volume of 10 mM Tris–HCl pH 7.0; 60 mM KCl, 60 mM NH₄Cl, seven mM magnesium acetate, seven mM β-mercaptoethanol, 0.25 mM EDTA. Isolated ribosomes were stored at −80 °C. For dissociation reaction, approximately 24 pmoles (one unit of OD₂₆₀) of ribosomes were mixed with different amounts of recombinant Aim23p, IF3, or Aim23ΔNΔC in the above-indicated buffer. Mixtures were incubated at 37 °C for 30 min, and then applied on 15–40% continuous sucrose gradients prepared on the same buffer. Samples were centrifuged for 18 h at 100,000×g, and then fractionated from top to bottom (45 fractions each of 250 μl were taken). Absorbencies of all fractions at 260 nm were measured.

In case of cross-linking, between the incubation with proteins and loading on the sucrose gradient, formaldehyde was added to ribosome preparations up to 1%, and the mixtures were incubated for 30 min on ice.

Molecular modeling

The homology model of the Aim23p complex with the E. coli 30S subunit was done with Modeller 9.17 (Sali & Blundell, 1993) (script may be found in Supplementary Information). For the building of this model, we have used the known structure of the bacterial 30S subunit complex with the cognate IF3 (Pioletti et al., 2001), as well as the sequence alignment of Aim23p with E. coli IF3 and other orthologs (Atkinson et al., 2012).

The folding of Aim23p N-terminal extension was done with the AbinitioFold protocol (Bonneau et al., 2001, 2002) based of fragments obtained from the Robetta web server (Kim, Chivian & Baker, 2004). Simulation was stopped after 180,000 decoys were collected. The homology modeling with new conformation of N-terminal extension was done with Modeller 9.17 (Sali & Blundell, 1993). Conformation of the C-terminal extension was equilibrated with FloppyTail protocol from Rosetta (Kleiger et al., 2009). The spatial structure visualization was done with the PyMOL Molecular Graphics System, version 2.0 (Schrödinger, LLC, New York, NY, USA).

Full-length Aim23p is undesirable for E. coli cells due to its terminal extensions

As it has been already mentioned in “Introduction,” mammalian mtIF3 possesses some functional activity in E. coli cells (Ayyub et al., 2018). On the other hand, we have previously demonstrated that E. coli IF3 may partially rescue the growth defects of the yeast strain lacking Aim23p (Kuzmenko et al., 2014). Moreover, Aim23p was shown to bind the small subunit of bacterial ribosome in vitro (Atkinson et al., 2012). Taken together, these findings have allowed us to hypothesize that Aim23p might be at least partially functional in E. coli cells as IF3.

To verify this hypothesis, we have constructed three plasmids for further delivery into E. coli cells, coding for either cognate IF3 (positive control), Aim23p, or Aim23p without its mitochondria-specific N- and C-terminal extensions (Aim23ΔNΔC). This last construct was designed in order to specifically check possible effects of Aim23p terminal extensions on bacterial translation: theoretically, these protein parts, being mitochondria-specific, might not be needed for protein biosynthesis in E. coli. Cloned genes of Aim23p and Aim23ΔNΔC did not contain sequences coding for mitochondrial targeting signal. Thereafter, we have disrupted E. coli infC gene coding for IF3. This gene contains the promoter for the expression of the downstream gene (Wertheimer, Klotsky & Schwartz, 1988), so we have removed only first 153 nucleotides of infC gene from the bacterial genome. Since IF3 is indispensable for bacteria, before disruption we have transformed E. coli with the plasmid bearing infC gene under control of glucose-repressible promoter. Finally, we have delivered the above-described plasmids into bacterial cells and disrupted the genomic copy of infC gene. The scheme of bacterial strains engineering is presented in Fig. 1A.

Then, we down-regulated the expression of infC gene in these strains with glucose and measured their growth rates. The resulting curves may be found in Fig. 1B. The strain bearing wild-type infC gene on the plasmid grows normally, with fast entering the logarithmic phase and reaching the plateau. The strain carrying empty vector shows no growth at first 10 h of incubation which is easily explained by the absence of infC gene. However, slow growth has been detected afterwards, probably as a result of glucose-repressed promoter leakage which, in turn, allows minimal amount of IF3 to be synthesized. If E. coli cells contain Aim23ΔNΔC, the corresponding strain’s growth curve is almost identical to that of the strain containing an empty vector. This clearly indicated the impossibility of Aim23ΔNΔC to functionally substitute IF3 in bacterial cells. The most interesting case is definitely the bacterial strain bearing the full-size Aim23p. This strain, although reaching finally the level of the strain containing an empty vector, enters the logarithmic phase measurably slower than the latter. This means that the full-size Aim23p, but not Aim23ΔNΔC, negatively affects the viability of E. coli cells. It is rather possible that the terminal extensions of Aim23p may somehow interrupt the bacterial translation.

Terminal extensions provide the ability of Aim23p to fix an unusual state of E. coli ribosomes

The above-described unusual effect of Aim23p in E. coli cells has led us to study the interaction of Aim23p with E. coli ribosomes in vitro. It is well known that adding cognate IF3 to purified bacterial ribosomes shifts the equilibrium of the ribosome dissociation reaction making the dissociated state preferable (Gottleib, Davis & Thompson, 1975). Based on this, we have purified ribosomes from E. coli cells, incubated them with the recombinant IF3 (positive control), or Aim23p, or Aim23ΔNΔC, fractionated the reactions by sucrose gradient centrifugation and analyzed the corresponding sedimentation profiles by measuring the optical densities of the fractions at 260 nm. The results of our experiment are presented in Fig. 2A. First of all, sedimentation profile of the ribosome sample with no proteins added was characterized by clear UV peaks of 30S and 50S subunits, as well as the whole 70S ribosomes, with the latter being most pronounced. Adding IF3, as expected, led to the complete dissociation of ribosomes into subunits (disappearance of the 70S peak and significant increase of the 30S and 50S peaks) while adding of Aim23ΔNΔC gave no effect on the sedimentation profile. This was also expected: in our in vivo experiments, this protein could not substitute for the cognate IF3 in E. coli cells (see Fig. 1). The profile of sedimentation has been curiously changed with adding the full-size Aim23p. This protein caused a fusion of 50S and 70S peaks with the formation of a single wide peak with maximum UV absorbance corresponded to approximately 60S sedimentation coefficient. At the same time, the 30S peak was increased, but to the less extent than in case of the full dissociation promoted by IF3. The most logical explanation of this phenomenon would be that Aim23p cannot promote the normal ribosome dissociation at concentrations used in the experiment (20:1 molar ratio in relation to the ribosomes concentration) but instead binds it and fixes this unusual state of ribosomes. The appearance of this “60S state” might be linked somehow to the Aim23p slight toxicity for E. coli cells observed by us (see Fig. 1B). It should be noted, finally, that such action of Aim23p on the bacterial ribosomes is definitely mediated by its mitochondria-specific terminal extensions.

The same ribosomal fractions were analyzed for presence of recombinant proteins by Western-blot hybridization. We used the homemade antibodies against 6-His-tagged recombinant Aim23p, and, luckily, they had a significant cross-reactivity with the 6-His-tag (data not shown). Thus, we were able to detect Aim23p, Aim23ΔNΔC, and IF3 since all recombinant proteins used in our experiments were 6-His-tagged. Results of this experiment can be found in Fig. 2B. E. coli IF3 was indeed detected only in 30S fractions while the Aim23p version without terminal extensions was not seen in either ribosomal fraction; instead, Aim23ΔNΔC was found in the very first fractions with no ribosomes. This explains the impossibility of Aim23ΔNΔC to promote dissociation of bacterial ribosomes: this protein binds neither 70S ribosomes nor their separate subunits. Interestingly, full-size Aim23p behaves in all the contrary way compared to its version lacking terminal extensions. This protein is detectable in nearly all fractions containing ribosomes, either assembled or in the form of subunits. Maximum amount of Aim23p is bound to free 30S subunits, and there is almost equal distribution of the protein between fractions corresponding to 50S and “60S” peaks. Such non-canonical manner of binding ribosomes might be one of the reasons why Aim23p promotes formation of their “60S” state.

“60S state” of E. coli ribosomes is not a stable structure

The observed peak at 60S zone of the sedimentation profile may be explained in at least two ways. The first explanation is that Aim23p binds E. coli ribosomes and causes changes in their structure so that their sedimentation coefficient decreases to 60S. An alternative explanation is that, upon Aim23p binding, the ribosomal subunits become more flexible relative to one another. This, in turn, allows their reciprocal movements without full dissociation. In this case, the observed 60S peak might reflect the changed dynamic equilibrium between associated and dissociated states of the ribosome, rather than formation of a stable structure. In order to distinguish between these two possibilities, we have repeated our ribosomes fractionation experiment with additional cross-linking step (with the help of formaldehyde) after incubation with proteins. We did not use Aim23ΔNΔC here since this protein is unable to bind E. coli ribosomes (Figs. 2A and 2B). The results of cross-linking experiment are presented in Fig. 2C. After the formaldehyde treatment, the sedimentation profiles of ribosomes incubated with Aim23p or IF3 were almost identical, and there was no 60S peak of the Aim23p-bound ribosomes. If 60S peak would reflect the formation of a stable ribosomal structure, this structure would be fixed by cross-linking. Thus, our results clearly indicate that the observation of 60S peak is the consequence of some dynamic processes caused by Aim23p binding. One may speculate that these processes are the very first stages of 70S ribosomes dissociation which cannot continue normally due to the unusual manner of Aim23p interaction with the ribosomes.

After obtaining these intriguing results, we decided to analyze the dose-dependency of the Aim23p effect on E. coli ribosomes. The resulting profiles of ribosomes sedimentation after adding Aim23p at different concentrations are presented at Fig. 3A. If Aim23p concentration is 2.5 times more than in previously described experiment (i.e., 50:1 molar ratio in relation to the ribosomes concentration), then the peak of “60S state” is almost not observed. Instead, one can see a normal 50S peak which is slightly moved toward the increase of the sedimentation coefficient. At the same time, the 30S peak is meaningfully increased relative to the situation when the “60S state” is clearly observed. When Aim23p concentration is increased twice more (up to 100:1 molar ratio in relation to the ribosomes concentration), the resulting profile is identical to that in case of IF3 adding to ribosomes. One can hypothesize that these results reflect the consecutive stages of 70S ribosomes dissociation through “60S state” to free 30S and 50S subunits.

Aim23p and E. coli IF3 act jointly to dissociate bacterial ribosomes in vitro

The discovered “60S state” of bacterial ribosomes might be the result of the decreased ribosome stability. In other words, the equilibrium of the dissociation reaction in this case may be slightly shifted toward free subunits without full dissociation. This, in turn, means that such state of the ribosome should be subjected to dissociation easier than the normal 70S state. In order to check this hypothesis, we performed in vitro dissociation experiments with Aim23p and IF3 being simultaneously added to ribosomes. While Aim23p was added in concentration sufficient for “60S state” fixation (10:1 molar ratio in relation to the ribosomes concentration; see Fig. 3B), the amount of IF3 used was not enough for full ribosome dissociation (15:1 molar ratio in relation to the ribosomes concentration). If both proteins were presented in the reaction together (each at the same concentration as alone), the complete dissociation was detected. This phenomenon could be explained as follows. If “60S state” appears as a result of Aim23p action, the minimal amount of the free 30S subunits is immediately formed (this was also seen in our previous experiments, see Figs. 2A and 3A). Adding a little amount of the cognate IF3 leads to the fixation of the 30S subunits in their free state and further shifts the reaction equilibrium toward the dissociated state of the ribosome. Thus, Aim23p and IF3 may act jointly to promote the dissociation of the bacterial ribosomes. To verify this, we performed a Western-blot analysis of the ribosomes fractions corresponding to the free subunits and to the whole ribosomes in presence of Aim23p, or IF3, or both proteins together. The results are presented at Fig. 3C. IF3 in this experiment has been found to bind exclusively free 30S subunits but not 70S ribosomes, exactly as expected. Aim23p, however, is detected both in free 30S subunits and in the 70S ribosomes fractions (which fits well to our results presented in Fig. 2B), and this does not qualitatively depend on presence or absence of IF3 in the reaction. This explains well the joint action of these two proteins resulting in the ribosomes dissociation which cannot be achieved when using Aim23p and IF3 at the same concentrations separately.

Molecular modeling points on the importance of Aim23p terminal extensions for protein interaction with E. coli ribosomes

A very interesting question rises from the above-described results: in which manner does Aim23p interact with bacterial ribosome and what is the role of its terminal extensions in such an interaction? To shed light on this problem, we have performed molecular modeling.

Previously (Atkinson et al., 2012), sequence alignment of Aim23p with E. coli IF3 and other orthologs has been done. On the base of this data, as well as the known structure of 30S complex with the cognate IF3 (Pioletti et al., 2001), we have built the homology model of Aim23p complex with E. coli 30S ribosomal subunit with the help of Modeller 9.18 (script may be found in Supplementary Information). In the resulting model, Aim23p eventually has a long and extended N-terminal tail (Fig. S2). The size of this tail was comparable with the size of the 30S subunit, and the model could not provide valuable information about N-terminal extension function. We have suggested that the N-terminal extension is somehow structured and have built the corresponding model with the Rosetta AbInitio protocol. From 18,398 decoys of N-terminal extension, the top 10 had an alpha-helical structure with RMSD less than 10 Å. This observation reflects the fact that the N-terminal extension does not possess a certain spatial structure but probably has a mobile helical packaging. The N-terminal extension model with the best Rosetta score was used to rebuild the new homology model of 16S RNA and the Aim23p complex. The resulting model surprisingly revealed a strong interaction of the N-terminal extension with the C-terminal domain of the Aim23p core part and with the long 3′ terminal helix of 16S RNA. Additional distance restraints between centers of mass from 15 to 40 Å were applied to sample distance between Aim23p’s N-terminal extension and C-terminal domain. As a result, the top models have confirmed the interaction of the N-terminal extension with the C-terminal domain, while interaction with 16S RNA does not look favorable. The best models of packed C-terminal extension showed interaction with the N-terminal domain. Thus, in silico modeling points to the possible mode of Aim23p interaction with 30S subunit where the terminal extensions of the protein “press down” the core Aim23p part to the ribosome. This may be the reason for the importance of the Aim23p terminal extensions for binding bacterial ribosomes (see Fig. 2). The summary of molecular modeling is presented in Fig. 4.