Glucose depletion causes a severe but reversible inhibition of translation initiation by an unidentified mechanism (Ashe et al., 2000
). Formaldehyde cross-linking of cells before polysome analysis has been used to provide mechanistic insight into translational regulation (Nielsen et al., 2004
). Using this methodology, we compared glucose starvation with amino acid starvation, a stress that inhibits translation initiation by activating the eIF2α kinase Gcn2p to give lower TC levels (Hinnebusch, 2005
). Importantly, the abundance of the translation factors tested in this analysis is not appreciably altered by glucose starvation (Fig. S1, available at http://www.jcb.org/cgi/content/full/jcb.200707010/DC1
). Both glucose and amino acid starvation cause an accumulation of 80S monosomes via polysome run-off (). This is characteristic of the inhibition of translation initiation (Ashe et al., 2000
) and is accompanied by a movement of the ribosomal proteins Rps3p and Rpl35p toward the 80S fractions ().
Figure 1. Glucose starvation caused the resedimentation of specific translation factors. (A and B) Sucrose density gradient analysis on extracts from strain yMK36 grown in YPD (A) or SCD (B) and resuspended in YPD (+glucose), YP (−glucose; A), SCD (more ...)
Both eIF3 and eIF2 interact with the 40S ribosomal subunit in a 43S complex before mRNA recruitment (Kapp and Lorsch, 2004
) and are therefore detected in 40S fractions (). Additional protein is detected in polysomal regions and likely represents the formation of initiation complexes on heavily translated polysome-bound mRNAs. After amino acid starvation, a reduced level of eIF2α was observed cosedimenting with the 40S subunit (). This is consistent with the translational consequence of amino acid starvation: reduced TC levels, which would ultimately produce 40S ribosomal subunits with decreased associated eIF2 (Hinnebusch, 2005
). In contrast, eIF3 is still associated with the 40S subunit after amino acid starvation (). eIF3 is also maintained on the 40S ribosome after eIF2 depletion (Nielsen et al., 2004
). Therefore, it seems that a partial 43S complex can still form when eIF2 function has been abrogated.
After glucose starvation, we observed maintained or even slightly increased levels of eIF3 and eIF2 with the 40S ribosomal subunit accompanied by decreases in polysomal fractions (). Therefore, in contrast to amino acid starvation, glucose starvation does not decrease the level of eIF2 with the 40S ribosomal subunit. Any increase in eIF2 and eIF3 with the 40S subunit after glucose starvation is likely to result from the stress-induced polysome run-off. This causes polysomal 43S complexes in unstressed cells to relocate to the 40S region.
The most striking observation from the polysome analysis is that after glucose starvation, there is a marked increase in the levels of eIF4G and Pab1p sedimenting in submonosomal fractions (, fractions 1–3; +/− glucose). Similarly, the proportion of eIF4E in these fractions becomes enriched, although submonosomal eIF4E in glucose replete extracts somewhat masks this effect (). This demonstrates a reduced association of these factors with ribosomes and contrasts noticeably with the sedimentation of these factors after amino acid starvation (). Therefore, the key translational components present in the closed loop mRNP appear to resediment away from ribosomal fractions after glucose starvation. Given that there is a low affinity interaction between eIF4A and eIF4G in yeast (Neff and Sachs, 1999
) and that eIF4A is one of the most abundant translation initiation factors (present at roughly 40 times the number of eIF4G molecules; von der Haar and McCarthy, 2002
), it is unsurprising that the sedimentation pattern of eIF4A remains submonosomal regardless of the stress condition (). Even if a weak eIF4A–eIF4G interaction persisted after stress, the considerable level of free eIF4A would mask changes in ribosome-associated material.
Overall, these data suggest a reorganization of the mRNP closed loop translation complex after glucose starvation, whereby the cosedimentation of eIF4E, eIF4G, and Pab1p with ribosomal complexes is compromised. Amino acid starvation produces a contrasting effect, which is consistent with a fundamentally different mode of action. Indeed, the observed decrease in eIF2 cosedimentation with the 40S ribosomal subunit agrees with the established model for amino acid starvation.
As part of our analysis, we assessed the localization of translation initiation factors in response to both glucose and amino acid starvation. We show that glucose starvation induces genomic GFP fusions of eIF4E, eIF4G1, eIF4G2, and Pab1p to accumulate as cytoplasmic granules (). In contrast, the localization of eIF3b, eIF4AI, eIF2α, or eIF2Bγ is unaffected by glucose starvation. Previously, we have shown that eIF2α and eIF2Bγ localize to a large cytoplasmic focus (Campbell et al., 2005
). Irrespective of the stress condition, we still observe this defined localization.
Figure 2. Glucose starvation induces cytoplasmic granules of eIF4E, eIF4G1, eIF4G2, and Pab1p. (A) Confocal microscopic images of yMK strains 885 (eIF4E-GFP), 1172 (eIF4G1-GFP), 1214 (eIF4G2-GFP), 1185 (Pab1p-GFP), 1170 (eIF4AI-GFP), 883 (eIF2α-GFP), or (more ...)
Strikingly, those factors forming stress-induced granules (eIF4E, eIF4G1/2, and Pab1p) are identical to those with altered gradient sedimentation properties (). The translation initiation factor granules occur with an approximate frequency of four to five granules per cell (see ). In correlation with the cosedimentation analysis, amino acid starvation failed to induce cytoplasmic granules for any of these factors after 30 min ().
Figure 4. EGP-bodies represent a distinct subpopulation of mRNP foci. (A) Epifluorescent microscopic images of glucose-starved yMK strains 1302 (Dcp1p-GFP and eIF4G1-RFP), 1303 (Dcp1p-GFP and eIF4E-RFP), and 1344 (Dcp1-GFP and Pab1p-RFP). Arrowheads identify bodies (more ...)
It is intriguing that glucose starvation causes eIF4E, eIF4G, and Pab1p to relocalize to cytoplasmic bodies, whereas amino acid starvation does not elicit this response or induce the marked resedimentation of eIF4E, eIF4G, and Pab1p. We attribute this to either variation in the severity of the imposed stresses or fundamental differences in their modes of action. Interestingly, an exposure of cells to 10 min of glucose starvation induced a shift in eIF4E, eIF4G, and Pab1p of lower magnitude than that observed for cell extracts from cells starved for 30 min. In accordance with this observation, 10 min of glucose starvation also failed to induce cytoplasmic granules of eIF4E, eIF4G, and Pab1p (Fig. S2, available at http://www.jcb.org/cgi/content/full/jcb.200707010/DC1
). Collectively, these data imply a requirement for the large-scale release of eIF4E, eIF4G, and Pab1p from polysomes to induce visible cytoplasmic aggregations. Short periods of glucose starvation are insufficient to induce such an effect as a result of an apparent lag in factor release. Amino acid starvation induces minimal eIF4E, eIF4G, and Pab1p resedimentation and aggregation into cytoplasmic granules after either 30 or 60 min (, , and S2). These data further corroborate the fundamental difference between these nutritional stresses and their mechanism of inhibition of translation initiation.
An obvious question is whether eIF4E, eIF4G, and Pab1p localize to the same cytoplasmic bodies after glucose starvation. To investigate this, we generated dual-tagged strains bearing specific genomic GFP and RFP fusions. As shown in , we compared eIF4G1-GFP with eIF4E-RFP, eIF4G1-GFP with Pab1p-RFP, and eIF4E-GFP with Pab1p-RFP. For all of these experiments, we observed >90% colocalization ( and see ).
As eIF4E, eIF4G, and Pab1p colocalize to the same cytoplasmic granules after glucose starvation, a key question is whether they relocate to these granules as mRNPs (i.e., bound to mRNA). To test the ability of eIF4E, eIF4G, and Pab1p to interact with each other as well as the cap and poly(A) tail structures after glucose starvation, we used cap and poly(A) affinity chromatography approaches (). The levels of eIF4E and copurifying eIF4G on the cap affinity column remain largely unchanged in extracts from glucose-starved or unstarved cells (, left). Similarly, glucose starvation does not alter the level of Pab1p, eIF4G, or eIF4E associated with the poly(A) column (, right). Therefore, the interaction of these factors with mRNA is likely to persist after glucose starvation. We also used a strategy that enables a specific mRNA to be followed in live cells (Brodsky and Silver, 2000
; Teixeira et al., 2005
). Here, the PGK1
mRNA has multiple U1A-binding sites inserted into the 3′ untranslated region and can be visualized using fluorescence microscopy when coexpressed with a plasmid bearing a U1A-GFP fusion. When this mRNA localization system is used during the translational repression induced by glucose starvation, PGK1
transcripts formed distinct cytoplasmic GFP foci, which precisely colocalize with the eIF4E-RFP granules (, top). Controls using either the PGK1
reporter plasmid or the U1A-GFP fusion plasmid in isolation generate no GFP foci.
Figure 3. mRNA colocalizes with the translation initiation factor granules. (A) Immunoblots on affinity purifications of extracts from glucose-starved (−glucose) or unstarved (+glucose) cultures. 7-methyl-GTP-Sepharose (left) and poly(A)-Sepharose (more ...)
These results suggest that the dramatic effect of glucose starvation on translation initiation does not result from a reduced capacity to form mRNA closed loop complexes. Furthermore, the combination of live cell imaging and affinity chromatography experiments suggests that mRNA accompanies eIF4E, eIF4G, and Pab1p to cytoplasmic granules. Thus, it appears that there is a relocalization of closed loop mRNP complexes to cytoplasmic granules after severe translational repression.
After stresses such as glucose starvation, mRNA can become localized to P-bodies (Sheth and Parker, 2003
). Originally in yeast, translation initiation factors were not identified in P-bodies (Brengues et al., 2005
; Teixeira et al., 2005
). However, while this manuscript was in preparation, this question was reevaluated, and the authors now suggest that eIF4E, eIF4G, and Pab1p do enter P-bodies (Brengues and Parker, 2007
). In our experiments, we also investigated the relationship between P-bodies and the granules containing eIF4E, eIF4G1, and Pab1p. We simultaneously visualized eIF4E-, eIF4G1-, or Pab1p-RFP versus Dcp1p-GFP as a marker for P-bodies in living cells. After a 30-min glucose depletion, eIF4E, eIF4G1, and Pab1p were found in most Dcp1p-containing bodies (). In addition, Dcp1p cosediments with these translation factors across sucrose gradients, and there is an increase in eIF4G, eIF4E, and Pab1p coimmunoprecipitation with Dcp1p in glucose-starved versus unstarved extracts (Fig. S3, available at http://www.jcb.org/cgi/content/full/jcb.200707010/DC1
). Strikingly, however, Dcp1p was not always found in the translation initiation factor granules (). Furthermore, we identified a mean of four to five granules per cell for each of the translation initiation factors, which is approximately twice that observed for Dcp1p (). This value for Dcp1p correlates well with the two to three granules per cell previously published for another core P-body component, Dhh1p (Sheth and Parker, 2003
Overall, these data suggest that after glucose depletion for 30 min, a mean of two to three P-bodies form per cell, which also contain eIF4E, eIF4G1, and Pab1p. Intriguingly, a mean of approximately two additional bodies form per cell that contain eIF4E, eIF4G1, and Pab1p but do not harbor Dcp1p. Similar results were obtained when a second marker for P-bodies, Dcp2p, was used ( and not depicted). This is consistent with observations that Dcp1p colocalization with PGK1 mRNA is incomplete in that PGK1 mRNA granules exist that do not contain Dcp1p (, bottom). Overall, these results suggest that previously unrecognized cytoplasmic bodies exist that contain the translation initiation factors eIF4E, eIF4G, and Pab1p but do not contain the key P-body markers Dcp1p and Dcp2p. We have termed these bodies EGP-bodies, as thus far their only known protein constituents are eIF4E, eIF4G, and Pab1p.
Distinct possibilities exist as to the mode of EGP-body formation, which have important functional implications. First, EGP-bodies may form via a P-body maturation process in which core P-body components such as Dcp1p and Dcp2p are lost. Alternatively, EGP-bodies could arise de novo having never contained core P-body components.
To investigate these possibilities, we undertook detailed time course experiments following individual cells after glucose starvation, simultaneously visualizing eIF4E-RFP as a marker for EGP-bodies and Dcp1p-GFP as a marker for P-bodies (). The complex experimental setup dictates that the earliest possible time point for image acquisition is 10 min after glucose starvation. At this time point, it can be seen that in accordance with previously published data, Dcp1p has accumulated in P-bodies (; Teixeira et al., 2005
). In contrast, eIF4E enters P-bodies after 20–25 min of glucose starvation, and, ultimately, most P-bodies recruit eIF4E (). Perhaps the most striking observation from these time course experiments concerns the EGP-bodies. The EGP-bodies harboring just eIF4E and not Dcp1p always arise spontaneously in the cell and have never been observed to accumulate as a result of the loss of Dcp1p from eIF4E-containing P-bodies (). This defined population of de novo EGP-bodies is quantified in , in which there is a gradual increase in the formation of EGP-bodies starting after 20–25 min. Overall, therefore, we conclude that P-bodies recruit eIF4E after prolonged glucose starvation but do not mature into EGP-bodies by losing core P-body components. In fact, EGP-bodies arise spontaneously as independent entities within the cell.
Figure 5. EGP-bodies form independently of P-bodies. (A) Epifluorescent real-time 2D deconvolved projections generated from continuous z-sweep acquisition of a representative glucose-starved cell of strain yMK1303 (Dcp1p-GFP and eIF4E-RFP). Exponential cells were (more ...)