Our results highlight the importance of quantitative, dynamic, and replicated measurements in understanding the true relationship between changing mRNA and protein. The high mass accuracy mass spectrometry used here in combination with isobaric tags allowed us to make a number of observations: (1) The magnitude of transcript induction is significantly more predictive of changes in protein level than previously thought, at least during the response to NaCl. (2) Transcript reduction did not reduce levels of the corresponding proteins; this result has not been reported before and the breadth of the effect came as a surprise. (3) The temporal relationship between transcript and protein changes is not linear but rather determined by a more complicated function. (4) There is evidence for pervasive PTR, even though the magnitude of that PTR is small. The biological implications of these observations are discussed below.
The correlation between increased transcripts and their encoded proteins during NaCl acclimation is significantly higher than previously measured, due not only to increased measurement precision but also to dynamic considerations that allowed comparison of the appropriate time points. The high protein–mRNA correlation for increased transcripts confirms that the purpose of transcript elevation is to modulate protein abundance. Although we find evidence for pervasive PTR, most of the variance in observed protein increase remains explained by mRNA changes (). Notably, however, nearly 20% of all changing proteins do not correlate with underlying mRNA changes. Thus, while increases in mRNA are a good predictor of protein change under the conditions studied here, the absence of mRNA induction is not necessarily conclusive.
In stark contrast, transcript reduction did not alter protein levels under these conditions. This result has not been widely reported previously, although analysis of available data sets generally shows fewer and smaller changes in reduced proteins compared with those increasing in abundance, across several different environmental comparisons (Blomberg, 1995
; Ideker et al, 2001b
; Griffin et al, 2002
; Li et al, 2003
; Washburn et al, 2003
; Picotti et al, 2009
; Soufi et al, 2009
; Fournier et al, 2010
). However, Fournier et al (2010)
observed a decrease in RPs after rapamycin treatment that was significantly delayed (~6 h) compared with the reduction in RP transcripts. In that case, the late protein decrease may be due to ribosome consumption through ribophagy (Kraft et al, 2008
), rather than translational repression. The lack of RP reduction soon after rapamycin-dependent transcript reduction is therefore consistent with our results, and suggests that the lack of correlation between transcript reduction and protein abundance may be common to other stress conditions. This raises major implications for the interpretation of transcriptomic data, since reduced mRNA abundance is often used to infer protein reduction and dispensability.
Reduced levels of abundant transcripts, particularly those affecting ribosome biogenesis, has been proposed to reduce the costly synthesis of encoded proteins in proportion to growth demand (Waldron and Lacroute, 1975
; Kief and Warner, 1981
; Nomura, 1999
; Jorgensen et al, 2002
; Rudra and Warner, 2004
; Regenberg et al, 2006
; Brauer et al, 2008
). Indeed, our mathematical modeling revealed that changes in growth rate explain the lack of protein change, not just for long-lived RPs but for all proteins from reduced transcripts. However, our results strongly suggest that transcript reduction is not the driving force behind decreased protein synthesis. Reduced translation initiation (inferred from the increased monosome:polysome ratio) occurs much before transcripts drop in abundance (; Uesono and Toh, 2002
). We also observed that final levels of most RP transcripts were at or above unstressed levels, even though division occurred at half the initial rate. This observation underscores the discordance between RP transcript abundance and growth rate during osmotic stress.
Instead, the reduced abundance of these transcripts may serve to redirect translational capacity to newly made mRNAs. The timing of transcript reduction occurs when translation profiles begin to recover as cells resume growth and coincides with maximal levels of increasing transcripts. The temporal relationship between increasing and decreasing transcripts holds across many environmental shocks, even though the kinetics of transcript change is not well correlated with the duration of cell-division arrest during these transitions (Gasch et al, 2000
). We suggest that the coordination of transcript reduction with transcript induction avoids competition for translational machinery, simply through the temporary removal of high-abundance transcripts when translation is resuming.
In this model, cells may be limited for translation factors, particularly initiation factors that regulate translation in response to stress (Proud, 2007
; Gandin et al, 2008
; Park et al, 2011
), but another possibility is competition for translating ribosomes. Nearly 90% of ribosomes in growing cells are actively translating proteins (Warner, 1999
; Arava et al, 2003
; von der Haar, 2008
), leaving little capacity to synthesize new proteins during adversity. Under this scenario, the level of transcript reduction might be tuned to ribosome demand at increased mRNAs. A simple calculation () based on the number of ribosome per cellular transcript (Arava et al, 2003
), transcript counts per cell (Lipson et al, 2009
), and our own measurements of mRNA fold-change suggests that at 30 min after NaCl treatment, the fraction of ribosomes available solely due to transcript reduction (24±5% of translating ribosomes) is approximately the estimated maximum needed to translate the increased transcripts (32±6% translating ribosomes). Although our estimate does not capture known translational regulation at specific transcripts (Uesono and Toh, 2002
; Law et al, 2005
; Melamed et al, 2008
; Halbeisen and Gerber, 2009
; Warringer et al, 2010
), it supports the hypothesis that transient transcript reduction is linked to translational capacity. Our polysome analysis is consistent with this model, since failure to repress transcripts in the dot6
Δ mutant led to continued polysome association of aberrantly abundant mRNAs. These results support a recent theoretical study by Scott et al (2010)
, suggesting that the relationship between continuous growth rate and nutrient quality in Escherichia coli
is due to ribosome allocation.
Figure 7 Estimated fraction of translating ribosomes made available due to transcript reduction. The fraction of 171 000 translating ribosomes before stress that becomes available due solely to transcript reduction (blue bars) was estimated as described in Materials (more ...)
Our model is also consistent with the observed dynamics of gene expression change. Through direct observations and mathematical modeling, we demonstrate that the transient burst in transcript abundance serves to accelerate protein change. It is likely not a coincidence that this response peaks as resuming translational machinery is available to new transcripts. The mRNA response follows a second-order dynamic system, resulting in a pulse of gene expression change before cells acclimate. This pulse is likely due in part to the interplay between transient alterations in transcription and mRNA stability during stress response and acclimation (Garcia-Martinez et al, 2007
; Shalem et al, 2008
; Molin et al, 2009
). In addition, feed-forward signaling loops can also produce pulse-like responses (Mangan and Alon, 2003
; Alon, 2007
; Kaplan et al, 2008
), and such signaling motifs likely contribute here as well (Gasch, 2002a
). The outcome of transient transcript changes is rapid alteration in protein changes, with larger mRNA pulses producing faster protein acclimation. More broadly, the complex relationships between these processes highlight the importance of considering dynamic observations in the study of living systems.