The geometry of worm growth predicts differential age-specific expression of cytoplasmic and membrane proteins. As the worm grows, demand for, and production of, all cell components naturally increases. However, because of the special balloon-like features of C. elegans
, the demand for membrane components should decrease relative to the demand for cytoplasmic components. To test this idea, we examined previously published transcriptomic data that gives the relative expression levels of 17
869 genes in four larval stages (L1–L4) and young adults (Jiang et al. 2001
), and studied the ‘age-response’ of these genes, asking whether their relative expression increases (late genes) or decreases (early genes) with age, discarding those that had an unclear age-response (see §2 for details). We then classified the remaining genes into cellular compartments using Gene Ontology categories so as to identify membrane-related genes (3860) and cytoplasm-related genes (334) with the remainder being discarded (see §2 for details). We then asked, for each of the two GO classes, how many genes increased and decreased with age. We found that the number of cytoplasmic genes that increased with age was 2.37× greater than the number that decreased while for membrane genes the same ratio was 0.62; the difference between these ratios is highly significant (p
-test). Assuming that gene expression reflects protein levels this result suggests that the relative production of cytoplasmic proteins, increases with age while that of membrane proteins falls, which is consistent with a balloon-effect.
Amino acid usage is driven by geometry rather than age per se
. When we compare the early and the late genes we find a striking difference in amino acids that they use (a
). Why is this? We showed that as the worm grows there is a change in the kinds of proteins—cytoplasmic or membrane—that it makes. Furthermore, it has long been known that the proteins of some intracellular compartments are especially rich in certain kinds of amino acids (Cedano et al. 1997
). For example, many membrane-related proteins are rich in hydrophobic amino acids in order to keep them anchored in the lipid bilayer. Perhaps, then, this compartmental shift accounts for the temporal shift in amino acid use. Consistent with this idea, we find that the change in amino acid use between cytoplasmic and membrane proteins explains 77 per cent of the change in use between late and early proteins (b
, one-tailed Pearson correlation, p
, one-tailed Spearman rank). Or, to put it another way, almost four-fifths of the change in amino acid use of worms as they grow can be explained by the different amino acid content of cytoplasmic versus membrane proteins. This same difference also accounts for the shift in the hydrophobicity (Kyte & Doolittle 1982
) of the amino acids used in early versus late proteins seen in a
=0.012, two-tailed Pearson correlation, p
=0.011, two-tailed Spearman rank) since in a multiple correlation analysis in which age-response—log (late/early)—is the dependent, and hydrophobicity and compartment—log (cytoplasmic/membrane)—are independent variables, hydrophobicity is no longer significant (two-tailed partial regression p
values: compartment p
=0.000018, hydrophobicity p
=0.50, interaction p
=0.65; and without interaction, compartment p
=0.000010, hydrophobicity p
Figure 2 The consequences of the changing cellular geometry of the worm for its amino acid economy. (a) Change in amino acid proportions in late versus early proteins. Based on all proteins with values in Jiang et al. (2001). (b) Association between the change (more ...)
Although the correlation between the shift in amino acid usages seen in the compartment and age-responses is striking, it does not follow that the former actually drives the latter. Amino acid usage might vary across ages for some altogether different, but hidden reason. An alternative explanation, for example, is that selection for metabolic efficiency varies with age and has shaped the age-specific composition of proteins. Metabolic efficiency is a particularly strong candidate hypothesis as it is known generally to shape protein composition (Akashi & Gojobori 2002
; Swire 2007
). But it does not explain the age-response in this dataset (for proof, see electronic supplementary material information IV). However, although we can reject metabolic efficiency, it is not possible more generally—given that our data are not experimental but entirely correlative—to exclude definitively the agency of some hidden variable or to establish the direction of causality. Nevertheless, we think that cellular geometry is the driver. In part, this is because, as mentioned above, membrane proteins have the amino acid composition that they do for good functional reasons—to keep them anchored in the lipid bilayer—and so there is simply no need to invoke other unknown age-dependent processes to explain their composition. But it is also because a closer look at our data provides further evidence in support of our explanation.
First, while there is a very large shift in amino acid usage between cytoplasmic and membrane proteins, the shift between early and late proteins is much smaller (the former exceeding the latter by 4.4×, b
). Looking at this another way, when we compare cytoplasmic with membrane proteins, we find that the relative frequency of the typical amino acid differs by 30 per cent; but when we compare late and early proteins we find that this difference is 7 per cent. Second, we can control directly for the effects of age-response and compartment. We control for the effects of age-response by calculating the shifts in amino acid usage between cytoplasmic and membrane proteins within each of the early or late subsets (c
). These shifts are highly correlated, which implies that the difference in amino acid usage between cytoplasmic and membrane proteins is largely independent of age-response (ragecontrolled2
Spearman). Analogously, we control for the effects of compartment by calculating the shifts in amino acid usage between early and late proteins within each of the cytoplasmic or membrane subsets (d
). These shifts, by contrast, are only very weakly correlated, which implies that the difference in amino acid usage between these late and early proteins largely depends on compartment (rcompartmentcontrolled2
=0.021). In order to examine the robustness of these results, we resampled our dataset 100
000 times and found that ragecontrolled2
was constant and high while rcompartmentcontrolled2
was variable and low (see figure S1 in the electronic supplementary material). Given these results, we conclude that the changing geometry of the growing worm, rather than its increasing age, predicts the amino acid composition of the proteins that it is manufacturing at any time. Indeed, there is little evidence for an age effect independent of the changing geometry.
The geometry of growth predicts amino acid pool sizes. If, during growth, the amino acid composition of the worm's proteins changes, then so too must the demand for particular amino acids. Is this change in demand reflected in the pool sizes of amino acids? To answer this question, we measured the relative pool sizes of 14 standard amino acids using 1H NMR spectroscopy. (The remaining six were not estimated, see §2.) We sampled worms at seven ages: each of the four larval stages, young adulthood (66 hours after bleaching eggs), middle age (83 hours) and old age (159 hours). All 14 amino acids showed strong changes with age, usually in a consistent direction (). Once again, we used four different regression models—linear, log, power and exponential—and for each amino acid estimated the slope of relative concentration against time. We only fitted the first five ages because we wanted to compare our results directly to Jiang et al.'s transcriptomic data.
Figure 3 Dynamics of free pool size in 14 standard amino acids during the course of postembryonic growth. Relative pool sizes measured using 1H NMR spectrophotometry and semi-automated fitting of individual metabolites. ((a) Ile; (b) Val; (c) Leu; (d) Phe; (e (more ...)
We have already shown how amino acid usage changes during growth (b). In fact, the y-axis values of this plot are estimates, albeit imperfect ones, of changes in amino acid demand during growth. Similarly, the slopes describing how the relative pool sizes of amino acids change during growth are, potentially, estimates of changes in their supply. a shows that the change in supply and demand for the 14 amino acids is positively correlated (linear: r2=0.39; p=0.0087, p=0.0061; see the electronic supplementary material for other models). This suggests that amino acid pool sizes are dynamically regulated in response to the changing demands of protein synthesis. b shows that this regulation seems to be driven by the increasing proportion of cytoplasmic—and decreasing proportion of membrane proteins—produced as the worm grows since the change in supply of the amino acids is also positively correlated with log (cytoplasmic/membrane) as in the x-axis of b (linear: r2=0.41; p=0.0067, p=0.00092; see the electronic supplementary material for other models). In other words, just as the changing cellular geometry of the growing worm predicts the amino acid composition of the proteins that it is manufacturing at any time, it also predicts the size of its amino acid pools.
Association between changes in free amino acid concentration with the change in amino acid proportions: (a) in late versus early proteins; (b) in cytoplasmic versus membrane proteins. For further details of models and statistics, see text.