Secretory quality control constrains functional selection-associated protein structure innovation

Abstract

Biophysical models suggest a dominant role of structural over functional constraints in shaping protein evolution. Selection on structural constraints is linked closely to expression levels of proteins, which together with structure-associated activities determine in vivo functions of proteins. Here we show that despite the up to two orders of magnitude differences in levels of C-reactive protein (CRP) in distinct species, the in vivo functions of CRP are paradoxically conserved. Such a pronounced level-function mismatch cannot be explained by activities associated with the conserved native structure, but is coupled to hidden activities associated with the unfolded, activated conformation. This is not the result of selection on structural constraints like foldability and stability, but is achieved by folding determinants-mediated functional selection that keeps a confined carrier structure to pass the stringent eukaryotic quality control on secretion. Further analysis suggests a folding threshold model which may partly explain the mismatch between the vast sequence space and the limited structure space of proteins.

Fig. 1. In vivo functions of CRP are evolutionarily conserved.

Acute liver failure (n = 11 mice/group; n = 5 rats/group) and sepsis (n = 8 mice/group; n = 12 rats/group) were induced in wild-type and CRP KO mice (a) or rats (b) by i.p. injection of acetaminophen or LPS. c Human CRP (huCRP) was administrated into wild-type mice with acute liver failure (n = 11 for vehicle; n = 10 for huCRP treatment) or sepsis (n = 10 mice/group). CRP KO aggravated, whereas huCRP administration alleviated both diseases. These results reveal consistent in vivo functional phenotypes of mouse, rat, and human CRP. Data are presented as mean ± SEM; ***p < 0.001, two-tailed Student’s t test, two-sided.

Fig. 2. The native structure and associated activities of CRP are evolutionarily conserved.

a The pentameric assembly of purified mouse, rat, and human CRP was examined with electron microscopy and single-particle analysis. The diameters of pentameric class averages of the mouse (181 particles), rat (209 particles), and human CRP (208 particles) are 10.98, 10.97, and 10.24 nm, respectively. b The subunit structures of mouse and rat CRP were generated with SWISS-MODEL using subunit A from the crystal structure of human CRP (PDB 1B09) as the template. c The binding of mouse, rat, and human CRP to immobilized PC-KLH were examined with ELISA in the presence (left; n = 3 independent experiments) or absence of calcium (right; n = 6 independent experiments for mouse CRP; n = 3 independent experiments for rat and human CRP). Neither their native structures nor calcium-dependent PC-binding activities exhibited significant difference. Data are presented as mean ± SEM.

Fig. 3. The unfolded structure and associated hidden activities of CRP are evolutionarily distinct.

The stability of mouse, rat, and human CRP was examined with heat- (a) (n = 10 independent experiments for mouse CRP; n = 6 independent experiments for rat CRP; n = 9 independent experiments for human CRP) and urea-induced unfolding (b) (n = 4 independent experiments). The melting temperature (Tm) of mouse CRP was much lower than that of rat and human CRP. Accordingly, calcium exhibited little effect on urea-induced unfolding of mouse CRP, but markedly inhibited that of rat and human CRP. c The refolding of urea-unfolded CRP was examined following dilution (n = 3 independent experiments for mouse CRP; n = 4 independent experiments for rat and human CRP). Rat CRP efficiently refolded, whereas mouse and human CRP did not. d The hidden activities of the major functional motif of unfolded CRP, i.e., CBS, were examined with competitive ELISA against 6 ligands. Relative potency for each ligand was defined as: binding of mCRP in the absence of CBS divided by that in the presence of CBS. Each data point represents a potency index of a different ligand (n = 3 independent experiments). Mouse CBS was most active, while rat CBS was least active. These results indicate that mouse CRP is most prone to unfold and possesses the strongest hidden activities. By contrast, rat CRP is most resistant to unfolding and possesses the weakest hidden activities. e Wild-type mice with acetaminophen-induced liver failure were treated with human CRP at the indicated dosages (n = 14 mice for vehicle; n = 6 mice for 1.25 mg/kg dosage; n = 9 mice for 2.5 mg/kg dosage; n = 5 mice for 5 mg/kg dosage). Human CRP reduced liver injury at dosages of 1.25 and 2.5 mg/kg, but showed little effect at the dosage of 5 mg/kg. The loss of protection of human CRP at the highest dosage might be due to strong proinflammatory responses evoked by excessive mCRP. Data are presented as mean ± SEM; *p < 0.05; **p < 0.01; ***p < 0.001, one-way ANOVA with Tukey post hoc.

Fig. 4. Structural constraints show little effect in shaping CRP evolution.

a Phylogenetic tree of CRP amino acid sequences from 11 species was constructed by the Neighbor-Joining method using MEGA-X. The percentage of replicate trees in which the associated taxa clustered together in the bootstrap test (500 replicates) are shown next to the branches. The tree is drawn to scale, with branch lengths in the same units as those of the evolutionary distances used to infer the phylogenetic tree. CRP of 11 species was expressed in E. coli cells, and their expression levels in supernatants of cell lysates (n = 3 independent experiments) (b), PC-binding activities in supernatants of cell lysates (n = 2 independent experiments for Limulus and Xenopus; n = 3 independent experiments for other species) (c), and culture media (n = 2 independent experiments for Limulus and Xenopus; n = 3 independent experiments for other species) (d) were determined with immunoblotting or ELISA. The obtained values were then normalized to that of human CRP. CRP of 6 species with extracellular PC-binding activities was further examined with their thermal stability in culture media (n = 3 independent experiments) (e) and cell lysates (n = 3 independent experiments) (f) upon heating to 70 °C, and their capacities to refold in culture media after urea treatment (n = 3 independent experiments for Equus and Homo; n = 4 independent experiments for other species) (g). These measured parameters showed little evolutionary correlation among the 11 nodes of CRP. Data are presented as mean ± SEM.

Fig. 5. Folding determinants shape the evolutionary pattern of CRP.

Conservation scores of CRP were calculated with Consurf using 121 available amino acid sequences of different species. a Conservation scores were mapped onto the crystal structure of the human CRP subunit (PDB 1B09). Side chains of residues critical to PC-, C1q-, and FcγR-binding are shown (binding-decisive) (left), and their conservation scores are compared with scores of other CRP residues (right). b Strands of the hydrophobic core that folds spontaneously are indicated (left). The effects of mutations on CRP folding were evaluated (right). c Side chains of residues critical to calcium binding and disulfide bonding (folding-decisive) were shown with mapped conservation scores. Binding-decisive residues were significantly more variable than folding-decisive residues and those not directly involved in binding and folding (others). Mutating residues within the hydrophobic core did not abrogate its spontaneous formation (Step 1; 10 mutants examined, Supplementary Table 1), whereas mutating residues involved in disulfide bonding or calcium binding abrogated the subsequent nonspontaneous folding (Step 2; 8 mutants examined, Supplementary Table 1). Each data point represents the relative folding of a different mutant (n = 3 independent experiments). d The effects of mutations at folding determinants on CRP secretion by prokaryotic E. coli (44 mutants examined; Supplementary Table 2) and eukaryotic COS-7 cells (16 mutants examined; Supplementary Table 2). Though these mutations could not be secreted by COS-7 cells, they were still able to be efficiently secreted by E. coli cells. Each data point represents the relative secretion of a different mutant (n = 3 independent experiments). e Most part of CBS is buried within the native structure of human CRP. The crystal structure of CRP subunit (1B09) is shown in both Ribbon (left) and Surface modes (right). CBS is colored in purple with the side chain of Cys36 also indicated. f The sequence logo of CBS generated with WebLogo 3. Sequences of mouse, rat, and human CBS are also shown. The three pattern-decisive residues are highly conserved, whereas the rest are rather variable allowing fine adjustment of CBS activities. g IDPs (disorder contents of soluble protein or that of the extracellular region of membrane proteins >50%) in human (98), rodents (mouse and rat; 33), other multicellular organisms (drosophila, nematode, and arabidopsis; 22), and yeast (34) were retrieved from DisProt database (Supplementary Table 3). Ratios of secretory IDPs with a signal peptide are much lower even when corrected for the ratio of secretory proteins versus the entire proteome (36%). h Ratios of disordered sequences in extracellular/luminal (non-cytoplasmic; median ratio = 0) versus cytoplasmic portions of all membrane IDPs (median ratio = 1) from all species regardless of their overall disordered contents or cellular localization (Supplementary Table 4; n = 130; left). The sequence lengths of the extracellular/luminal and cytoplasmic portions of these membrane IDPs are also shown (right). Disordered sequences are significantly enriched in cytoplasmic over extracellular/luminal portions of membrane proteins, whereas the overall lengths of the two portions are comparable. In box plots, centerline represents the median; box limits represent upper and lower quartiles; whiskers represent 1.5× interquartile range; points represent outliers. *p < 0.05; ***p < 0.001, Kolmogorov–Smironv tests, two-sided.

Fig. 6. The proposed folding threshold model of protein evolution.

Functional constraints play a major role in shaping protein evolution, yet a folding threshold imposed by the eukaryotic quality control must be satisfied first before functional selection can be effective for a large portion of secretory and luminal proteins. Therefore, mutations that disrupt the original fold without giving rise to an alternative stable fold are equivalent to functional knockout. Only those that meet the threshold will be selected further leading to functional improvement and/or adaptation.