Conformational exchange divergence along the evolutionary pathway of eosinophil-associated ribonucleases

SUMMARY

The evolutionary role of conformational exchange in the emergence and preservation of function within structural homologs remains elusive. While protein engineering has revealed the importance of flexibility in function, productive modulation of atomic-scale dynamics has only been achieved on a finite number of distinct folds. Allosteric control of unique members within dynamically diverse structural families requires a better appreciation of exchange phenomena. Here, we examined the functional and structural role of conformational exchange within eosinophil-associated ribonucleases. Biological and catalytic activity of various EARs was performed in parallel to mapping their conformational behavior on multiple timescales using NMR and computational analyses. Despite functional conservation and conformational seclusion to a specific domain, we show that EARs can display similar or distinct motional profiles, implying divergence rather than conservation of flexibility. Comparing progressively more distant enzymes should unravel how this subfamily has evolved new functions and/or altered their behavior at the molecular level.

eTOC Blurb

Although proteins rely on atomic-scale motions to promote biological and catalytic function, the evolutionary conservation of dynamic events between structural homologs remains poorly understood. In the current work, Bernard et al. reveal important functional and evolutionary divergent features that govern conformational exchange within structurally homologous eosinophil ribonucleases.

Graphical Abstract

INTRODUCTION

Evidence supporting the importance of conformational flexibility in protein function has led to the evolution of the structure-function paradigm to encompass the ensemble representation of protein structures ^1–3. Conformational dynamics occurs over a wide range of timescales, ranging from fast side-chain motions to slower loop and domain motions ^4,5. These atomic-scale movements facilitate the sampling of conformational substates, which correspond to energy minima in the conformational landscape of proteins ⁶. Indeed, conformational dynamics occurring on the timescale of the catalytic turnover have been shown to be critical for optimal biological function in a variety of enzyme systems, such as E. coli dihydrofolate reductase ⁷, cyclophilin A ⁸, adenylate kinase ⁹, HIV-1 protease ¹⁰, DhlA haloalkane dehydrogenase ¹¹, and bovine RNase A (BtRA) ^12,13. Further, mutational perturbations of the observed conformational exchange within functionally important regions were shown to significantly affect catalytic turnover in many of these systems ^7,8,12,14. However, understanding substate-transitioning movements remains limited to discrete and often unrelated individual protein systems in the sequence and evolutionary space.

Since they form a continuum, decoupling the individual contributions of structure and dynamics represents one of the major challenges to identify the role of conformational motions in protein function. One way to address this is to comparatively characterize the conformational landscape between structural and functional enzyme homologs. This provides a better understanding of how conservation of conformational dynamics contributes to the observed diversity in biochemical and biological function within an enzyme subfamily. Since pancreatic-type ribonucleases (RNases) share the same structural fold and catalyze the endonucleolytic cleavage of RNA ^15,16, they represent an ideal model system to characterize the role of conformational motions in enzyme function (Figure 1). In addition to preserving the core catalytic function of transphosphorylation ^15,17, canonical RNase members have been subjected to high rates of evolutionary sequence substitutions ¹⁸. This allowed various family members to develop a remarkable diversity of unrelated biological functions that rely on different molecular mechanisms of action ^18–24. Distinct RNase subfamily members are thus involved in host defense ²⁵, angiogenesis ^20,26, neuroprotection ²⁷, vascular homeostasis ²⁸, and digestion ¹⁶. Interestingly, while most family members use the same active-site mechanism to cleave RNA substrates with varying levels of affinity and efficiency, their novel biological roles do not always require preservation of RNA cleavage potential ²⁹. Indeed, while ribonucleolytic function is absolutely required for some members of the family to carry out these other biological roles, the core RNA catalytic function also appears purely accessory in other members ^16,30. Recent studies have also illustrated the conservation of amino acid networks whose conformational exchange on the millisecond time frame correlates with the rate of enzyme catalysis ³¹. Further, conservation of global dynamic properties was observed for RNase members grouped into phylogenetic subfamilies, linking unique conformational exchange events to biological function ³².

Figure 1. Overall fold of the pancreatic-type ribonuclease.

Cartoon representation of human eosinophil-derived neurotoxin (EDN), also known as Homo sapiens RNase 2 (HsR2, PDB entry 1GQV). The characteristic V-shaped kidney structure shows two opposite domains termed V₁ and V₂, defined by two opposing antiparallel β-sheets formed by strands β2-β3-β6-β7 for V₁ and strands β1-β4-β5 for V₂ ¹⁷. Secondary structure elements are annotated, and conserved disulfide bridges are shown in yellow. Catalytic residues His15, Lys38 and His129 are labelled in blue.

Although phylogenetic and conformational clustering provides a measure of their distinct biological roles ³², catalytic and biological functions remain significantly diverse within subfamilies. For instance, human RNase 2 (also known as eosinophil-derived neurotoxin, or EDN), displays potent antiviral ^33,34 and chemotactic activities ³⁵, while human RNase 3 (also known as eosinophil cationic protein, or ECP), exhibits cytotoxic ³⁶, helminthotoxic ³⁷, and antibacterial activities ^29,38,39. Both RNases 2 and 3 arose as a result of a gene duplication event that occurred when Old and New World monkeys evolutionarily diverged from each other ⁴⁰. RNase 2 exhibits distinct millisecond conformational exchange and was shown to be nearly 90 times more catalytically active than RNase 3 ⁴¹.

Despite our current understanding linking the ribonucleolytic function with conformational exchange in distant members of the broad pancreatic-type RNase superfamily ^{17,32,42–44}, it remains unclear how phylogenetically close subfamily members have evolved conformational exchange to preserve similar biological function(s). Using closely related members of this subfamily as model systems, we performed systematic characterization of the conformational flexibility, catalytic activity, and biological function of various eosinophil-associated RNases (EARs). In addition to human RNases 2 and 3 (HsR2 and HsR3), we broadened the evolutionary space to include representative orthologs from primates, i.e. RNase 3 from the Bornean orangutan (Pongo pygmaeus, PpR3), Sumatran orangutan (Pongo abelii, PaR3), crab-eating macaque (Macaca fascicularis, MfR3), as well as RNase 2 from the three-striped night monkey (Aotus trivirgatus, AtR2). We integrated NMR experiments (¹⁵N-CPMG, ¹⁵N-CEST) and molecular dynamics (MD) simulations to quantify the conformational exchange experienced by EARs spanning the nanosecond (ns) to second (s) timescales. We measured kinetic parameters, antibacterial activity against Gram-positive and Gram-negative bacteria, and performed cytotoxicity assays against HeLa cells to characterize and compare the catalytic and biological functions of these enzymes. Our results illustrate greater diversity in the conformational landscape than the diversity in catalytic and biological activities for RNases within this phylogenetic subfamily.

Despite the obvious limitations of comparing flexibility profiles between ligand-free homologs, our results nevertheless demonstrate that conformational exchange is largely concentrated in the V₁ domain of evolutionarily related EARs. This contrasts with the conformational exchange profiles observed between structurally similar yet phylogenetically (and functionally) more distant members of the pancreatic-type RNase superfamily ³², further supporting the idea that V₁ domain flexibility is functionally preserved in EARs. Our observations also evoke rapid divergence of conformational exchange within the eosinophil-associated RNase subfamily, which may fine-tune the biochemical and biological functions of these enzymes along their evolutionary trajectory.

RESULTS

Phylogenetic analysis

To characterize the evolutionary relationships between the homologous sequences within the RNase superfamily in the context of functional and dynamic comparison, we performed the phylogenetic clustering of RNase sequences from different organisms. Results show classification of the canonical RNases into distinct subfamilies, as shown in Figure 2 and corroborated by earlier studies ^{21,22,32,45,46}. This observation supports the fact that several RNase subfamily clusters diverged before species differentiation, although individual members may still be evolving for specific target functions within species. RNases 2 and 3 diverged only in Old World monkeys, and thus are relatively new genes ²⁴. The owl monkey RNase (AtR2) belongs neither to RNases 2 nor to RNases 3; it is rather an undifferentiated eosinophil RNase. RNase 3 sequences from primates display high sequence similarity, yet the elevated rate at which eosinophil RNases have evolved ²³ has made them dissimilar enough to serve as effective models to probe the effect of dynamical variation on their function (Table 1).

Figure 2. Phylogenetic classification of pancreatic-type RNases.

Multiple sequence alignment of selected members of the RNase superfamily depicted in Figure S1 was used for phylogenetic clustering. Branching of the eight human RNases is identified using distinct colors. Inset shows the branch corresponding to the EAR subfamily. The six RNases characterized in this work (HsR2, HsR3, AtR2, PpR3, PaR3, and MfR3) are indicated using an asterisk in the inset (*).

Table 1. Sequence identity and similarity between RNases investigated in this study.

Sequence identity and similarity scores are given before and after the slash, respectively. Scores were calculated using NCBI BLAST.

	BtRA	HsR2	AtR2	HsR3	MfR3	PpR3	PaR3
BtRA		35%/46%	34%/46%	28%/45%	27%/41%	28%/44%	30%/42%
HsR2	35%/46%		69%/76%	65%/75%	64%/73%	70%/78%	76%/82%
AtR2	34%/46%	69%/76%		68%/75%	62%/72%	69%/78%	68%/75%
HsR3	28%/45%	65%/75%	68%/75%		86%/92%	85%/91%	74%/85%
MfR3	27%/41%	64%/73%	62%/72%	86%/92%		81%/90%	68%/81%
PpR3	28%/44%	70%/78%	69%/78%	85%/91%	81%/90%		84%/89%
PaR3	30%/42%	76%/82%	68%/75%	74%/85%	68%/81%	84%/89%

Biochemical and biological activities of EARs

The premise for the comparison of simian eosinophil RNases is that their biological activities must be similar amongst each other and to that of their human counterpart. To test this hypothesis, we characterized the catalytic and biological (antibacterial and cytotoxic) activities of the selected RNases. We quantified the endonucleolytic cleavage reaction of EARs using uridylyl-(3′−5′)adenosine (UpA) as substrate, one of the smallest and most characterized consensus dinucleotide sequences (Table 2). This substrate preserves the scissile phosphate subsite (P1) flanked by a pyrimidine base in B1 and an adenine base essential for proper binding to the B2 subsite in RNases ¹⁵. For comparison, kinetic parameters of HsR2–3 and the prototypical bovine RNase A (BtRA) were also quantified for reference. While these enzymes can accommodate larger substrates with better affinity due to additional molecular interactions ¹⁷, the choice of the UpA dinucleotide substrate facilitates comparison between the different RNases and limits bond cleavage to a single catalytic event per substrate molecule. Results show that all RNases cleave UpA, as expected from canonical members of this superfamily ¹⁶. All RNase 3 orthologs exhibit a k_cat of ~1 s⁻¹ and a K_M of ~1–5 mM, resulting in catalytic efficiencies (k_cat/K_M) three to four orders of magnitude weaker than that of the prototypical bovine RNase A (BtRA) (Table 2). HsR2 has the highest k_cat (~88 s⁻¹) among EARs and a catalytic turnover only 20 times lower than BtRA. As previously reported ²⁴, AtR2 exhibits lower catalytic efficiency than all other EARs.

Table 2. Comparison of the kinetic and biological properties of RNases.

The Michaelis constant (_KM) and turnover number (k_cat) were obtained using the Michaelis-Menten equation, as described in experimental procedures. The isoelectric point (pI) was estimated using ProtParam on the ExPASy server.

Enzyme	pI	kcat (s−1)	kcat relative to RNase A	KM (μM)	KM relative to RNase A	kcat/KM (M−1s−1)	kcat/KM relative to RNase A	Antibacterial activity	Cytotoxicity
BtRA	8.64	471±26	1	130±35	1.00	3.62*106	1	-	-
HsR2	9.20	87.83±3.90	0.186	520±54	4.00	1.69*105	0.047	+/−	+
HsR3	10.47	0.97±0.06	0.0021	1230±220	9.46	789	2.18*10−4	++	++
MfR3	10.60	1.000±0.142	0.0021	2650±69	20.38	377	1.04*10−4	++	+
PpR3	10.01	1.28±0.12	0.0027	1380±310	10.62	928	2.56*10−4	+	+
PaR3	9.53	Not saturated		Not saturated		416	1.15*10−4	+	+
AtR2	9.20	0.52±0.10	0.0011	4960±1450	38.15	105	2.90*10−5	-	+

While catalytic activity is common to all canonical RNases, RNase 3 orthologs are typically defined by their antibacterial activity, which was shown to be lacking even in RNase 2 ⁴⁷. The antimicrobial activity of EARs, which was particularly well characterized for HsR3, is known to proceed via a mechanism that involves protein anchoring and aggregation at the cell membrane, with subsequent membrane solubilization through a carpet-like mechanism ^19,39. In bacteria, amyloid-like protein aggregation was shown to induce cell agglutination, which is essential for the antimicrobial activity of ECP at the cell surface ⁴⁸. Interactions with outer membrane lipopolysaccharides (LPS) were also shown to play a role in cell agglutination in Gram-negative bacteria ^19,49. Interestingly, this molecular mechanism is independent from the ribonucleolytic activity of HsR3 ²⁹. We investigated this biological role of eosinophil RNases against both Gram-negative Escherichia coli and Gram-positive Staphylococcus aureus (Figure 3). Our results show that all Old World monkey eosinophil RNase 3 orthologs sharing an elevated pI (HsR3, MfR3, PpR3, and PaR3) cause a dose-dependent reduction in the survival rate of both Gram− and Gram+ bacteria, with a stronger effect against E. coli than S. aureus. HsR3 and MfR3 display the most potent toxicity, with a survival rate lower than 1.1% of the negative control in presence of 5 μM RNase. Other RNase 3 orthologs are less potent, but still efficient, with 28.9% (42.1%) and 40.8% (68.4%) of E. coli (S. aureus) survival at 5 μM PpR3 and PaR3, respectively. Surprisingly, HsR2 displays antibacterial activity against E. coli, leading to a survival rate of 45.2% at 5 μM concentration. We note that the maximal concentration used was greater than the one reported by Rosenberg and Dyer ²⁹, and no antibacterial effect was observed at 2 μM, consistent with their observations. We did however observe a significant increase in the bacterial survival rate (~200% of the negative control) for E. coli incubated with 0.5 μM HsR2. No antibacterial activity was detected for HsR2 against S. aureus. Finally, AtR2 did not significantly affect the survival rate of either bacterial strain, consistent with previous observations ²⁴.

Figure 3. Antibacterial activity of simian eosinophil RNases.

Normalized counts of colony-forming units (CFUs) of exponentially growing gram-negative E. coli or gram-positive S. aureus incubated for three hours in presence of various amounts of eosinophil RNases. Bovine RNase A (BtRA) was used as a negative control, and human RNase 3 (HsR3) was used as a positive control. Each condition was monitored at least in triplicate. Individual data points are shown as blue scatter plots on each bar graph. Significance was assessed using OneWay ANOVA with * for p < 0.05, ** for p < 0.01, *** for p < 0.001, and **** for p < 0.0001.

To characterize the cytotoxic effects of RNases, we performed cytotoxicity assays against HeLa cells using increasing enzyme concentrations (Figure 4). To the best of our knowledge, this is the first study reporting the direct cytotoxic effects of these simian RNases against eukaryotic cells. All eosinophil RNases tested decreased HeLa cell survival in a concentration-dependent manner. As expected, RNase A did not alter the ability of HeLa cells to metabolize the Cell Titer Blue assay reagent, while HsR3 significantly reduced cell viability at 50 μM. These results are in accordance with previous observations demonstrating the ability of HsR3 to induce a significant reduction in HeLa cell viability at 20 μM and 40 μM ³⁶. Our results show significant cytotoxicity induced by HsR3, even at 10 μM concentration (Figure 4). Interestingly, while AtR2 did not exhibit antibacterial activity, it is cytotoxic to an extent comparable to that of other eosinophil RNases. The cytotoxic potency observed herein correlates with respective protein pI values, with the highest cytotoxic effects for RNases exhibiting pI values above 10 (HsR3, MfR3, and PpR3), an intermediate response within 9 < pI < 10 (HsR2, PaR3, and AtR2), and no ability to kill HeLa cells (even at 50 μM) for RNase A (pI < 9). These observations are further supported by recent studies evaluating the cytotoxicity of various mouse eosinophil cationic RNases (mEARs), which show that high pI enzymes such as mEAR2 and mEAR5 are the most cytotoxic among all tested mEARs ⁵⁰.

Figure 4. Cytotoxicity of simian eosinophil RNases.

A) Viability of HeLa cells as a function of increasing concentrations of simian RNases assessed with the Cell Titer-Blue Cell Viability Assay from Promega. Bovine RNase A (BtRA, black) was used as a noncytotoxic negative control. Results are expressed as percentage of control (non-treated cells) and represent the mean±SEM of at least 3 independent experiments performed in triplicate. B) Close-up view of the dashed rectangle found in panel A.

Conformational exchange behavior of eosinophil RNases

Conformational exchange of the distant loop 1 on the rate-limiting product release step was previously demonstrated to modulate catalytic function in BtRA ^44,51. While such a direct link between conformational exchange and catalytic turnover has not been confirmed in other RNases, the presence of concerted motions on functionally relevant timescales was also reported among several human RNases, including HsR2 and HsR3 ^{17,32,43,52–54}. Here, we characterized the dynamic profiles of inter-species RNases orthologs over motional and functional timescales ranging from nanoseconds (ns) to seconds (s). We performed ¹⁵N-CPMG and ¹⁵N-CEST NMR experiments to characterize conformational exchange on the relevant catalytic timescale (k_cat) (Table 2). CEST experiments, which measure exchange for dynamic events ranging between ~2.5 ms and ~50 ms (k_ex ~20 to ~400 s⁻¹) ^55,56, are used to complement CPMG relaxation dispersion experiments, which measure exchange occurring over the ~300 μs to ~10 ms time frame (k_ex ~100 to ~3000 s⁻¹) ⁵⁷. MD simulations were used to probe the ns-μs timescale motions. We also used Markov State Model (MSM) analysis ⁵⁸ to obtain equilibrium-weighted estimates of the timescale of conformational exchange processes in the simulation ensembles.

Figure 5 shows the three-dimensional distribution of conformational motions on the ns-s timescales with residues experiencing conformational exchange depicted as spheres. Figure 5A illustrates the difference between the two types of analyses performed on the various datasets, i.e., considering the exchanging residue location alone or combining that information with the exchange rate or timescale. A summary of residues experiencing conformational exchange is presented in Tables S3–S5. In contrast to other members of the vertebrate RNase superfamily ³², residues undergoing conformational exchange on the CPMG timescale are located primarily in the V₁ domain in all eosinophil RNases. These include residues of loop 4 and β-strands 2, 3, 6 and 7 (Figure 5B). Additional exchanging residues are also observed in α-helix 3 of all eosinophil RNases except HsR3, in loop 5 of PpR3, PaR3 and AtR2, and in loop 7 of AtR2. A comparison of the exchange rates (Figure 5B, represented as color gradient) showed similar rates for HsR3, MfR3, PpR3, and AtR2, with a majority of exchanging residues exhibiting k_ex values ranging between 200 and 500 s⁻¹. A few isolated residues experienced faster exchange rates (yellow spheres) in MfR3 and PpR3. Exchanging residues in HsR2 exhibit faster exchange rates, ranging between 1000–1400 s⁻¹, consistent with the higher k_cat of this enzyme relative to other eosinophil RNases. PaR3 exchanges on a broader range (500–1900 s⁻¹), with no apparent clustering of residues sharing similar k_ex. Most of the residues identified as undergoing conformational exchange using CEST (Figure 5C) were also observed as undergoing conformational exchange on the slower end of the CPMG timescale (Figure 5B), consistent with their overlapping time range. This network of interconnected residues are clustered in the V₁ domain of AtR2, HsR3, MfR3, and PpR3. The lack of residues experiencing conformational exchange on the CEST timescale in HsR2 and PaR3 is consistent with the faster k_ex values determined from CPMG relaxation dispersion experiments.

Figure 5. Conformational exchange experienced by simian eosinophil RNases.

Subpanel A depicts the two different analysis methods used to characterize conformational exchange within each RNase, using HsR3 ¹⁵N-CPMG data as an example. The left panel depicts exchanging residues as grey spheres to highlight their location, whereas the right panel includes the exchange rate. Residues undergoing conformational exchange are represented as spheres and were identified using dual fits of ¹⁵N-CPMG relaxation dispersion NMR data (B - first column), ¹⁵N-CEST experiments (C - second column), and Markov State Model analysis implied timescales (D - third column). Catalytic residues are depicted as sticks. In B and C, exchanging residues are colored according to the exchange rate k_ex, whereas residues are colored according to the implied timescale of the exchange process in D. The magnitude of k_ex values and implied timescales are represented using the color gradient legends on the bottom. Residues whose fits give a k_ex with a standard deviation larger than the actual value are depicted as black spheres. Black loops correspond to unassigned residues due to line broadening in the NMR spectra. Only residues with an implied timescale slower than 50 ns (see STAR Methods) are depicted as spheres in the third column. Representative curves for each technique are displayed in Figures S5 and S6.

We also characterized the ns-μs timescale motions by determining MSM timescales or decorrelation times of backbone and side-chain torsion dynamics from the MD ensembles for the different eosinophil RNases, in an effort to tease out comparably slow, metastable dynamics rather than fast, thermal fluctuations. Figure 5D shows residues experiencing conformational exchange on the ns-μs timescale as spheres on the three-dimensional structure of the six eosinophil RNases. Exchanging residues vary in number and location across the different RNases. HsR3 and HsR2 displayed the largest (65) and the smallest (26) number of exchanging residues, respectively. Residues of loop 4 experience conformational exchange in all RNases except MfR3, which showed no exchange in this region on the faster timescale. The maximum likelihood estimates of loop 4 motion timescales were observed to span from 60–100 ns for PaR3 (fastest) to 100–750 ns for PpR3 (slowest). Conformational exchange was also observed to varying extents for residues of loop 6 across the different RNases, most noticeably in MfR3, with many residues exchanging on a timescale neighboring ~850 ns.

We further determined the structural changes associated with exchange observed from the MSM analysis. Briefly, MSM implied timescales describe relaxation times of molecular exchange processes. As implied timescales are derived from the eigenvalues of the MSM transition matrix, the corresponding eigenvectors describe the structural process that gives rise to a specific implied timescale ⁵⁹. Therefore, exchange processes can be uniquely identified in terms of their structures ⁶⁰, and be depicted by drawing representative frames (e.g., of minimum and maximum of the eigenvector) from the MD simulation. Figure 6 shows a noteworthy process for His82, a residue in the V₂ β-sheet known to be essential for substrate binding in the B₁ subsite of HsR2 and HsR3. Our results show that rotamer dynamics is conserved for all eosinophil RNases, as this applies to the structurally equivalent Tyr82 in all other eosinophil RNases (AtR2, PaR3, PpR3, and MfR3). Additionally, we identified an exchange phenomenon on the 0.1-to-1 μs range in a cryptic pocket between α1, α2 and β1 at conserved residue Phe43. However, this event was only observed for AtR2 and HsR3, pointing towards a pattern that was not evolutionarily conserved. We note that due to limited sampling, the MSM analysis is restricted to characterizing single residue dynamics on the ns-μs timescale in only one of potentially many large-scale metastable protein conformations. While our sampling is too sparse to predict slower timescale motions or provide high confidence timescales (Figure S7), our analyses indicate convergence of the per-residue MSM implied timescales for all RNases.

Figure 6. Examples of residue dynamics experienced on the 100-ns timescale.

A) Overall structure of HsR3 showing locations of amino acids analyzed in subsequent panels. B-C) Representative structures of MSM metastable states illustrating between which structures the identified exchange process occurs (left) at positions 82 (B) and 43 (C). In both panels, MSM implied timescales (right) indicate the relaxation time corresponding to that process, resolved for the different eosinophil RNases. Confidence intervals (shaded areas) are computed from Bayesian sampling of the posterior ⁸⁰. In panel B, two rotamers of His82 (in HsR2, HsR3 [shown structure], PpR3, and MfR3) or Tyr82 (in AtR2 and PaR3) (located on β4) are configurations between which the MSM exchange process occurs. It can be observed in all investigated RNases. In panel C, Phe43 (on β1, embedded between α1 and α2) is conserved in all RNases but shows conformational exchange on the 100-ns regime only for AtR2 and HsR3. Structure renders were created with VMD ⁸¹.

Quantitative characterization of conformational exchange properties of eosinophil RNases

A comparison of the location of residues undergoing conformational exchange revealed qualitative similarities in the dynamical patterns of eosinophil RNases on the catalytically relevant timescales (Figure 5). To quantitatively characterize this similarity in the conformational exchange profiles, we calculated the pairwise cosine similarity between all RNases (Figure 7). The conformational exchange profile for bovine RNase A (BtRA), determined previously from CPMG experiments ¹³, was used for comparison with eosinophil RNases. Characterization of the pairwise cosine similarity between the different RNases showed higher similarity between HsR3, MfR3, PpR3, and AtR2, with the cos(θ) ranging from 0.583 to 0.730 on the CPMG timescale (Figure 7A, left panel). PaR3 displayed a cos(θ) of ~0.45 with other eosinophil RNases, even though most of its exchanging residues are located in the V₁ β-sheet like other eosinophil RNases. HsR2 and BtRA displayed the lowest cosine similarity scores with other RNases, highlighting the distinctly different exchange profiles of these enzymes relative to other RNases.

Figure 7. Pairwise comparison of eosinophil RNase exchange profiles.

Pairwise cosine similarity values for location (left panels) and exchange timescale (right panels) calculated based on comparison of residues along the consensus sequence, color-coded based on strong (white) to weak (dark) similarity. Pairwise cosine similarity values are shown for A) ¹⁵N-CPMG, B) ¹⁵N-CEST, and C) MSM implied timescales.

A comparison of the cosine similarities based on the magnitude of the exchange rates (k_ex) showed smaller values, approaching 0, highlighting a lack of similarity of the exchange rates between RNases (Figure 7A, right panel). Among RNases that displayed exchange on the CEST timescale (Figure 7B, left panel), highest similarities were observed between HsR3, MfR3, and PpR3 (between ~0.55 and ~0.7). HsR2 showed lowest correlations with other RNases. No correlations were observed involving PaR3, which showed no exchange on the CEST timescale. A comparison of the exchange timescales (Figure 7B, right panel) showed a similar trend as that observed for residue locations, albeit with small variations in the similarity scores. The cosine similarity values are smaller for most pairwise comparisons, with similarities > 0.5 observed between only three enzyme pairs (AtR2 with HsR3, HsR3 with MfR3, and MfR3 with PpR3), suggesting a diversity in the conformational exchange profiles on this timescale. On the ns-μs timescale probed by MSM analyses, similarities between the location of the metastable dynamics (Figure 7C, left panel) vary between ~0.25 and ~0.67. A comparison of the magnitude of the timescales lowers the similarity values down to at most 0.578, going as low as 0.086 for the MfR3-PpR3 pair (Figure 7C, right panel). This once again corresponds to a lack of similarity for the exchange on this timescale. Overall, more enzymes tend towards dissimilarity with their homologs than the reverse, and this holds true at all timescales.

Characterizing the amplitude of atomic fluctuations on the ns-μs timescale

We characterized the amplitude of conformational fluctuations on the faster timescale by computing the root mean square fluctuations for the top ten modes (RMSF₁₀) from molecular dynamics simulation trajectories as a function of the consensus sequence. A comparison of the RMSF10 of Cα atoms for the six eosinophil RNases (Figure 8A) showed qualitative similarity in the dynamical patterns for all members of this subfamily on the faster timescale, consistent with previous observations ³². Large fluctuations were observed in loop 4 (consensus positions 57–70), loop 6 (consensus positions 88–96) and loop 7 (consensus positions 114–124), in addition to the N- and C-terminal regions. Interestingly, AtR2, the undifferentiated member of this subfamily, showed additional fluctuations in the loop 1 region (consensus positions 17–22). To quantitatively characterize the similarity between the eosinophil RNase sequences, we calculated the pairwise Pearson’s correlation coefficients between the different members (Figure 8B). Our results showed average correlations between eosinophil RNases of 0.7±0.1, similar to previous observations ³². Large correlations were observed between most eosinophil RNases, with the exception of MfR3, which showed the smallest correlations with other RNases.

Figure 8. Amplitude of atomistic fluctuations on the ns-μs timescale.

A) Root mean square fluctuations as a function of the consensus sequence for members of the eosinophil associated RNases. The calculated RMSFs represent the Cα displacements of the top ten quasi-harmonic modes of eosinophil RNases. The consensus residue numbering represents the indexing that includes gaps, corresponding to insertions/deletions, in sequences. B) Pairwise Pearson’s correlation coefficients for the eosinophil RNases calculated based on comparison of RMSF₁₀ values for each of the positions without an insertion/deletion in any of the six RNases.

Intrinsically disordered regions within eosinophil RNases

Although EARs adopt the well-defined structural fold of pancreatic-type RNases, a large class of other biologically active proteins are known to act in a completely disordered fashion or contain intrinsically disordered regions (IDRs) that play important biological roles in the cell ⁶¹. Due to the structural requirements of catalysis, enzymes have long been considered an exception to the IDR rule. Even though the active sites are highly ordered, enzymes also retain functionally important regions (especially the surface loops), which sample multiple conformations resembling the IDRs and intrinsically disordered proteins (IDPs). The function of IDRs and IDPs relies heavily on sampling of the conformational landscape, including the functionally relevant conformations, which is also the case for surface loops in proteins such as EARs. Therefore, conformational flexibility and intrinsic disorder represents an important foundation of the protein structure-function continuum model ⁶¹. To investigate whether our conformational exchange analysis specifically correlates with IDRs in eosinophil RNases, we used the IUPred2 server to generate and compare their disorder profiles (Figure S8) ^62,63. Except for N- and C-terminal regions (which include unstructured leader sequences), the disorder prediction score correctly highlights the ordered nature of these small globular proteins. Most of the profiles indicate that EARs have few regions of disorder; however, as noted before, there is some degree of correlation between the loops showing high conformational flexibility and disorder ⁶⁴. Furthermore, at the family level, the conservation of dynamical surface loops is similar to the preservation of disordered regions in families of IDRs and IDPs, further indicating that both flexibility and disorder are a part of the protein structure-function continuum model.

DISCUSSION

The importance of conformational exchange in enzyme function has been illustrated for a variety of model systems ^{7–9,13,65–67}. While these studies demonstrated the role of conformational motions on the function of discrete enzymes, they pertain to proteins that are distant in the sequence space, and thus do not account for the diversity or similarity of conformational profiles observed between evolutionarily closer members within protein families. Other studies have attempted to address this by probing the conformational properties of homologous sequences within protein families ^32,68,69. Enzymes clustered into phylogenetic subfamilies that share similar biological functions were shown to display conservation of dynamical properties ³². However, these observations still do not account for the notable differences in the function observed between members within phylogenetic subfamilies. Identification of factors driving this difference in function will provide insights into the evolutionary trajectory of these enzymes. To characterize the role of conformational exchange on evolutionary divergence, here we determined the catalytic, biological, and dynamical properties of six members of the eosinophil RNase subfamily using a combination of biochemical, biophysical, and computational approaches. To the best of our knowledge, this is also the first study reporting on the molecular, dynamical, and functional characterization of eosinophil homologs from Macaca fascicularis (MfR3), Pongo pygmaeus (PpR3), Pongo abelii (PaR3), and Aotus trivirgatus (AtR2).

Eosinophil RNases form one of the subfamilies of pancreatic RNases and include human RNases 2 (EDN) and 3 (ECP), which evolved as a result of a gene duplication event after the divergence of Old World monkeys ⁴⁷. A comparison of their catalytic activities shows that HsR2 exhibits significantly higher affinity for the UpA substrate and is >80 times more catalytically active than all other eosinophil RNases (Table 2). Previous studies reported distinct antibacterial activities between human and owl monkey eosinophil RNases ²⁴. Our results further revealed a correlation between antibacterial activities and protein charge. RNases with a pI > 10 display the largest antibacterial activities, while RNases with pI between 9 and 10 exhibit lower activities (Figure 3). Further mutagenesis experiments are necessary to confirm this charge dependence on antibacterial activity. All eosinophil RNases exhibit cytotoxic activities against HeLa cells (Figure 4), in line with previous observations on human HsR2 and HsR3 ³⁶. These results indicate that cytotoxic activities of eosinophil RNases do not maintain the charge-dependent gradient in activity observed with antibacterial assays.

We characterized the timescale and amplitude of conformational motions sampled by eosinophil RNases over the ns-s timescales by combining observations from MD simulations, in addition to NMR ¹⁵N-CPMG and ¹⁵N-CEST experiments. Consistent with previous reports ^17,32, our results show that eosinophil RNases preserve 3D location similarities for residues undergoing conformational exchange on the functionally relevant μs-ms catalytic timescale (Figure 5A and 5B). In contrast to other members of the RNase superfamily ³², this 3D localization largely correlates with the V₁ domain (Figure 1), suggesting evolutionarily pressure to maintain conformational exchange within this domain to preserve biological function(s) promoted by EARs. A comparison of conformational exchange timescales illustrates differences in exchange profiles for HsR2 and PaR3. Residues of AtR2 and all other RNases 3 (except for PaR3) that displayed slower conformational exchange in CPMG measurements were complemented by similar observations from CEST measurements on the ms-s timescale (Figure 5C). Quantitative characterization of the similarity in the conformational exchange patterns between the eosinophil RNases showed higher cosine similarity scores for AtR2 and all RNases 3 (except for PaR3), while HsR2 displayed the largest variations relative to other RNases (Figure 7). Similar trends with diminished similarity scores were observed from CEST experiments, with the exception of PaR3, which showed no exchange on this timescale. Characterization of the faster ns-s timescale motions using MSM analysis also highlighted distinct exchange profiles for the different eosinophil RNases. A comparison of the amplitude of conformational motions showed higher similarities between eosinophil RNases, consistent with previous observations ³².

Our results also illustrate that higher pairwise cosine similarity scores between RNases (Figure 7) do not correlate with the proximity in the evolutionary tree of eosinophil RNases (Figure 2). This is best exemplified by AtR2, the undifferentiated eosinophil RNase. AtR2 shares a higher cosine similarity with RNases 3 on all timescales relative to PaR3, a post-differentiation enzyme. In addition, cosine similarity scores between HsR2 and other eosinophil RNases were just as low as that between these enzymes and BtRA, which diverged early in the evolutionary tree (Figure 2). These results hint at a rapid divergence of conformational exchange patterns, in contrast to a strict conservation along the evolutionary pathway. Such divergence has been described previously, albeit on much more evolutionarily distant enzymes ⁷⁰. Our observations may also suggest a potential limitation of the analysis, which compares similarity across each consensus residue position, and is not sensitive to any similarities that may arise from the comparison of through-space amino acid networks.

Current observations are obviously limited by the fact that the present study was conducted on the ligand-free EARs in similarly controlled in vitro experimental conditions. Not only is the physiologically crowded molecular environment within eosinophils expected to influence protein dynamics, but prior work on pancreatic-type RNases also demonstrated significant changes in conformational profiles upon ligand binding ⁷¹. Additionally, selected EARs may have evolved distinct conformational behavior in the context of specific substrates or molecular targets that would undoubtedly influence their dynamic behavior. Further studies are necessary to investigate the effect of molecular perturbations and ligand binding on the coupling between conformational exchange and biological function within individual EAR members. Nevertheless, the extent to which eosinophil RNases are subjected to conformational exchange may be linked to the rapid molecular evolution of members within this enzyme subfamily ²³. Indeed, proteins undergoing extensive conformational exchange have been shown to also possess high evolvability, defined by “the ability to rapidly adopt (within a few sequence changes) new functions within existing folds” ⁷². Following this definition, evolvability is hard to evaluate without recourse to extensive mutagenesis.

Another feature of highly evolvable enzymes is the ability to perform more than one reaction with a weak efficiency relative to specialized enzymes of the same fold that catalyze the same reactions ⁷³. Numerous examples of such highly flexible, generalist enzymes have been reported, primarily through the characterization of the mechanisms allowing successful protein engineering ^67,73–78. Our observations for AtR2 show a larger number of exchanging residues at all timescales than all other eosinophil RNases investigated, while at the same time exhibiting lower chemical and biological activities. These two properties are associated with enzyme evolvability, suggesting that these motions may be necessary to preserve function. Subsequent evolution from an undifferentiated protein that might be similar to AtR2 could have led to the ‘freezing out’ of some unproductive motions to give rise to the conserved clusters found in simian RNases 3. Similarly, simian RNases 2 may have evolved independently to acquire their antiviral activity, and the larger number of ‘frozen out’ unproductive motions might have led to the increase in activity relative to other eosinophil RNases. The immune system being in a constant arms race with rapidly evolving pathogens such as viruses, bacteria, and other parasites, it must be able to adapt for species to carry on over the course of evolution ⁷⁹. High evolvability is thus a desirable feature for host defense proteins such as eosinophil RNases.

In conclusion, individual side-chain nanosecond fluctuations as well as clusters of residues exchanging on slower timescales (μs-s) seem to be conserved within eosinophil RNases sharing chemical and biological functions, but the specifics of the large-scale movements on all timescales investigated tend towards rapid divergence rather than conservation. Our results show how enzymes from the eosinophil RNase subfamily retain flexibility throughout evolution, and how these movements change from one member to the next, thus providing a possible mechanism by which they remain adaptable towards targets which are themselves changing rapidly. We thus demonstrate how rapidly conformational fluctuations can differ between closely related proteins.

STAR METHODS

RESOURCE AVAILABILITY

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Nicolas Doucet (nicolas.doucet@inrs.ca).

Materials availability

Plasmids generated in this study are available upon request.

Data and code availability

All NMR chemical shift assignments for MfR3, PpR3, PaR3, and AtR2, including Cα and Cβ resonances, were deposited in the in the Biological Magnetic Resonance Bank (BMRB) (http://www.bmrb.wisc.edu) under accession numbers 27542, 27544, 27545, and 27546, respectively.

This paper does not report original code. Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

EXPERIMENTAL MODEL AND SUBJECT DETAILS

Cell lines

Human cervix adenocarcinoma HeLa cell line was acquired from ATCC (American Type Culture Collection). It was grown in MEM medium supplemented with 10% FBS, as recommended by the supplier. Cells were maintained at 37°C in a humidified atmosphere of 5% CO₂. When confluence reached ~80%, a 1:4 subcultivation ratio was used to inoculate new growth flasks.

Method Details

Phylogenetic analysis

Pancreatic-type RNases comprising hominid, bovine, and mouse sequences, were selected for the analysis. A list of all sequences is provided in the supporting information (Table S1). Multiple sequence alignment of selected sequences was performed using ClustalΩ ⁸³. The phylogenetic tree was inferred using the maximum likelihood approach implemented in PhyML ⁸⁸ and the JTT substitution model. Reliability of the internal branches was estimated using aLRT test (SH-like) ⁹⁰. The phylogenetic tree was visualized using FigTree v1.4.2 [http://tree.bio.ed.ac.uk/software/figtree/].

Protein expression, refolding, and purification

Unlabeled bovine RNase A (BtRA) was purchased from BioBasic (Markham, ON, Canada). The human RNase 3 (HsR3) gene was produced as previously described ⁴³. E. coli codon-optimized genes of Macaca fascicularis RNase 3 (MfR3, UniProtKB entry P47779), Pongo pygmaeus RNase 3 (PpR3, P47781), Pongo abelii RNase 3 (PaR3, H2NKI8) and Aotus trivirgatus RNase 2 (AtR2, O18937) were synthesized by ATUM (Newark, CA, USA) and cloned into pJexpress 414 vectors. Unlabeled, ¹⁵N- or {¹³C, ¹⁵N}-labeled protein samples were produced as described previously ⁹¹, with the following changes. After induction with IPTG, bacterial cultures were grown overnight at 30°C and the refolding step was performed at pH 9.1 for HsR2 and AtR2. Identical purification fractions were identified by circular dichroism. Protein concentration was determined using extinction coefficients of 17,460; 17,460; 14,940; 11,960; 13,450; and 13,450 M⁻¹cm⁻¹ for HsR3, HsR2, MfR3, PpR3, PaR3, and AtR2, respectively, as estimated by Expasy ProtParam.

Steady-state kinetics

Steady-state kinetics data were obtained on a PerkinElmer λ35 UV-Vis spectrophotometer (Waltham, MA, USA). RNase activity was measured in a 50 mM sodium acetate buffer containing 1 mM EDTA at pH 5.5 and 25°C. Dinucleotide substrate UpA was obtained from Dharmacon at GE Healthcare (Lafayette, CO, USA). The concentration of UpA was measured using an extinction coefficient of 24,600 M⁻¹cm⁻¹, as supplied by Dharmacon. Change in UpA absorbance was measured in 1- or 2-mm path length cuvettes at 286 nm for 10 min. Enzyme and substrate concentrations used are indicated in Table S2, and all measurements were performed in duplicate. Water was incubated with 1% DEPC for 1 h and autoclaved. All instruments used were cleaned with DEPC-treated water before initiating the experiment to inhibit environmental contaminant RNases from interfering with the kinetic assays. Buffer was prepared using DEPC-treated water. The data were fitted to the Michaelis-Menten equation using SigmaPlot.

Antibacterial assays

Escherichia coli DH5α strain (Gram-negative) and Staphylococcus aureus Newman strain (Gram-positive) bacteria were grown to mid-exponential phase and resuspended in 10 mM sodium phosphate buffer at pH 7.4. Bacteria were then incubated at 37°C for three hours in the presence of 0.5, 2.0, or 5.0 μM bovine RNase A (BtRA), HsR3, MfR3, PpR3, PaR3, AtR2, or HsR2, each performed in triplicate ^38,92,93. BtRA was used as a non-bactericidal enzyme (negative control), while HsR3 was used as positive control ^40,92. Bacterial suspensions were then diluted and plated as 4 × 20 μL droplets on LB-agar plates as per the Miles-Misra technique ⁹⁴, and allowed to grow overnight at 37°C. Colony-forming units (CFU) were then counted the following day.

Cytotoxicity assays

Culture of the human HeLa cell line was achieved using DMEM supplemented with 10% Foetal Bovine Serum (FBS) as culture media. HeLa cells were seeded in 96-well plates at a density of 50,000 cells/well and incubated overnight at 37°C. Cells were then serum deprived for 2 h prior to RNase treatment performed in serum-free culture media with a final volume of 100 μL per well. Cells were treated with RNases at concentrations ranging from 0 to 10 μM for 48 hours. Cell viability was then evaluated using the Cell Titer-Blue Assay from Promega (Madison, WI, USA) following the supplier’s protocol. Data were recorded on a Tecan Infinite M1000 Pro microplate reader (Männedorf, Switzerland) set at 560 nm and 590 nm for excitation and emission wavelengths, respectively. Results were expressed as percentage of control (non-treated cells) and represent the mean±SEM of at least 3 independent experiments performed in triplicate.

Solution NMR experiments

NMR spectra were recorded at 298 K on 100–400 μM protein samples in 15 mM sodium acetate buffer, pH 5.0, and 10% ²H₂O. All spectra were performed on a Varian INOVA spectrometer (Palo Alto, CA, USA) at a working ¹H frequency of 500 or 800 MHz (MfR3 and PpR3), or on a Bruker Avance III spectrometer (Billerica, MA, USA) at a working ¹H frequency of 600 MHz (equipped with a 5 mm TCI cryoprobe) or 800 MHz (equipped with a 5 mm TXI probe) (PaR3 and AtR2). 2D {¹H, ¹⁵N}-HSQC, 3D-HNCACB, 3D-CBCA(CO)NH assignment experiments were performed for all nonhuman proteins. Chemical shift resonance assignments are presented in Figures S2–S3 for 2D {¹H, ¹⁵N}-HSQCs and have been deposited in the Biological Magnetic Resonance Bank (BMRB) (http://www.bmrb.wisc.edu) under accession numbers 27542 (MfR3), 27544 (PpR3), 27545 (PaR3), and 27546 (AtR2). Relaxation-compensated ¹⁵N-Carr-Purcell-Meiboom-Gill (¹⁵N-CPMG) dispersion experiments ⁹⁵ were recorded in an interleaved fashion with τ_cp delays of 0.625, 0.714 (x2), 1.25, 1.667, 2.0, 2.5 (x2), 3.333, 5.0, and 10.0 ms, using a total relaxation period of 40.0 ms, as described earlier ^13,42,43. ¹⁵N-Chemical Exchange Saturation Transfer (¹⁵N-CEST) experiments ⁵⁵ were performed using 15 and 25 Hz B₁ irradiation fields. Series of 114 to 124 2D datasets were acquired, corresponding to ¹⁵N offsets interspaced by 0.25 ppm (~20 Hz) that span the whole resonance array of each analyzed protein and for both B₁ fields. All spectra were processed using NMRPipe ⁸⁷, and peak picking was performed using CcpNmr AnalysisAssign ⁸⁴ and Sparky [T. D. Goddard and D. G. Kneller, SPARKY 3, University of California, San Francisco, https://www.cgl.ucsf.edu/home/sparky/]. ¹⁵N-CPMG-derived R_2,eff values were fitted to the full single-quantum CPMG equation ^57,96 using GraphPad Prism. ¹⁵N-CEST intensity profiles were analyzed using the Chemex scripts [https://github.com/gbouvignies/chemex; ⁵⁵].

NMR chemical shift assignments of the four simian eosinophil RNases

To determine the atomistic behavior of the various eosinophil RNases, we assigned chemical shifts of protein backbone atoms using {¹H, ¹⁵N}-HSQC, HNCACB, and CBCA(CO)NH heteronuclear NMR experiments. All HSQC spectra (Figures S2–S3) show well-dispersed resonances, indicating well-folded proteins. Assigned non-proline backbone amides of each protein amounted to 120 out of 121 (MfR3), 121 out of 122 (PpR3), 115 out of 121 (PaR3), and 118 out of 121 (AtR2) resonances identified. All chemical shift assignments for MfR3, PpR3, PaR3, and AtR2, including Cα and Cβ resonances, were deposited in the BMRB under accession numbers 27542, 27544, 27545, and 27546, respectively. Due to resonance line broadening, we note that residues from the putative loop 5 (Ser74 to Val78) could not be assigned in PaR3, suggesting exchange on the intermediate regime for this loop. This is further supported by the observed weakness in signal amplitude for upstream residues His72 and His73 in PaR3. Similarly positioned residues Cys71, His72, His73, and Ser74 in AtR2, or Gly75, Ser76, and Gln77 in HsR2, show analogous behavior ^52,53. Since both AtR2 and HsR2 share a single residue difference with the homologous loop 5 of PaR3, this observation is consistent with a similar exchange regime.

Homology modeling

Structures of simian RNases PpR3, PaR3, MfR3, and AtR2 were determined through homology modeling using MODELLER v9.13 with human RNases 2 (PDB entry 1GQV) and RNase 3 (PDB entry 1QMT) structures as templates. The four simian RNase sequences share high sequence identities with human RNases 2 and 3 (Table 1). Models were generated using the automodel option and structures with the lowest DOPE scores for each enzyme were selected for each of the four enzymes.

Computational simulations and analyses

Crystal structures of human RNases 2 and 3 (PDB entries 1GQV and 1QMT, respectively) were used as starting structures. For other RNases with no available crystal structures, homology models were used as initial coordinates (see above). We validated our models by comparing NMR ¹⁵N chemical shifts with the computationally derived ones using the CS2BACKBONE module of PLUMED 2.4.4 ⁸⁹, as shown in Figure S4. All simulations were performed using the graphical processing unit (GPU) enabled version of AMBER ⁸² v14 (pmemd program) and the ff14SB force field. Each enzyme was placed in a rectangular box with SPC/E water and neutralized through the addition of counter ions. The minimum distance between the enzyme and box edge was set to 10 Å. System equilibration, comprising a series of energy minimization and equilibration steps, was performed at 300 K using the protocol described previously ^32,97. Simulations were performed using timesteps of 2 fs under constant energy conditions (NVE) with periodic boundary conditions. Long-range electrostatic interactions were computed using the particle mesh Ewald (PME) method and a cut-off of 8 Å. All simulations were performed for a total of 1 μs for each enzyme. Root mean square fluctuations of the top ten modes (RMSF₁₀) was calculated using the trajectory (conformations) obtained for each enzyme using the CPPTRAJ program [https://github.com/Amber-MD/cpptraj]. Translational and rotational diffusion of individual structures was removed by superposing all structures within a trajectory to a reference.

Markov State Model (MSM)

MSM analysis ^59,98–100 was performed to estimate the conformational exchange timescales using PyEMMA ⁵⁸. For each residue, we computed backbone and side-chain torsion angles and used them separately as input features for per-residue MSMs ¹⁰¹. We thereby assume that the degrees of freedom that govern the motions of the single amino acids are defined by their constant local environment such that larger scale protein motions are negligible for the timescales derived here. The continuous angle trajectories per residue were discretized with 150 cluster centers (k-means clustering) to obtain time series of discrete states. The connection between changes of torsion angles and MSM implied timescales is exemplified in Figure S6. Model validation was performed by comparing results for different numbers of discrete states (50, 100, 150) and by ensuring implied timescale convergence. Hidden Markov Models (HMMs) that fulfilled these measures were manually selected to estimate the implied timescales presented below. Model estimation was performed at a lag time of 10 ns.

Cosine similarity comparisons

Cosine similarity comparisons were performed using the above sequence alignments for the dimensionality, residue-dependent parameters extracted from NMR experiments, or MSM analyses as the vector components, and the following formula:

$$\cos \theta =\frac{\overrightarrow{A}\cdot \overrightarrow{B}}{\Vert \overrightarrow{A}\Vert \Vert \overrightarrow{B}\Vert }=\frac{\sum _{i=0}^N{A}_i{B}_i}{\sqrt{\sum _{i=0}^N{A}_i^2}\sqrt{\sum _{i=0}^N{B}_i^2}}$$

This technique uses a vector representation of each protein, with a dimensionality equal to the number of positions in the consensus sequence (134 when comparing only eosinophil RNases, 140 when accounting for RNase A). The consensus sequence represents an indexing corresponding to the multiple sequence alignment of RNases that includes gaps representing insertions/deletions in sequences and facilitates easier comparison of sequences of different lengths. The value of any residue-dependent parameter can be used as the vector components. Calculating the cosine of the angle θ between these vectors (cos(θ)) through a dot product thus gives an easily obtainable quantitative way to compare proteins, with values close to 1 (i.e., parallel vectors) corresponding to similar proteins. To clarify these comparisons, the effects of location and exchange rate have been bisected. For pairwise comparison of the exchanging residue location, we used a binary encoding (1 or 0) for exchanging and non-exchanging (or alignment gap) residues as they are identified by CPMG and CEST experiments as well as MSM analyses. Residue-specific k_ex values obtained either from CPMG or CEST and MSM-implied timescales were then directly compared to assess timescale magnitudes.

QUANTIFICATION AND STATISTICAL ANALYSIS

All statistical analysis and software used can be found in the figure legends and method details. For cytotoxic and antibacterial assays, statistical tests used are expressed as percentage of control (non-treated cells). They represent the mean ± SEM of at least 3 independent experiments performed in triplicate. Significance was assessed using OneWay ANOVA with * for p < 0.05, ** for p < 0.01, *** for p < 0.001, and **** for p < 0.0001.

KEY RESOURCES TABLE.

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Bacterial and virus strains
Escherichia coli BL21(DE3) cells Genotype: F(-) ompT gal dcm lon hsdSB(rB- mB-) λ(DE3 [lacI lacUV5-T7 gene 1 ind1 sam7 nin5])	BCCM/GeneCorner, obtained from the lab of Eric Déziel (INRS)	Cat#LMBP1455,https://bccm.belspo.be/catalogues/genecorner-host-details?NUM=LMBP%201455
Escherichia coli DH5α cells Genotype: supE44 ΔlacU169 (φ80 lacZΔM15) hsdR17 recA1 endA1 gyrA96 thi-1 relA1	DSMZ (German Collection of Microorganisms), obtained from the lab of Eric Déziel (INRS)	Cat#DSM6897, https://www.dsmz.de/collection/catalogue/details/culture/DSM-6897
Staphylococcus aureus Newman strain	ATCC, obtained from the lab of Eric Déziel (INRS)	Cat#25904, https://www.atcc.org/products/25904
Chemicals, peptides, and recombinant proteins
Bovine RNase A (BtRA)	BioBasic	RB0473
Deposited data
Aotus trivirgatus RNase 2 (AtR2)	This paper	Biological Magnetic Resonance Bank (BMRB) entry 27546
Macaca fascicularis RNase 3 (MfR3)	This paper	Biological Magnetic Resonance Bank (BMRB) entry 27542
Pongo abelii RNase 3 (PaR3)	This paper	Biological Magnetic Resonance Bank (BMRB) entry 27545
Pongo pygmaeus RNase 3 (PpR3)	This paper	Biological Magnetic Resonance Bank (BMRB) entry 27544
Experimental models: Cell lines
Human HeLa cells	ATCC	CCL-2, https://www.atcc.org/products/ccl-2
Oligonucleotides
UpA dinucleotide	Dharmacon/GE Healthcare	N/A
Recombinant DNA
Homo sapiens RNase 3 (HsR3)	43	N/A
Aotus trivirgatus RNase 2 (AtR2)	ATUM	GeneID 150351
Macaca fascicularis RNase 3 (MfR3)	ATUM	GeneID 150353
Pongo abelii RNase 3 (PaR3)	ATUM	GeneID 150355
Pongo pygmaeus RNase 3 (PpR3)	ATUM	GeneID 150354
Software and algorithms
AMBER 14	82	https://ambermd.org/
ClustalΩ	83	https://www.ebi.ac.uk/Tools/msa/clustalo/
CcpNmr AnalysisAssign	84	https://ccpn.ac.uk/software/
Chemex	55	https://github.com/gbouvignies/chemex
CPPTRAJ	85	https://github.com/Amber-MD/cpptraj
ESPript 3.0	86	https://espript.ibcp.fr/
FigTree 1.4.2	N/A	http://tree.bio.ed.ac.uk/software/figtree/
GraphPad Prism	N/A	https://www.graphpad.com/
IUPred2	62,63
MODELLER 9.13	N/A	https://salilab.org/modeller/
NMRPipe	87	https://www.ibbr.umd.edu/nmrpipe/
PhyML	88	http://www.atgc-montpellier.fr/phyml/
PLUMED 2.4.4	89	https://www.plumed.org/
ProtParam	N/A	https://web.expasy.org/protparam/
PyEMMA	58	http://emma-project.org/latest/
SigmaPlot	N/A	https://systatsoftware.com/sigmaplot/
Sparky	T. D. Goddard and D. G. Kneller, SPARKY 3, University of California, San Francisco	https://www.cgl.ucsf.edu/home/sparky/
VMD	81	http://www.ks.uiuc.edu/Research/vmd/

Highlights.

• Evolutionary conservation of flexibility between functional homologs remains elusive

• Allosteric control of unique family members is required to achieve target specificity

• Conformational and biophysical characterization of homologous EARs was performed

• Profiles within family imply divergence rather than conservation of flexibility

INCLUSION AND DIVERSITY

One or more of the authors of this paper self-identifies as a member of the LGBTQIA+ community. We support inclusive, diverse, and equitable conduct of research.