Designed leucine‐rich repeat proteins bind two muramyl dipeptide ligands

Abstract

Designed protein receptors hold diagnostic and therapeutic promise. We now report the design of five consensus leucine‐rich repeat proteins (CLRR4–8) based on the LRR domain of nucleotide‐binding oligomerization domain (NOD)‐like receptors involved in the innate immune system. The CLRRs bind muramyl dipeptide (MDP), a bacterial cell wall component, with micromolar affinity. The overall K _{d app} values ranged from 1.0 to 57 μM as measured by fluorescence quenching experiments. Biphasic fluorescence quenching curves were observed in all CLRRs, with higher affinity K _d1 values ranging from 0.04 to 4.5 μM, and lower affinity K _d2 values ranging from 3.1 to 227 μM. These biphasic binding curves, along with the docking studies of MDP binding to CLRR4, suggest that at least two MDPs bind to each protein. Previously, only single MDP binding was reported. This high‐capacity binding of MDP promises small, soluble, stable CLRR scaffolds as candidates for the future design of pathogen biosensors.

Abbreviations

LRR

leucine‐rich repeat

CLRR

consensus LRR

TPR

tetratricopeptide repeat

ANK

ankyrin

ARM

armadillo

MDP

muramyl dipeptide

NOD

nucleotide‐binding oligomerization domain

NLR

NOD‐like receptor

NLR

NLRP, NOD‐LRR and pyrin domain

PRR

pattern‐recognition receptors

SPR

surface plasmon resonance

SEC

size exclusion chromatography

1. INTRODUCTION

Innovations in structure determination, computational technology, and experimental techniques have created a new and growing class of designed receptors that can be used as biosensors. ¹ Biosensors have the potential to play significant roles in medicine, agriculture, food safety, homeland security, and environmental monitoring. ² Traditionally, antibodies have been the gold standard in biosensor design due to potent antigen binding. ³ Designed protein scaffolds have been used to address the drawbacks associated with antibodies—their large size and complexity lead to difficult and expensive production. ⁴ , ⁵ , ⁶ Scaffolds can address these drawbacks because they are typically smaller in size, have increased environmental stability, lack reactive cysteine residues, utilize cost‐efficient bacterial overexpression methods, and can be designed for potent binding of a wide variety of ligands. ¹

Repeat proteins have emerged as leading candidates in the field of alternative protein scaffolds. ¹ These proteins contain repeating structural motifs where the number of amino acids in each repeat varies, usually consisting of 20 to 40 residues. ⁷ Each of the repeat protein families has a signature structural and sequence motif. Local sequence contacts of repeat proteins allow them to form elongated structures that provide large, regular surface areas for ligand binding. Local sequence contacts, allow for the design of diversity in ligand binding. ⁸ , ⁹ , ¹⁰ To date, the tetratricopeptide repeats (TPRs), ¹¹ ankyrin (ANK) repeats, ¹² armadillo (ARM) repeats, ¹³ and leucine‐rich repeats (LRRs) ¹⁴ , ¹⁵ , ¹⁶ have been used in novel binding scaffolds.

Nucleotide‐binding oligomerization domain (NOD)‐like receptors (NLRs) play a key role in early detection and defense against infectious diseases, a leading cause of death worldwide. ¹⁷ , ¹⁸ Crucial to the innate immune system, pattern‐recognition receptors (PRRs) are responsible for the body's non‐specific defense mechanisms by recognizing pathogen‐associated molecular patterns. ¹⁹ , ²⁰ , ²¹ NLRs represent a class of cytoplasmic PRRs that rely mainly on their LRR motif for recognition of ligands with huge chemical diversity, such as bacterial cell‐wall peptidoglycans, bacterial RNA, uric acid crystals, and viruses. ¹⁸ , ²⁰ , ²² , ²³ Because NOD2 is a well‐described receptor associated with bacterial peptidoglycan recognition, ²⁴ , ²⁵ our consensus design focused on the NOD subgroup of NLRs.

Based on the innate immunity NOD proteins, we have designed a series of proteins that require only the LRR domain of NODs to bind the bacterial cell wall fragment, muramyl dipeptide (MDP), ¹⁶ a fragment that generates an innate immune response (Figure 1). The isolated LRR domain of native NOD2 binds surface‐immobilized 6‐amino‐MDP with an affinity of 210 nM, while the full‐length NOD2 bound with an affinity in the range of 1–30 nM, as measured by surface plasmon resonance (SPR). ²⁶ , ²⁷ We now describe the design, synthesis, and characterization of a series of proteins that each bind at least two MDPs, with the highest affinity value of 40 nM as measured by fluorescence quenching experiments. Our consensus leucine‐rich repeat (CLRR) proteins are designated CLRRx, where x represents the number of internal repeats.

FIGURE 1.

Structure of muramyl dipeptide (MDP, N‐acetyl‐α‐d‐glucosamine‐d‐Lac‐l‐Ala‐d‐Glu‐NH₂)

2. RESULTS AND DISCUSSION

2.1. Design of consensus leucine‐rich repeat (CLRR) sequence

The consensus design approach retains amino acids that appear most frequently at each position. This approach typically leads to improved thermodynamic stability in the designed proteins over their natural counterparts. ²⁸ , ²⁹ The NOD subgroup includes NOD1–5, and class II major histocompatibility complex transactivator (CIITA) proteins. ³⁰ Analysis of the sequences showed that the most common number of internal repeats in NOD proteins was seven (Figure 2).

FIGURE 2.

Stacked column graph displaying the most common number of internal repeats (based on gene structure) in mammalian NOD protein families, and number of internal repeats for each individual protein

Each internal repeat has one β‐strand and one α‐helix. Internal repeats represent the core tandem repeats, while capping repeats contain short terminal sequences that flank the internal repeats, protecting the extended hydrophobic core from solvent exposure and aggregation (Figure 3c). ³¹ , ³² A typical NOD LRR contains a series of internal repeats that have a β‐sheet and an α‐helix, and the last internal repeat is capped by a β‐strand at the C‐terminus; they do not have an N‐terminal cap. ³³ , ³⁴

FIGURE 3.

The process of consensus design for the CLRR proteins. (a) The canonical motif for LRR. (b) The LRRs of the human NOD2 protein aligned with the canonical motif. (c) The first, second and third most common amino acids at each position from analysis of 811 repeats from 75 proteins, with their % frequency listed in the row below for internal repeats (green) and the β‐strand capping repeats (blue). The first three residues of the β‐strand capping repeat are greyed out because they are derived from the β‐strand of the internal repeat sequence. (d) The final CLRR consensus sequence of one repeat with internal repeats in green and capping repeat in blue

Six genes encoding for the human NOD proteins were obtained from the HUGO Gene Nomenclature Committee (HGNC) database. ³⁵ Individual tandem repeats were confirmed through the gene structure, and these LRR sequences were submitted to blastp through the National Center for Biotechnology Information (NCBI) database to find all homologous mammalian proteins. ³⁶ This resulted in 75 NOD protein sequences containing 811 individual LRR sequences (Table S1). Previously, 28 proteins from both the NLRP and NOD subgroups were used in the consensus design of CLRR2. ¹⁶ The addition of 47 proteins to the original database used in the design of CLRR2 led to significant differences in the consensus sequence. ¹⁶ Amino acids in the canonical motif and hypervariable positions remained the same.

The 811 sequences obtained through the NCBI database were manually aligned to the canonical LRR motif, xLxxLxLxxNxL(x)_nL(x)₈ (Figure 3a). ³⁷ For example, each repeat of human NOD2 protein was aligned with the canonical motif (Figure 3b). Because the first repeats were not significantly different from the internal repeats, no N‐terminal cap was designated separately as in the original CLRR2. ¹⁶ Next, the last repeats of each protein were analyzed separately to define the capping repeat. The C‐terminal cap was shortened from 28 in the original design to 13 residues in the current CLRR design, retaining the β‐strand and eliminating the α‐helix. ¹⁶ The sequences were then analyzed for the amino acid residue frequency at each position (Figure 3c). The most frequent amino acids, those with greater than 30% frequency, and some with even lower % frequency, were kept for the consensus sequence, which was then optimized for folding and stability as follows (Figure 3d).

The physical properties were optimized, first in the internal repeats, particularly positions with a high preference for charged residues. Positions 3, 6, 14, 15, and 22 showed preference for negatively charged Asp (D) or Glu (E), and positions 18 and 26 showed preference for positively charged Lys (K), so these residues were retained, except that at position 3, Lys (14%) was chosen instead of Glu (16%) (Figure 3c,d).

Proteins tend to aggregate and precipitate as a function of their isoelectric point (pI). ³⁸ Initially, three proteins with Ala (A) at position 25 were designed (pI ~6.3), ³⁹ but they dimerized and aggregated easily. Since a strong preference was not indicated at position 25, Ala was replaced with Glu to maintain a suitable isoelectric point (pI ≤ 5.5) for good protein solubility and negative net charge throughout (Figure 3d). ³⁹

The remaining hypervariable positions were assessed for preference of amino acid physical properties, such as size, polarity, and propensity for secondary structure. We made the following alterations to the consensus sequence. A Pro (P) was introduced at position 28 (Figure 3d), because a kink in the structure before the β‐sheet was found in the LRR‐domain of the ribonuclease inhibitor (RI) crystal structure (PDB ID: 1A4Y). ⁴⁰ Similarly, a Gly (G) was retained at position 13 to maintain a kink before the α‐helix, ⁴⁰ and Gly (G) is the most frequent residue at position 13 (Figure 3d). Trp (W) residues were introduced in the internal repeat at hypervariable position 8, and retained in the capping repeat at position 6, to enable fluorescence studies. Overall, 7 out of 28 residues of the CLRR2 internal repeat ¹⁶ were different upon redesign with the larger database to give the current CLRRs—Ser4, Asp6, Trp8, Gln11, Gly13, Glu25, and Pro28 (Figure 3d).

Four out of 13 residues in the shortened cap of the CLRRs—Leu2, Lys3, Arg8, and Thr11, were changed from CLRR2 because these had higher frequencies in the expanded database. ¹⁶ The 10 residues (positions 4–13) of the β‐strand capping repeat had higher conservation than the internal repeats (Figure 3c). Capping repeats are known to have higher levels of amino acid conservation. ³² The first three residues in the capping repeat are derived from the N‐terminal β‐strand of the internal repeats in order to connect the last α‐helix with the capping repeat β‐strand that protects the hydrophobic core (Figure 3c). ³² This analysis of amino acid conservation and physical properties resulted in the current CLRR repeat sequence (Figure 3d). The consensus repeats have an average of 49% sequence identity to the rabbit NOD2 LRR (PDB ID: 5IRM). ⁴¹

Proteins with 4 to 8 internal repeats (CLRR4–CLRR8), plus one capping repeat each, were chosen for synthesis and characterization because of their similar lengths to the 7 and 5 repeat NODs, which are the first and second most common number of repeats found in nature respectively (Figure 2). The CLRR4 protein (161 residues) has 4 full‐length internal repeats, no N‐terminal repeat, and one C‐terminal β‐strand capping repeat. CLRR2 (176 residues) also had 4 full‐length repeats: two internal, plus N‐ and C‐terminal repeats. ¹⁶

The computed physical and chemical parameters of the CLRR proteins studied are summarized in Table 1. Size exclusion chromatography (SEC) overestimates the MW of elongated proteins. ³⁴ Because repeat proteins have an elongated structure compared to the Biorad globular protein standards, they have been shown to elute slightly faster, and thus give slightly higher apparent molecular mass. ³⁴ CLRRs fairly consistently showed an apparent MW on average 10% higher than the calculated values (Table 1). The ε values scale with the number of Trp residues (Table 1). The pI decreases with the number of repeats because each internal repeat carries one negative charge (Table 1).

TABLE 1.

Computed physical and chemical parameters for the CLRR proteins

Name	CLRR4	CLRR5	CLRR6	CLRR7	CLRR8
Length (amino acid residues)a	161	191	221	251	281
Molecular weight (Da)a	17,978	21,142	24,392	27,556	30,719
Apparent MW by size‐exclusion (Da)b	20,000	23,300	26,700	30,500	33,500
ε at 280 nm (M−1·cm−1)a,c	33,460	38,960	44,460	49,960	55,460
pIa	5.50	5.40	5.32	5.27	5.22
Net chargea	–6	–7	–8	−9	−10

^aComputed using the ExPASy ProtParam tool (web.expasy.org/protparam/).

^bSize‐exclusion chromatography (SEC).

^c ε = extinction coefficient. ³⁹

DNA sequences for the CLRRs were synthesized using Klenow extension and subsequent PCR amplification (Figure S1). Sequences were cloned into plasmid pProExHtam under the control of the lac operon, and the proteins were overexpressed using E. coli BL21 (DE3) cells. CLRRs with the expected molecular weights were observed by SDS‐PAGE in both the soluble and insoluble fractions (Figure S4a). This was an improvement over the CLRR2 protein that was only present in the insoluble fraction and had to be solubilized. ¹⁶ Soluble fractions of the proteins were purified by Ni‐NTA and size‐exclusion chromatography, and refolded by dialysis out of urea into PBS before characterization (Figure S6).

The CLRR proteins were properly folded and monomeric. All proteins eluted from the analytical size‐exclusion column as single smooth peaks, indicating high purity and monomeric state (Figure 4a). Experimental molecular mass values calculated from the standards were ca. 1.1 times higher than the theoretical values for the molecular weight of monomeric protein, as expected for elongated proteins (Table 1). ³⁴ MALDI‐TOF mass spectrometry of CLRR7 and CLRR8 further confirmed their molecular mass and monomeric state (Figure S5). Purified proteins stored at 4°C remained stable for 2 weeks without showing signs of aggregation, loss of structure, or loss of activity. This is a significant improvement over the 24 hr stability of the previous CLRR2. ¹⁶

FIGURE 4.

Protein characterization. (a) Analytical size‐exclusion chromatography of the purified proteins. The void volume was 6.40 mL. (b) Normalized CD spectra of the CLRRs taken at 24°C after purification and dialysis. (c) Thermal melting curves of CLRRs obtained by monitoring the CD mean residue ellipticity, [θ]_mrw, versus temperature. Each curve was fitted to a two‐state model

2.2. Secondary structure

CD spectroscopy was used to assess the secondary structure of the CLRRs. Before refolding, CLRRs exhibited maxima in the 190–200 nm region, and minima at 205 nm and 220 nm (Figure S6a). The maxima near 190 nm and the minima at 205 and 220 nm indicated that the proteins were partially folded, not random coil. ⁴² Because the LRR motif contains both α‐helical and β‐sheet regions, the signature CD signals of these secondary structures were analyzed. The minima at 205 nm, instead of a typical α‐helical minimum at 208 nm, suggested that the proteins were not completely folded (Figure S6). Therefore, the proteins were refolded with decreasing concentrations of urea buffer into PBS. After refolding, the minima near 205 nm were slightly red‐shifted to 208 to 210 nm, while the minima at 220 nm remained the same (Figure 4b). The deeper minima at 208 nm compared to the minima at 220 nm indicated that CLRRs contained slightly more α‐helical content than β‐sheet content (Figure 4b). ⁴² Refolded CLRRs displayed unfolding curves typical to mixed α‐helical and β‐sheet containing proteins (Figure 4c).

A typical NOD LRR internal repeat has an 8‐residue β‐strand, a 5‐residue loop, a 12‐residue α‐helix, and another 3‐residue loop (Figure 3a). ³⁷ The CD spectra were analyzed using the BeStSel tool (bestsel.elte.hu/index.php). ⁴² For all CLRRs, the estimated α‐helical content averaged 42%, while the β‐sheet content averaged 13% (Table 2). ⁴² The α‐helical contents of the CLRR4–8 are comparable to the designed ribonuclease inhibitor (RI) LRRs (37% α‐helix). ⁴⁰ Because of the high content of Trp residues (one per repeat), unpredictable contributions to the CD spectra may create inaccuracies in the helix/sheet content estimates.

TABLE 2.

The biophysical characteristics of the CLRR proteins

Name	CLRR4	CLRR5	CLRR6	CLRR7	CLRR8
α‐Helical content (%) a	40	44	43	38	43
β‐Sheet content (%) a	10	14	9	17	16
T m (°C)	45 ± 1	44 ± 0	42 ± 1	44 ± 0	58 ± 1

^aComputed using BeStSel tool (bestsel.elte.hu/index.php). ⁴²

2.3. Homology model and surface charge

The rabbit NOD2 was chosen to model the CLRR structures (PDB ID: 5IRM). ⁴¹ The CLRRs share sequence identities of approximately 49% with the rabbit NOD2 LRR domain. ⁴¹ The homology model of CLRR4 is predicted to have a structure very similar to the LRR domain, with a β‐sheet on the concave face, and a series of α‐helices on the convex face (Figure 5a) observed in the model. The Ramachandran plot showed mostly favorable backbone Φ and Ψ angles (Figure 5b). The QMEAN structure scores of the homology model of CLRR4 mostly align with known structural metrics (Figure S7). Statistics show good local similarity with the target, PDB ID: 5IRM, and good fit of normalized QMEAN square with non‐redundant PDB structures and favorable geometrical properties (Figure S7). The α‐helical content of the model was 38%, and the β‐sheet content was 11%, in agreement with the CD analysis using BeStSel (Table 2). The CD spectra and homology model confirm that designed CLRRs retain structures analogous to native LRRs.

FIGURE 5.

Homology model of CLRR4. (a) Structure of CLRR4 shown as a cartoon and colored by QMEAN score (blue–favorable, red–outlier, Figure S7). (b) Ramachandran plot with Φ, Ψ coordinates. Each residue (dot) is colored by favorable (blue, 85.5%) or outliers (red, 14.5%)

Surface property mapping of the CLRR4 homology model showed ordered patterns of alternating charged and hydrophobic patches. The convex surface shows stabilization of the α‐helices by a network of possible salt bridges on the surface (Figure 6a). The potential salt bridges are intra‐helix i‐to‐i + 4, or inter‐repeat along the convex face, although some of these static distances are too long (Figure S8). Near the loops at the C‐termini of the helices, salt bridges are i‐to‐i + 1 or inter‐repeat. The concave surface shows, from bottom to top, a distinct hydrophobic patch of tryptophans (green), a negative patch of aspartates (red), a patch of serines, and a positive patch of lysines (blue) (Figure 6b).

FIGURE 6.

Physicochemical surface map of CLRR4. (a) Convex α‐helix surface map, and (b) concave β‐sheet surface map. Surfaces are shown as physicochemical properties with blue (positive charges), red (negative charges), and green (hydrophobic)

2.4. Folded protein stability

Thermal unfolding was monitored by the change in CD signal at 217 nm near the combined α‐helix and β‐sheet minima. At 80°C, loss of signal at the minima indicated transition to the denatured state (Figure S6b). The normalized signal changes were plotted as a function of temperature (Figure 4c). Individual unfolding curves with data points are given in Figure S9. The transition temperatures (T _m) were approximately the same for CLRR4–CLRR7, and significantly higher for CLRR8 (Table 2). The unfolding curves displayed a broader transition than would be expected for a simple two‐state unfolding mechanism. Because the CLRRs displayed a partially folded state prior to dialysis out of urea, it is likely that folding intermediates exist. CLRR8 appears to have a broader folding transition, that is, the slope is different from the other CLRRs (Figure 4c). Since it has more segments to fold, and it folds more stably, there may be more noticeable intermediates to give the final folded structure. The previously designed CLRR2 also did not fit well to a simple two‐state unfolding model. ¹⁶ The CD unfolding of other α‐helix and β‐sheet containing repeat proteins shows that capping repeats and internal repeats do not unfold in a concerted manner. ⁴³ , ⁴⁴

Typically, designed homogeneous repeat proteins show linear increases in thermal stability with length. ⁴⁵ , ⁴⁶ , ⁴⁷ Natural repeat protein T _m values are found in discrete intervals of repeat numbers; TPRs are found in intervals of 3 repeats, ANKs in intervals of 6 repeats, ⁴⁸ and ARMs in intervals of 12 repeats. ⁹ ANKs show increased stability in step intervals of 6 repeats rather than a linear increase in stability. ⁴⁸ Plückthun and coworkers showed a similar phenomenon, where designed LRR proteins showed a step increase in stability from 8 to 12 repeats, while the 10 repeat had the same stability as the 8 repeat. ⁴⁹ We observed a T _m plateau from 4 to 7 repeats and step increase in stability at 8 repeats, so LRRs may show the same increase in stability in intervals of 4 as those of Plückthun (Table 2). ⁴⁹ The 14°C higher T _m of CLRR8 shows that 8 repeats may be a plateau for increased thermal stability (Figure 4c). Further studies on shorter and longer LRR proteins, such as 2, 3, 10, 12, 16 etc. repeats, would help confirm this step interval.

2.5. Fluorescence binding to muramyl dipeptide (MDP)

MDP is a known natural ligand for the NOD2 protein of the NLR family (Figure 1). ²⁵ Our previously designed CLRR2 showed an MDP binding affinity of 2.0 μM, measured by fluorescence quenching of the Trps on the concave surface. ¹⁶ This surface is the binding site for angiogenin on the ribonuclease inhibitor NLRP, ⁴⁰ so MDP was expected to bind to the concave surface. Each repeat of the CLRRs contains a Trp residue on the concave binding surface (Figure 6b) to measure MDP binding affinity by fluorescence quenching. Fluorescence was recorded up to 100 mM of MDP for CLRR4 and CLRR6‐8 (Table S3). When the fluorescence data were fitted to a single curve (Figure 7, dashed lines), the apparent binding affinities (K _{d app}) ranged from 1.0 to 57 μM, but the fits were poor and we noticed a significant inflection point in the data (Table 3, Figure 7).

FIGURE 7.

Fluorescence quenching of Trp residues in the CLRRs is shown as a function of total MDP concentration. Overall apparent affinity (K _{d app} dashed lines), and individual binding curves, (K _d1 and K _d2 solid lines), with error bars from two measurements at each concentration are shown for: (a) CLRR4, (b) CLRR5, (c) CLRR6, (d) CLRR7, and (e) CLRR8. (f) Compilation of the overall apparent affinity (K _{d app}) curves fitted to Equation (4) for CLRR4–8 (Table 3). Concentrations of MDP up to 100 mM are not shown in (a)–(e) because each curve plateaued above the highest concentration shown. Raw fluorescence data is given in Table S3

TABLE 3.

Fluorescence binding affinities of CLRRs for MDP

K d (μM)	CLRR4	CLRR5	CLRR6	CLRR7	CLRR8
K d1	0.04 ± 0.01	0.3 ± 0.2	1.1 ± 0.5	4.5 ± 1.3	0.06 ± 0.02
K d2	140 ± 23	211 ± 10	227 ± 40	50 ± 2	3.1 ± 0.5
K d app	57 ± 20	29 ± 10	46 ± 11	6.5 ± 0.5	1.0 ± 0.2

For each CLRR, increasing MDP concentrations gave a biphasic curve that showed a first plateau after about 50 μM for CLRR4–7, and after about 1 μM for CLRR8. The second phase plateaued after 300–600 μM for CLRR4–7, and after 5 μM for CLRR8 (Figure 7). This indicated binding of at least two MDP ligands, and the two curves were fitted independently to Equation (4) (Figure 7a–e). The K _d1 values ranged from 0.04 to 4.5 μM, and K _d2 values ranged from 3.1 to 277 μM (Table 3). The K _d1 values of CLRR4 and CLRR8 are in a similar range as the K _d value of the full‐length NOD2 protein binding to immobilized MDP (1–30 nM). ²⁵ Unlike the measurements from full‐length NOD2, the binding of our proteins were measured to free MDP in solution, rather than to MDP that was modified by converting its primary alcohol to a primary amine in order to immobilize it on a surface for SPR. ²⁵

The CLRRs have, on average, a sequence identity of 49% to the NOD2 protein (Figure 5). Although the CLRRs have fewer internal repeats than the rabbit NOD2, which shows 10 internal repeats in the crystal structure (PDB ID: 5IRM), ⁴⁰ this series of designed proteins preserved the necessary biophysical properties for an MDP‐binding scaffold. The K _{d app} of CLRR8 (1.0 μM) for MDP is comparable to the K _{d app} of CLRR2 (2.0 μM), although CLRR4 and 5 are closer in number of repeats to CLRR2. ¹⁶

The biphasic K _d1 values for CLRR4 (40 ± 10 nM) and CLRR8 (60 ± 20 nM) protein are nearly as tight as the MDP binding of a 10‐repeat NOD2‐LRR‐maltose binding protein fusion. ²⁵ , ²⁶ Furthermore, MDP‐LRR affinities have not been analyzed as biphasic binding curves previously. ¹⁶ , ²⁵ Biphasic binding curves with inflection points indicate multiple binding sites with potential for high and low affinities sites, as well as cooperative or non‐cooperative binding activities. ⁵⁰ , ⁵¹ We note that the tightest binding affinities were observed for CLRR4 and CLRR8, the same step interval observed for increased stability.

2.6. Modeling of multiple MDP binding sites

Homology modeling techniques have been successfully used in previous studies to model novel modifications to LRRs. ²⁶ , ²⁷ , ⁵² Previous molecular docking on LRRs have indicated that MDP binds to the horseshoe (or β‐strand concave) region, and interacts with residues that are polar, charged, and aromatic. ²⁶ Molecular docking simulations were performed to determine if CLRR4 was binding multiple MPD ligands. A single MDP molecule was docked flexibly to CLRR4 with the binding site unbiased; the nine lowest energy structures were saved. MDP docked in only two locations—site 1 is on the loops connecting the α‐helices and β‐strands below the tryptophans (5/9 times), and site 2 is on the β‐sheet concave surface (4/9 times) (Figure S10). The docking energy was −6.2 kcal/mol for pose 1, and it was −5.6 kcal/mol for pose 2. While site 2 is the expected binding site on the concave surface of the β‐sheet, site 1 was unexpected. This result of two binding sites for CLRR4 is consistent with the biphasic binding curve results found by fluorescence (Figure 7, Table 3). Combined, these data strongly suggested binding of at least two MDP ligands by the LRR region of NOD proteins.

To model binding of two MDP ligands, pose 1 and pose 2 were each fixed individually to CLRR4, and a second MDP was docked to this complex (Figures S11 and S12). The model shows two binding pockets below (site 1) and above (site 2) the tryptophan residues, as in the initial independent docking (Figure 8). The two poses are separated by a wall of tryptophans, which do not interact directly with either pose. Site 1 is a trough formed by the bottom loop side chains of hydrophilic Asn, Gln, and Thr residues (Figure 8). The bottom of the trough is formed by the backbones of Asn, Gln, and Gly residues (Figure 8). Site 2 is formed by electrostatic side chains of Lys and Asp, hydrophilic Ser, and backbones of Lys and Ser residues (Figure 8). The complete list of interacting residues is given in Table S4.

FIGURE 8.

Model of CLRR4 with two MDPs bound. The protein backbone is shown as a ribbon, colored as internal repeats (teal) and β‐strand capping repeat (dark blue), with Trp residues (brick red) forming a hydrophobic patch between the two binding sites. Pose 2′ (pink) was docked to the complex with pose 1 (orange) already fixed to the homology model of CLRR4. The side chains of residues interacting with each MDP are shown (teal and dark blue)

When each pose was fixed in its binding site first and a second MDP was docked, the second MDP with the lowest energy bound in the empty site, not in a new site (Figures S11–S13). The docking of a second MDP to an MDP‐CLRR4 complex indicates that binding of MDP in pose 1 allows the binding of MDP in pose 2′, which is in the same binding site as the independently docked pose 2 (Figure 8). The affinity of pose 2′ (−5.1 kcal/mol) is slightly lower than independently bound pose 2 (Figure S11). Conversely, the binding of MDP in pose 2 allows the binding of MDP in pose 1′, and the affinity of pose 1′ (−6.3 kcal/mol) is roughly the same as the independently bound pose 1 (Figure S12). The molecular modeling results indicate two MDP binding sites, without providing insight into whether binding is cooperative or independent. The interaction interfaces found provide a theoretical model for the binding modes of MDP to the CLRRs. These results are in accord with the modeling of MDP binding to the LRR region of full‐length native NOD2 in previous work. ¹⁶ , ²⁶

While our model is based on a different template NOD2 structure and is sequentially different than that by Lauro et al., ²⁶ we also observe MDP preference for binding in the β‐strand rich concave region. This work uses similar molecular docking techniques, and the results indicate favorable binding and key interactions with physiochemically similar residues as reported by Lauro et al. ²⁶ We expand on this by docking two ligands to determine sequential and dual binding, while also exploring the entire structure as potential binding regions for MDP and analyzing all pose positions and interactions. Our model provides a foundation for future redesign of our scaffolds to bind other ligands.

3. CONCLUSIONS

Consensus design of the CLRRs generated a series of stable, folded proteins ranging in length from 4‐ to 8‐repeats. To compare MDP binding with previous work, we fit the data to a single binding isotherm, K _{d app}. The longest repeat protein receptor, CLRR8, had an overall binding affinity K _{d app} of 1.0 μM for MDP, a bacterial cell wall glycopeptide fragment. This represents a 2‐fold increase in binding affinity over our previous designed CLRR2. ¹⁶ Biphasic analysis of the fluorescence binding data showed that the shortest repeat protein receptor, CLRR4, had a 40 ± 10 nM affinity for binding one of two MDP ligands. This is close to the tightest binding interaction between a NOD2‐based protein and MDP. ²⁵ Subsequent molecular docking to a homology model showed two very distinct binding sites for the two MDP ligands bound to CLRR4, and the two MDP ligands bound CLRR4 with different independent affinities. The affinities are also distinct when docking a second MDP after one MDP is already bound. This series of CLRRs bind MDP with high capacity.

The nature of repeat proteins allowed us to design this scaffold efficiently, while maintaining the stability and affinity of their larger protein counterparts. Increased stability of designed proteins facilitates the development of a biosensor. Because repeat proteins are modular and present a large surface area for binding interactions, this protein lays the foundations for a scaffold with the potential for specificity towards a variety of ligands. Different repeat lengths, number of repeats, secondary structure, and modification of individual residues allow for the future design of diversity in ligand specificity. ⁸ , ⁹ , ¹⁰ The protein scaffold can be engineered with substitutions at hypervariable positions, and different repeat lengths to tailor a specific binding interaction. Together, these qualities make our scaffold attractive for designing molecular recognition motifs.

4. MATERIALS AND METHODS

4.1. Consensus design

The six genes known to encode for the NOD subfamily of NLR proteins in humans were obtained from the HGNC database. ³⁵ The gene sequence was translated to its corresponding protein sequence and input into the NCBI protein basic local alignment search tool (BLAST) using the blastp algorithm. ³⁶ For each translated sequence, mammalian sequences encoding for the same protein were selected, excluding synthetic and predicted sequences. The resulting 75 protein sequences (Figure 3.1, Table S1) were manually aligned with the canonical motif of xLxxLxLxxNxL(x)_nL(x)₈. ²² From this alignment, the three most prevalent amino acids at each position were counted in Excel (Figure 3c). Amino acids at each position were further selected for conservation of physical properties: size, hydrophobicity, polarity, charge, and the overall pI to obtain the final consensus sequence.

4.2. DNA synthesis and cloning

The sequence encoding for one internal repeat and the capping sequence containing the restriction sites: BamHI, BglII, and HindIII, was designed using Klenow technology with four overlapping oligomers (Figure S1). Oligonucleotides were obtained from Integrated DNA Technologies, Inc. (Figure S2). Restriction sites were added to the consensus sequence to facilitate cloning multiple repeats. The final proteins include Gly‐Ser at the beginning of the first repeat from the BamHI restriction site, and Arg‐Ser between internal repeats from the BglII restriction site (Figure S3). A HindIII site was added to the DNA sequence after the C‐terminal capping repeat (Figure S3). The sequence was spliced into plasmid pProExHtam by ligation at the restriction sites for BamHI and HindIII (Figure S3). The single‐repeat DNA sequence was changed from Ala (A) to Glu (E) at position 5 in the consensus sequence by PCR using the Q5® Site‐Directed Mutagenesis Kit (New England Biolabs) (Figure S3). Longer variations of the protein were synthesized by digesting and ligating shorter coding sequences expressing designed proteins to form longer coding sequences expressing larger proteins (Figure S3). The identities of each of the sequences constructed were confirmed by Sanger sequencing (Virginia Tech Bioinformatics Institute).

4.3. Protein expression and purification

All proteins were expressed in BL21 (DE3) E. coli cells. Competent cells were purchased (New England Biolabs) and transformed with the CLRRx‐pProExHtam plasmid (1 μl in 50 μl cells) by electroporation (1.7 kV, 1 pulse, Bio‐Rad MicroPulser) at ambient temperature. Cell cultures were grown for 16 hr overnight at 37°C in Luria broth (LB) media (50 ml), with shaking at 180 rpm. The resulting cells were diluted 1:100 in fresh LB media (1 L) and grown under the same conditions until optical density 0.6 was reached, about 3 hr. Then protein expression was induced with 1 M isopropyl β‐D‐1‐thiogalactopyranoside (IPTG, 0.5 ml) and allowed to express for 4 hr. Cells were harvested by centrifugation at 5000 rpm for 15 min at 4°C. Pelleted cells were resuspended in lysis buffer (50 mM Na₂HPO₄, 300 mM NaCl, 0.1% Tween 20, pH 8.0), and stored at −80°C. For purification, cell pellets were thawed in a 37°C water bath for 10 min, and the cells were lysed using a microtip and Mison sonicator at 30% amplitude for 30 s in 10 s pulse intervals. Protease inhibitor (1/4th of a tablet, Pierce™ Protease Inhibitor Mini Tablets) and DNase I (1 μl, RNAase‐free, New England Biolabs) were added and the lysed cells were incubated on ice for 30 min. Lysed cells were centrifuged at 16,000 rpm for 30 min, and the supernatant was collected. The collected protein supernatant was incubated on Ni‐NTA beads (500 μl) for an hour, washed with 2 × 12 ml of 30 mM imidazole wash buffer (50 mM Na₂HPO₄, 300 mM NaCl, 0.01% Tween 20, pH 8.0), and eluted using 300 mM imidazole in lysis buffer (6 ml). The soluble fraction contained 4–13 mg protein per 1 L of culture after purification.

4.4. Gel electrophoresis

Crude cell extracts were prepared in a 1:1 vol/vol ratio of SDS buffer to protein, and analyzed on a 15% SDS‐PAGE with the 212 to 6.5 kDa protein marker from New England Biolabs (Figure S4). Protein bands were visualized by Coomassie blue.

4.5. Size‐exclusion chromatography

Proteins were further purified on an S75 Superdex 16/600 prep grade column on an ÄKTAprime plus chromatography system (Figure S4). Proteins were filtered through a 0.22 μm filter prior to injection of 5 ml of Ni‐NTA eluent. The mobile phase was 50 mM Na₂HPO₄, 150 mM NaCl buffer pH 8.0 (PBS) at a flow rate of 1.0 ml/min. Pure proteins (100 μl) were injected on an S75 Superdex 10/300 analytical grade column with PBS at a flow rate of 0.4 ml/min to determine the purity and molecular weight of the proteins. Biorad gel filtration standards (1,350 Da to 70,000 Da) were used to generate a calibration curve. The sizes of purified proteins were estimated from the retention times (Table S2). Each analytical SEC showed a single peak, and the purity was estimated to be >95% for each protein.

4.6. Dialysis

Protein fractions collected from the preparative size‐exclusion column were dialyzed serially to complete proper folding. Proteins were dialyzed (3,000 MW cut‐off) at 1 hr intervals, in 1 L each of decreasing urea concentrations (8.0 M, 6.0 M, 4.0 M, 2.0 M, and 0 M) with the last step done overnight for equilibration. Dialyzed samples were stored at −4°C for up to 3 weeks until used.

4.7. Mass spectrometry

Protein molecular weights of CLRR7 and CLRR8 were confirmed using mass spectrometry. Protein samples were diluted to 1 μM and precipitated with 4 portions of acetone (total volume ca. 1 ml) at −20°C. Samples were vortexed and incubated for 10 min, then centrifuged at 13,000 rpm for 10 min. The pellet was resuspended in acetonitrile:water (1:1) and filtered through a C₁₈ resin ZipTip (Milipore Sigma). This protein sample was mixed in a 1:1 ratio with α‐cyano‐4‐hydroxycinnamic acid matrix and analyzed on the ABSciex 4,800 MALDI‐TOF/TOF mass spectrometry system (Figure S5).

4.8. Circular dichroism

CD spectra of protein samples (5–10 μM in PBS) were obtained using a Jasco J‐815 CD Spectrophotometer. Far‐UV (190–250 nm) spectra were recorded at 24°C, with 1 nm bandwidth, 2 nm data pitch, and a data integration time of 1 s (Figure 4b, Figure S6). The average of three accumulations was averaged to give mean residue ellipticity, [θ]_mrw, for each sample. For CD thermal melting curves, samples were heated from 20–90°C in 2°C increments with 5 min equilibration time before each measurement (Figure 4c). Molar ellipticity (θ) was monitored at α‐helix (208 nm) and β‐sheet (ca. 217 nm) minima for each protein. We used 208 nm instead of 220 nm because it is more sensitive for the α‐helix, and more intense in the folded proteins. The signal was converted into fraction unfolded (Equation 1):

$$\text{Fraction unfolded}=\frac{\mathit{\theta }-{\mathit{\theta }}_{\mathit{N}}}{{\mathit{\theta }}_{\mathit{D}}-{\mathit{\theta }}_{\mathit{N}}}$$ (1)

where θ was measured at each temperature, θ _N was measured at 20°C, and θ _D was measured at the highest temperature measured, where protein is fully unfolded. The denaturation curve was generated using the Boltzman equation, by plotting the fraction unfolded versus temperature and fit to Equation (2):

$$y=\frac{1.0-0.0}{1+e^{\left(T-T_{\mathrm{m}}\right)/\mathit{dT}}}+1.0$$ (2)

where 0.0 is the initial fraction unfolded, 1.0 is the final fraction unfolded, T is the fraction unfolded at a given temperature, and T _m is the temperature at the halfway point (y = 0.5).

4.9. Fluorescence quenching

Protein samples (8 μM in PBS) were incubated with MDP at concentrations ranging from 0.0 to 100 mM in a total volume of 600 μl. The concentrations and fluorescence for each titration are given in Table S3. Duplicate samples were incubated in a 96‐well plate for 30 min prior to collecting fluorescence spectra (Varian Cary Eclipse Fluorescence Spectrophotometer). Samples were excited at 295 nm with a slit width of 10 nm, and emission was measured from 310 to 380 nm with a slit width of 20 nm. The fluorescence signal was converted into ligand fraction bound (Equation (3)):

$$r=\frac{F_0-F}{F_0-F_{\min }}$$ (3)

where r is equal to the fraction of bound ligand, F ₀ is the fluorescence signal without MDP, F is the signal at a given ligand concentration, and F _min is the fluorescence signal at saturation. The binding curve was generated by plotting the fraction bound versus MDP concentration, and fitted to a Langmuir isotherm equation collectively for the K _dapp, and individually for K _d1 and K _d2 (Equation (4)):

$$r=\frac{r_{\mathrm{b}}·\left[L\right]}{K_{\mathrm{d}}+\left[L\right]}+c$$ (4)

where r is the measured fraction of CLRRx‐MDP complex by fluorescence, r _b is a constant. K _d is the dissociation constant (K _{d app}, K _d1, or K _d2) and [L] is the concentration of MDP. ⁵³ For K _{d app} r _b = 2.0 and c = 0.0, for K _d1 r _b = 1.0 and c = 0.0, and to fit the second binding phase for K _d2 r _b = 1.0 and c = 1.0.

4.10. Homology modeling of CLRR4

A homology model of CLRR4 was generated with Schrödinger‐Maestro v. 2019–3 ⁵⁴ using PDB ID: 5IRM ⁴¹ as a template. PDB ID: 5IRM is an X‐ray crystal structure of the rabbit NOD2 protein, which is a member of the NOD‐like receptor family. ⁴¹ The sequence of CLRR4 had 49% identity, 63% similarity, and an E‐score of 6E‐25 to the PDB ID: 5IRM template structure (Figure S7). After generation of the CLRR4 homology model, it was energy minimized using Schrödinger‐Maestro v. 2019–3 Prime, with default parameters and the OPLS3e force field. ⁵⁴ The energy minimized CLRR4 model was then validated for structural parameters: backbone Φ and Ψ angles, comparison to experimentally solved protein structures, and favorable free energy ⁵⁵ using the Swiss Model Structure Assessment server (Figure 5). ⁵⁶ Structure validation metrics indicate an energetically favorable structure that was used in subsequent molecular docking experiments.

4.11. Molecular docking of MDP to CLRR4

Molecular docking of MDP (Figure 1) to the CLRR4 energy‐minimized structure was performed. The MDP structure was built in ChemDraw and imported into Schrödinger‐Maestro v. 2019–3 ⁵⁴ in order to correct for stereochemistry and saved as a .pdb file. Autodock Tools (v. 1.5.6) ⁵⁷ was utilized to preprocess and create pdbqt files for CLRR4 and MDP for molecular docking. Autodock Vina ⁵⁸ was used to dock MDP to the CLRR4 structure, and to dock a second MDP to the CLRR4 structure with MDP poses previously bound. Grid box size (40 Å × 48 Å × 40 Å) and center (38.04, 34.596, 7.679), with a 1.000 Å grid space, was selected to encompass the entire CLRR4 structure and bound ligands in the docking search space. The same grid box size and center was used in all docking experiments. Nine poses were created per docking run. Initial docking to CLRR4 (protein‐only) revealed two areas of clustering of MDP. Pose 1 and pose 2, the lowest energy poses in each dominant cluster observed in the above CLRR4 protein‐only docking, were selected for subsequent docking of an additional MDP ligand. The pose (pose 1 or pose 2) and CLRR4 structure were considered the “receptor,” and an additional MDP ligand was docked to that protein‐MDP structure in order to observe a second ligand docking. Schrödinger‐Maestro's Interaction Fingerprints tool ⁵⁴ was used to analyze and categorize CLRR4‐MDP interactions. Script files and input files for molecular docking may be accessed on our public Open Science Framework page at https://osf.io/82n73/.

AUTHOR CONTRIBUTIONS

Christina S. Kim: Formal analysis; investigation; methodology; validation; visualization; writing‐original draft; writing‐review & editing. Anne M. Brown: Funding acquisition; investigation; project administration; software; validation; visualization; writing‐review & editing. Tijana Z. Grove: Conceptualization; formal analysis; funding acquisition; project administration; resources; supervision; writing‐review & editing. Felicia A. Etzkorn: Formal analysis; methodology; project administration; resources; supervision; writing‐original draft; writing‐review & editing.

Supporting information

Appendix S1: Supporting Information. NCBI reference sequences and multiple sequence alignment (Kim_LRR_ProtSci_MSA.xlsx), detailed DNA synthesis and cloning schemes, primer sequences, representative SDS‐PAGE and SEC, calibration table for SEC, MALDI‐TOF of 4CLRR7 and 4CLRR8, denatured and unpurified protein CD spectra, scores for homology modeling, figure of residues stabilizing α‐helixes, individual unfolding curves, additional docking figures and interacting residues table (Kim_LRR_ProtSci_SI.pdf). Model files and docking parameters can be found on our Open Science Framework Page (https://osf.io/82n73/).

Kim CS, Brown AM, Grove TZ, Etzkorn FA. Designed leucine‐rich repeat proteins bind two muramyl dipeptide ligands. Protein Science. 2021;30:804–817. 10.1002/pro.4031