Structure-Function Analysis of the Non-Muscle Myosin Light Chain Kinase (nmMLCK) Isoform by NMR Spectroscopy and Molecular Modeling: Influence of MYLK Variants

Abstract

The MYLK gene encodes the multifunctional enzyme, myosin light chain kinase (MLCK), involved in isoform-specific non-muscle and smooth muscle contraction and regulation of vascular permeability during inflammation. Three MYLK SNPs (P21H, S147P, V261A) alter the N-terminal amino acid sequence of the non-muscle isoform of MLCK (nmMLCK) and are highly associated with susceptibility to acute lung injury (ALI) and asthma, especially in individuals of African descent. To understand the functional effects of SNP associations, we examined the N-terminal segments of nmMLCK by 1H-15N heteronuclear single quantum correlation (HSQC) spectroscopy, a 2-D NMR technique, and by in silico molecular modeling. Both NMR analysis and molecular modeling indicated SNP localization to loops that connect the immunoglobulin-like domains of nmMLCK, consistent with minimal structural changes evoked by these SNPs. Molecular modeling analysis identified protein-protein interaction motifs adversely affected by these MYLK SNPs including binding by the scaffold protein 14-3-3, results confirmed by immunoprecipitation and western blot studies. These structure-function studies suggest novel mechanisms for nmMLCK regulation, which may confirm MYLK as a candidate gene in inflammatory lung disease and advance knowledge of the genetic underpinning of lung-related health disparities.

Fig 1. Selection of the N-terminal segments of nmMLCK1.

The segment of 1-494aa was initially selected for protein expression within the N-terminal sequence of nmMLCK1, containing three ALI-associated SNPs and two phosphorylatable Tyr sites, Y464 and Y471. This sequence generates a protein of ca. 53 kDa, within a suitable range of size for practical bacterial protein expression and survived preliminary NMR trials. Included in this ca. 500aa protein are three immunoglobulin C-2 type (IGc2) domains and a low-complexity region (preceding the 3rd IGc2 domain) as predicted by SMART. Subsequently, a shorter 1-264aa segment of ca. 28 kDa was also generated spanning the three ALI-associated SNPs (two IGc2 domains) and exhibited advantages for NMR-based structural determination.

Fig 2. SDS-PAGE characterization of ¹⁵N-labeled N-terminal segments of nmMLCK1.

(A) Five ¹⁵N-labeled 1-494aa proteins (ca. 3 μg each). (B) Two ¹⁵N-labeled 1-264aa proteins (ca. 3 μg each).

Fig 3. Comparison of the HSQC spectra of ¹⁵N-labeled 1-494aa segments of nmMLCK1.

(A) Wild type and P147; (B) wild type and H21-P147; (C) wild type and H21-P147-A261; (D) wild type and A261; (E) H21-P147 and H21-P147-A261; (F) H21-P147 and P147. By superimposing the spectra of the SNP variants (the 2nd set of spectra, in red color) onto wild type or other SNP variants (the 1st set of spectra, in blue color), we have identified characteristic changes in the HSQC signals corresponding to each individual SNP mutation versus the wild type, which are indicated along red (S147P), pink (P21H) and brown (V261A) lines, respectively (see insets for more details). The characteristic patterns associated with these three SNPs suggest that these SNP mutations are indeed distant to each other in the tertiary structure and therefore cause only minor or minimal local conformational changes and independent changes in the HSQC signals.

Fig 4. Comparison of the HSQC spectra of ¹⁵N-labeled 1-264aa/ 1-494aa segments of nmMLCK1.

(A) and (B), the HSQC spectra of 1-264aa segments of nmMLCK1: (A) wild type; (B) P147 SNP variant. (C) to (F), the superimposition of the HSQC spectra (the 1st set of spectra shown in blue color and the 2nd in red color): (C) wild type, 1-494aa and 1-264aa; (D) P147, 1-494aa and 1-264aa; (E) 1-264aa, wild type and 147P; (F) zoom in of the squared region in (E). The HSQC spectra of the 1-264aa segments exhibited a better dispersion of ¹H-¹⁵N chemical shifts, with less degenerate, better resolved signals than 1-494aa segments. Superimposition of the spectra of the 1-264aa segments onto those of their corresponding 1-494aa segments demonstrated that the spectra of 1-264aa segments are a recapitulation of the subsets of those of their corresponding 1-494aa segments, suggesting that the shorter segments each possess a structure similar to the corresponding part of their longer counterparts. The same characteristic pattern of signal changes observed for 1-494aa segments is recapitulated with better resolution by superimposition of the spectra of the 1-264aa 147P SNP mutant onto those of the 1-264aa wild type segment, suggesting that the same structural difference exist for the 1-264aa wild type and the P147 SNP mutant.

Fig 5. Local conformational changes in the 31-253aa segment of nmMLCK1 [template: titin Z1Z2 (PDB ID: 1ya5A)] upon S147P SNP mutation and energy minimization.

(A) The S147P SNP site localized in a potentially flexible loop connecting the 1st IGc2 domain and the 2nd IGc2 domain (with side chains of some neighboring residues shown in ball and stick model). (B) to (D), a closer look at the loop connecting the 1st IGc2 domain and the 2nd IGc2 domain after energy minimization (131-154aa shown in ball and stick model): (B) the loop in the wild type; (C) the loop in the P147 SNP variant; (D) overlay of the two loops (structural changes caused by the P147 SNP mutation shown in pink).

Fig 6. Potential involvement of S147P and P21H SNP sites in nmMLCK1 in phosphorylation-dependent binding of 14-3-3 proteins.

(A) Sequence alignment of human nmMLCK1 (wild type and SNPs) with the consensus of 14-3-3 binding modes and selected protein kinase substrates as well as its murine ortholog. The “S” or “T” in red indicates a phophorylation site, with underline indicating a predicted site (black underlined) or a confirmed site (red underlined). The “R” or “K” in blue and the “P” in green indicate their potential involvement in key binding recognition. The “P” indicates a “P” that may be missing and hence nonessential. The “X” in black indicates any amino acid. The “X” indicates an “X” that may be missing and hence nonessential. The consensus of GSK3 substrate (SXXXS, the 2nd S representing a pre-existing pSer or pThr) is shown in repeat in order to align with multiple potential phosphorylation sites in nmMLCK1. While not shown, additional multiple alignments of S147, S18, and other nearby Ser residues including S16, S26, S145 and S154, are possible with the highly variable consensus substrate RXXXS of the AGC group of protein kinases that include PKA, PKG and PKC families, suggesting a complex regulation of nmMLCK via these SNP-embracing loops by different kinase-mediated phosphorylation and subsequent binding to 14-3-3 proteins. (B) Molecular modeling of the 1-252aa segment of nmMLCK1 [template: deleted in colorectal cancer (DCC) (PDBID: 3lafA)] revealing localization of S147P and P21H SNP sites in separate loops at the two ends of the single, 1st IGc2 domain (with side chains of some loop residues of interest shown), despite that the modeling of loop conformations may be of poor quality. (C) Immunoprecipitation (IP) of Flag-tagged nmMLCK1 wild type using Flag-M5 anitibody followed by western blot using pan-14-3-3 antibody indicating the binding of 14-3-3 proteins to nmMLCK1 before and after S1P stimulation. Note: In the IP result shown, the bottom 14-3-3 bands correspond to 24 kDa and the top 14-3-3 bands correspond to 27 kDa.

Fig 7. Molecular modeling of the N-terminal segments of nmMLCK1 (30-813aa) and nmMLCK2 (30-744aa) [template: titin I65-I70 (3b43A)].

(A) nmMLCK1 (green); (B) nmMLCK2 (red); (C) superimposition of nmMLCK1 (green) and nmMLCK2 (red). The positions of V261 (V261’ in nmMLCK2), Y464 and Y471 (in nmMLCK1 only) are indicated as appropriate. V261 in nmMLCK1 localizes immediately preceding the 1st alpha-helix of an alpha-beta fold which contains multiple Pro-rich motifs that may be implicated in binding to partner proteins containing SH3 domains. V261’ in nmMLCK2 localizes within the 1st beta-strand following the 2nd IGc2 domain. Y464 and Y471 are exposed at the back surface of the structures shown for nmMLCK1. A well-folded, the 2nd IGc2 domain-interacting, alpha-beta structure (ca. 253-405aa) that embraces the V261A SNP site and multiple Pro-rich loops, together with the 3rd IGc2 domain (ca. 405-505aa), in nmMLCK1, is replaced by a relatively loosely-folded structure composed of less helixes, more beta strands and more accessible Pro-rich loops (ca. 253-440aa) in nmMLCK2.