Identifying amino acids in CNTN3 that are required for APP binding

In an initial experiment, we compared the binding of the full APP ECD and the APP E1 domain, expressed as AP fusions, to each of the six CNTN ECDs, without their GPI anchor motifs, expressed as Fc fusions (Fig 1B). In general agreement with Osterfield et al. [22], both APP probes bound with similar efficiencies, and the relative affinities for the different CNTNs was: CNTN3 ≃ CNTN4 > CNTN5 >> CNTN1, CNTN2, and CNTN6. The one discrepancy with Osterfeld et al. [22] is with CNTN5: Fig 1B shows weak binding whereas Osterfeld et al. [22] saw no binding. Based on this experiment, we focused on CNTN3 for in-depth structure-function studies and we used the APP-E1-AP probe to assess binding to the CNTN3 derivatives.

Deletion of the six N-terminal Ig domains of CNTN3 had no effect on APP-E1-AP binding, thereby localizing the binding interface to the four Fn domains, in agreement with Osterfield et al. [22] (Fig 1C). Deletion of different combinations of Fn domains (in the absence of the Ig domains) revealed an essential role for Fn(2), with Fn(1) providing a several-fold enhancement of the binding signal (Fig 1C). Fn(3) and Fn(4) were found to be dispensable for binding. The signal intensities in this and subsequent binding assays are summarized in the figures and in S1 Table, where -, +/-, +, ++, +++, and ++++ represent the range of intensities from undetectable (-) to maximal (++++). Based on this analysis, alanine substitution mutagenesis experiments were conducted with a target Fc fusion protein containing only the first two Fn domains of CNTN3 (CNTN3-Fn(1–2)-Fc).

To define the amino acids within CNTN3 Fn(2) that mediate APP binding, we designed two rounds of alanine substitution mutagenesis guided by the crystal structure of Fn(2) PDB ID 5I99 [25]. In the first round of mutagenesis, 22 clusters of 1–4 amino acids each were substituted with alanine to narrow down the region(s) of interest. These amino acids were chosen based on their surface location in the Fn(2) crystal structure. Among the 22 mutant Fn(1–2)-Fc proteins, ten were expressed with yields roughly comparable to wild type (WT) Fn(1–2)-Fc (Fig 1F). These ten proteins are assumed to be correctly folded, and they are labeled alanine substitution (AS)1-10 in Fig 1D, upper sequences. AS2, AS6, AS7, AS8, and AS9 exhibited greatly reduced binding to APP-E1-AP, suggesting a role for one or more amino acid side chains in these clusters (Fig 1E, left panel and S1A Fig).

In the second round of mutagenesis, each of the 13 amino acids that were mutated in CNTN3-Fn(1–2)-Fc AS2, AS6, AS7, AS8, and AS9 was individually mutated to alanine (Fig 1D, lower sequences). Twelve of the 13 individual alanine substitution mutants were secreted with yields roughly comparable to WT (Fig 1F). E716A was produced with a poor yield and was not further studied. Among the 12 single alanine mutants, V752A showed undetectable binding, and R714A, R760A, and R764A showed greatly reduced binding (Fig 1E, right panel and S1B Fig), indicating an important role for each of these four amino acids in APP binding. It is interesting that three of these four amino acids are arginine.

Identifying amino acids in APP that are required for CNTN3 binding

We applied a similar strategy to localize the CNTN3 binding site within APP E1. We first tested the binding properties of the two subdomains of E1 –GFLD and CuBD–and observed that CuBD alone conferred binding to CNTN3 with essentially no contribution from GFLD (Fig 2A). APP CuBD contains a Type 2 non-blue copper center in which His147, His151, and Tyr168 bind the central copper ion [26].

To further localize the amino acids within APP CuBD responsible for CNTN3 binding, nine clusters of 1–5 alanine substitution mutations were constructed in nearby surface residues in the APP-CuBD-AP fusion protein, based on the CuBD and E1 crystal structures PDB ID 2FKL and PDB ID 3KTN [26, 27] (Fig 2B AS 1–9, upper sequences). All nine variants were secreted with yields roughly comparable to WT (Fig 2D). In the binding assays with WT CNTN3-Fn(1–2)-Fc, APP-CuBD-AP mutants AS2 and AS8 showed no binding, AS1 and AS3 showed strongly reduced binding, and AS4 showed modestly reduced binding (Fig 2C, left panel and S1C Fig). Interestingly, mutating His147 (in AS5), which is one of the copper-liganding residues, did not affect binding.

In the second round of mutagenesis, each of the 12 amino acids that were mutated in APP-CuBD-AP AS1, AS2, AS3, AS4, and AS8 was individually mutated to alanine (Fig 2B, lower sequences). All 12 mutants were expressed with yields roughly comparable to WT (Fig 2D), and, among these, H137A and E183A showed undetectable binding, and K132A, Q138A, E139A, R140A, M141A, K178A, and R180A showed a several fold reduction in binding (Fig 2C, right two panels and S1D and S1E Fig).

Spatial locations of amino acids that are essential for APP-CNTN3 binding

The locations of the alanine substitution mutations were mapped onto ribbon diagrams of CNTN3 Fn(2) and APP CuBD, color-coded by their effect on binding (Fig 3). The four CNTN3 Fn(2) substitutions that most severely disrupt binding reside on one face of the Fn(2) domain (Fig 3A). Similarly, the two APP CuBD substitutions that most severely disrupt binding reside on adjacent strands of a beta-sheet on one face of the CuBD (Fig 3B). Five of the six APP CuBD substitutions with several-fold lower binding efficiency (Fig 3B, color-coded blue) reside adjacent to these two critical positions, and the sixth resides further away on the same face of the CuBD. As the mutations with the greatest effect on CNTN3 Fn(2) and APP CuBD binding reside on one face of each of these domains, the most parsimonious explanation for the mutagenesis data is that these two faces interact.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Amino acids that are critical for APP-CNTN3 binding displayed on the 3-dimensional structures of CNTN3-Fn(2) and APP-CuBD.

(A) Locations of alanine substitutions in CNTN3-Fn(2) PDB ID 5I99 [25], color coded by their effect on APP-E1-AP binding. (B) Locations of alanine substitutions in APP CuBD PDB ID 2FKL [26], color coded by their effect on CNTN-Fn(1–2)-Fc binding. The cylinders and arrows show the 90 degree rotations that relate the images in adjacent panels. Only data from CNTN-Fn(1–2)-Fc or APP-CuBD-AP fusion proteins that were secreted well are shown. In the leftmost panels, side chains are labeled for those amino acid that are most critical for binding, i.e., those for which alanine substitution severely impairs binding (categories: +/- (magenta) and–(red)).

https://doi.org/10.1371/journal.pone.0219384.g003

Constructs

Fc and AP fusions were cloned into the pRK5 vector for expression in HEK/293T cells [37]. For the Fc-fusion proteins, the open reading frame consisted of (i) the native signal peptide or the mouse Frizzled8 signal peptide, (ii) the full ECD or the subdomain of interest, (iii) a 22 amino acid linker corresponding to the region immediately distal to the cysteine-rich region of mouse Frizzled8, and (iv) the Fc region of human IgG1 starting from the hinge [38]. For the AP-fusion proteins, the open reading frame consisted of: (i) the native signal peptide or the mouse Frizzled8 signal peptide, (ii) the full ECD or the subdomain of interest, (iii) a glycine/serine linker followed by a myc epitope tag, and (iv) the coding sequence of human placental alkaline phosphatase (PLVAP) lacking the GPI-anchoring sequences [39].

CNTN and APP inserts

Full-length CNTN2, CNTN4.1, CNTN5, and APP ECDs were amplified from a mixture of prenatal mouse embryo and neonatal mouse brain cDNA by nested PCR. Mouse CNTN1, CNTN3, and CNTN6 ECD-Fc constructs were a generous gift of Dr. Davide Comoletti. CNTN3 and APP domain deletions and alanine substitutions were generated by PCR using the cloned ECD sequences as templates.

Cell culture and transfection

HEK/293T cells (ATCC CRL-11268) were transiently transfected with polyethylenimine (PEI). For small-scale transfections, HEK293T cells were grown in 12-well plates. 24 hours after transfection, wells were washed twice with 1 ml serum-free medium (SFM), and then cultured for an additional 24 hours in 0.5 ml SFM. The serum-free conditioned medium (SFCM) was centrifuged at 5000xg for 5 minutes, and 20 ul of the supernatant was used for immunoblotting. Plasmids that produced ~1 ug/ml protein in the SFCM were used for large-scale transfection. For large-scale transfections, HEK/293T cells were plated in 10 cm dishes and processed as described above, except that (i) washes were performed three times with 5 mls SFM, (ii) 5 mls SFCM was generated per plate, and (iii) SFCM production was continued with a fresh 5 ml of SFM for an additional day, if more protein was needed. SFCM was filtered through a 0.2 um filter, and stored at 4°C with sodium azide added to 0.01%.

Binding assays

For estimating Fc fusion protein concentrations, 1 ml of SFCM was captured on Protein-G sepharose beads (Thermo-Fisher), and analyzed by SDS-PAGE and Coomassie Blue staining. Protein concentrations were compared to a bovine serum albumin (BSA) dilution series. Relative protein concentrations for WT and mutant Fc- and AP-fusion proteins were determined using the LI-COR fluorescent immunoblotting system (LI-COR Biosciences), and the volumes of SFCM adjusted accordingly to equalize the concentrations across WT and mutant versions in the binding assays. For binding assay, the wells of a Protein-G coated 96-well plate (Thermo-Fisher Scientific/Pierce) were washed twice with PBS, 100 μl of volume-adjusted conditioned medium containing each Fc-fusion proteins was added per well, and then the plate was rotated horizontally for 1.5–2 hours at room temperature. The medium was removed and the wells washed 5 times in PBS with 0.5% Triton X-100, 0.1% bovine serum albumin (PBSTA). 100 μl of volume-adjusted conditioned medium containing the AP-fusion protein was added per well, and the plate was rotated for 1.5–2 hours at room temperature, and then washed 5 times with PBSTA. Five minutes before use, a 1:1 mixture of the BluePhos Microwell Substrate (Kirkegaard and Perry Laboratories) was prepared and 100 μl was added to each well. Absorbance at 620 nm was recorded at 5 minutes intervals at room temperature using a SpectraMax M3 Microplate Reader (Molecular Devices) for 60–90 minutes. Plates were photographed after the incubation was complete.