Cholesterol recognition motifs in the transmembrane domain of the tyrosine kinase receptor family: the case for TRKB

Cholesterol is an essential constituent of cell membranes. Recently, the discovery of cholesterol recognition aminoacid consensus (CRAC) on proteins indicated a putative direct, non-covalent interaction between cholesterol and proteins. In the present study, we evaluated the incidence of CRAC motifs, and its inverted version (CARC), in the transmembrane region (TMR) of tyrosine kinase receptor family (RTK) in Caenorhabditis elegans (nematode); Drosophila melanogaster (fruit fly); Danio rerio (zebrafish); Mus musculus (mouse), and Homo sapiens (human) by using in silico methods. We found CRAC motifs across all species analyzed, whereas, CARC motifs, comprising the extracellular side of the TMR, were only found in vertebrates. On top of its contribution to structural properties of cells, cholesterol (or its’ metabolites) regulates a myriad of cellular functions culminating in increased neuronal plasticity in the central nervous system. The tropomyosin-related kinase B (TRKB), a member of RTK family, is also a core participant in neuronal plasticity process. Therefore, focusing on TRKB receptor, we observed high homology in TRKB.TMR and conservation of CARC motifs across multiple organisms. Upon the recognition of conserved CARC motif in the TRKB, we compared the effect of point mutations in CARC on structural changes in the TMR of human TRKB. The alignment of wild-type and mutant TMR indicates small changes across the 5 mutations analyzed (Y434F, Y434C, Y434A, V439K and R428A), indicated by the root-mean-squared deviation (RMSD) values for the superimposed structures, average=2.448Å. The TMR point mutation also impacted the physicochemical properties of the models, measured by solvent accessibility. The mutation of the Y, V or R residues in the CARC motif increased the solvent access in the juxtamembrane portions of the TMR, ranging from 150-500%. In conclusion, it is plausible to consider a potential role for CARC/CRAC motifs in the function of RTK family TMR. The presently described CARC motif in TRKB.TMR may exert its effect by regulating the receptor capability to interact with the polar heads of membrane phospholipids.


INTRODUCTION
The human brain contains 23% of the body's total cholesterol, mostly in the myelin sheath of oligodendrocytes [1,2] . The blood brain barrier prevents lipoprotein or cholesterol transport to the brain, therefore local de novo synthesis takes place. In mouse brain, cholesterol synthesis peaks during the second postnatal week and then decreases significantly, without being influenced by sex or blood cholesterol concentration [3,4] . During early development, neurons produce cholesterol autonomously [4][5][6] . In later stages, cholesterol is synthesized by glial cells. However, it is not defined whether it is a constant or regulated production [4,7] .
Cholesterol, which can be localized on both leaflets of the plasma membrane [8] , induces changes in the membrane physical properties such as fluidity [9] and curvature [10] or interacts with transmembrane domain to regulate protein function [11,12] . Cholesterol is a core constituent of micro-domains known as lipid rafts, which serve as signaling platforms for several pathways [13][14][15] . In the nervous system, cholesterol interaction with membrane proteins influences several crucial events such as exocytosis of synaptic vesicles [16] , synaptic activity, connectivity, plasticity, signal transduction, transmission, and cell survival [17][18][19] . The neurotrophic receptor subfamily is one of the most prominent representatives of tyrosine kinase receptor family (RTK), given its crucial role in regulating neural plasticity [20] . The tropomyosin-related kinase receptors (TRK) comprise three members TRKA, TRKB and TRKC, which are phosphorylated in different tyrosine residues in the intracellular portion upon activation by their high-affinity ligands: NGF, BDNF and Thus, the aim of the present study was to evaluate the incidence of cholesterol interacting motifs (CRAC and CARC) in the RTK family. Given the promiscuous nature of CRAC motifs, we focused on the RTK TMR, where such interaction has a higher probability of occurrence.
Transmembrane domains are crucial for proper positioning in the lipid bilayer of biological membranes [11] . Interaction of transmembrane domains of embedded integral proteins with the lipid component of the bilayer provides a diffusion barrier, and contains environment to maintain electrochemical properties [32] . Upon the recognition of CARC in TRKB, we determined the impact of a series of mutations in the transmembrane region physicochemical properties, measured as solvent accessibility to the aminoacid residues, and conformation in the human TRKB.TMR.

Alignment and solvent accessibility of models
The models of human TRKB.TMR (in this case with extra 10aa residues in each terminal) were generated in the RaptorX server [37] , using wild-type (WT) and mutated FASTA sequences.
The point mutation Y434F was chosen to exclude only the -OH group, but protect the aromatic ring structure of tyrosine. Y434A mutation was induced to examine the effects of the aromatic ring of tyrosine removal. Finally, Y434C mutation was introduced based on evidence showing the association of this particular change with hyperphagic obesity syndrome, which also causes defective memory, learning, and nociception [38,39] . The V439K mutation was chosen aiming to change an apolar/neutral to a basic aminoacid residue, and the same reason was considered for the R428A mutation. The models were aligned with WT for comparison, and the root-mean-squared deviation (RMSD) between the backbone carbon structure calculated by the server. The relative solvent accessibility per residues was calculated using PyMOL (v2.0 Schrödinger, LLC), and normalized by WT values.

Data mining
The presence of CRAC motifs were found throughout all the species analyzed [human (11

Homology of TRKB (and TMR) across species
Upon the identification of CARC motif in TRKB (human and mouse: REHLSVYAVVV; zebrafish: RVAVYIVV), the homology between TRKB.TMR sequences of several species (human, chimpanzee, mouse, rat, dog, chicken and zebrafish), listed in table 2, was examined using UniProt [33] . Over 90% of homology was found in both TRKB.TMR and full length TRKB of human, chimpanzee, mouse, rat and dog. The homology results of paired comparison between the species analyzed are organized in figure 2, where green gradient highlights the percentage of similarity between the sequences. For comparative purposes, we also determined the homology of the full length TRKB among those species. The homology results of paired comparisons are also organized in figure 2, with red gradient highlighting the percentage similarity between sequences.

Alignment and solvent accessibility of models
We proceeded with human TRKB.TMR to analyze the solvent accessibility following point mutations on tyrosine, valine or arginine in this region. Based on the assessment of solvent accessibility, we detected a prominent increase in the exposed surface to water (ranging from 150-500%) at the two juxtamembrane regions (figure 3a-e). On the other hand, the point mutations induced small morphological changes when the wild-type and mutant models were aligned, as indicated by the RMSD values (figure 3f-j). The relative solvent accessibility used in the present study is an indication of the surface exposure of the aminoacid residues in a polypeptide sequence [40] . Although simple, this parameter indicates putative conformational alterations in proteins that may translate into functional changes. Thus, even mild/neutral mutations such as Y to F in the CARC motif alters the physicochemical properties of the TMR.

DISCUSSION
In the present study, we evaluated the incidence of cholesterol-recognition motifs in the receptor tyrosine-kinase family (RTK) from mouse, human, zebrafish, fruit fly, and nematode C. elegans . We found that while the bona fide CRAC, located in the C-terminal portion of transmembrane domain [11] , is found in all the species analyzed, its inverted version -CARC -was only observed in vertebrates. Furthermore, we found that TRKB, a tyrosine-kinase receptor crucial for neuronal plasticity, contains a CARC sequence in the TMR. This motif is conserved across different species, and point mutations of the tyrosine, valine or arginine residues alter the solvent accessibility of the juxtamembrane regions of TRKB.
Cholesterol can interact with membrane proteins in several ways, one of its most prominent effect involves a direct post-translational modification on members of the Hedgehog pathway; particularly described in Drosophila sp. [41,42] . In this model organism, cholesterol is also found to regulate membrane depolarization through transient receptor potential (TRP) channels [43] , and serves as a precursor for ecdysteroids, which in turn regulate several steps of the fly development [44] . In nematodes, such as C. elegans , cholesterol is only obtained from diet, although these worms can modify the basic steroid structure into derivatives [45] . In both organisms, cholesterol appears to play a major role as signaling molecule with post translational modifications of proteins as the main mechanism [46] .
Although neurons synthesize the absolute minimum necessary cholesterol, glial production and release of lipoproteins supply neuronal demand during development and in adulthood [47] . In particular, apolipoprotein E (APOE) is synthesized primarily by astrocytes and glial cells [4,48] .
Glia-derived cholesterol stimulates synapse formation and synaptic efficacy [49,50] . In the presynaptic plasma membrane, the cholesterol-rich lipid rafts are necessary for SNARE-dependent exocytosis of vesicles with high content of cholesterol. At the postsynaptic level, such rafts organize the disposition of receptors, protein scaffold and signaling cascades [49,50] [51,52] .
Since it is difficult to directly visualize cholesterol binding motifs and crystallize the transmembrane domains containing such sequences, two consensus motifs with predictive value have been defined through in silico methods [27] : CRAC and its 'inverted' version -CARC [28,29] .
The non-covalent binding of cholesterol to specific motifs in proteins has been the focus of various recent studies. For example, cholesterol modulates docking of NMDA receptors into lipid rafts [53] , and the function of vanilloid receptors (TRPV1, a member of the TRP family) [54] , thus interfering in synaptic plasticity. Increased cholesterol concentration enhances the plasticity and the flexibility of 5HT1a dimers and adrenergic receptors [55,56] . Given the opposed dispositions of CARC and CRAC motifs, it is possible to assume the co-existence of both in the same transmembrane domain, and their potential interaction with two cholesterol molecules in a tail-to-tail configuration [27] . However, none of the analyzed TMR of RTK family members in the present study displayed co-existing CARC and CRAC motifs.
Interestingly, we only observed the occurrence of CARC motifs on zebrafish, mouse and human RTK family, and TRKB.TMR is highly conserved among vertebrates, similarly to full length TRKB. However, the TRKB.TMR CARC sequence from chicken differs in the juxtamembrane residue from the other species compared (see table 2). It is plausible to consider two optional scenarios, the role of R/K (charged, basic residues) is fulfilled by glutamate (E), which is also charged at pH 7, although negatively; or by glutamine (N), which is not charged but carries a basic amino group. Additionally, for the second possibility it is also necessary to relax the proposed '5-residue rule' between the Y and the juxtamembrane residue [57] , since N is located 6 residues apart from the central Y.
As mentioned above, TRKB plays a crucial role in several aspects of neural plasticity. The activation of this receptor is associated with the reopening of visual critical period [58] , and the formation, retention and recall of memory [59,60] . Increased activity of TRKB is linked to the occurrence of epilepsy, while blockade of such signaling normalizes the condition [61,62] . Whole genome sequencing performed on patients with developmental and epileptic encephalopathy (DEE) links a de novo missense variant of TRKB ( NTRK2 ) to the disease pathology [38] . Patients carrying cysteine at residue 434 instead of tyrosine, located inside TRKB CARC motif, were observed to have a similar phenotype varying from severe global developmental delay or intellectual disability to visual and feeding impairments [38] . Therefore, it is possible to speculate that such Y434C mutation, interfering with TRKB-cholesterol interaction, leads to impairment of adequate TRKB activity and function.
Neurotrophins induce homodimerization of TRK monomers to activate downstream kinases [63] . However, through a p75NTR-dependent mechanism, TRKA and TRKC receptors may induce cell death in the absence of their ligands, NGF or NT-3, respectively. On the other hand, TRKB does not act as death-inducing receptor when coupled to p75NTR [64] . In this sense, it is suggested that TRKA and TRKC interaction with p75NTR is a necessary step for moving the complex to lipid rafts, where the p75NTR-dependent death signal is triggered through its cleavage by gamma secretase [20] . TRKB was not observed to display such co-occurrence with p75-NTR in lipid rafts. The association of TRKA and TRKC with p75NTR in rafts involves the transmembrane domain of these receptors, given that a chimeric TRKB carrying TRKA transmembrane sequence requires BDNF for survival signal [20] . Curiously, p75NTR possesses a classical CRAC domain in its TMR: VVVGLVAYIAFKR; conserved in human, mouse, rat and chicken.
TRKB is found in lipid rafts only upon activation by BDNF [22] . Interestingly, when cholesterol is sequestered, TRKB translocation to lipid rafts is impaired, and BDNF-dependent potentiation is prevented [22] . However, loss of cholesterol in hippocampal cultures is associated with increased baseline activity of TRKB [65] . These opposite outcomes might be due to a differential modulation exerted by cholesterol, depending on the challenge to TRKB receptor (basal vs BDNF-stimulated), cell type/origin and stage of differentiation. Another possibility poses that cholesterol affects TRKB activity in a bell-shaped manner, where higher as well as lower concentrations of cholesterol impedes, instead of promoting, TRKB phosphorylation. In fact, the decrease of cholesterol levels by beta-cyclodextrin was found to differentially modulate the neurite growth of hippocampal and cortical cultured neurons. In hippocampal cells, the decrease of cholesterol levels induces an increase in neurite length and number, while no effect was observed in cortical cells. Interestingly, cultures of hippocampal cells show higher levels of cholesterol than the cortical counterparts [66] . Altogether, it is a provocative idea to consider TRKB as a 'sensor' of cholesterol levels in the cell membrane via CARC. Thus, TRKB would trigger the synaptic maturation or neurite growth only if the cholesterol levels are ideal for such changes. This assumption, however, remains to be tested.
Our data indicates increased solvent accessibility to the juxtamembrane regions of mutant TRKB in all the residues analyzed. The changes varied from 150-500% above the levels found in wild-type sequence. The increase in the hydrophilicity of juxtamembrane could implicate a decrease in the BDNF-induced movement of TRKB towards the lipid raft compartments, where the changes in TMR orientation may be altered, due to membrane thickness [67,68] .
According to in silico analysis of cholesterol interaction with CARC motifs, the Y to F, or Y to W mutations do not interfere with the motif interaction with cholesterol. However, these analyses were performed in multipass or multi-subunits transmembrane receptors, such as nicotinic cholinergic [27] or GPCR receptors [55,56] . Given that TRKB, as well as virtually all members of the RTK family, possess a single transmembrane domain, small changes in this region may impact their function, while not detected in the multipass proteins or to the direct interaction with cholesterol. The exception was found to be Q19238 (Putative tyrosine-protein kinase F09A5.2) from C. elegans , with 2 transmembrane domains, although none is CRAC/CARC-positive.
The insights in the present study could serve as a primary step for experimental testing of the impact of mutations in CRAC/CARC motifs in TMR of RTK family. However, we are limited by only considering the role of CRAC/CARC motifs in the TMR of RTKs. Given the promiscuous properties of these motifs, it is plausible to assume multiple false positive CRAC/CARCs alongside proteins, making the data mining and putative in silico or in vitro analysis difficult to perform.
Therefore, more studies focused on refining the algorithms for detecting these motifs are necessary.