Structure-based optimization of a PDZ-binding motif within a viral peptide stimulates neurite outgrowth

Protection of neuronal homeostasis is a major goal in the management of neurodegenerative diseases. Microtubule-associated Ser/Thr kinase 2 (MAST2) inhibits neurite outgrowth, and its inhibition therefore represents a potential therapeutic strategy. We previously reported that a viral protein (G-protein from rabies virus) capable of interfering with protein–protein interactions between the PDZ domain of MAST2 and the C-terminal moieties of its cellular partners counteracts MAST2-mediated suppression of neurite outgrowth. Here, we designed peptides derived from the native viral protein to increase the affinity of these peptides for the MAST2–PDZ domain. Our strategy involved modifying the length and flexibility of the noninteracting sequence linking the two subsites anchoring the peptide to the PDZ domain. Three peptides, Neurovita1 (NV1), NV2, and NV3, were selected, and we found that they all had increased affinities for the MAST2–PDZ domain, with Kd values decreasing from 1300 to 60 nm, while target selectivity was maintained. A parallel biological assay evaluating neurite extension and branching in cell cultures revealed that the NV peptides gradually improved neural activity, with the efficacies of these peptides for stimulating neurite outgrowth mirroring their affinities for MAST2–PDZ. We also show that NVs can be delivered into the cytoplasm of neurons as a gene or peptide. In summary, our findings indicate that virus-derived peptides targeted to MAST2–PDZ stimulate neurite outgrowth in several neuron types, opening up promising avenues for potentially using NVs in the management of neurodegenerative diseases.

The stabilization of homeostasis and the promotion of neuron survival are major challenges in the treatment of neurodegenerative diseases and neuron protection after nerve damage. The PI3K 8 /AKT-signaling pathway mediates cell survival and promotes resistance to apoptosis by inactivating components of the apoptotic machinery (1,2). The PI3K/AKT-signaling pathway also controls neuronal morphology and the dynamic formation of neurites during development. In mature neurons, the PI3K/AKT-signaling pathway activates neurite extension and branching (referred to hereafter as neurite outgrowth).
Neurite outgrowth is strongly repressed by inhibitory factors, such as phosphatase and tensin homolog deleted on chromosome 10 (PTEN), an upstream inhibitor of the PI3K/AKT/ mechanistic target of the rapamycin cascade-signaling pathway (3)(4)(5). PTEN silencing restores neurite outgrowth in various types of neurons (6 -9) and promotes branching (8,10,11) in cells and in vivo. Neurite outgrowth in human neuroblastoma cells (SHSY5Y) could be restored by the silencing of an AGC kinase widely expressed in the nervous system, the microtubule-associated serine-threonine kinase 2 (MAST2) (12). The similar impacts of gene silencing observed for both PTEN and MAST2 probably reflect the occurrence of protein-protein interactions between these proteins, involving the PDZ domain (post-synaptic density protein 95, Drosophila disc large tumor suppressor, and zona occludens 1) of MAST2. Valiente et al. (13) showed that the PDZ domain of MAST2 (MAST2-PDZ) binds the C-terminal PDZ-binding motif (PBM) of PTEN, greatly increasing the phosphorylation of PTEN by MAST2 at sites thought to control both the cellular distribution and activity of PTEN. Through the formation of this binary complex, MAST2 and PTEN act in synergy to inhibit the signal transduction pathways required for neuronal survival, neurite outgrowth, and regeneration (13)(14)(15). Specific disruption of the complex formed by MAST2-PDZ and the PBM of PTEN would therefore be expected to block this inhibition, stimulating neuron survival, neurite outgrowth, and regeneration.
We previously showed that virulent strains of the rabies virus (RABV) promote the survival of infected neurons, through the binding of the C terminus of the viral G-protein to MAST2 via a PDZ-PBM interaction resembling that between PTEN and kinase (16). We have shown that the RABV G-protein of virulent strains promotes neurite outgrowth by activating the AKTsignaling pathway in a PBM-dependent manner (16,17). Using NMR, we determined the high-resolution structures of the PDZ/PBM complexes formed by MAST2 and PTEN and by MAST2 and RABV G-protein (18,19). The PBMs of RABV G-protein and PTEN both display an unconventional mode of binding to MAST2-PDZ involving two anchor sites and similar binding affinities enabling these proteins to compete at the binding site of MAST2-PDZ. Finally, we have shown that the expression of the RABV G-protein in human neuroblastoma cells modifies the cellular distribution of PTEN in a PBM-dependent manner (18,19). Thus, the C-terminal sequence of RABV G-protein promotes neuron survival by disrupting the complex formed between MAST2-PDZ and PTEN. MAST2-PDZ appears to be a promising target for neuroprotective therapy. We aimed to design short-interfering polypeptides capable of efficiently promoting neurite outgrowth by targeting MAST2-PDZ.
In this study, we first identified the minimal core of the viral G-protein required for a peptide to trigger neurite outgrowth in different types of neurons after lentiviral vector-mediated transduction. Then, using the original PBM of the RABV G-protein (NV1-PBM) as our starting material, we performed a rational search for sequence modifications increasing the affinity of the PBM for MAST2-PDZ. We investigated the efficacy of three peptides, Neurovita1 (NV1), Neurovita2 (NV2), and Neurovita3 (NV3), by comparing their capacities to trigger neurite extension and branching in neurons transduced with the corresponding lentiviral vectors. The affinity and specificity profiles of the NV-PBMs were checked with a collection of 222 human PDZ domains (20). Finally, NV3-PBM was delivered to the cells as a peptide after fusion to a cell-penetrating molecule (CPM), as an alternative to lentivector transduction with the aim to use it as a potential drug compound for neurodegenera-tive diseases, and the capacity of the peptide to stimulate neurite outgrowth was assessed.

Relevance of the in vitro biological models we chose to study Neurovita properties
The aim of the study was to analyze the biological properties of different sequences of Neurovita based on the basis of the strength of their interaction with the PDZ of MAST2. Thus, it is required that the selected cell models used in the biological assays do express MAST2. MAST2 expression was detected by qPCR in human neuroblastoma, mouse primary cells, and rat neuroscreen (NS) cells (Fig. S1A) and by immunocytochemistry using a specific MAST2 antibody in human neuroblastoma and rat NS (Fig. S1B). In addition, we confirmed that MAST2 silencing in the human SH-SY5Y neuroblastoma cell line used in our laboratory stimulates neurite outgrowth (Fig. 1A) as previously shown by Loh et al. (12).

Construction of the initial NV1 peptide
The infection of neurons with RABV led to the expression of five viral proteins: the N, M, G, NS, and L proteins. The G-protein (Fig. 1B, diagram of the full-length RABV G-protein) is a 524-aa type I glycoprotein. It is synthesized in the endoplasmic reticulum (ER) of the soma, and it is then glycosylated in the Golgi apparatus (see G-protein trafficking in Fig. 1B). The G-protein then migrates to the plasma membrane of the neurites, where it associates with the viral endoplasmic reticulum and M protein, for virion formation and egress by budding. The G-protein is attached to the ER and plasma membranes via its unique transmembrane domain (TM). The ectodomain (EC) of the G-protein is located within the luminal compartment of the ER initially, and then at the cell surface after fusion to the plasma membrane. During G-protein trafficking from the ER to the plasma membrane, the cytoplasmic domain (Cyto) containing the C-terminal PBM is always exposed to the cytoplasm of the somato-dendritic neuronal area, where it can interact with cellular partners, such as MAST2, via its PDZ domain. The formation of this complex triggers neurosurvival-signaling pathways and the localization of MAST2 to a submembrane area of the infected cells (16).
We identified the parts of the viral G-protein required for a short-interfering polypeptide to mimic the neurosurvival properties of the viral G-protein. We constructed a mini-RABV G-protein, NV1 (Fig. 1B, NV1 scheme). NV1 harbors a 99.6% deletion of the EC domain but retains the SP, TM, and Cyto-PBM domains, which are required for its processing. We investigated the role of the C-terminal PBM in the biological properties of NV1 by generating a polypeptide lacking four aa at the C terminus, NV1⌬ (Fig. 1B, NV1⌬ scheme). The properties of the NVs were investigated following the transduction of two types of neuronal cell lines: SH-SY5Y cells and NS. The NV1 and NV1⌬ proteins were detected by Western blotting (WB) with a specific Cyto-G antibody in SHSY5Y cell lysates 24 h after transduction (Fig. 1C) and in NS cell lysates 48 h after transduction (Fig. 1D). Neurite outgrowth in NV1⌬-transduced cells was similar to that in nontransduced cells (Fig. 1, C and D). By contrast, NV1 trig-Interfering peptides stimulating neurite outgrowth gered a significantly stronger neurite outgrowth by comparison with NV1⌬ in both types of cell (1.6 times stronger in SHSY5Y cells 30 h after transduction, and 3.1 times stronger in NS cells 72 h after transduction (Fig. 1, C and D, right panels, and Fig. S2 for NeuronJ drawings).
Neurite outgrowth was also triggered in neurons other than neuroblastoma and pheochromocytoma cells grown in the presence of a DF medium (NGF or Bt 2 cAMP for NS and SHSY5Y, respectively). The effects of NV1 were assessed in transduced human iPSc-derived motor neurons ( Fig. 2A) and in transduced E16 mouse primary cortical neurons (Fig. 2B). In these two models, where lentivirus transduced more than 80% of cells (data not shown), NV1 stimulated neurite outgrowth more efficiently than NV1⌬ (1.8-fold increase in human motor neurons and 1.2-fold increase in primary mouse cortical neurons). Thus, NV1 induces neurite outgrowth in a PBM-dependent manner in primary cultures of neurons, whether of mouse or human origin.
These results show that it is feasible to construct a polypeptide conserving the properties of the viral G-protein if this poly- triggers spontaneous neurite outgrowth in SH-SY5Y culture. This effect is not observed when cells were treated with pL.KO1shRNA (Control). Two-tailed unpaired Student's t test, ϭ p Ͻ 0.0001, was used. These results are representative of two independent experiments. B, structure of the RABV G-protein gene encoding an SP (19 aa), an EC, a TM (22 aa), and a Cyto (44 aa) together with a C-terminal PBM. Scheme on the right shows G-protein trafficking from the ER to the plasma membrane in RABV-infected cells. Diagram of the lentiviral vector transgene expression cassettes: NV1 contains the SP, two amino acids from the EC domain (2 aa), the TM, and the Cyto domain. NV1⌬ is identical except that it lacks the last 4 aa at the C terminus of the PBM. C, neurite outgrowth triggered by NV1 and NV1⌬ in SH-SY5Y cells 24 h post-transduction with the lentivectors. Left panel, expression of NV1 and NV1⌬ assessed by WB with a RABV G-protein Cyto-specific antibody. The expected masses of NV1⌬ and NV1 are 9.0 and 9.5 kDa, respectively. Expression of tubulin is used as a loading control. To note, NV1 and NV1⌬ bands, obtained in the same membrane but on noncontiguous lanes, were spliced (dashed line). Right panel, measurement of neurite outgrowth in SH-SY5Y cells expressing NV1 or NV1⌬ 30 h post-transduction. Images of 30 fields (mean of 300 cells per field) were taken with a UV microscope for each condition.

Interfering peptides stimulating neurite outgrowth
peptide includes the cytoplasmic part of the G-protein and the last 4 aa from the C terminus of the PBM.

Optimization of NV-PBMs sequences to improve their binding to MAST2-PDZ
We previously solved the high-resolution structures of the C-terminal sequences of 13 residues of PTEN and RABV G-protein bound to MAST2-PDZ and calculated their affinities (1.99 and 1.26 M, respectively) (19). We discovered that, unlike most PBMs, the PBMs of PTEN and of viral G-protein have an unusually large interaction surface (700 Å 2 ), encompassing not only the 4 aa at the C terminus of the PBM (-QTRL-COOH for G-protein and -ITKV COOH for PTEN), but also a second anchoring region located at the N terminus of the assayed peptides ( NH2 SW-for G-protein and NH2 PF-for PTEN) (Fig. 3A). The sequences interacting with MAST2-PDZ are NH2 SWESHKSGGQTRL COOH and NH2 PFDEDQHTQITKV COOH for G-protein and PTEN, respectively. They contain three contiguous regions, the N-terminal anchor (2-aa residues in blue in Fig. 3), a linker (7-aa residues in green), and the conventional C-terminal anchor (4-aa residues in orange).
We began our search for peptides with a greater affinity for MAST2-PDZ by focusing on the 13 residues at the C-terminal end of the viral G-protein ( SW Cyto13 QTRL ) and human PTEN ( PF PTEN ITKW ). Using our high-resolution NMR structures of MAST2-PDZ/peptide complexes, we rationally designed sequences likely to increase the enthalpic contribution and decrease the entropic cost of binding to MAST2-PDZ. Isothermal titration calorimetry was performed in binding experiments with the MAST2-PDZ domain; the thermodynamic quantities associated with peptide binding were compared with the data obtained for the original sequences (18,19).
We focused on optimizing the seven-residue linker (positions Ϫ4 to Ϫ10) tethering the N-and C-terminal anchors of viral G-protein onto MAST2-PDZ, which has been shown to be a sensitive target for improving binding affinity (18,21). By contrast, we kept the C-terminal anchor -QTRL COOH , which maintains specificity, intact, together with the N-terminal dipeptide NH2 SW-, which significantly increases binding affinity.
We began by maintaining linker length at seven residues and mutating exposed residues to increase the degree of hydrophobic contact of the two original linker sequences (Ser(-6) and Lys(-7) for SW Cyto13 QTRL or Asp(-8), and Asp(-10) for PF PTEN ITKV ) (Fig. 3B).
These modifications increased affinity modestly with respect to SW Cyto13 QTRL , by a factor of 3 at best (Fig. 3B). This was not unexpected because, for all assayed sequences with a linker of this length, any decrease in the free enthalpy of binding is strongly compensated by a simultaneous decrease in the associated entropy. The mean compensation between these two thermodynamic quantities was estimated at 80% from the slope of the plot of the corresponding data displayed in Fig. 3C.
By analyzing the kinetic behavior of potential candidates with seven-residue linkers, we previously showed (18) that the flexibility of linker sequences strongly affects their affinity for MAST2-PDZ. However, the high-resolution 3D structures of the PDZ/PBM complexes provided no evidence of contact between MAST2-PDZ and the linker present in the Cyto13 of viral G-protein or PTEN. This discrepancy was probably mostly due to the conformational entropy of the potential ligand in the unbound state, because the polypeptides most disordered when free in solution were the poorer binders. We therefore reduced linker length from 7 to 5 aa and designed seven new sequences, for which we compared binding abilities (lines 5 to 11 in Fig.  3B). The deletion of 2 aa from sequence 442 to generate sequence 441 led to a significant increase in the entropy of binding with no marked change in enthalpy contribution, resulting in a 4-fold increase in affinity. For these shorter sequences, the compensation between the enthalpy and the entropy terms remained at 80% (Fig. 3C). The binding affinities of the new candidates were also modulated by the substitution of one or both of the doublet glycine present at positions Ϫ4 and Ϫ5. The Gly to Gln change at position Ϫ4 yielded the candidate with the highest affinity, PBM 454.
We retained the two sequences with the highest affinities from this second step in peptide design: PBM 441 (which we named NV2-PBM) and PBM 454 (which we named NV3-PBM). Affinity was 10 times higher than that of the starting sequence (NV1-PBM) for NV2-PBM (K d of 120 nM) and 20 times higher for NV3-PBM (K d of 60 nM). The kinetics of

Interfering peptides stimulating neurite outgrowth
binding for NV1-PBM (original sequence) and NV3-PBM (sequence with the highest affinity) were compared in stoppedflow experiments. For NV3-PBM, we were able to obtain an accurate value for the association rate constant (1.4 ϫ 10 7 M Ϫ1 ⅐s Ϫ1 , a 1.6-fold increase with respect to NV1-PBM). We could obtain estimates only for the very slow dissociation process, with an approximate dissociation rate constant of 0.89 s Ϫ1 , more than 1 order of magnitude higher than the corresponding values for the other peptides tested, such as the original PBM, PTEN-PBM (9.7 s Ϫ1 ), and NV1-PBM (16.6 s Ϫ1 ). The residence time of NV3-PBM on the MAST2-PDZ domain was 1.1 s, a value 10 times greater than that for the initial NV1-PBM sequence.

Affinity profiles of the three NV sequences
We addressed the issue of PDZ-NV-PBM specificities by performing automated holdup assays (20). We used this highthroughput chromatographic approach to quantify PDZ-PBM interaction affinities at equilibrium in solution for a collection of 222 human PDZ domains, corresponding to 83% of the original PDZome (266 PDZ domains) overexpressed in a soluble form. The holdup screens probing the interaction of the PDZome with NV1-, NV2-, and NV3-PBM, were performed in triplicate. In Fig. 4, the PDZ-binding intensities (BI) of NV-PBMs are ranked in descending order from left to right. Using a stringent BI threshold of 0.2 to retain only high-confidence binding pairs, the holdup assay identified 5, 8, and 15 hits for NV1-PBM, NV2-PBM, and NV3-PBM, respectively. The PDZ domains recognized by NV1-PBM also bound to NV2-PBM. The NV2-PBM binders were also included in the set of proteins recognized by NV3-PBM. The number of binders detected increased with affinity for MAST2-PDZ, but the slope of BI decay in the binding profiles indicated that recognition specificity was roughly maintained from NV1-PBM to NV3-PBM. Furthermore, this affinity-based ranking demonstrated that all three NV-PBMs targeted MAST2-PDZ preferentially among the 222 PDZ domains tested.

Comparison of the biological activities of NV constructs in NS cells
Mini-G sequences encoding NV2-PBM or NV3-PBM were constructed as described (Fig. 5A). They were delivered to NS

Interfering peptides stimulating neurite outgrowth
cells by lentivirus vectors. The viability of the three lentivirus vectors encoding NV1, NV2, and NV3 was demonstrated by quantifying NV transcripts in the transduced NS cells (Fig. 5B, left panel). Each viral vector produced similar amounts of the corresponding NV1, NV2, and NV3 polypeptides, as shown by WB (Fig. 5B, right panel). We then compared the effects on neurite outgrowth of NV1, NV2, and NV3 in NS cells, 72 h after transduction ( Fig. 5C and Fig. S2 for NeuronJ drawings). Peptide rankings for the stimulation of neurite outgrowth (NV3 Ͼ NV2 Ͼ NV1) were identical to those obtained in vitro for affinity for MAST2-PDZ.
We then analyzed whether the greater capacity of NV3 and NV2 to stimulate neurite outgrowth was associated with a greater complexity of neurite arborization. We performed Sholl analysis (22) to determine how many times the neurites of a given cell crossed concentric circles of increasing radius centered on the cell soma. The results of the analysis are expressed as the mean number of neurites crossing each circle of increasing radius from 0 to 250 m for NS cells transduced with each type of construct and for nontransduced cells (Fig. 5D). The expression of the various NVs was associated with a significant increase in neurite arborization complexity relative to that for nontransduced cells, even though the increase between arborization triggered by NV1 and NV2 stays modest. As already shown above for the neurite outgrowth phenotype, NV3 yielded the best increase in arborization.

Practical impact on drug design
The NVs were expressed in neurons via lentiviral vector transduction. However, this type of vectorization is not optimal for human therapeutic applications; therefore, alternative ways to deliver NV need to be considered. Peptide delivery is a possible option, only if two criteria are respected: (a) the active sequence is coupled with a CPM to facilitate passage across the outer membrane of the neurons, and (b) the entire polypeptide size remains small to facilitate its production.
We reduced the size of the most active peptide NV3 by restricting the peptide to its cytoplasmic domain (NV3-Cyto). A new lentiviral vector, NV3-Cyto, was designed to determine whether the removal of the part of the sequence responsible for endomembrane anchorage (the SP, two amino acids of the EC, and the TM domain) could be tolerated (Fig. 6A). We assessed lentivector viability for NV3 and NV3-Cyto by performing RT-PCR on NS cells after transduction (Fig. 6B). NS cells transduced with NV3-Cyto displayed neurite outgrowth, but less than that observed with the parental peptide NV3 (Fig. 6C). Unlike cells transduced with the parental peptide, the cells transduced with NV3-Cyto were unable to achieve higher levels of branching (data not shown). The properties of NV3-Cytotransduced cells were similar to those of cells transduced with the initial construct, NV1 (as shown in Fig. 6C for neurite outgrowth and Fig. 6D for branching). Thus, NV polypeptides encoding the PBM sequence having a high affinity for MAST2-PDZ can still trigger neurite elongation, even if membrane anchorage is lost.
To directly deliver NV3-Cyto into neurons as a peptide, we coupled the NV3-Cyto construct and its control, NV3-Cyto⌬, to a CPM, as displayed in Fig. 7A. We used the variable part (VHH) of a single-chain llama antibody as the CPM, because this molecule has been shown to cross the blood-brain barrier and the plasma membrane of brain cells (23,24). Constructs encoding CPM-NV3-Cyto and CPM-NV3-Cyto⌬, as a control (CPM-Neurovita and CPM-Neurovita⌬, respectively), were genetically engineered (Fig. 7A) and expressed in bacterial periplasm, and the resulting proteins were purified. We have previously shown that when CPM-Neurovita was added to a mixed culture of humans neurons, astrocytes and microglia

Interfering peptides stimulating neurite outgrowth
CPM-Neurovita target neurons preferentially (25). Here, we analyzed whether CPM-Neurovita could be detected in NS by WB or flow cytometry after CPM-Neurovita was added to the culture medium (Fig. S3). Six hours after the addition of CPM-Neurovita to the culture medium, CPM-Neurovita could be detected by WB in the cell lysates at the expected molecular mass of 22.7 kDa (Fig. S3A). Thirty minutes after the addition of CPM-Neurovita to the culture medium, CPM-Neurovita entry was detected by flow cytometry when the cells had been previously permeabilized (Fig. S3B, left panel), whereas no staining was observed on nonpermeabilized cells (Fig. S3B, right panel). This indicates that CPM-Neurovita associates with some cellular area made accessible to the antibody by the permeabilization treatment such as the cytoplasm or the submembrane area.
When added to the culture medium of NS cells, CPM-Neurovita (2.35 pg/cell) triggered stronger neurite outgrowth than the one observed for CPM-Neurovita⌬-cells or nontreated cells (Fig. 7B). Time-lapse recordings over a period of 12 h (starting 18 h to after treatment up to 30 h post-treatment) with Incucyte Technology were used to compare the speed of neu-rite growth. The number of neurites per neuron (Fig. 7C) and the speed of growth (Fig. 7D) were significantly greater in NS cells treated with CPM-Neurovita than in cells treated with CPM-Neurovita⌬.
Thus, NV efficacy is qualitatively maintained when the NV sequence is delivered as a cell-penetrating interfering peptide.

Discussion
Protein-protein interaction domains, including PDZ domains governing signaling pathways critical for cell homeostasis, are emerging as promising targets for therapeutic applications (26 -30). These signaling pathways can be manipulated with short-interfering peptides, which disrupt these interactions. Viruses are obligate intracellular parasites, and their multiplication depends on cellular resources. The mimicry of PBM by viral proteins is a simple strategy by which several viruses take control of the cellular machinery. In particular, this mimicry enables some viruses to modify the commitment of infected cells to death or survival (16,31,32). Identification of the viral sequences involved and elucidation of their mechanisms of action could potentially facilitate the design of efficient interfering peptides. We have identified a viral protein, the RABV G-protein, that promotes neurite outgrowth by disrupting interactions between the PDZ of the MAST2 kinase and its endogenous ligands, such as PTEN, via its C-terminal PBM. In this study, we investigated whether we can leverage the robust strategy evolved by a virus for ensuring its own propagation in the brain to develop a short-interfering peptide that promotes neurite outgrowth by specifically targeting MAST2-PDZ.
We hypothesized that NV peptides derived from the RABV G-protein and competing with the endogenous ligand for binding to the MAST2-PDZ domain would promote neurite outgrowth following their delivery into cells. We first showed that short peptides encoding the Cyto-domain of the G-protein reproduce the properties of RABV. However, the presence of certain parts of the viral RABV G-protein involved in association with endomembranes, such as the SP and TM domain, were necessary to ensure their correct cellular targeting. The resulting interfering peptide, NV1, encodes a PBM bound to MAST2-PDZ with an affinity in the micromolar range similarly to PTEN, which is one of the endogenous ligands of MAST2-PDZ. We suggested that increasing the affinity of the interfering peptide for MAST2-PDZ would increase the peptide potency to stimulate neurite outgrowth. We used structural, kinetic, and thermodynamic data for a large number of PDZ/peptide complexes to design peptides with a higher affinity for MAST2-PDZ, with the viral and one endogenous partner of MAST2 (i.e. RABV G-protein and PTEN) as the starting material. One of the original features of our approach relates to the specificity of MAST2-PDZ recognition. In addition to increasing the affinity of NV peptides for MAST2-PDZ, we also wished to maintain this specificity for MAST2-PDZ to prevent undesirable additional interactions with other endogenous proteins. PDZ binding selectivity seems to have been achieved throughout evolution (33,34). Extensive scans of random peptides against a large collection of PBM domains led to the definition of discrete sets of PDZ classes, conserved from worms to humans, encompassing the last seven amino acid residues of Interfering peptides stimulating neurite outgrowth potential PBM candidates (35). Most attempts to improve the strength of a given PDZ-PBM interaction have focused on modifications to the pre-existing stretch of seven amino acid residues at the C terminus of the PBM, position by position. Such efforts have generally resulted in a loss of the specificity of this interaction pair with respect to competing PDZ or PBM molecules (16, 36 -38). Indeed, we previously showed that one mutation at position Ϫ3 in the PBM sequence of the RABV G-protein is sufficient to extend cellular protein recognition patterns and to switch cell fate from survival to death (16). We therefore decided to keep the C terminus QTRLϾ intact in the new NVs.
Some PDZ domains make use of amino acid sequences beyond the canonical PDZ domain fold, which are referred to as extended PDZ domains. Similarly, some PDZ ligands require extension sequences beyond the C-terminal PBM, including some more than 10 residues from the last C-terminal residue, for complex formation with the PDZ domain of the partner protein (38). We have shown that both RABV G-protein and PTEN make use of a PBM extension up to position Ϫ12 for binding to MAST2-PDZ (18,19). The unusually large interaction interface between MAST2-PDZ and G-protein or the PTEN C terminus involves not only the four-amino acid canonical PBMs, but also a two-amino acid N-terminal anchor (SW and PF for G-protein and PTEN, respectively). This N-terminal anchor increases MAST2-PDZ-binding affinity and probably plays a crucial role in narrowing the specificity of interaction between MAST2 and its partners (18). This finding is reminiscent of observations from other studies indicating that PTEN also uses an N-terminal extension to bind the third PDZ domain of PAR-3, improving the specificity of PTEN for its target and its subcellular distribution at the junctional membrane of polarized cells (39). We suggested that in this case the N-terminal anchor might be involved in PBM-PDZ specificity and might increase the affinity of the PBM for MAST2-PDZ. We therefore did not modify this part of the molecule either.
By conserving the two viral anchoring sites, NH2 SW-and -QTRL COOH , from virulent RABV strains, which promote neuron survival, we were able to conserve the specificity of NV peptides for MAST2-PDZ, as illustrated by high-throughput PDZome screening.
We increased the affinity of NV peptides for MAST2-PDZ mostly by decreasing the entropic cost of binding by modulating the flexibility of the central linker connecting the two binding sites of NV peptides, thereby greatly increasing the residence time of NV on MAST2-PDZ. After two rounds of optimization, we obtained peptides with an affinity 20 times higher (K D ϭ 60 nM) than that of the initial viral or endogenous peptides (1300 nM). Such an affinity, in the nanomolar range, is particularly high for a PDZ/ligand system, especially for molecules targeting a single PDZ domain. For example, interfering peptides targeting the Disheveled PDZ and functioning as an

Interfering peptides stimulating neurite outgrowth
inhibitor of the Wnt/␤-catenin-signaling pathway or targeting the PDZ1 of PSD95 and functioning as an inhibitor of nitric oxide (NO) synthesis have K d values in the micromolar range (40 -42). A few examples of nanomolar affinity (3-16 nM) have been reported for multivalent inhibitors targeting multi-PDZcontaining proteins, such as those targeting proteins of the MAGUK family (43).
The selected peptides, NV1, NV2, and NV3, are capable of inducing the neurite outgrowth of NS cells with an efficiency related to their affinity for MAST2-PDZ. These data strongly support our hypothesis that NV activity can be increased by enhancing the strength of interaction with the cellular partner.
Given their capacity to promote neuron survival, NVs could have therapeutic applications for diseases of the nervous system in which neuron survival is compromised. The different ways to deliver NVs will be determined by the nature of the disease. For chronic diseases, it may be advantageous to deliver NVs in gene form, via a viral vector, such as a lentivirus or an adeno-associated virus, which can transduce neurons. The delivery of a peptide, rather than a gene, may be more appropriate when a fast mode of action is required, for example in the case of an acute nerve injury. We therefore tried to adapt NV sequences for peptide production, by reducing the peptide size, to facilitate production and delivery.
We showed that for the optimized peptide NV3 that the processing and translocation were dispensable. Therefore, we were able to halve the size of the peptides, while maintaining the activity of the original construct. However, the 20-fold improvement in affinity initially achieved was counterbalanced by imperfections of targeting. This design was feasible only for NV3, the peptide with the highest affinity for the target. Indeed, our first attempt to generate a similar construct from the NV1 peptide resulted in a total loss of neurite outgrowth stimulation (data not shown). We subsequently demonstrated the feasibility of replacing the viral vector delivery system by fusing a shortened NV3 peptide to a CPM.
In this study, resulting from a combination of biophysical and cellular approaches, we showed for the first time that a virus-derived peptide could be optimized to trigger neurite outgrowth in several types of neurons. This work opens up a number of very promising perspectives for the use of NV as a potential drug candidate in neurodegenerative diseases.

Cell lines, plasmids encoding RABV G-protein sequences, lentivectors, and CPM-Neurovita
Neuroscreen cells, a subclone of rat pheochromocytoma 12 (PC12) neuronal cells, were obtained from Cellomics and cultured according to the supplier's instructions. NS cells were grown in nondifferentiating (NDF) medium (Roswell Park Memorial Institute 1640 (RPMI 1640) medium, Thermo Fisher Scientific France, catalog no. 11875093; 5% fetal bovine serum (FBS); 10% horse serum; 1% 200 mM glutamine, Thermo Fisher Scientific France, catalog no. 25030149 and 1% 100ϫ penicillin/ streptomycin (Pen-Strep), Thermo Fisher Scientific France, catalog no. 15140148). The differentiating (nondifferentiating) medium had the same composition with the addition of 200 ng/ml of NGF (Sigma, catalog no. N2513). A clone of the human SH-SY5Y neuroblastoma cell line, which was originally used to decipher the role of the kinome, including the role of MAST2, in neurite outgrowth, was obtained from Dr. S. H. Loh and was grown as described previously (12 Cortical neurons were prepared from E16 Swiss mice as described by Vitry et al. (44). Embryonic mouse brains were dissected in Hanks' balanced salt solution (HBSS) (Life Technologies, Inc., catalog no. 14175) supplemented with 10 mM HEPES (Life Technologies, Inc., Inc. catalog no. 15630), 0.05% glucose (Sigma, catalog no. G8769), and 50 g/ml gentamicin (Life Technologies, Inc., catalog no. 15750). Meninge-free brain cortices were collected in HBSS supplemented with 400 units/ml DNase (Roche Applied Science, catalog no. 04536282001) and incubated with the Liberase HD enzymatic mixture (Sigma, catalog no. 5401054001). The digested tissues were triturated mechanically, and the enzymatic reaction was stopped by adding FBS. The suspension was washed in HBSS and centrifuged for 5 min at 200 ϫ g, and the resulting cell pellet was suspended in Neurobasal medium (Life Technologies, Inc., catalog no. 21103) supplemented with 25 nM 2-mercaptoethanol (Sigma, catalog no. M7522), 50 g/ml gentamicin, 5 mM HEPES, 2 mM Glutamax (Life Technologies, Inc., catalog no. 35050), N2 (Life Technologies, Inc., catalog no. 17502), and B27 (Life Technologies, Inc., catalog no. 17504). Viable cells were used to seed the wells at a density of 10 4 to 3 ϫ 10 4 cells per cm 2 on plates coated with poly-D-lysine (Sigma, catalog no. P6407) and laminin (Sigma, catalog no. L2020). The cells were allowed to sediment and to adhere for 20 min at room temperature. They were then transferred to an incubator containing 9% CO 2 in air maintained at 37°C, for 40 min, before the addition of 10 g/ml fluorodeoxyuridine (Sigma, catalog no. F0503) and 25 g/ml uridine (Sigma, catalog no. U3003). The iPSc clones (clone 2.1 (45) and clone 3.1, unpublished) 9 were stimulated to differentiate into motor neurons as described by Maury et al. (46).
CPM-Neurovita is a 209-aa-long (22.7 kDa) chimeric molecule consisting of the following: (i) Neurovita; (ii) a CPM formed by a Neuro-Tag domain (a RABV-derived peptide from the RABVvirG-Protein ectodomain that recognizes the acetylcholine receptor ␣7 subunit), and by the variable part (VHH) of a llama single-chain antibody capable of crossing biological membranes and the blood-brain barrier (23); and (iii) a StrepTag sequence to label the molecule. CPM-Neurovita⌬ is the disabled form of the molecule with a deletion of the last 4 aa of the PBM. The GenBank TM accession numbers are KJ784393 (CPM-Neurovita) and KJ784394 (CPM-Neurovita⌬). The genes were synthesized chemically (MWG Operon, Germany) and inserted into the pASK-IBA2 plasmid (IBA, BioTAGnology, Germany) under the control of the tetracycline promoter. The ompA signal sequence added by this plasmid directs the expressed protein into the periplasmic space of the bacteria and is cleaved during translocation. The recombinant plasmids were used to transform XL1-blue Escherichia coli bacteria (Stratagene). Recombinant bacteria were identified by PCR, and glycerol stocks were generated. CPM-Neurovita peptides were expressed and purified from the bacterial periplasm as described by Lafaye et al. (50). The periplasmic extract was further purified by immunoaffinity chromatography with a mouse high-affinity anti-StrepTag IgG1 mAb, C23.21 (kindly provided by Pierre Lafaye). The molecules were stored in 0.25ϫ PBS at Ϫ80°C. CPM expression was monitored by Western blotting with the C23.21 mAb, with quantification by Odyssey Li-Cor IR chemo luminescence (Li-Cor) with a standardized VHH-StrepTag preparation. Interfering peptides stimulating neurite outgrowth et al. (12) and Prehaud et al. (16). Briefly, SH-SY5Y or NS cells were used to seed 24-well plates (CellBind plastic ware, Corning, catalog no. 83-3337) containing nondifferentiating medium, at a density of 40,000 cells per well. They were cultured overnight at 37°C. The medium was then replaced with DF medium, and the cells were incubated for a further 6 h. They were then infected with 30 ng of p24 lentiviral vector or treated with 2.35 pg/cell CPM-Neurovita or CPM-Neurovita⌬ in DF medium for 1 h. The cells were washed once with DF medium, and incubated in DF medium for 24 h at 37°C. Thirty hours (SH-SY5Y) or 72 h (NS) post-transduction, the cells were fixed by incubation with 3% paraformaldehyde (PFA) in phosphatebuffered saline (PBS, Thermo Fisher Scientific France, catalog no. 14040141) for 20 min at room temperature and treated for 5 min with 0.1% Triton X-100 and 50% normal goat serum (NGS) in PBS. Neuron-specific anti-␤III-tubulin antibody (Promega France catalog no. G7121) and Hoechst 33342 (Thermo Fisher Scientific France, catalog no. 62249) were used to stain the neurite processes and the nuclei, respectively. Alternatively, the cells were also stained with crystal violet, which preserves the neurite processes. SH-SY5Y human neuroblastoma cells (30 fields, mean of 300 cells per field) and NS cells (36 fields, mean of 324 cells per field) were imaged with a Leica DM 5000B UV microscope equipped with a DC 300FX camera (ϫ40 or ϫ20 objective). The images were analyzed using ImageJ 1.38X Software (Wayne Rasband, National Institutes of Health) and its plug-in NeuronJ (51). The effects of bias were minimized by performing the experiment blind, with relabeling of the samples.

SH-SY5Y and NS cells-Neurite
Motor neurons-Motor neurons were plated at a density of 10,000 cells per well (48-well plates). Gene transfer was performed with lentiviral vectors (6 ng of p24 per well) the next day.
Transduction was obtained in 80 -90% of cells. Nontreated conditions was used as a quality control of the culture. Cells were fixed 3 days later (4% PFA in PBS for 20 min at room temperature). The cells were washed in PBS and incubated for 1 h at room temperature in PBS containing 2% NGS, 1% bovine serum albumin (BSA, Sigma, catalog no. A2153), and 0.1% Triton X-100 (Sigma, catalog no. X100). Cells were then incubated overnight at 4°C with anti-␤III-tubulin antibody (TUBB3, Biolegend France catalog no. 801201), washed three times in PBS, and incubated for 1 h at room temperature with Alexa Fluor 555-conjugated anti-mouse IgG2a antibody (Thermo Fisher Scientific, catalog no. A-21137). Nuclei were stained with 1 mg/ml Hoechst H33342 (Thermo Fisher Scientific France, catalog no. 62249). Motor neurons (500 cells) were analyzed with the Cell Insight platform (Cellomics, Thermo Fisher Scientific France), using the Neuronal Profiling Bio-application suite. Neurite outgrowth phenotype was determined in triplicate experiments.
Mouse cortical neurons-Mouse cortical neurons (3 days in vitro) were plated on 96-well dark-sided cell-binding plates (Corning, catalog no. 3340). They were infected with 10 ng of p24/well of lentiviral vector 2 h after seeding. The medium was changed 12 h after infection. Three days post-infection, the medium was carefully removed, and the neurons were fixed by incubation with 4% PFA for 20 min at room temperature. Plates were washed three times with PBS and the cells were permeated by incubation with 0.3% Triton X-100 (Sigma, catalog no. X100) for 10 min at room temperature. Analysis of neuronal ␤III t51ulin immunofluorescence and Hoechst staining were performed as described above for SH-SY5Y/NS cells. Primary cortical neurons (10 fields, mean of 650 cells per field) were analyzed with the Cell Insight platform (Cellomics, Thermo Fisher Scientific France), using the Neuronal Profiling Bio-application suite. Neurite outgrowth phenotype was determined in triplicate experiments.

Time-lapse experiments in NS cells
We determined the neurite-bearing cell phenotype and assessed the speed of neurite outgrowth by using NS cells in NDF to seed collagen-coated 24-well plates (CellBind plastic ware, Corning, catalog no. 83-3337) at a density of 40,000 cells per well. The cells were cultured overnight at 37°C. The NDF medium was replaced by DF medium containing only 10 ng/ml NGF and 7% FBS (52), 24 h after seeding, and the cells were incubated for a further 3 h. They were then treated with 700 ng/well CPM-Neurovita or CPM-Neurovita⌬ in DF medium. Cells were incubated in DF medium for 48 h at 37°C, under an atmosphere containing 5% CO 2 in an IncuCyte Live-Cell analysis system. Images were recorded every hour over a 2-day period. For live-cell imaging on IncuCyte, NS cells (nine fields mean of 43 cells per field) were recorded on bright-field images and analyzed with Fiji/ImageJ (59) and the NeuronJ plug-in. The percentage of neurite-bearing cells (53) and the speed of neurite outgrowth and elongation were monitored over 12 h, from 18 to 30 h after treatment.

Arborization analysis
Sholl analysis (22,54) was used to assess dendritic arborization. This analysis was performed as described by Sahay et al. (22) and Lioy et al. (55). Measurements were obtained from 8-bit converted images with the Sholl analysis plugin (51). Branching complexity was quantified by counting the intersection of neurites with each of 10 concentric circles of increasing radius drawn around the cell body. We picked 30 neurons at random from each group for analysis. Experiments were performed at least three times. Bias was minimized by performing the experiments blind after sample relabeling.

Reverse transcription and quantitative PCR
The procedures used for RT-qPCR have been described elsewhere (56). Basically, cDNA was synthesized from 1 g of total RNA with the SuperScript II reverse transcriptase (Thermo Fisher Scientific France, catalog no. 18064014). RT-qPCR was performed in triplicate with an ABI Prism 7500 fast sequence detector system with GoTaq PCR master mix (Promega France catalog no. M7133). After normalization against 18S rRNA, the relative abundance of mRNA was obtained by calculating the difference in threshold cycles between the test and control samples (mock value set at 1), as according to the ⌬⌬C T method. The primers used were as follows: Lenti-1 (AGACGCCATCCACG-CTGTTTTGACCTCCATAG); Lenti-2 (CCAAAACACAAT-GGAAAAACCAGAAGGGGTAC); Lenti-3 (GTGTTGCGT-AGGTTCTGATCGATTGACTCTTC); 18S (forward, CTT Interfering peptides stimulating neurite outgrowth AGA GGG ACA AGT GGC G, and reverse, ACG CTG AGC CAG TCA GTG TA); and MAST2, Quantitect primers (Qiagen, Germany) QT00042574.

WB
Proteins were suspended in RIPA buffer (Sigma catalog no. R0278) supplemented with Complete ULTRA EDTA (Roche Applied Science, catalog no. 05892970001), PhosSTOP (Roche Applied Science, catalog no. 04906845001), and nuclease (Thermo Fisher Scientific France, catalog no. 88701). Protein extracts were diluted in NuPAGE LDS buffer (Thermo Fisher Scientific France, catalog no. NP0007) containing the NuPAGE-reducing agent (Thermo Fisher Scientific France, catalog no. NP0004). Protein samples (15 g/l) were then loaded onto NuPAGE Novex 12% BisTris gels (Thermo Fisher Scientific France, catalog no. NP0341) and subjected to electrophoresis in MOPS SDS running buffer (Thermo Fisher Scientific France, catalog no. NP0001). The protein bands were blotted onto PVDF Immobilon-FL membranes (Millipore France, catalog no. IPFL00010). Membranes were blocked with 5% BSA in PBS-T buffer (PBS with 0.1% Tween 20) and incubated overnight at 4°C with the primary antibody in blocking buffer containing 0.1% Tween 20 (Sigma, catalog no. P1379). The membrane was washed with 0.1% Tween 20 in PBS, and the antigen/ antibody complex was detected with HRP-conjugated secondary antibody. Membranes were washed with PBST and visualized with the ECL Western blotting kit (Thermo Fisher Scientific France, catalog no. 35055) on a G-Box (SYNGENE USA). The primary antibodies were the anti-␤III-tubulin (Promega France catalog no. G7121) and an anti-RABV G-cytodomain (produced in-house, NIV biological collection) antibody. The secondary antibody was an HRP-conjugated anti-mouse IgG antibody (GE Healthcare, UK, catalog no. NA9310).

MAST2 immunofluorescence in NS or SH-SY5Y cells
Neuronal cells were seeded on poly-L-lysine collagen-coated glass slide (3eϩ04 cells per slide) in a 24-well cell culture plate. Twenty four hours later, cells were rinsed with PBS buffer (three times) and fixed with 100% methanol at Ϫ20°C for 20 min. Then cells were washed three times with PBS, blocked with 5% goat serum in PBS for an hour at room temperature, rinsed three times with PBS, and incubated overnight at 4°C with the anti-MAST2 antibody (1:500, Thermo Fisher Scientific France, catalog no. PA5-75974) diluted in 5% goat serum in PBS. The day after, the glass slides were washed three times with PBS and incubated with the secondary antibody conjugated with AF555 (1:500, Thermo Fisher Scientific France, catalog no. A27039) and Hoechst 33342 (Thermo Fisher Scientific France, catalog no. H3570) for 1 h at room temperature. Then the slides were washed three times with PBS and mounted with prolong gold (Thermo Fisher Scientific France, catalog no. P36982). Cells were imaged with a Zeiss LSM70 confocal UV microscope.

Penetration of CPM-Neurovita in NS
The flow cytometry procedures, have been described in Prehaud et al. (47,57). Briefly, NGF-treated (6 h) NS cells, which had been seeded on 6-well plates (0.5.eϩ06 cells per well), were treated with 20 ng of CPM-Neurovita or mock-treated with PBS buffer for 30 min at 4°C. Then, cells were washed three times with PBS buffer before being permeabilized to detect both the molecules bound to the membrane and the molecules in the cytoplasm. The CPM-Neurovita molecules were detected with an anti-Strep Tag mAb C23.2 (IP collection, Dr. Pierre Lafaye) plus a Cy5-conjugated anti-mouse antibody (The Jackson Laboratory).
For WB analyses, 1eϩ06 NS cells cultivated in NDF medium were processed as above after being treated with 9 g of CPM-Neurovita or mock-treated with PBS buffer for 6 h at 37°C. Then cells were washed three times with PBS buffer, lysed with RIPA buffer, and processed for WB using the anti-Strep tag antibody C23.2.

MAST2 silencing
SH-SY5Y were silenced for the expression of MAST2 by using specific lentiviral shRNA MAST2 (TRCN0000001733, TRCN0000001735, and TRCN0000001736) and pL.KO1shRNA (RHS4080) as control. The lentiviral shRNAs were found among the collection of the Open Biosystems Expression Arrest TM TRC library (Broad Institute, Massachusetts Institute of Technology, and Harvard, sold by Thermo Fisher Scientific). Lentiviral shRNA viral stocks were produced as described previously. Briefly, SH-SY5Y (74 ϫ 10 6 cells in 12 ml of medium) was infected with 900 ng of p24 equivalent of each lentiviral shRNA vector stock for 2 days. Then an aliquot of the cells was taken, and total RNA was extracted, and gene expression was monitored by RT-qPCR. Neurite outgrowth was assayed in cultures exhibiting a knockdown of expression of Ͼ80%. Silencing experiment was conducted on triplicate individual samples.

Statistical analysis
ANOVA with Bonferroni's multiple comparison post hoc test and Student's t test were conducted with Graph Pad Prism version 7 or 8 software.

Isothermal titration microcalorimetry
ITC measurements were performed as described previously (16) with VP-ITC and ITC200 calorimeters (Malvern). MAST2-PDZ (10 to 30 M) was titrated at 298 K by 20 -45 consecutive injections of 2-7-l aliquots of the peptides (110 -280 M). Raw data were normalized and corrected for the heats of dilution of the peptides. Equilibrium dissociation constants were determined by fitting nonlinear curves to the corrected based on a model with one set of sites, with Origin7.0 software (OriginLab).

Stopped-flow experiments
The kinetics of association between MAST2-PDZ and the NV3 peptide were studied at 20°C with SFM300 stopped-flow Interfering peptides stimulating neurite outgrowth equipment with fluorescence detection (BioLogic). Typically, 150 l of MAST2-PDZ solution was rapidly mixed with an equal volume of peptide solution. For each peptide, three concentrations (2.5, 5, and 10 M) were tested at a constant MAST2-PDZ-peptide ratio of 1:10 to respect pseudo-firstorder conditions. The maximal concentration of MAST2-PDZ was 1 M. A mixing time of 15 ms was used (total flow rate of 20 ml⅐s Ϫ1 ), resulting in a dead time of 2.5 ms. Data points recorded between 0 and 3 ms were excluded from the fitting process. The flow cell was illuminated with a monochromatic UV light at 295 nm (bandwidth 8 nm), and emitted fluorescence was collected perpendicularly to the incident beam through a 325-nm highpass filter. Kinetic data were acquired and analyzed as described previously (18). We stored and processed 10 -12 averaged traces independently. Averaged kinetic traces from stoppedflow experiments were fitted independently with the following mono-exponential equation: F(t) ϭ A⅐e Ϫkobs⅐t ϩ offset. For each peptide concentration, the mean values of amplitude, the observed rate constant, and the corresponding standard deviations were calculated from the fitted values deduced from each series of averaged kinetic data. Intrinsic association and dissociation rate constants, k on and k off , were determined from the linear regression of the variation of k obs as a function of peptide concentration.

Hold-up
As described by Vincentelli et al. (20), 222 PDZ domains clones from a complete collection of 266 PDZome clones were successfully expressed. The automated holdup assay protocol was performed on a Tecan freedom Evo200 machine with 384well plates. A 41 M stock solution of biotinylated Neurovita peptide fixed on streptavidin beads was incubated with a 4 M His-MBP-PDZ domain solution. After incubation, the liquid phase of the plates was collected and filtered, and the concentration of unbound PDZ domains in the filtered flow-throughs was measured by microfluidic capillary gel electrophoresis with a LabChip GX II machine (Caliper, PerkinElmer Life Sciences). BIs were calculated as described in the supporting information of Ref. 20.