Discussion
Structure-based drug design allows the optimization of a chemical structure in which the structure of a protein is used as the basis to identify or design new chemical compounds that are predicted to bind to a target (Aparoy, Reddy & Reddanna, 2012). Alternatively, ligand-based approach is used in the absence of the protein structure and relies on knowledge of molecules that bind to the biological target (Aparoy, Reddy & Reddanna, 2012). To disrupt the Cav2.2–CRMP2 interaction, we had neither a structure of the complex nor any drug target data. Thus, based solely on the sequence of the CBD3 peptide, we resorted to first principles to identify the structural motif that triggers this interaction. Recently, we tackled a similar problem whereby using MDS, we identified the residue anchoring the interaction between a peptide derived from the disordered N-terminal of Kv2.1 and syntaxin1a, which in combination with syntaxin crystal structure led us to the discovery of a first-in-class small molecule neuroprotectant (Yeh et al., 2019).
In this study, we developed a novel pipeline that leverages the computational power of MDS to identify the most stable conformational motif of a peptide, i.e., CBD3, and then used it to develop a pharmacophore model to generate peptidomimetics – a medicinal chemistry approach where parts of the peptide are successively replaced by non-peptide moieties until a non-peptide small molecule is discovered (Perez, 2018). Peptides present desirable medicinal properties like predictable metabolism, good efficacy, safety, and tolerability; however, they are chemically and physically unstable due to rapid proteolysis and inadequate membrane permeability (Fosgerau & Hoffmann, 2015). Notably, editing peptide sequences to develop peptidomimetic analogs creates a promising class of therapeutics that can have inherent advantages, including oral administration, good membrane penetration ability, and enhanced biological activity (Smith, 2015). Considering these strengths, our model predicted first-in-class compounds to disrupt the Cav2.2–CRMP2 interaction. Specifically, we screened more than 27 million commercially available compounds in the ZINC database using the open access server ZincPharmer and identified 77 suitable compounds for experimental testing. The primary in vitroscreening identified 9 compounds that inhibit Ca+2influx by more than 50% relative to control. Furthermore, analysis of the shared pharmacophores among the 9 compounds permitted us to predict active chemotypes that when screened against MDS of CRMP2-derived peptides allowed us to fully rationalize active from inactive peptides. The latter provides a strong rationale for our method and its potential application for other targets. From a biological point of view, we show that by disrupting the Cav2.2–CRMP2 interaction, CBD3063: (i) inhibits Ca2+ influx in rat DRG neurons, (ii) decreases the functional activity of N-type Ca2+channels, (iii) reduces membrane expression of Cav2.2, (iv) does not affect the activity of other voltage-gated ion channels, (v) reduces sensory neuron excitability, and (vi) is antinociceptive in rats with a spared nerve injury model of pain.
Cav2.2 channels are almost expressed exclusively in neuronal tissue (Nowycky, Fox & Tsien, 1985) and are abundant at presynaptic nerve terminals where they trigger the release of neurotransmitters such as glutamate, calcitonin gene-related peptide, and substance P (Evans, Nicol & Vasko, 1996) via physically interacting with the synaptic release machinery (Zamponi, 2003). For this reason, Cav2.2 channels are critical determinants of increased neuronal excitability (Yang et al., 2018) and neurotransmission that accompany chronic neuropathic pain (Cizkova et al., 2002). In 2018, the Dolphin laboratory (Nieto-Rostro, Ramgoolam, Pratt, Kulik & Dolphin, 2018) reported a mouse expressing Cav2.2 channels with an extracellularly accessible hemagglutinin epitope tag engineered into their pore forming Cav2.2 α1 subunit (Cav2.2_HA) permitting, for the first time, identification of endogenous Cav2.2 channels in the plasma membrane of sensory neurons. These mice revealed disease-associated changes in the subcellular distribution of Cav2.2 in the pain pathway that confirmed the importance of these channels as suitable targets for development of novel pain therapies. Consistent with these findings, the therapeutic potential of targeting Cav2.2 has been demonstrated in Cav2.2 deficient mice which have reduced responses to mechanical stimuli, radiant heat, and chemical-induced inflammatory pain (Hatakeyama et al., 2001; Kim et al., 2001) and in nociceptive neurons specifically expressing the exon 37a variant of Cav2.2 (Cav2.2e[37a]) mice that display increased N-type Ca2+ currents and open channel probability when compared to neurons that only express the exon 37b variant of (Cav2.2e[37b]) (Bell, Thaler, Castiglioni, Helton & Lipscombe, 2004). In these mice, in vivo silencing of Cav2.2e[37a] prevented the development of mechanical allodynia and thermal hyperalgesia, demonstrating that targeting Cav2.2e[37a] channels by using splice isoform-specific gene silencing is an effective means for controlling the transmission of thermal and mechanical stimuli in pain conditions.
Despite compelling genetic evidence of the importance of Cav2.2 in pain, clinical development of N-type Ca2+ channel blockers have proven to be challenging. Although modulators of Cav2.2–α2δ interaction (i.e., Gabapentin and Pregabalin) are recommended as first-line treatment for neuropathic pain (Attal et al., 2006), they only partially alleviate chronic pain, are implicated in overdose deaths (Kuehn, 2022), and cause a litany of side effects (Zamponi, Striessnig, Koschak & Dolphin, 2015). A Cav2.2-selective, state-dependent inhibitor – N-triazole oxindole (TROX-1) – was reported by Merck but cardiovascular and motor impairment hampered its further development (Abbadie et al., 2010). Another study reported a sulfonamide-derived, state-dependent inhibitor of Cav2.2, but this compound was limited by structural liabilities of this class of compounds (Shao et al., 2012). Targeting other interacting partners of Cav2.2 can also result in reversal of pain symptoms. For example, the β3 auxiliary subunits interact with Cav2.2 channels (Ludwig, Flockerzi & Hofmann, 1997; Scott et al., 1996) to speed up their activation, increase their membrane localization, and increase neurotransmitter release (Richards, Butcher & Dolphin, 2004; Welling et al., 1993). Expression of β3 protein increases in neuropathic pain (Li et al., 2012) and β3-null mice exhibit suppressed pain responses due to decreased Cav2.2 currents (Murakami et al., 2002). Consistent with these findings, we reported that targeting the Cav2.2–β interaction reduces currents through Cav2.2 channels, inhibits spinal neurotransmission, and alleviates neuropathic pain (Khanna et al., 2019). Collectively, these studies converge on the idea that targeting auxiliary subunits of Cav2.2 channels is beneficial for pain management. Along these lines, accumulating evidence points to CRMP2 as a new auxiliary subunit of Cav2.2 channels (Striessnig, 2018).
Over the past decade, we have shown that interrupting the interaction between Cav2.2 and CRMP2 with CBD3 peptides is efficacious in reversing pain. We reported that disrupting Cav2.2 –CRMP2 binding with tat-CBD3 or myr-tat-CBD3 peptides does not affect memory, motor functions, or anxiety/depression, and does not produce any addictive behaviors (Brittain et al., 2011b; Francois-Moutal et al., 2015). Likewise, interrupting this interaction has neuroprotective effects (Brittain et al., 2011a; Brustovetsky, Pellman, Yang, Khanna & Brustovetsky, 2014; Ji et al., 2019). Tat-CBD3 (Brittain et al., 2011a; Brustovetsky, Pellman, Yang, Khanna & Brustovetsky, 2014) and R9-CBD3 (Ji et al., 2019) also disrupts the CRMP2–NMDAR interaction. As a result of this, both peptides attenuate NMDAR-mediated currents, have neuroprotective effects against glutamate-induced Ca2+ dysregulation (Brittain et al., 2011a; Brustovetsky, Pellman, Yang, Khanna & Brustovetsky, 2014) and protect against neurotoxicity caused the toxic fragment of amyloid-β (Aβ)25-35 (Ji et al., 2019). These studies demonstrate that interfering with Cav2.2-CRMP2 binding is safe, beneficial, and does not produce unwanted side effects.
We previously reported that a single amino acid point mutant of CBD3 (tat-CBD3A6K) decreased R- and T-type Ca2+ currents in DRG neurons (Piekarz et al., 2012), and that tat-CBD3 peptide also decreased T-type currents (Piekarz et al., 2012) in sensory neurons. Importantly, in the present study, CBD3063 did not exert an effect on any other voltage-gated calcium channel, which argues for selectivity of our peptidomimetic compound for Cav2.2. This is in line with our previous observations that the activity of low-voltage-activated Ca2+channels is independent of CRMP2 (Cai, Shan, Zhang, Moutal & Khanna, 2019). We have shown that Cdk5-mediated CRMP2 phosphorylation at residue S522 increases its binding to Cav2.2, which leads to an increase in calcium influx (Brittain, Wang, Eruvwetere & Khanna, 2012). To discard a potential effect of CBD3063 on CRMP2 phosphorylation, we measured pS522 CRMP2 and found that CBD3063 did not affect this posttranslational modification. This data correlates with our previous findings that interrupting CaV2.2-CRMP2 interaction with myr-tat-CBD3 does not affect CRMP2 phosphorylation (Francois-Moutal et al., 2015). Along the same lines, we investigated the regulation of sodium channels by CBD3063 since we previously described that phosphorylation and SUMOylation of CRMP2 regulatesNav1.7channel trafficking and activity (Dustrude, Moutal, Yang, Wang, Khanna & Khanna, 2016; Dustrude, Perez-Miller, Francois-Moutal, Moutal, Khanna & Khanna, 2017; Dustrude, Wilson, Ju, Xiao & Khanna, 2013). In the present study, we found that CBD3063 does not directly affect currents through sodium channels, or indirectly through an effect on CRMP2 phosphorylation or CRMP2 SUMOylation (data not shown), indicating that this regulation is unaffected by CBD3063 treatment. These data, together with the lack of effect on K+ currents, suggest that targeting this interaction with CBD3063 does not result in inhibition of other ion channels relevant for pain signaling or in dysregulation of CRMP2 posttranslational modifications.
Despite the documented success of CBD3 peptide in achieving analgesia without side-effects, peptides are nevertheless hampered by (i) short half-life caused by their poor in vivo stability (Bruckdorfer, Marder & Albericio, 2004) which may be attributed to the presence of numerous peptidases and excretory mechanisms that inactivate and clear peptides, and (ii) negligible bioavailability caused by digestive enzymes that are designed to break down amide bonds of proteins and cleave the same bonds in these peptides. To circumvent some of these problems, in the present work we utilized our pharmacophore modelling to generate small molecule peptidomimetics to improve upon the biological activity of the CBD3 peptide by mimicking the chemical features responsible for bioactivity with enhanced drug-like properties. As a result, in contrast to the short-lived actions of CBD3 peptides, the peptidomimetic developed here (i.e., CBD3063) exhibited prolonged antinociceptive profile following a single intrathecal administration as well as long-lasting (>14 days) reversal of mechanical allodynia with repeated intrathecal dosage in rats with chronic neuropathic pain. The latter also points to a lack of tolerance to CBD3063. The prolonged antinociceptive effect of CBD3063 could be attributed to the significant decrease of action potential firing in sensory neurons that could potentially translate into a decrease in spinal cord excitability and neurotransmitter release (not tested here). These findings are congruent with those observed with an adenoviral form of a mutant CBD3 (i.e., AAV6-CBD3A6K), which reduced the firing of dorsal horn neurons and reversed mechanical allodynia and thermal hyperalgesia for up to 6 weeks following intraganglionic delivery of the virus in rats with tibial nerve injury (Yu et al., 2019a).
Here, we have presented a translational workflow that uses structural modeling to direct the resolution of protein-protein interactions involving poorly characterized disordered domains, resulting in both mechanistic insights and the identification of an analgesic. This translational workflow led to the discovery of CBD3063 as a first-in-class, CRMP2-based peptidomimetic, which selectively regulates Cav2.2 to achieve analgesia. We note that although CBD3063 exhibits promising drug-likeness (QED score, Table 1 ) it does have marginal predicted blood-brain barrier penetration (BBB score,Table 1 ). To improve to CNS exposure, CBD3063 can be further optimized to lower the topological polar surface area by, for example, modifying and/or removing the acetamide and urea bonds. Another option to increase hydrophobicity is to introduce aromatic group (e.g., in the acetamide area) or to strategically position fluorine atoms to lower capacity of N-H donor (e.g., as 3-fluro-2-aminopyridyl moiety). Overall, we intend to explore targeted changes in CBD3063 to improve its predicted ADME properties while maintaining its current biochemical profile.