Spirocycles in Drug Discovery

Medicinal chemists are constantly in search of molecules that explore new chemical space whilst looking for novelty amongst the plethora of patented molecules. Three-dimensionality is also key in maintaining solubility of new drug compounds.

Oxetanes have long been known in medicinal chemistry as very useful bioisosteres or surrogates for carbonyl or gem-dimethyl motifs. They reduce lipophilicity and therefore increase solubility when compared to their gem-dimethyl counterparts and reduce metabolism and are less likely to be covalent binders in comparison with the corresponding carbonyls. They have also been shown to be excellent hydrogen bond acceptors,   Penny 1

Whilst oxetanes have been thoroughly explored in drug discovery as a single motif, their composition in spirocycles has been relatively unexplored. In 2008, Carreira et al compared parent amino heterocycles (eg. azetidines, pyrrolidines and piperidines) with the corresponding spirocycle incorporating oxetanes.

Penny 2

In all cases, the spirocycles were stable at pH 1-10 and were considerably less basic than their parent compounds due to their conformation. All have lower logD than their gem-dimethyl and carbonyl analogues. In the cases shown above, the spirocycles also exhibited far lower intrinsic clearance than the carbonyl or parent compounds.

Morpholine moieties are also regularly found in drug discovery programs and in final drug candidates, however, the spirocyclic surrogate seems to fall under the radar despite it being much more soluble, less lipophilic and more metabolically stable.

Penny 3

In addition, the oxygen lone pair is shown to be 1.3 angstroms further out than in morpholine and can therefore be regarded as an elongated morpholine (similar to piperidone but much more metabolically stable). This spirocycle could be used to probe chemical space within a binding site or could in fact have a stronger interaction within the binding site due to probing deeper within a given pocket of the site.                          Penny 4

Carreira et al later looked at the differences between morpholines, piperidines, piperizines and thiomorpholines compared with their spirocylic analogues in terms of their physico and biochemical properties. In general, the spirocyclic compounds had higher solubility, lower logD and were intrinsically more stable in human and liver microsomes compared to the parent compounds.

Penny 5

The authors proceeded to synthesise the antibacterial compound Ciprofloxacin with the piperazine motif replaced with either a piperizine-like spirocycle (Compound A) or the morpholine-like spirocycle (Compound B). Both compounds showed comparable MIC and most interestingly, neither A nor B showed any sign of metabolism in human microsomes whereas Ciprofloxacin showed slight metabolism.

Penny 6

Mykhailiuk et al published an interesting paper very recently which is what initially sparked this blog. They comment on the distinct lack of spirocyclic compounds in FDA approved molecules despite the number of patents surrounding such molecules slowly increasing in recent years. The authors describe spirocycles as an “overlooked motif for drug discovery”.  Their work surrounds the replacement of 2-substituted piperidines with 2-azaspiro[3.3]heptane surrogates. These compounds are not synthetically challenging as one might imagine and they originate from commercially available, relatively cheap starting materials with synthesis easily scalable to 50g of product in a single batch. They also demonstrate that the synthesis has scope and can tolerate a wide range of substituents.

The synthesis of the spirocyclic analogue of FDA-approved local anaesthetic Bupivacaine was completed by this group to compare this motif in a drug discovery setting. The results showed that despite their analogue exhibiting slightly higher plasma protein binding and in-vitro metabolism, in-vivo, their analogue showed faster onset and similar duration to Bupivacaine with lower systemic toxicity than the FDA approved drug.                               Penny 7Whilst spirocycles remain unrepresented in FDA approved molecules, it is clear that interest both within academia and industry is growing. With studies like these being published more frequently demonstrating the improved drug-like properties of these compounds when compared to their monocyclic counterparts, it is likely that it won’t be long before these motifs are seen in approved drugs.

Blog written by Penny Turner


Angew. Chem. Int. Ed., 2010, 49: 9052-9067

Angew. Chem. Int. Ed., 2008, 47: 4512-4515

Angew. Chem. Int. Ed., 2010, 49: 3524-3527

Angew. Chem. Int. Ed., 2017, 56: 8665-8869

Dopaminergic neurons display antigens: autoimmunity in Parkinson’s disease pathology

Parkinson’s disease (PD) is a chronic, progressive degenerative disorder of central nervous system.  Its pathology is characterised by selective loss of dopaminergic neurons in the nigrostrial pathway, and clinical manifestations are exhibited as motor impairments, including resting tremor, bradykinesia, and rigidity.  Current medications offer symptomatic relief but, to date, do not address the dopaminergic neuronal death. The lack of understanding of the etiology of this selective cell death still remains the major stumbling block in the development of neuroprotective therapies. Current research implicates a number of key molecular mechanisms compromising the function and survival of this specific subset of neurons, and these involve abnormal protein accumulation and phosphorylation, mitochondrial dysfunction, oxidative damage and deregulated kinase signalling.  Although the current hypothesis focuses on the toxic aftermath of α-synuclein protein deposits, an alternative theory pioneered by Dr. Sulzer’s group within the Department of Neuorology at Columbia University, implies a role of the immune system in PD pathology, more specifically suggesting that Parkinson’s is in fact an autoimmune disease.

As mentioned, pathological features of PD include the loss of nigrostriatal dopamine neurons and the formation of Lewy bodies rich in fibrillar α-synuclein. This 140-amino-acid protein is abundantly expressed at a relatively high level throughout the brain (Iwai et al., 1995) and is thought to play physiological roles in the regulation of the dopamine transporter (Sidhu et al., 2004). However, misfolding of α-synuclein into protofibrils and higher-order oligomers (Uversky et al., 2002) leads to a toxin gain of function (Martin et al., 2006), which is associated with the pathogenesis of neurodegeneration (Giasson et al., 2000). Furthermore, this protein has been genetically linked to the early onset of familial PD (Kruger et al., 1998). However, the mechanism by which α-synuclein causes neurodegeneration remains unclear.

In addition to α-synuclein dysfunction, PD pathology is also characterised by a sustained microglial reaction throughout the disease progression Imamura et al., 2003). Microglial cells are the resident immune cells in brain and play a major part in the neuroinflammatory response (Soulet and Rivest, 2008). On an epidemiological level, the contribution of an inflammatory response in neurodegeneration is evidenced by the decreased risk of falls in PD patients on administering the non-steroidal anti-inflammatory drug (NSAID), ibuprofen (Gagne and Power, 2010). On a cellular and molecular level, the significant elevation in inflammatory cytokines has been found in both the cerebrospinal fluid and postmortem brain of PD patients (Mogi et al., 1994). These cytokines have been reported to induce the death of dopaminergic cells (Vivekanantham et al., 2015), and thus facilitating neurodegeneration in PD.

Dr. Sulzer’s group have recently identified the presence of antigens displayed on dopaminergic neurons in post-mortem brain tissues. It has long been argued that brain cells are safe from immune cell attack because they do not display these molecular markers for immune cell target recognition.  However, these new findings indicate that they can in fact be targeted.  Abnormal processing of self-proteins can produce epitopes, which are presented by major histocompatibility complex (MHC) proteins to be recognised by specific T cells that have escaped tolerance during thymic selection (Marrack and Kappler, 2012). Such actions by the acquired immune system have been implicated in autoimmune disorders, including type-1 diabetes. Although PD has not before been linked to autoimmunity, it does demonstrate altered protein processing. As previously described, activation of microglia and elevated cytokine levels have described in PD patients, indicating a role of the innate immune system.  But what evidence is there to implicate the acquired immune system?

Rationale for targeting the adaptive arm of the immune system as a therapeutic strategy in PD was initially provided by Brochard, et al (2009). It was found that CD8+ and CD4+ T cells, but importantly not B cells, infiltrate the substantia nigra (SN) in the 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine (MPTP) mouse model of PD during the course of neuronal degeneration, which is consistent with postmortem human PD specimens. Further investigation concluded that T cell-mediated dopaminergic toxicity is almost exclusively arbitrated by CD4+ T cells (Brochard et al., 2009).  The Sulzer group from Columbia University, more recently, reported antigen presentation by MHC class I expression in dopamine neurons in the SN of adult human PD brains (Sulzer et al., 2017).  This was the result of activation by cytokines released from microglia.  CD8+ T cells kill neurons that present the appropriate combination of MHC class I and peptide (Cebrián et al., 2014).  On comparison of PD patients to age-match healthy controls, the group then identified two antigenic regions in α-synuclien (Fig. 1a).  The first near the N terminus elicited an apparent class II restricted IL-5 and IFNγ response (Fig. 1b–d). The second antigenic region was near the C terminus required phosphorylation of amino acid residue S129 that resulted in a markedly higher IL-5 responses in patients with Parkinson’s disease than in healthy controls. The Y39 antigenic region is noticeably close to the α-synuclien mutations that cause Parkinson’s disease (A30P, E46K, H50Q, G51D, A53T; Hernandez et al., 2016), and phosphorylated S129 residues, found in the second antigenic region, are present at high levels in Lewy bodies of patients with Parkinson’s disease (Fujiwara et al., 2002). Finally, blood tests have revealed that people with Parkinson’s show an immune response to these antigens, while people who don’t have the condition do not (Sulzer et al., 2017).


Figure 1α-synuclein autoimmune responses are directed against two regions. a, Sequence of α-synuclein. Antigenic regions are highlight with dashed line with amino acids Y39 and S129 shown in bold. b-d, Magnitude of responses expressed as SFC per 106 PBMC’s per peptide and participant combination. Left, response to all overlapping native α-synuclein 15-mer peptides in patients with PD (n=733) and control (n=372).  Right, response against specific 15-mers.  Grey shading indication the antigenic region containing Y39. e-g,  Magnitude of responses. Left, response to all native phosphorylated S129 α-synuclein 15-mer peptides in patients with PD (n=150) and control (n=72). Right, response against specific S129 peptides. Closed circles, patients with PD (n=19, † indicates peptides; n=25, all other peptides); open circles, control (n=12).  Two-tailed Mann-Whitney U-test; NS, not significant.

These findings are the first time the immune system has been associated with a major pathological role in Parkinson’s. They present an argument for the classification of PD as an autoimmune disorder.  However, what isn’t clear is which comes first: does the immune response directly causes neuron death, or does the disease result in a heightened immune response?  If suppression of this autoimmune response does indeed stop disease progression, these findings could provide an attractive target for therapeutic intervention.

Blog written by Victoria Miller


Brochard et al., J. Clin. Invest. (2009) 119, 182–192.

Cebrián et al., Nat. Commun. (2014) 5, 3633.

Fujiwara et al., Nat. Cell Biol. (2002) 4, 160–164.

Gagne & Power, Neurology (2010), 74, 995–1002.

Giasson et al., Science (2000), 290, 985–989.

Hernandez et al., J. Neurochem. (2016) 139, 59–74.

Imamura et al., Acta Neuropathol. (2003), 106, 518–526.

Iwai et al., Neuron (1995), 14, 467–475.

Kruger et al., Nat. Genet.(1998), 18, 106–108.

Marrack & Kappler, Cold Spring Harb. Perspect. Med. (2012) 2, a007765.

Martin et al., J. Neurosci. (2006), 26, 41–50.

Mogi et al., Neurosci. Lett. (1994), 165, 208–210.

Sidhu et al., FEBS Lett. (2004), 565, 1–5.

Soulet & Rivest, Curr. Biol. (2008), 18, R506–508.

Sulzer, Nature (2017) 0.

Uversky et al., J. Biol. Chem. (2002), 277, 11970–11978.

Vivekanantham et al., Int. J. Neurosci. (2015), 125, 717–725.

The Necessary Nitrogen Atom

In the modern drug discovery process, medicinal chemists strive for high-impact design elements during multiparameter optimisation of lead compounds into efficacious drug candidates. One of such well-known design elements, often referred to as “the magic methyl effect”, “the methyl walk” or “the methyl scan”, involves a replacement of a H atom with a Me group and, as the result, can lead to profound potency improvement (>100-fold).

The authors of the recently published perspective (DOI: 10.1021/acs.jmedchem.6b01807) discuss another common design element for multiparameter optimisation: substitution of a CH group with a N atom in aromatic and heteroaromatic ring systems. Such replacement, when going from a simple benzene ring to pyridine, results in profound changes in molecular and physicochemical properties, such as distribution of electron density, polar surface area, basicity, which in turn affects lipophilicity and solubility. In more complex chemical structures, these changes can impact upon a number of intra- and intermolecular, orbital, steric, electrostatic, and hydrophobic interactions, such as lone pair, dipole−dipole, hydrogen bonding, metal coordination, van der Waals, σ-hole, σ*S−X, and π-system interactions, which in turn can translate into modified pharmacological profiles.

The authors then illustrate an extensive number of drug discovery case studies from the recent literature, where a replacement of a CH group with a N atom resulted in ≥10-fold improvement in at least one key pharmacological parameter, as shown, for example, in Figure 1. The replacement of the C7 CH group in (1) with a N atom to give (2) resulted in a 300-fold improvement in biochemical potency, Cdc7 IC50 = 2700 and 9.0 nM for (1) and (2), respectively. This potency improvement was attributed to the different conformational preferences of the two analogues. The biaryl dihedral angle of >150° in indole (1) greatly differs from that of aza-indole (2) (dihedral angle of 0°), facilitated by a steric clash of H7 and H6′ in (1) and lone pair repulsion between N7 and N2′ in (2).


Figure 1. An example of a necessary nitrogen atom on potency improvement, likely due to different conformational preferences.

The effects of such replacement on basicity, lipophilicity, polar surface area, and hydrogen bonding capacity are relatively predictable; whereas effects on aqueous solubility, passive permeability, efflux profiles, active transport, protein binding, and metabolic stability can be more whimsical and counterintuitive. In some examples, no rational explanation could be given for the observed effect.

With regards to changes in potency, as the authors are cautious to point out, there is an approximately equal probability of increasing or decreasing potency by exchanging CH groups and N atoms, based on a matched molecular pair analysis (MMPA) of available data. However, these findings are very similar to those discussed for the magic methyl effect.

This perspective is not a manual for what pharmacological improvements will be realized upon the substitution of a CH group with a N atom, but rather as an extensive exemplar of the dramatic improvements that can be achieved under certain circumstances. The systematic N atom scan (N-scan) should be exploited where appropriate, particularly in cases where the preferred binding pose of the ligand is not known.

Some recommendations for the N-scan tactics are summarised below. The newly installed N atom might:

  1. Engage in a hydrogen bond with specific residues of the target receptor or receptor-bound water molecules that need to be satisfied
  2. Remove unfavorable van der Waals interactions the replaced CH group made with the target receptor
  3. Form unfavorable electrostatic interactions with an antitarget receptor
  4. Have a positive effect on the binding conformation of the ligand
  5. Mask a hydrogen bond donor in the ligand
  6. Reduce the basicity or HBA strength of an existing N atom in the ligand
  7. Evenly distribute the polar surface area of the ligand
  8. Properly tune the lipophilicity of the ligand
  9. Be shielded by other substituents or functionality in the ligand
  10. Stabilize chemically labile functionality in the ligand
  11. Replace a metabolically labile CH group in the ligand

Medicinal chemists have to constantly juggle the design of biologically active molecules with drug-like properties according to the rules of Lipinski and Veber, in order to achieve good permeability and absorption levels, which in turn lead to high oral bioavailability. These properties are particularly important when considering therapeutic targets located in the central nervous system (CNS) behind the blood−brain barrier (BBB). Keeping the number of hydrogen bond acceptors down is one of such requirements; however, in some cases, it may turn out that the introduction of one (or two) extra N atoms may be necessary.

Blog written by Irina Chuckowree


TMEM16A: 2 pores, or not 2 pores

Two groups (Lim et al. 2016 & Jeng et al. 2016) with companion papers in the Journal of General Physiology have tried to answer the question, does TMEM16A have 1 or 2 Cl conducting pores. They have done this with functional studies using covalently linked mouse TMEM16A (mTMEM16A) subunits over expressed in HEK293T cells with one of the two subunits carrying mutations that change the functional properties of TMEM16A ion channels.

Functional members of the TMEM16 family were known to consist of 2 identical subunits (Fallah et al., 2011; Sheridan et al., 2011; Tien et al., 2013) however, with Brunner et al. (2014) solving the crystal structure of the phospholipid scramblase TMEM16 family member from the fungus Nectria haematococca (nhTMEM16) confirmed that TMEM16 molecules adopt a homodimeric architecture. With each subunit harbouring a hydrophilic groove, the “subunit cavity”, located at the periphery of the dimer that is exposed to the lipid bilayer (Figure 1A). The location of the Ca2+ binding site in the hydrophobic part of the phospholipid bilayer offers a plausible explanation for the observed voltage dependence of calcium activation in TMEM16A ion channels, as Ca2+ has to cross part of the transmembrane electric field to reach their binding site.

However, because of the unique architecture of the subunit cavity, forming a half-channel that is exposed to lipids on one side, a potential alternative arrangement of subunits in ion channels of TMEM16A and B was envisioned. In this alternative arrangement, the 2 exposed half-channels can theoretically form a single enclosed aqueous pore that would be completely surrounded by protein residues, akin to other know channel architecture (Figure 1B). In such an arrangement, the Ca2+ binding site and the residues lining the ion conduction path would be in close proximity, and it could thus be expected that any changes in the pore or the Ca2+ binding site in one of the subunits may affect the activation and conduction properties of the entire protein. In contrast, in the case of the separated 2 pore ion conduction pathways, the same mutation may only affect activation and conduction in one of the 2 pores (Figure 1A).

Roy 1

Figure 1. A) Schematic representation of TMEM16A containing two pores that are independently regulated by Ca2+. B) Hypothetical alternative arrangement of subunits resulting in a single pore. Lim et al. 2016 J. Gen. Physiol. 148:5, 375-392.

In CLC proteins, this question was addressed by kinetic analysis of single channel recordings which are known to consist of 2 independent ion transport pathways (Miller 1982). But, because the low single-channel conductance of TMEM16A precludes such a strategy, Lim et al. (2016) studied macroscopic currents of the over expressed, covalently linked subunits of mTMEM16A in HEK293T cells with one of the 2 subunits carrying mutations that change the functional properties of the channel and therefore attempting to answer whether mTMEM16A ion channels consist of 1 or 2 pores, and if 2, whether the 2 pores function independently.

Recordings were performed in the inside-out patch configuration and not whole cell, which under some of their test conditions was potentially non-physiological. They showed that these linked proteins were stable and dimeric, and that the covalent link between the 2 subunits does not significantly alter the functional properties of the TMEM16A protein. The covalent linked dimers showed the established Ca2+ and voltage dependent gating, Ca2+ binding cooperativity and chloride selectivity of TMEM16A ion channels. However, they have also shown a biphasic Ca2+ activation is evident upon careful correction of the irreversible rundown that becomes more severe at higher Ca2+ concentrations with a predominant Ca2+ activation at low and a second shallow step at high Ca2+ concentration, for the linked WT-WT dimer the EC50 for Ca2+ activation was 0.209 µM and 724µM respectively. The second activation lacks any voltage dependence and might thus reflect the interaction of Ca2+ with an unknown low-affinity site located at the cytoplasmic part of the channel. Their results suggest that exposure to 1mM Ca2+ does not change the high anion over cation selectivity of the channel, nor its conductance, but that it results from an increase in the open probability.

More importantly they have also demonstrated that both subunits act independently with respect to Cl permeation and gating characteristics. They have shown WT linked dimers have the same properties as wild type TMEM16A channels and when they link WT-WT and WT-mutated channels, the signature of the linked dimer retains the functional signature of each subunit of the dimer, inferring 2 independent conducting pores in the dimer. Also, besides the unaltered anion selectivity and conductance of the construct containing only a single activatable subunit, provides additional evidence for the spatial separation of both pores. Functional independence is also corroborated by experiments on constructs where 2 subunits show different potency of Ca2+ activation and where each activation step retains the signature of the non-concatenated counterparts (Figure 2).

Roy 2

Figure 2. Rundown-corrected concentration-reponse relation of the WT-E702Q concatemer at 80mV. The solid line is the best fit to the triphasic Hill equation. Dashed lines indicate the first activation of WT (green) and E702Q (orange) at 80mV. Lim et al. 2016 J. Gen. Physiol. 148:5, 375-392.

Their experiments with mutant containing dimers thus provide strong functional evidence for independent activation of 2 separate ion conduction pores in the covalently linked dimeric mTMEM16A channel. Although the results presented in this study suggest that activation of different subunits of TMEM16A opens distinct pores, the exact mechanism of TMEM16A activation is still a subject of much speculation. Although, their evidence implies that the TMEM16A ion channel may contain two pores, and Ca2+ activation of individual subunits opens the pore associated with that activated subunit (Figure 3).

Roy 3

Figure 3. Cartoon summarizing the functional properties of TMEM16A. Ion conduction pores in the dimeric protein are indicated in light blue. Ca2+ is displayed as dark blue, and Cl- is displayed as red spheres. Lim et al. 2016 J. Gen. Physiol. 148:5, 375-392.

As yet no structural information on either mouse or human TMEM16A protein has been published, neither do we have, at the moment, the resolution to see TMEM16A crystal structure in either the open or closed state. However, Lim et al. (2016) have provided functional evidence that mTMEM16A ion channel has 2 independently functional pores, at high Ca2+ concentrations increases their open probability, potentially giving us a condition for high resolution crystallography to confirm whether TMEM16A has 2 pores, or not 2 pores.

Blog written by Roy Fox


Brunner et al. (2014) Nature 516:207-212.

Fallah et al. (2011) Mol. Cell. Proteomics. 10:M110.004697.

Jeng et al. (2016) J. Gen. Physiol. 148:5, 393-404.

Lim et al. (2016) J. Gen. Physiol. 148:5, 375-392.

Miller et al (1982) Proc. Natl. Acad. Sci. 299:401-411.

Ni et al. (2014) PLoS One. 9:e86734.

Sheridan et al. (2011) Exp. Physiol. 97:177-183.

Tien et al., (2013) Proc. Natl. Acad. Sci. USA. 110:6353-6357.