Protein aggregation-inhibition: a therapeutic route from Parkinson’s disease to sickle cell anemia

Abstract

Protein aggregation is implicated in multiple diseases, so-called proteinopathies, ranging from neurodegenerative disorders such as Alzheimer’s disease and Parkinson’s disease (PD) to type 2 diabetes mellitus and sickle cell disease (SCD). The structure of the protein aggregates and the kinetics and mechanisms of aggregation have been the object of intense research over the years toward the development of therapeutic routes, including the design of aggregation inhibitors. Nonetheless, the rational design of drugs targeting aggregation inhibition remains a challenging endeavor because of multiple, disease-specific factors, including an incomplete understanding of protein function, the multitude of toxic and non-toxic protein aggregates, the lack of specific drug binding targets, discrepant action mechanisms of aggregation inhibitors, or a low selectivity, specificity, and/or drug potency, reflected in the high concentrations required for some inhibitors to be effective. Herein, we provide a perspective of this therapeutic route with emphasis on small molecules and peptide-based drugs in two diverse diseases, PD and SCD, aiming at establishing links among proposed aggregation inhibitors. The small and large length-scale regimes of the hydrophobic effect are discussed in light of the importance of hydrophobic interactions in proteinopathies. Some simulation results are reported on model peptides, illustrating the impact of hydrophobic and hydrophilic groups in water’s hydrogen-bond network with an impact on drug binding. The seeming importance of aromatic rings and hydroxyl groups in protein-aggregation-inhibitor-drugs is emphasized along with the challenges associated with some inhibitors, limiting their development into effective therapeutic options, and questioning the potential of this therapeutic route.

Keywords:

• Proteinopathies
• protein aggregation inhibitors
• hydrophobic interactions
• drug design
• disease

Introduction

Protein aggregation is implicated in several human pathologies, ranging from sickle cell disease (SCD) (Pauling et al. 1949; Ingram 1956, 1957; Noguchi and Schechter 1985; Eaton and Hofrichter 1990), a red blood cell disorder, to various neurodegenerative diseases (NDs) (Aguzzi and O’Connor 2010; Sweeney et al. 2017), such as Alzheimer’s, Huntington’s, or Parkinson’s disease (PD), commonly known as proteinopathies or protein conformational diseases. Although the root cause of most NDs, including PD, is not exactly known, compelling evidence (Spillantini et al. 1997, 1998; Feany and Bender 2000; Masliah et al. 2000; Bucciantini et al. 2002; Soto 2003) posits a relationship with protein misfolding and aggregation, resulting in the formation of abnormal protein aggregates, generally referred to as amyloids (Trojanowski and Lee 2003; Aguzzi and O’Connor 2010; Sweeney et al. 2017). SCD, on the other hand, is probably the most well-known proteinopathy, coined the first molecular disease by Pauling et al. (1949), being associated with the reversible aggregation of a mutated form of hemoglobin (i.e. sickle cell hemoglobin) most common in some parts of sub-Saharan Africa (Williams 2016) and known to confer protection against malaria (Ferreira et al. 2011).

Several therapeutic routes have been concomitantly explored both in SCD and NDs in general. A common route includes the reduction of the monomeric precursor protein. This is because the kinetics of aggregation of both amyloid and HbS fibers depends on the concentration of the precursor as well as on the cell environment. Hydroxyurea, for instance, the most widely used drug to treat SCD, increases the levels of fetal hemoglobin (HbF), which does not polymerize (Charache et al. 1995; Mehanna 2001; Vinjamur et al. 2018). The increase in HbF decreases the concentration of HbS, enhancing the delay time that precedes fiber growth, ultimately reducing the percentage of sickled erythrocytes.

Preventing or reducing protein aggregation in SCD (Eaton and Bunn 2017; Tisdale et al. 2020; Eaton 2022) and NDs through a direct disruption of the aggregation mechanism is another possible therapeutic strategy since it could potentially prevent the diseases’ progression without compromising the biological function of the protein. Ultimately, this route requires the rational design of drugs that can hinder or delay the aggregation process and, thus, a molecular knowledge of the aggregation mechanism of the proteins involved in the disease.

Herein, we provide an overview of the main features of two very different proteinopathies, PD (section “Parkinson’s disease”) and SCD (section “Sickle cell disease”), along with some of the molecules, including small peptides, that have been found to exhibit some aggregation inhibitory activity and their proposed action mechanisms (section “Protein aggregation inhibitors”).

Protein aggregation, much like protein folding, is deeply connected with hydrophobic interactions (Kauzmann 1959; Tanford 1980; Dill 1990), which in turn are influenced by the surrounding environment, driving the conformational search toward a thermodynamic stable or native state. Hence, hydrophobic interactions are also discussed (section “Hydrophobic effect and protein aggregation”) concerning protein aggregation and their influence on the design of aggregation inhibitors (section “Hydrophobic effect and drug design”).

The main goal of this review is twofold: (i) expose putative similarities among aggregation inhibitors proposed for distinct proteinopathies, including their seeming “universal” or non-specific action mechanism, and (ii) point out some of the main challenges concerning drug design associated with proteinopathies, with emphasis on the harnessing of the hydrophobic effect and the understanding of the action mechanism of some inhibitors and its relationship with specific chemical groups long reported to be pivotal to protein aggregation inhibition.

Parkinson’s disease

PD is the second most prevalent disease among NDs, next to Alzheimer’s disease, and the most common movement disorder. PD’s etiology has been associated with the formation of cytotoxic oligomers (Breydo et al. 2012; Cremades et al. 2017), of which α-synuclein (α-syn) (Maroteaux et al. 1988) is the primary component. These prefibrillar transient oligomers accumulate in intracellular inclusions called Lewy bodies and Lewy neurites (Spillantini et al. 1997, 1998), and are now (Goldberg and Lansbury 2000; Caughey and Lansbury 2003; Karpinar et al. 2009; Winner et al. 2011) thought to be the main culprit, as opposed to the misfolded monomer and mature (insoluble) fibrils, in the loss of nigral dopaminergic neurons (Feany and Bender 2000; Masliah et al. 2000). Despite the cell toxicity mechanism remains elusive, this is believed to be connected with perturbations in the neuronal membrane (Lashuel et al. 2002; Winner et al. 2011; Goedert et al. 2017). In addition, α-syn aggregates spread through a prion-like mechanism across brain cells, accelerating disease’s progression (Masuda-Suzukake et al. 2013; Ma et al. 2019; Cascella et al. 2022).

α-Synuclein aggregation

α-syn is a 140 amino acid intrinsically disordered protein (IDP) mainly expressed in the central nervous system and whose function is not completely understood, although several putative functions have been put forward (Fink 2006; Cheng et al. 2011; Breydo et al. 2012; Lashuel et al. 2013; Lee et al. 2008; Norris et al. 2004). α-syn is comprised of three distinct domains (see Figure 1), the N-terminal (N-term), a membrane-binding domain that tends to form α-helices, encompassing amino acids 1–60 (Davidson et al. 1998; Eliezer et al. 2001); the so-called non-amyloid-β component (NAC) (Uéda et al. 1993), a highly hydrophobic and amyloidogenic domain comprising amino acids 61–95; and the C-terminal (C-term) domain, a more disordered region comprised of the amino acids 96–140 (Breydo et al. 2012). Besides PD, α-syn aggregates are at the center of dementia with Lewy bodies (Baba et al. 1998) and multiple system atrophy (Spillantini et al. 1998; Tu et al. 1998), jointly known as synucleinopathies or Lewy body diseases (Trojanowski and Lee 2003; Uversky 2007; Goedert et al. 2017).

Figure 1. Membrane-bound α-syn structure (2kkw.pdb) (Rao et al. 2010) showing its three regions, N-term (positively charged), the NAC (highly hydrophobic), and the C-term (negatively charged). The last amino acid of the N-term domain (Lys-60) and the first amino acid of the C-term domain (Lys-96), encompassing the NAC, are shown as vdW spheres. The chain colors correspond to the following secondary structures: α-helix (orange), turn (yellow), and coil (ice blue).

The structure of α-syn amyloid polymorph fibrils has been assessed by cryo-electron microscopy and NMR (Tuttle et al. 2016; Li, Ge, et al. 2018; Li, Zhao, et al. 2018; Ni et al. 2019) (see Figure 2) including membrane-bound α-syn (Ulmer et al. 2005; Rao et al. 2010) (Figure 1). In addition, the structure of toxic oligomers (Cremades et al. 2012; Bengoa-Vergniory et al. 2017) has been assessed through multiple experimental techniques, including circular dichroism, small-angle X-ray scattering, and atomic force microscopy-infrared spectroscopy (Hong et al. 2008; Ruggeri et al. 2015; Zhou and Kurouski 2020).

Figure 2. (a) α-syn monomer extracted from the α-syn experimental (NMR spectroscopy) protofibril reported by Tuttle et al. (2016) (2n0a.pdb). (b) α-syn dimer extracted from the same experimental protofibril; cartoon representation showing a β-sheet-rich region in the NAC (amino acids 61–95) domain; some β-sheet is also visible in the N-term region. The last amino acid of the N-term domain (Lys-60) and the first amino acid of the C-term domain (Lys-96), encompassing the NAC, are shown as vdW spheres. The chain colors correspond to the following secondary structures: 3₁₀ helix (orange), β-sheet (red), turn (yellow), and coil (ice blue).

The transient (Zurlo et al. 2021) and heterogeneous aggregational nature of α-syn poses a challenge to a comprehensive understanding of the aggregation mechanism(s) and kinetics (Arter et al. 2020), and their relationship with the onset of PD (Breydo et al. 2012; Cremades et al. 2017; Cremades and Dobson 2018). Nucleation–polymerization and nucleation–conversion–polymerization kinetic mechanisms have been proposed to describe the formation of soluble oligomeric species and the growth into insoluble mature fibrils (Cremades et al. 2017). The latter foresees a conversion stage where disordered oligomers with little or no stable β-sheet structure convert into more orderly and stable oligomers before they grow into fibrils (Cremades et al. 2017). Furthermore, the trigger behind the nucleation stage is thought to be associated with a conformational transformation (i.e. misfolding) of the natively unfolded protein into a more aggregation-prone partially folded intermediate (Uversky et al. 2001; Uversky and Eliezer 2009). The latter induces the formation of β-sheet-rich structures, a hallmark of the cytotoxic oligomers, nearly absent in the native α-syn. Further, the kinetics of aggregation is promoted by its binding with lipid membranes (Galvagnion et al. 2015, 2016), in addition to the temperature, pH, or osmolytes.

In addition to the wild type α-syn, several pathogenic missense mutations (Polymeropoulos et al. 1997; Krüger et al. 1998; Zarranz et al. 2004) (e.g. A53T, A30P, and E46K) have been implicated in early-onset familial PD, as opposed to the most common sporadic form of PD, with some mutated forms accelerating the kinetics of oligomerization (Conway et al. 1998; El-Agnaf et al. 1998; Li et al. 2001; Greenbaum et al. 2005; Ghosh et al. 2013).

Hydrophobic interactions (section “Hydrophobic effect and protein aggregation”), although insufficient to promote the formation of secondary and tertiary structures, are thought to be pivotal to IDP’s aggregation. Thus, several domains within the NAC region have been shown to be key to the aggregation process (Bodles et al. 2001; Giasson et al. 2001; Du et al. 2003; El-Agnaf et al. 2004; Bertoncini et al. 2005; Madine et al. 2008; Eugene et al. 2014; Rodriguez et al. 2015). That is to say that oligomerization decreases when specific changes in this region are promoted, opening the way for the development of drugs targeting these NAC sub-domains or stabilizing conformations of monomeric α-syn where intramolecular interactions shield the NAC region (Bertoncini et al. 2005; Dedmon et al. 2005; Esteban-Martín et al. 2013). In addition, small domains in the NAC flanking regions have also been associated with α-syn’s aggregation mechanism and function (Afitska et al. 2017; Doherty et al. 2020; Landeck et al. 2020; Stephens et al. 2020; Tripathi 2020; Farzadfard et al. 2022; Ulamec et al. 2022), including the domain comprised by amino acids 46–53 where the abovementioned missense mutations, implicated in familial PD, are found. The fact that these mutations are located in the N-term region suggests this segment could be equally important to the aggregation process. Figure 2(b) shows a β-sheet-rich domain in this region of the N-term. This is also consistent with molecular dynamics (MD) results which indicate that the dimerization process already involves regions beyond the NAC segment (Guzzo et al. 2022). Some of the abovementioned studies (Doherty et al. 2020; Ulamec et al. 2022) identified putative master-controller sequence motifs in the N-term region, namely, a segment formed by residues 36–42 coined “P1”, that may regulate α-syn aggregation. These segments, could, therefore, constitute key alternative (to NAC segments) targets to inhibit the aggregation of α-syn (Doherty et al. 2020; Tripathi 2020; Horsley et al. 2022).

Several other proteins, IDPs and globular proteins, are connected with the formation and deposition of aggregates implicated in different NDs. Well-known examples include aggregates of mutated forms of the Huntingtin protein, forming intranuclear inclusions in Huntington’s disease (Davies et al. 1997), an autosomal dominant inherited disease, and the amyloid-β peptide and the tau protein, which form, respectively, extracellular amyloid plaques and intracellular neurofibrillary tangles in the brain, implicated in Alzheimer’s disease (Aguzzi and O’Connor 2010; Sweeney et al. 2017). Although not discussed in this work, many of the drugs discussed in section “Protein aggregation inhibitors” (e.g. polyphenols) were found to inhibit aggregation of the proteins and peptides implicated in these diseases.

Sickle cell disease

At the other end of proteinopathies’ spectrum is SCD, an autosomal recessive inherited disorder that affects hemoglobin, the protein responsible for the transport of O₂ and CO₂ in the red blood cells. Opposite to IDPs, hemoglobin (normal adult Hb; HbA) has well-defined secondary, tertiary, and quaternary structures (Perutz et al. 1960). HbA is an allosteric protein, existing in low-oxygen affinity and high-oxygen affinity quaternary conformational states, known as the T-state (tense or deoxyhemoglobin) and the R-state (relaxed or oxyhemoglobin), respectively. HbA is composed of four polypeptide chains, two α subunits (α-globin) and two β subunits (β-globin); each α-globin is formed by 141 amino acids and a Heme group whereas β-globin is formed by 146 amino acids and a Heme group (see Figure 3(a)).

Figure 3. (a) Normal deoxygenated (T-state) hemoglobin, deoxy-HbA (2dn2.pdb) (Park et al. 2006), showing the four heme groups and the Glu-β6 amino acid in β₁ and β₂, replaced by Val-β6 in HbS; (b) sickle-cell deoxy-HbS dimer (2HBS.pdb) (Harrington et al. 1997) showing the Val-β6 (grey spheres) in 2β₂, lodged in a hydrophobic cavity in 1β₁ formed by Ala-β70, Phe-β85, and Leu-β88 (blue spheres) (Wishner et al. 1975; Dykes et al. 1979; Padlan and Love 1985a, 1985b).

HbS aggregation

SCD is caused by a monogenic mutation in the β-globin gene that results in the substitution of a surface glutamic acid (charge −1) for valine (neutral and hydrophobic) at the 6th position of the β-globins of HbA (Pauling et al. 1949; Ingram 1956; 1957; Noguchi and Schechter 1985; Eaton and Hofrichter 1990). This mutation, while not significantly changing the conformation of HbS (Perutz et al. 1951), relative to HbA, reduces the solubility (Eaton and Hofrichter 1990) of deoxygenated sickle cell hemoglobin (deoxy-HbS) from 7.0 g·cm⁻³ to 1.7 g·cm⁻³, inducing the (reversible) aggregation of deoxy-HbS into 14-stranded helical fibers (Eaton and Hofrichter 1990; Dykes et al. 1979; Harrington et al. 1977; Padlan and Love 1985b, 1985a; Eaton and Bunn 2017; Tisdale et al. 2020; Perutz 1970; Wishner et al. 1975; Edelstein 1980; Dykes et al. 1978). These fibers are ∼20 nm in diameter and distort the red blood cells into a stiff, non-deformable sickle-like shape, disrupting microcirculation and causing hemolysis (Kato et al. 2018). Vaso-occlusion is responsible for pain crisis and organ failure, which ultimately can lead to death (Piel et al. 2017).

HbS fibers involve a lateral contact where Val-β6 has its hydrophobic side chain lodged in a hydrophobic cavity of a neighbor HbS tetramer (see Figure 3(b)). Further, the structure of the fibers has been shown to be similar to that of the deoxy-HbS crystal (Nagel et al. 1980). The kinetics of polymerization of HbS is characterized by a delay time and an exceedingly large dependence of the HbS intra-cellular concentration (Adachi and Asakura 1978; Eaton and Hofrichter 1990, 1987; Hofrichter et al. 1974). This delay time has been shown to be highly correlated with HbS supersaturation (Cellmer et al. 2016). The kinetics is thought to proceed via a double nucleation mechanism (Ferrone et al. 1985a, 1985b) involving a stochastic homogeneous nucleation stage (Galkin et al. 2007) in which the formation of HbS fibers occurs, followed by a heterogeneous nucleation stage involving the nucleation of additional polymers on the surface of the original fibers (Samuel et al. 1990). A kinetic model based on classic nucleation theory, which accounts for such a mechanism, aimed at probing the effectiveness of potential anti-sickling drug candidates was recently reported (Lu et al. 2019). Another recent kinetic study found the HbS polymerization process to be a rather rapid and inefficient process, namely, fiber growth, re-opening the window for drugs directly targeting aggregation (Castle et al. 2019).

Whereas in vitro polymerization is commonly studied in a high concentration phosphate buffer, allowing reducing the solubility of HbS (Itano 1953) and therefore the amount of HbS required to observe polymerization, a similar structure (Wang et al. 2000) and double nucleation mechanism is seemingly found. Thus, although some differences have been reported (see for instance Chen et al. 2004 and references therein) polymerization is characterized by a lag time, similar to that observed at physiological conditions (Adachi and Asakura 1978, 1979). A similar solubility decrease is found in 2,3-bisphosphoglycerate (2,3-DPG) solutions (Poillon et al. 1986, 1995; Poillon and Kim 1990); 2,3-DPG is an allosteric effector found in erythrocytes that enhances oxygen delivery, shifting the hemoglobin R-T equilibrium to the T conformational state, thus, enhancing aggregation by increasing the concentration of deoxy-HbS. In addition, 2,3-DPG stabilizes the fibers by decreasing the solubility of HbS (Eaton and Bunn 2017).

Upon oxygenation, the fibers disassemble without a delay period and most (5–10% erythrocytes sickle irreversibly, Dean and Schechter 1978a) erythrocytes recover their biconcave disk shape (Eaton and Hofrichter 1990). However, the rate of uptake of O₂ of deoxygenated sickled erythrocytes (82 ± 4.7 ms) is slower than that of normal erythrocytes (135 ± 17.6 ms), reflecting the time of depolymerization, among other possible factors (Harrington et al. 1977; Dean and Schechter 1978a). These sickling–unsickling repeated cycles possibly damage the erythrocytes’ membrane leading ultimately to extra- and intravascular hemolysis (Eaton and Hofrichter 1990; Kato et al. 2017).

Concerning the molecular nature of the aggregation, this is believed to be primarily associated with hydrophobic interactions because of the nature of the abovementioned lateral contact (Eaton and Hofrichter 1990; Cao and Ferrone 1997; Wang and Ferrone 2013). A negative free energy around −3 kcal·mol⁻¹ along with a nearly zero enthalpy was found for the polymerization (i.e. gelation) process, in a 0.15 M potassium phosphate solution at 37 °C (Ross et al. 1977). The solubility was found to decrease with increasing temperature, with a minimum (0.16 g cm⁻³) at 37 °C, increasing at higher temperatures. A monomer–polymer contact binding free energy of −7.5 kcal·mol⁻¹ was also reported by Cao and Ferrone (1997) from nucleation theory and kinetic rate measurements. Assuming each Val-β6 and hydrophobic cavity in each β-globin contributes (Nozaki and Tanford 1971) ∼1.5 kcal·mol⁻¹, Cao and Ferrone (1997) estimated hydrophobic interactions to contribute 80% of the monomer-polymer binding free energy (Cao and Ferrone 1997). This estimate, however, assumes the host pocket as well as the other Val-β6 not involved in the lateral contact, also contributes ∼1.5 kcal·mol⁻¹, an assumption that can be questioned because the pocket also exists in HbA and should be largely dewetted; recent MD simulations support this view (Galamba 2019).

Wang and Ferrone (2013) studied through light scattering experiments, the aggregation of deoxy-HbS and several nonpolymerizable species, including deoxy-HbA, below the solubility for polymer formation. A positive binding free energy was found at 1 mM and 302 K for deoxy-HbA (4.0 kcal·mol⁻¹) and deoxy-HbS (1.8 kcal·mol⁻¹); the dimerization of deoxy-HbA involves, however, the (weaker) axial contact, as opposed to the lateral contact, which involves the β6 mutation site. Furthermore, HbS association was found to be entropically favored and enthalpically disfavored, consistent with the view that deoxy-HbS aggregation is impelled by hydrophobic interactions.

Nonetheless, it has also been suggested that electrostatic interactions could be at least equally important (Kuczera et al. 1990; Harrington et al. 1997; Galamba and Pipolo 2018; Galamba 2019). Kuczera et al. (1990) in fact have long suggested that “for a fiber-like dimer structure, it is not the stabilizing hydrophobic interaction of Val in HbS that is the dominant factor, but the loss of the destabilizing interaction of Glu in HbA.” Galamba and Pipolo (2018) provided evidence, based on MD simulations, that aggregation could be triggered by the formation of hydrogen-bonded ion pairs (i.e. salt bridges) between several surface residues (Lys, Asp, Glu) as well as heme groups. A subsequent study (Galamba 2019) provided evidence that the absence of Glu-β6 was more important than the presence of Val-β6, consistent with Kuczera et al. (1990). In particular, it was shown that the Glu-β6→Val-β6 mutation favors aggregation through the elimination of strong electrostatic repulsions involving Glu-β6 and several residues, noteworthy, Asp-β73 and Glu-β90, as well as heme, whereas a mild attractive potential energy was found to be connected with Val-β6 (Galamba 2019). Asp-β73 was found to be especially relevant, being associated with strong electrostatic repulsions with Glu-β6 in the deoxy-HbA dimer, while forming a major attractive residue pair with Val-β6 in deoxy-HbS (see Figure 4), supporting the view that damping of electrostatic repulsions involving Glu-β6 in deoxy-HbA could explain the polymerization of deoxy-HbA at high potassium phosphate concentrations (Eaton and Hofrichter 1990).

Figure 4. (a) Glu-β6 (HbA) and (b) Val-β6 (HbS) interactions with every residue from adjacent HbA and HbS tetramers, respectively. The residues forming the hydrophobic pocket in HbS-1, Ala-β70, Phe-β85, and Leu-β88, are shown in orange in (b). The most important residues in the 1β₁ (blue) and Glu-β6 and Val-β6 in the 2β₂ (red) polypeptides from a MD snapshot are shown below. The potential energy, and not the electrostatic and vdW energy, are plotted in (a) and (b), respectively. Reprinted from Galamba (2019).

Interestingly, Asp-β73 is a point mutation in HbC-Harlem (Glu-β6→Val-β6; Asp-β73→Asn-β73) which differs from HbS in this single amino acid (i.e. Asp-β73→Asn-β73) (Bookchin et al. 1967). This mutation was shown to have a pronounced effect in the aggregation of HbC-Harlem, relative to HbS, resulting in the formation of crystals as opposed to polymers, seemingly through a similar mechanism, although with a much slower kinetics than HbS gelation (Adachi and Asakura 1980; Ivanova et al. 2001). Adachi et al. (2003) related some of these differences with the formation of a hydrogen bond between Asp-β73 and Thr-β4, further influencing the hydrophobic interaction between Val-β6 and the hydrophobic pocket in deoxy-HbS.

A binding free energy of −14 kcal·mol⁻¹ and −4 kcal·mol⁻¹ was found, respectively, for the HbS and HbA lateral contact dimers in the abovementioned MD studies (Galamba and Pipolo 2018; Galamba 2019). These studies further suggested that possible aggregation inhibitors could target several salt bridges found in the HbS dimer, instead, or in addition to the hydrophobic contact.

MD simulations of sickle and normal hemoglobin and hemoglobin fibril models were also recently reported (Lu et al. 2016; Maity and Pal 2021; Olagunju et al. 2022). Maity and Pal (2021) argued that the presence of hydrophobic residues without a bulky side chain at β6 in hemoglobin explained the stability of the fibrils, consistent with the experiments by Adachi et al. (1993), which showed that some substituents in the β6 position, such as phenylalanine and tryptophan, polymerized less readily compared to deoxy-HbS and that when oversaturated polymerization occurred without the delay time observed for HbS. Adachi et al. argued that the difficulty of insertion of the bulky side chains of phenylalanine and tryptophan into the hydrophobic acceptor pocket on an adjacent tetramer could inhibit nuclei formation prior to polymerization. Interestingly, however, phenylalanine, tryptophan and derivatives, and analogues (Noguchi and Schechter 1977, 1978; Dean and Schechter 1978a; Poillon 1982) are themselves aggregation inhibitors, although it is not known whether these interact with the amino acids that form the hydrophobic pocket that lodges Val-β6.

Olagunju et al. (2022)) found through MD simulations that both electrostatic and hydrophobic interactions involving the mutation site are important in the HbS aggregation. The authors suggested that a potential aggregation inhibitor could, in addition to target HbS–HbS interactions involving Val-β6, aim to interrupt an electrostatic contact involving Lys-β17 and the Glu-β90 of a neighbor tetramer. This same contact was identified as being among the lowest energy contacts in the deoxy-HbS and deoxy-HbA dimers (see SI of Galamba and Pipolo 2018; Galamba 2019).

While not an exhaustive road map through the drugs developed for NDs and SCD, in what follows we aim at providing a broad perspective on some of the early and most recent molecules shown to have aggregation inhibitory activity, and their putative action mechanism, in the context of SCD and NDs, especially PD. In addition, we aim to establish some contact between aggregation inhibitors found for such different diseases, involving archetypes of an IDP and an allosteric globular protein.

Protein aggregation inhibitors

Parkinson’s disease

A potential disease-modifying drug for PD, that is, one beyond the symptomatic treatment, should aim at reducing the α-syn expression, aggregation, membrane affinity and cytotoxicity, and/or propagation (Lashuel et al. 2013). In addition, reversing α-syn aggregation could be pivotal to treating several synucleinopathies because of delayed diagnosis. Despite several molecules being found to induce at least one of these responses in vitro and/or in vivo, these either have some major drawback, ruling out their therapeutic potential, or have not passed clinical trials. Hence, the most effective drug still used today against PD is l-3,4-dihydroxyphenylalanine (aka levopoda or l-dopa) (Abbott 2010), a dopamine precursor developed in the late 1960s. l-dopa increases the dopaminergic flux in the striatum, without, therefore, curing the disease nor influencing non-dopamine-associated symptoms (Jenner 2003; Huot et al. 2013; Stoker and Barker 2020).

Thus, several alternative therapeutic routes continue to be actively explored (Fields et al. 2019; Stoker and Barker 2020). These include, for instance, drugs that target non-dopaminergic neurotransmitters (Barone 2010; Blair and Dhillon 2017), associated with symptoms not alleviated by l-dopa and other related drugs, neuroprotective therapies (Rosenblad et al. 1999), aggregation inhibitors, gene therapies, or cell-based treatments (Cyranoski 2017). The latter, as discussed in the next section (section “Sickle cell disease”), were already successfully applied in the cure of SCD, although several challenges persist.

Concerning protein aggregation inhibitors, covered herein, although there are still no approved drugs that preclude or even delay the formation of α-syn oligomers (Stoker and Barker 2020), major advances have been achieved in recent years both in unraveling the structure of the multiple putative cytotoxic oligomers and in the design of new drugs targeting aggregation-inhibition. Molecules that showed aggregation-inhibition potential in PD and other proteinopathies encompass peptides (Madine et al. 2008; El-Agnaf et al. 2004; Horsley et al. 2022; Armiento et al. 2020; Cheruvara et al. 2015; Torpey et al. 2020; Kim et al. 2009), peptidomimetics and macrocycle peptides (Wrasidlo et al. 2016; Dougherty et al. 2017), antibodies (El-Turk et al. 2016), heat-shock proteins (Luk et al. 2008; Gao et al. 2015), and small organic molecules, including natural products (Stefani and Rigacci 2013; Doig and Derreumaux 2015; Bu et al. 2016; Young et al. 2017; Giorgetti et al. 2018; Javed et al. 2018). Here, we discuss various small molecules and peptide-based aggregation inhibitors developed through in silico, in vitro, and in vivo model studies.

Small molecule drugs

Several small organic molecules were found to exhibit anti-amyloid activity, although their action mechanism is not always completely understood. Some of the most explored small molecule anti-amyloid agents, including natural products (Khatoon et al. 2018), are catecholamines (Conway et al. 2001; Rochet et al. 2004; Li et al. 2005; Norris et al. 2005; Mazzulli et al. 2006, 2007; Herrera et al. 2008; Latawiec et al. 2010), phthalocyanines (Lee et al. 2004; Lamberto et al. 2009, 2011; Valiente-Gabioud et al. 2016), and polyphenols (Giorgetti et al. 2018; Borah et al. 2021; Mohammad-Beigi et al. 2019; Palazzi et al. 2018; Meng et al. 2009, 2010; Masuda et al. 2006; Caruana et al. 2011; Singh et al. 2013; Gautam et al. 2017; Ahmad and Lapidus 2012; Bieschke et al. 2010; Gautam et al. 2014; Rao et al. 2008; Ono and Yamada 2006; Arbo et al. 2020; Yang et al. 2023; Zhang et al. 2018), and among the latter, flavonoids (Hong et al. 2008; Jia et al. 2019; Meng et al. 2010, 2009; Tinku et al. 2021; Yang et al. 2023; Zhu et al. 2004). Many other small molecule drugs (Dong et al. 2019; Masuda et al. 2006; Mohankumar et al. 2020; Perni et al. 2017, 2018; Pujols et al. 2018; Saffari and Amininasab 2021; Tatenhorst et al. 2016; Tóth et al. 2014; Vittorio et al. 2020), were, however, reported to inhibit the aggregation of α-syn. In addition, many such anti-amyloid drugs are not protein specific, exhibiting anti-amyloid activity for different proteins involved in several proteinopathies (Masuda et al. 2006; Ehrnhoefer et al. 2008; Bieschke et al. 2010; Doig and Derreumaux 2015; Valiente-Gabioud et al. 2016; Giorgetti et al. 2018). Some of these small molecules stabilize the monomeric form, inducing the formation of disordered nontoxic oligomers, whereas others prevent aggregation into toxic and nontoxic oligomeric species either by interacting with specific domains or by displacing the protein away from the lipid membrane. In addition, several molecules were shown to disrupt preformed fibrils (Ehrnhoefer et al. 2008; Meng et al. 2010; Pujols et al. 2018; Saffari and Amininasab 2021; Zhu et al. 2004). The mechanism of several drugs (e.g. polyphenols) generally considered neuroprotective agents does not necessarily involve, however, protein aggregation inhibition alone, being also or exclusively associated with other PD pathogenic events such as increased oxidative stress and defective mitochondrial function (Moon and Paek 2015).

Among the most studied catecholamines is dopamine (Figure 5), in addition to several oxidation derivatives and analogs (Li et al. 2005; Herrera et al. 2008; Latawiec et al. 2010), which have been shown to inhibit aggregation through different mechanism (Conway et al. 2001; Rochet et al. 2004; Norris et al. 2005; Mazzulli et al. 2006, 2007). A non-covalent aggregation-inhibition mechanism of dopamine, in particular, was associated with nonspecific interactions with the ₁₂₅YEMPS₁₂₉ sequence region in the C-term tail, and with long range electrostatic interactions involving E83 in the NAC domain (Mazzulli et al. 2007; Herrera et al. 2008).

Figure 5. Molecular structure of several α-syn aggregation-inhibitors: dopamine, OleA, EGCG, Curcumin, SynuClean-D, and Fasudil.

Oleuropein and derivatives (Mohammad-Beigi et al. 2019), including oleuropein aglycone (OleA) (Palazzi et al. 2018; Mohammad-Beigi et al. 2019; Borah et al. 2021) (Figure 5), a phenolic compound found in olive oil, were shown to stabilize less aggregation-prone conformations of monomeric α-syn, favoring the growth of stable nontoxic aggregates (Palazzi et al. 2018; Mohammad-Beigi et al. 2019). A cytotoxicity reduction was also connected with a reduction of the propensity of oligomers to bind to cell membrane components by interacting with the membrane-binding N-term domain of α-syn (Palazzi et al. 2018; Borah et al. 2021). In addition, it was argued (Palazzi et al. 2018) that OleA stabilized the NAC and C-term regions of α-syn, preventing long-range and hydrophobic interactions between these sub-domains, which could favor aggregation.

Interestingly, however, long-range (tertiary) interactions involving the NAC and C-term regions were found in other studies to prevent, not promote, aggregation (Bertoncini et al. 2005; Dedmon et al. 2005; Esteban-Martín et al. 2013). Furthermore, truncation of the C-term region has been shown to induce an acceleration of fibril formation in vitro (Crowther et al. 1998; Breydo et al. 2012; Farzadfard et al. 2022). This illustrates the apparent wide spectrum of monomeric α-syn conformations which might be involved in aggregation pathways of toxic and nontoxic oligomers.

Epigallocatechin gallate (EGCG) (Figure 5), a polyphenol within the group of flavonoids, and an aggregation inhibitor not specific to α-syn, was also suggested to induce the formation of less toxic disordered oligomers (Ehrnhoefer et al. 2008) and remodel mature α-syn fibrils into less toxic aggregates (Bieschke et al. 2010). Another study, however, reported that EGCG binds to the oligomers without changing either the secondary structure or its size distribution. In this study, the EGCG-induced toxicity reduction was linked with a decrease in the oligomers’ membrane affinity (Lorenzen et al. 2014).

The molecular mechanism underlying the aggregation-inhibition of α-syn by several flavonoids (e.g. baicalein) was shown to be associated with the restriction of conformational changes as well as the stabilization of α-syn’s monomeric and oligomeric species (Zhu et al. 2004; Hong et al. 2008; Meng et al. 2009, 2010). Furthermore, flavonoids with three vicinal hydroxyl groups exhibited enhanced inhibitory effects on α-syn aggregation (Meng et al. 2009); this was related with the flavonoids anti-oxidant activity, although limitations of this correlation were pointed out (Meng et al. 2009). In addition, the oxidized species (e.g. baicalein quinone) rather than the polyphenol (e.g. baicalein) were found to be the main aggregation inhibitors; aggregation inhibition of baicalein is significantly reduced under anaerobic conditions (Zhu et al. 2004; Meng et al. 2009). A complex mechanism encompassing the auto-oxidation of baicalein and other flavonoids and the subsequent covalent bonding to α-syn, through the formation of a Schiff base between the quinone of baicalein and a Lys of α-syn were, in fact, found to be the key factors for the inhibition of α-syn aggregation (Zhu et al. 2004; Meng et al. 2009). Caruana et al. (2011) also suggested that the main factors underpinning α-syn self-assembly inhibition and destabilization are the existence of aromatic elements that bind to α-syn monomers/oligomers, and neighbor hydroxyl groups on a single phenyl ring.

A drawback of some potential drugs found in small molecule libraries, in particular several aggregation inhibition polyphenols (e.g. curcumin; see Figure 5), quinones, and catechols, is the fact that these might be pan-assay interference compounds (PAINS) (Baell and Holloway 2010; Baell and Walters 2014), that is, molecules that give false positives in high-throughput screening assays for several possible reasons. Since the action mechanism of these molecules is not always completely understood, neither as PAINS nor as aggregation inhibitors, and because of their nonspecificity, it is difficult to predict whether molecules such as curcumin can both be PAINS and effective aggregation inhibitors (Gautam et al. 2014, 2017).

In addition, polyphenols at dietary concentrations have been connected with the prevention and attenuation of PD through alternative mechanisms, including oxidative stress (i.e. a reduction of reactive oxygen species) and neuroinflammation reduction (de Andrade Teles et al. 2018; Carecho et al. 2022).

Another class of compounds explored, concerning anti-amyloid activity, are phthalocyanines, which suppress aggregation through the interaction of the aromatic rings with aromatic amino acids via π–π interactions (π-stacking) (Valiente-Gabioud et al. 2016). This mechanism led to the suggestion that aromatic interactions could be key players in the aggregation mechanism of α-syn (Lamberto et al. 2009, 2011), in spite of a relatively reduced number of aromatic amino acids. α-syn has only four tyrosine (Tyr39, Tyr125, Tyr133, Tyr136), two phenylalanine (Phe4, Phe94), and no tryptophan amino acids, of which only Phe94 is in the NAC segment. The anti-amyloid activity of these compounds also depends on the type of metal ion coordinated to the tetrapyrrolic system (Valiente-Gabioud et al. 2016).

A series of pyridinyl-triazole derivatives were also recently reported to inhibit α-syn aggregation from in vitro screening and docking studies (Vittorio et al. 2020). Fasudil (5-(1,4-diazepan-1-ylsulfonyl)isoquinoline) (Figure 5), a small isoquinoline derivative, was shown to inhibit α-syn aggregation through direct binding to tyrosine residues Y133 and Y136 in the C-term region of α-Syn (Tatenhorst et al. 2016). SynuClean-D (2-hydroxy-5-nitro-6-(3-nitrophenyl)-4-(trifluoromethyl)nicotinonitrile) (Figure 5), an aromatic compound, was shown to reduce the in vitro aggregation of wild-type α-syn and the A30P and H50Q variants in a sub-stoichiometric molar ratio (Pujols et al. 2018). In addition, this compound was found to disrupt mature amyloid fibrils and prevent fibril propagation.

Squalamine (Perni et al. 2017) (Figure 6), a natural product isolated from the dogfish shark was shown to inhibit the aggregation of α-syn in vitro and in vivo by blocking the nucleation of α-syn. The mechanism of action of squalamine is not linked with specific protein–drug interactions but instead with a competition with α-syn for binding the membrane. The latter stimulates nucleation (Galvagnion et al. 2015) and, thus, the displacement of α-syn from the membrane hampers the first steps of the aggregation process. A later study by Perni et al. (2018) showed that the related compound, trodusquemine (Figure 6), interferes not only with the nucleation of α-syn but also with fibril-dependent secondary pathways. In addition, trodusquemine was shown to suppress the toxicity of α-syn oligomers in neuronal cells (Perni et al. 2018). These molecules are already relatively long, potentially binding to larger protein/cell domains relative to most small molecules previously discussed. In this sense, these are more similar to peptides, discussed in the next section.

Figure 6. Molecular structure of the α-syn aggregation inhibitors squalamine and trodusquemine.

While small molecules remain appealing because they have a good metabolic stability and can, in principle, more easily cross the blood–brain barrier, in addition to other reasons, including economic, they suffer in general from poor selectivity, specificity, and potency, regarding protein aggregation inhibition. Thus, alternative aggregation-inhibitors have been explored. Among these, we focus on some small peptide-based drugs (El-Agnaf et al. 2004; Wrasidlo et al. 2016; Torpey et al. 2020; Liang et al. 2021; Meade et al. 2021).

Peptide drugs

Peptide-based drugs re-emerged as promising alternatives to small molecules concerning proteinopathies because of their specificity and potency (Neddenriep et al. 2011; Cunningham et al. 2017; Armiento et al. 2020). The enhanced potency of small peptides is directly related with their larger interaction surfaces, allowing, in principle, to interfere with extended protein domains linked to the aggregation process (Cunningham et al. 2017). Linear peptide drugs, however, suffer from other drawbacks such as bioavailability and proteolytic instability (Armiento et al. 2020). This has stimulated the development of macrocycle peptides and peptidomimetics to overcome some of these limitations.

Anti-amyloid designed peptides are often β-sheet breakers or blockers, disrupting or inhibiting the formation of cross-β structures, a structural hallmark of toxic oligomers (Young et al. 2017; Serpell et al. 2000; Bartolini and Andrisano 2010; Sciarretta et al. 2006; Kim et al. 2009). Furthermore, in addition to aggregation inhibitors, since peptides are often modified segments of the amyloid proteins, they can provide insight into pivotal aggregation-prone domains (Neddenriep et al. 2011). Peptide modification strategies include, for instance, peptide termini modifications, insertion of prolines, which are potent β-sheet breakers, backbone modifications, or peptide cyclization (Sciarretta et al. 2006). In addition, polar amino acids can be inserted to enhance the solubility (El-Agnaf et al. 2004).

El-Agnaf et al. (2004) showed that modified peptides containing amino acid sequences ₆₈GAVVT₇₂ from the NAC inhibit aggregation into oligomers and mature amyloid fibrils. The peptides’ modifications included the insertion of RG and GR amino acids in the N-term and C-term. Aggregation inhibition was observed at (α-syn:peptide) 2:1, 1:1, and 1:2 molar ratios and the shortest peptide that inhibited α-syn aggregation had the central sequence ₆₉AVVT₇₂. These peptides were also observed to inhibit NAC aggregation suggesting that aggregation-inhibition is due to the binding of the peptides to their homologous sequence in α-syn.

Madine et al. (2008) showed that a peptide from NAC (₇₇VTGVTAVAQKTV₈₂), N-methylated in the C-terminus, disrupted the aggregation of α-syn. This peptide is a sub-domain of the 72–84 region of α-syn, absent in β-syn, which in spite of sharing 78% of similarity with α-syn, does not aggregate (Biere et al. 2000).

Kim et al. (2009) also proposed a small peptide (₇₂PGVTAV₇₇) able to block aggregation and to dissolve preassembled fibrils. This is a modified (T₇₂ → P₇₂) sub-domain of NAC (₇₂TGVTAV₇₇), part of the NACore (₆₈GAVVTGVTAVA₇₈) an 11 amino acid peptide later shown (Rodriguez et al. 2015) to play an important role in the aggregation and cytotoxicity of α-syn.

A 10 amino acid peptide (KDGIVNGVKA) was proposed by Cheruvara et al. (2015) from intracellular screening of a peptide library based on α-syn residues 45–54, involved in several familial PD mutations (E46K, H50Q, and A53T). The peptide was shown to inhibit the aggregation of α-syn and the associated toxicity. Interestingly, Torpey et al. (2020) found, through NMR, that although this peptide precludes oligomerization of the wild-type and several mutations associated with familial PD, it does not bind to the monomer neither to low-n (n < 4) oligomers. Thus, the aggregation-inhibition mechanism of this peptide is not completely understood.

A cyclic peptidomimetic (NPT100-18A) with a seemingly similar mechanism to squalamine was proposed by Wrasidlo et al. (2016). This peptidomimetic has the ability to displace α-syn from the membrane by interacting with domains in the C-terminus of α-syn.

Rezaeian et al. (2017) proposed two peptides, KISVRV and GQTYVLPG, which suppressed the aggregation of α-syn in vitro. These peptides were chosen based on their binding affinity to the amino acids 70–75 (VVTGVT) and 46–53 (EGVVHGVA) of α-syn. The first region is among the several regions of NAC found to be pivotal in the aggregation process. The second corresponds to the region where several missense mutated forms of α-syn (E46K, H50Q, G51D, A53T), implicated in familial PD, were found, and shown to induce an acceleration of the aggregation process, as previously discussed.

More recently, a small ubiquitin-related modifier 1 (SUMO1) derived peptide SUMO1 (15–55) (see Table 1), which targets two SUMO-interacting motifs within the N-term region flanking the NAC was shown to inhibit α-syn aggregation (Liang et al. 2021). Another recent study (Santos et al. 2021) proposed the use of α-helical peptides with a low affinity toward the monomeric form, avoiding perturbing the natural function of α-syn, while interrupting aggregation by binding to toxic oligomers and fibers. The PSMα3 peptide was found to have a high affinity toward a large number of binding sites in the oligomers, inhibiting aggregation (Santos et al. 2021).

Table 1. Examples of α-syn aggregation inhibition peptides.

Aggregation inhibition peptide	α-syn parent domain/target	Ref.
68GAVVT72	NAC	El-Agnaf et al. (2004)
77VTGVTAVAQKTV82	NAC	Madine et al. (2008)
72PGVTAV77	NAC	Kim et al. (2009)
45KDGIVNGVKA54	N-term	Cheruvara et al. (2015)
KISVRV	NAC	Rezaeian et al. (2017)
GQTYVLPG	N-term	Rezaeian et al. (2017)
SUMO1(15–55)^a	N-term	Liang et al. (2021)
PSMα3^b	Oligomers/fibrils	Santos et al. (2021)
K84s^c	Oligomers	Popova et al. (2021)
K102s^d	Oligomers	Popova et al. (2021)
GVLYVGS-Aib	N-term	Horsley et al. (2022)

Hydrophobic amino acids with aliphatic side chains are displayed in blue; hydrophobic amino acids with aromatic side chains are displayed in red. The second column identifies either the region of α-syn used in the peptide design and/or the believed target region/species.

^a ₁₅DKKEGEYIKLKVIGQDSSEIHFKVKMTTHLKKLKESYCQRQ₅₅.

^b MEFVAKLFKFFKDLLGKFLGNN.

^c FLVWGCLRGSAIGECVVHGGPPSRH.

^d FLKRWARSTRWGTASCGGS.

Popova et al. (2021) also found two synthetic peptides, through a high-throughput screening study (Table 1), 25 and 19 residues-long, that suppress α-syn aggregation. While the action mechanism was not disclosed, the peptides were shown to significantly suppress the first steps of oligomerization. The peptides were also shown to be specific to α-syn. Furthermore, the larger and more potent peptide was shown to reduce α-syn aggregation in human cells.

A recent study (Horsley et al. 2022) compared the aggregation inhibitory activity of two peptides, GVLYVGS-Aib and VAQKTV-Aib (L and D isomers; Aib stands for aminobutyric acid, a β-breaker), targeting respectively, the “P1” region (residues 36–42) of the N-term domain (Doherty et al. 2020), previously discussed, and the segment comprised by the residues 77–82 within the NAC domain. Horsley et al. (2022) showed that the D isomer of the peptide targeting the “P1” region was the most effective in preventing fibril formation, increasing the solubility of α-syn, and altering the conformation of the monomer into a less aggregation-prone state.

We now discuss the nature of some of the drugs, including amino acids and peptides, found to inhibit the aggregation of HbS, aiming at establishing some contact points between potential common drugs and aggregation inhibition mechanisms.

Sickle cell disease

Opposite to NDs, SCD can be potentially cured, either through hematopoietic stem cell transplantation (Eapen et al. 2019) or gene therapy (Ribeil et al. 2017), although several limitations and challenges, including economic, persist, preventing these treatments’ widespread (Eaton and Bunn 2017; Kapoor et al. 2018; Salinas Cisneros and Thein 2020; Tisdale et al. 2020). Thus, great interest remains in the development of drugs that can be used in the treatment of SCD.

In addition to hydroxyurea, long used in the treatment of SCD, other drugs became recently available. These include l-glutamine (Niihara et al. 1997, 1998; Ortiz de Montellano 2018), whose action mechanism, although thought to reduce the oxidative stress in the erythrocytes, remains largely unknown, and voxelotor (Vichinsky et al. 2019), an allosteric modulator (i.e. an oxygen affinity modifying drug) aimed at stabilizing the nonpolymerizing R quaternary structure of HbS. l-Glutamine was approved in 2017 by the Food and Drug Administration (FDA) for adult and pediatric patients 5 years and older. Voxelotor was approved by the FDA and the European Medicines Agency (EMA) and recently extended to the treatment of children of ages 4–11 years old. Nonetheless, in spite of representing important alternatives and/or potential co-adjuvants to hydroxyurea, these drugs have limitations (Tisdale et al. 2020; Henry et al. 2021). For instance, Henry et al. recently provided evidence that although voxelotor significantly reduces sickling, oxygen delivery to tissues is offset by increased hemoglobin O₂ affinity (Henry et al. 2021). The latter is the biggest potential drawback of allosteric modulators, in general. Nonetheless, allosteric modulators are among the most studied anti-sickling agents, and several other drugs were recently reported (Abdulmalik et al. 2005; Nakagawa et al. 2014; Oder et al. 2016; Oksenberg et al. 2016; Strader et al. 2019; Gopalsamy et al. 2021; Pagare et al. 2022). Other anti-sickling drugs that neither bind to HbS nor change the HbS oxygen affinity were also recently reported; the antisickling mechanism of this compound (SCD-101) is unknown (Swift et al. 2016).

Various SCD therapeutic strategies have evolved over the years (see Eaton and Bunn 2017; Telen et al. 2019; Tisdale et al. 2020 for recent reviews). These include (Eaton and Bunn 2017; Tisdale et al. 2020): (i) increase of HbF or the increase of the erythrocyte volume to decrease the intra-cellular HbS concentration, (ii) decrease the concentration of the allosteric effector, 2,3-DPG, increasing the solubility and decreasing the fibers’ stability, (iii) shift the allosteric equilibrium toward the R-state, and (iv) block protein–protein contacts by binding to HbS. In addition, several drugs aiming to reduce adhesion of erythrocytes to the vascular endothelium, decreasing transit times, have been investigated (Ataga et al. 2017; Eaton and Bunn 2017; Telen et al. 2019). Although most of these approaches started being explored in the 1970s (see Table 1 of Dean and Schechter 1978a), only recently there have been some important advances in the development of effective drugs falling within the purview of at least one of the above action mechanisms, as briefly discussed above.

Our main focus herein is on non-covalent stereospecific aggregation inhibitors that block protein–protein contacts. Again, our approach is not an exhaustive one, but rather to provide some examples of small molecules, including amino acids and peptides, that may share common features across the drugs studied in other proteinopathies.

Small molecule drugs

A foremost obstacle to the design of an effective anti-sickling molecule addressing the polymerization process, concerns the high concentration of HbS in the erythrocytes (Dean and Schechter 1978a) (∼5 mM) and therefore the putative high concentration required of any effective aggregation inhibition drug (Eaton and Bunn 2017). Nonetheless, the fact that a small change in the solubility of HbS can have a major impact in the lag period that characterizes the polymerization process, motivated the continuous exploration of this therapeutic strategy (Eaton and Bunn 2017). Furthermore, a recent study provided evidence that HbS fiber growth is a rather inefficient process (∼4% efficiency), as previously mentioned in the Introduction, where monomer addition and loss are nearly equal. This led to the suggestion that HbS fiber growth is a viable therapeutic target even at drug concentrations below the total hemoglobin concentration. Nonetheless, this concentration will depend further on drug selectivity, among other factors, since many drugs will bind to additional domains of HbS, not affecting or even enhancing polymerization.

Many small molecules (Abdulmalik et al. 2011; Abraham et al. 1983; Dampier 2015; Dash et al. 2013; Dean and Schechter 1978b; Iyamu et al. 2002; Mehanna 2001; Syed et al. 2019; Votano et al. 1984; Xu et al. 2017) including some amino acids (Noguchi and Schechter 1977; Dean and Schechter 1978a; Noguchi and Schechter 1978; Schechter 1980; Russu et al. 1986), were long found to decrease the deoxy-HbS aggregation in vitro by increasing the solubility of HbS. While these molecules, many reported in the 1970s and 1980s, failed their purpose as effective drugs for the treatment of SCD it is of interest to contrast some of these molecules with those proposed more recently to address SCD and other proteinopathies such as PD.

Examples of small molecules long shown to increase the HbS solubility include for instance (Dean and Schechter 1978a; Mehanna 2001): alkyl ureas (Elbaum et al. 1974, 1976, 1978), Hofmeister (lyotropic) salts (Poillon and Bertles 1979), aromatic compounds with a phenyl group and a pendant side chain terminating in a hydrogen bond donor/acceptor (e.g. NH₃⁺, COO^–, and OH) (Ross and Subramanian 1977), benzyl and phenoxy acids (Abraham et al. 1984), and clofibric acid and gemfibrozil (Abraham et al. 1983) (see Figure 7).

Figure 7. Molecular structure of several α-syn deoxy-HbS aggregation inhibitors: n-propyl urea, benzyl alcohol, clofibric acid, and gemfibrozil.

Ross and Subramanian (1977) provided a comprehensive analysis of several small molecules and concluded that deoxy-HbS aggregation inhibition (i.e. solubility increase) was promoted by the combination of a hydrophobic and a (single) hydrophilic group. In this regard, aromatic rings were more potent than aliphatic chains, and the hydrophilic group should be located on an aliphatic side chain attached to the aromatic ring to provide proper flexibility and/or distance to interact with deoxy-HbS. The latter was based upon the fact that molecules such as phenol, aniline, and salicylic acid do not inhibit aggregation. According to Ross and Subramanian, a similar logic should apply to alkyl ureas. They proposed an aggregation inhibition mechanism where the deoxy-HbS lateral contact was blocked through the interaction of the aromatic ring with the hydrophobic pocket (i.e. Phe-β85 and Leu-β88) whereas the hydrophilic group formed a hydrogen bond with the Asp-β73.

This mechanism is at odds with the later study by Adachi et al. (1993), which showed that phenylalanine and tryptophan in the Val-β6 position largely precluded aggregation, presumably due to stereo hindrance, as previously discussed. With regard to the requirement of an aromatic ring and a hydrophilic moiety at a given distance, this is common to many small molecules found to inhibit aggregation, although their specific action mechanism is not always understood.

Among amino acids, phenylalanine and tryptophan (see Figure 8) were the only amino acids found to exhibit aggregation inhibition activity in vitro (i.e. gelation inhibition) (Dean and Schechter 1978a; Schechter et al. 1978). Again, following Ross and Subramanian (1977), this was consistent with their findings regarding aromatic small molecules. A study in 1975, nonetheless, showed that 3.8 mM homoserine, asparagine, and glutamine but no other amino acid, reversed the erythrocytes sickling (Rumen 1975). This suggested at the time that these amino acids should inhibit sickling through a completely different (unknown) path than that observed in in vitro antigelling experiments.

Figure 8. Molecular structure of several α-syn deoxy-HbS aggregation inhibitors: phenylalanine, tryptophan, alizarin, piperine, capsaicin, and cubebin molecular structures. Phenylalanine and tryptophan exhibit some aggregation inhibition activity and were used as building blocks in aggregation inhibition peptides (Dean and Schechter 1978a; Schechter et al. 1978). Alizarin (hydroxyl anthraquinone) is a bioactive compound from the plant Rubia cordifolia. Piperine, capsaicin, and cubebin have been pointed out (Ameh et al. 2012) as possible aggregation inhibitors present in Niprisan (Iyamu et al. 2002) (drug Nix-0699) a product of the extracts of four different plants.

More recently, an experimental and simulation/docking study investigated alizarin (Syed et al. 2019), a hydroxyl anthraquinone (see Figure 8) found in the plant Rubia cordifolia (aka Indian madder). HbS polymerization was shown to decrease with the concentration of alizarin. The proposed mechanism involved the perturbation of the Val-β6 binding pocket through hydrogen bonding and hydrophobic interactions of HbS with alizarin. Many other natural products found in plants and long used in traditional medicine have also been shown to have some anti-sickling activity (Fall et al. 1999; Moody et al. 2003; Mpiana et al. 2007; Ameh et al. 2012; Imaga 2013). For instance, piperine, capsaicin, and cubebin (see Figure 8) have been pointed out (Ameh et al. 2012) as possible aggregation inhibitors present in Niprisan (Iyamu et al. 2002) (drug Nix-0699), a product of the extracts of four different plants.

A high-throughput assay encompassing 12,657 compounds of the Scripps ReFRAME drug repurposing library, where sickling times following deoxygenation to 0% oxygen of red cells from sickle trait individuals were assessed, was recently reported (Metaferia et al. 2022). Table 2 summarizes 50 compounds, among the 106 with antisickling activity, reported in that study.

Table 2. Set of 50 compounds, among the 106, reported by Metaferia et al. (2022) with antisickling activity (see Table 1 of Metaferia et al. 2022 for additional details).

Compound	Compound	Compound	Compound	Compound
Ammonium trichlorotellurate	Cisplatin	Curcumin	Butein	AM-103
Niferidil	Cloturin	AEG-33783	Ethacrynic acid	ODM-103
Mavatrep	EF-4040	VRT-532	TS-80	ON-09310
Manoalide	Cyclovalone	Licochalcone A	SC-1	P-7435
HR325	VML 530	NSC-150117	Tucaresol	AZD-4769
NSD-721	Peruvoside	AT-101	CBS-1114	Calcimycin
FT-11	ONO-7746	Darapladib	YM 57158	Vadadustat
Mitapivat	BMS-919373	Oxyfedrine hydrochloride	Allisartan	Metabromsalan
ARA DMAP	KUX-1151	Octenidine	Losartan potassium	IMD-0560
Pafuramidine	NT-113	WP-1066	Samixogrel	Laflunimus

Among the compounds depicted in Table 2, there are several polyphenols, including butein and curcumin, as well as some derivatives (e.g. cyclovalone) with reported neuroprotective effects, including for Parkinson, via an aggregation inhibitory mechanism (Tinku et al. 2021). Resveratrol derivatives were also recently studied concerning their potential as HbF inducers (Bosquesi et al. 2020). In addition, some of the compounds in Table 2 were already studied for SCD and later abandoned whereas other are currently being investigated. Tucaresol, for instance, a substituted benzaldehyde, was originally developed for treating SCD as an aggregation inhibitor by enhancing oxygen affinity, reducing hemolysis (Rolan et al. 1993; Arya et al. 1996), but later discontinued. Mitapivat, recently approved in the United States of America and Europe to treat hemolytic anemia, is also being investigated concerning SCD (Quezado et al. 2022; van Dijk et al. 2022; Xu et al. 2022). Mitapivat increases adenosine triphosphate (ATP) levels and decreases the 2,3-diphosphoglycerate (2,3-DPG) concentration, thus potentially reducing sickling, although other (unknown) mechanisms may play a role (Quezado et al. 2022). The large number of antisickling compounds reported by Metaferia et al. (2022) holds the promise that one or more compounds can be developed in a near future with one or more of the abovementioned action mechanisms.

Several potential drugs were also recently investigated through docking studies (Olubiyi et al. 2018; Das et al. 2020). A recent in silico drug repurposing study (Olubiyi et al. 2018) identified a series of compounds (praziquantel, losartan, ketoprofen, glipizide, rosuvastatin, atorvastatin, ergotamine, and risperidone) interfering in the Val-β6 lateral contact with the hydrophobic cavity in the neighbor HbS tetramer. This perturbation involved mostly the interaction of the drugs’ aromatic rings with the pocket formed by Ala-β70, Phe-β85, and Leu-β88. Unfortunately, assessing the aggregation inhibition activity of these and other compounds through docking or molecular simulations remains a challenging problem because of computational limitations.

Similar to other proteinopathies, the deoxy-HbS aggregation inhibitory effect of small peptides in SCD, now discussed, has long been probed.

Peptide drugs

Several peptides were long shown to exhibit anti-sickling activity (Dean and Schechter 1978b; Gorecki et al. 1980; Kubota et al. 1976; Schechter 1980; Schechter et al. 1978; Votano et al. 1977; Votano and Rich 1985). The latter were the object of a recent comprehensive review (Olubiyi et al. 2019). Herein, we highlight the main features that characterize the inhibition mechanism of peptides as opposed to small molecules, as well as putative resemblances with some of the peptide-drugs explored within the context of α-syn aggregation, discussed in section “Peptide drugs”.

Gorecki et al. (1980) found, based on a study of over 30 peptides, that the hydrophobicity of the side chains was the most important feature with respect to the HbS aggregation inhibition (i.e. antigelling activity). Votano and Rich (1985) later reported that “compounds containing bicyclic or multi-aromatic residues have a higher activity than those that carry a single aromatic or aliphatic side chain”. In addition, an increase in the apolar content of the aromatic residue and ring polarizability further enhanced the antigelling activity of such compounds. The peptides in these studies commonly comprised one to two Phe or Trp and a Gly, and were succinylated to enhance the solubility (Votano and Rich 1985); additional amino acids comprising the peptides included Arg, Nle, and Lys (Gorecki et al. 1980). Whereas the binding sites in HbS were not completely disclosed it was argued that peptide-HbS contacts should involve several amino acids near the hydrophobic cavity where Val-β6 is enclosed in the fibers. The fact that peptides containing several aromatic residues are more potent that those with a single Phe or Trp suggest that the aggregation inhibition mechanism should involve a larger domain than that comprised by the principal amino acids that form the Val-β6 binding pocket.

A pivotal advantage of peptide-based drugs concerning SCD is the possibility to enhance the protein–drug contact area, decreasing, in principle, the drug:HbS ratio. The low stereoselectivity of small molecules toward protein surfaces can translate into the interaction with several protein domains, which can disfavor, promote, or exert no significant effect on aggregation. In principle, larger peptides should depict a greater aggregation inhibitory activity, relative to single amino acid or di- and tri-peptides because of the increased contact area. However, some critical length should exist above which aggregation inhibition “saturates”. The latter, however, will depend on the contact area, which in turn depends on the chemical nature of the peptide. Several works (Kubota et al. 1976; Noguchi and Schechter 1978; Akbar et al. 2006) addressed this problem concluding that the aggregation inhibitory activity increased with the peptide size, although large concentrations were still required to observe effective antigelation effects (see also Olubiyi et al. 2019 for a broader discussion of this aspect).

A possible explanation for the large concentrations required for observing antigelation activity by small molecules and peptides involving aromatic rings (Phe and Trp) is the fact that these may not interact with the hydrophobic pocket where Val-β6 is inserted. The latter can be speculated based on the results of Adachi et al. (1993), previously discussed. Thus, such a pocket could only be blocked at high drug concentrations when interactions with neighbor sites imped the entrance of Val-β6 from a neighbor tetramer. An alternative explanation is that some of these peptides and or small molecules are neither selective enough nor the aggregation inhibition mechanism is associated with the Val-β6 binding pocket. The abovementioned studies, point, nevertheless, to a common direction, that is, aromatic amino acids are key to the blockage of protein–protein contacts behind nucleation (increasing the lag time), and heterogeneous nucleation, preventing the formation of fibers. In the following sections, we attempt to rationalize some of the above information regarding the role of hydrophobic interfaces, aromatic rings, and hydroxyl groups, all seemingly playing a part in protein aggregation inhibition in SCD, PD, among other proteinopathies.

Hydrophobic effect and protein aggregation

The design of a drug to oppose or delay protein aggregation could, in principle, aim at increasing the solubility of the protein, in principle, by weakening the hydrophobic effect. The hydrophobic effect comprehends two related phenomena, hydrophobic hydration, associated with the low solubility of apolar molecules and groups in water, and hydrophobic interactions, the spontaneous association of apolar molecules in water. The most remarkable feature of hydrophobic hydration is, perhaps, its system size dependence, with the hydration of small (R_S < ∼1 nm) and large radius (R_S > ∼1 nm) spherical solutes being dominated, respectively, by entropy and enthalpy (Pratt and Chandler 1977; Hummer et al. 1996; Lum et al. 1999; Chandler 2005; Ashbaugh and Pratt 2006; Berne et al. 2009; Jamadagni et al. 2011).

Hydrophobic hydration

Small length hydration

Hydration of small apolar molecules is governed by the formation of a suitable cavity to lodge the solute. For hard spheres, this probability is related to the hydration free energy (i.e. excess chemical potential) by $\Delta G_{\text{hyd}}=-k_{B}T\hspace{0.17em}\ln \hspace{0.17em}{p}_0(R)$, where R is the solvent accessible radius, given by $R=R_S+R_W$, and R_S and R_W are the radius of the solute and water, modeled as hard spheres, whereas $p_0(R)$ is the probability that a sphere of radius R randomly inserted in water is devoid of water molecules (Widom 1982; Garde et al. 1996; Ashbaugh and Pratt 2006).

Whereas solute–water interactions are favorable, the hydration enthalpy is smaller in magnitude than the hydration entropy and the process is, therefore, entropic. Although displaying differences relative to bulk water, the water structure (and dynamics) and, therefore, water–water interactions, produce a small net effect on the hydration enthalpy and entropy. For instance, a tetrahedral enhancement (Davis et al. 2012; Galamba 2013, 2014; Conti Nibali et al. 2020) of a sub-population (e.g. ∼70% for methane) of water molecules in the hydration layer of small apolar solutes has been found, along with an enhancement of the hydrogen bond strength between those water molecules and the water molecules that comprise the vertices of the imperfect “tetrahedrons” (Galamba et al. 2019; Tamoliūnas and Galamba 2020; Galamba 2021). This water population illustrated in Figure 9(a) is denoted here W-4W. This is opposed by water molecules (e.g. ∼30% for methane) that are not tetrahedral (i.e. water molecules closer to the solute than to the four nearest water neighbors (W-3W; see Figure 9(a)), which cannot, therefore, form up to four hydrogen bonds), where a weakening of the average water–water interactions with the third and fourth water neighbors is observed, making a positive contribution to the hydration enthalpy. These populations (i.e. tetrahedral and non-tetrahedral; see Figure 9(a) and the discussion in the next section) decrease and increase, respectively, with the solute size (Tamoliūnas and Galamba 2020). The excluded volume induced weakening of water–water interactions (i.e. hydrogen bond breaking), nonetheless, outweighs the tetrahedral enhancement, except for small solutes at low temperatures (Galamba 2021), making a small positive contribution to the enthalpy. Furthermore, notwithstanding these contributions to the hydration entropy and enthalpy, the water reorganization around the cavity is characterized by a nearly exact entropy–enthalpy compensation, thus, not contributing to the hydration free energy (Ben-Naim 1975, 1978; Lee 1991; Grunwald and Steel 1995; Graziano 2005).

Figure 9. (a) A water pentamer next to Ile-6 in an 11-mer isoleucine peptide (₁ILE₁₁) displaying a water in the C_β coordination sphere with at least four nearest water neighbors (W-4W) and a water with three or less nearest water neighbors (W-3W) because of the proximity of the solute; (b) number of water molecules in the first hydration layer (r_min ≤ 6.45 Å) of the C_β of, Ile-6 in an 11-mer isoleucine peptide (₁ILE₁₁), Ile-6 in an 11-mer serine peptide with a single (middle) isoleucine (₁SER₅-ILE-₇SER₁₁), and Ile-4 in NAC-term (₈₅AGSIAAATGFV₉₅), an 11-mer peptide comprised of the last 11 amino acids of NAC; (c) tetrahedrality of the W-4W water populations in the first hydration shell of the C_β of Ile and Ser, respectively, in isoleucine (₁ILE₁₁) and serine (₁SER₁₁) 11-mer peptides, compared with the tetrahedrality of bulk water; (d–g) water pair interaction energy distributions, P(W···W_n) for n = 1–4, for bulk water and W-4W and W-3W water populations in the first hydration shell of the C_β of Ile in isoleucine (₁ILE₁₁).

For long linear hydrophobic solutes such as n-alkanes, a behavior similar to small solutes is observed as these can be accommodated in water almost as an ensemble of small apolar molecules and the hydration free energy is a linear function of the carbon number (n_C) up to n_C ∼20 (Chandler 2005; Galamba 2021; Singh and Sharma 2022). Folding in long alkanes (n_C > ∼20) was recently shown to result in the violation of entropy and enthalpy convergence (Galamba 2021). This was demonstrated to result from the release of water molecules in the hydration layer of methylene groups upon folding, increasing the hydration entropy and reducing solute–water interactions, thus, lowering the enthalpy (Galamba 2021). Entropy convergence violation is observed for low curvature (R_S > ∼1 nm) hard spheres (Huang and Chandler 2000; Ashbaugh and Pratt 2006) and for many globular proteins upon unfolding (Robertson and Murphy 1997), because of the heterogeneous nature of protein–water interactions (Sedlmeier et al. 2011) and probably folding/unfolding differences (Galamba 2021).

Large length hydration

For low curvature solutes (R_S > ∼1 nm), hydration is governed instead by the formation of a solute–water interface, as opposed to density fluctuations and spontaneous cavity formation, resulting in a substantial loss of water–water HBs and, therefore, an enthalpy-dominated hydration free energy and a positive hydration entropy (Chandler 2005; Jamadagni et al. 2011). Hydrophobic hydration of such large surfaces is characterized by an increase of the local water fluctuations and isothermal compressibility, relative to bulk water (Chandler 2005; Jamadagni et al. 2011). Furthermore, a more favorable binding of hydrophobic molecules has been observed because of facilitated cavity formation next to the surface, favoring the binding of hydrophobic drug-groups (Jamadagni et al. 2011).

The above picture, nonetheless, applies to hydrophobic molecules or interfaces, as opposed to (amphiphilic) proteins and other biomolecules where chemical heterogeneities result in heterogeneous solvation, with water molecules next to hydrophilic groups sharing the hydration layer of hydrophobic side chains of neighbor amino acids, leading to quenching of density fluctuations (Chandler 2005; Jamadagni et al. 2011). In addition, topological (microscopic) irregularities influence protein hydration and density fluctuations. Hydrophobic hydration has, for this reason, been more difficult to characterize in biomolecules than in hard spheres, hydrocarbons, or flat planar self-assembled monolayers (Granick and Bae 2008; Jamadagni et al. 2011).

To illustrate some of the above points, Figure 9 displays some MD results on the size of the water populations next to hydrophobic and hydrophilic amino acids in 11-mer peptides along with the tetrahedrality (Errington and Debenedetti 2001) and energetics of water–water interactions. The peptides and water were described by the AMBER99sb (Hornak et al. 2006) and TIP4P-Ew (Horn et al. 2004) force fields, respectively. Further details about these simulations can be found elsewhere (Galamba 2022). Figure 9(b) shows a hydration increase by 15% of an Ile amino acid when sided by Ser amino acids in an 11-mer peptide compared to an Ile amino acid in an 11-mer Ile peptide. This increase is related both with the hydrophilic nature of Ser and with a reduction of the excluded volume. The latter is confirmed by an even larger hydration increase in a C-term segment of NAC (₈₅AGSIAAATGFV₉₅), denoted herein NACterm peptide (Bertoncini et al. 2005; Galamba 2022), where Ile is sided by a Ser and an Ala; although no significant folding is found for 11-mer peptides (Galamba 2022), hydration is also influenced by peptide structural fluctuations.

As previously discussed, excluded volume, either next to a hydrophobic or hydrophilic amino acid results in a tetrahedrality increase of a water population that retains four water neighbors (W-4W, Figure 9(a)). This can be seen in Figure 9(c) for Ile in an 11-mer Ile and NACterm peptides, and Ser (Ser-6), in an 11-mer Ser peptide. Figure 9(d–g) also confirms the water HB enhancement next to Ile within the W-4W population, compared to bulk water, more pronounced for the 3rd and 4th potential HB partners. The W-3W population shows an enhancement of only the first HB, whereas solute exclusion volume hinders the formation of more than two HBs, in average, within this population. The fact that such a population dominates in low curvature solutes (Figure 9(b)), along with the weak nature of dispersion interactions involved in hydrophobic group–water interactions, is responsible for a mild dewetting not found around molecules and groups that form HBs with water.

However, hydrophobic segments in proteins are generally limited in length and may not favor hydrophobic drug binding over protein–protein aggregation at therapeutically relevant concentrations. The latter is among the main obstacles to the rational drug design targeting protein–protein hydrophobic contacts.

Hydrophobic interactions

Hydrophobic interactions, the other related manifestation of the hydrophobic effect, are indirect, solvent-induced interactions associated with the dewetting of apolar groups and the formation of a liquid–vapor-like interface that culminates in the so-called hydrophobic collapse for sufficiently large hydrophobic surfaces (Stillinger 1973; Chandler 2005; Berne et al. 2009). This microscopic dewetting, while not opposed by solute–water interactions in ideal hard (i.e. repulsive) flat surfaces is outweighed by van der Waals interactions in real hydrophobic systems bringing water closer to the hydrophobic surface (Granick and Bae 2008; Jamadagni et al. 2011), resulting in a small (i.e. less than the size of a water molecule, Godawat et al. 2009), although important, width depletion region (Chandler 2005; Chattopadhyay et al. 2010). A similar phenomenon is not observed for small solutes, unless an extended surface cluster can form (Chandler 2005); even relatively large (e.g. neopentane) and long (e.g. dodecane) alkanes show small tendency to aggregate (Xavier and Galamba 2021).

Hydrophobic interactions are the main driving “force” responsible for protein folding (Kauzmann 1959; Dill 1990) and IDPs such as wild-type α-syn are thought not to fold because of their high net charge and low hydrophobicity. Hydrophobic interactions are also thought to be the main driving force in protein aggregation, as previously discussed for HbS and α-syn.

In light of the above discussion, small and large length-scale hydrophobic effects should be important to protein hydration, (mis)folding, and aggregation. In addition, water-mediated electrostatic interactions should also play an important part. For instance, rationalization of the hydrophobic interaction of Val-β6 with the host pocket in the neighbor HbS within the small length hydrophobic picture, would suggest a low tendency for aggregation since the hydrophobic pocket, which also exists in HbA, should be largely dehydrated in the monomer, as previously discussed. Thus, the main hydrophobic driving force should be associated with the dewetting of the side chain (isopropyl) of Val-β6. However, following the above discussion, other hydrophobic and electrostatic interactions resulting both from the departure of Glu-β6 and the insertion of Val-β6 should be important in triggering the aggregation process and the formation of a nucleus. Thus, although undoubtedly linked with the interaction of Val-β6 with the acceptor hydrophobic pocket (see Figure 10(a,b)) and neighbor hydrophilic and hydrophobic amino acids, a comprehensive understanding of the initiation of the homogeneous nucleation at deoxy-HbS saturation conditions, remains elusive.

Figure 10. (a) Deoxy-HbS monomer showing the hydrophobic pocket (yellow); (b) deoxy-HbS monomer showing Val-β6 (yellow); (c) α-syn. (blue) Hydrophobic amino acids with aliphatic side chains (Ala, Ile, Leu, Met, and Val); (red) hydrophobic amino acids with aromatic side chains (Phe, Trp, Tyr); (orange) remaining amino acids; (green) Hem in deoxy-HbS.

With respect to α-syn aggregation, again, in spite of copious differences relative to HbS, both hydrophobic length scales should be pivotal. For instance, NAC (35 amino acids) has 17 hydrophobic amino acids, mostly Val and Ala, but no more than four contiguous hydrophobic amino acids (₈₈IAAA₉₁), with the main hydrophobic domains being ₈₈IAAA₉₁, ₆₉AVA₇₁, ₇₆AVA₇₈, and ₉₄FV₉₅ (Figure 10(c)). The aggregation of the 11-mer domain (Bertoncini et al. 2005) NACterm (₈₅AGSIAAATGFV₉₅) of α-syn, which contains this longer hydrophobic domain (i.e. ₈₈IAAA₉₁) was recently studied through molecular simulations (Galamba 2022). A binding free energy of ∼−10 kJ mol⁻¹ was found (Galamba 2022). Urea, a protein denaturant, that also induces protein disaggregation at high concentrations (∼6–8 M) was shown to strongly inhibit aggregation through the occupation of the dewetted interpeptide region precluding hydrophobic collapse. Whether some of the small molecules discussed in section “Protein aggregation inhibitors”, including some polyphenols, inhibit aggregation by directly interacting with the NAC or other domains of α-syn and/or reduce density fluctuations next to hydrophobic groups remains unknown. Furthermore, the relationship between polyphenols and other small molecules that stabilize specific oligomers and the hydrophobic effect, seemingly responsible for the oligomerization process, is also not clear. Nevertheless, some chemical groups such as aromatic rings and hydroxyl groups seem to play a key role in the aggregation and aggregation inhibition of distinct proteins. This is now briefly discussed, including some of the limitations and challenges associated with these ubiquitous groups and compound families, in drug design.

Hydrophobic effect and drug design

A small molecule or peptide aimed at inhibiting aggregation in vivo through some protein–protein contact impediment must (i) cross the cell membrane in intracellular aggregation-related diseases and the blood–brain barrier in NDs, (ii) selectively bind to some key protein aggregation site(s), and (iii) inhibit aggregation or induce the formation of non-toxic oligomers, and/or disrupt pathological oligomers or fibrils without compromising the protein’s function(s). Furthermore, the drug should be soluble, have high selectivity, specificity and potency, and low toxicity, along with good bioavailability and metabolic stability. We pause to note that in Alzheimer’s disease and PD, the diseases further induce changes in the blood–brain barrier, altering drug permeability, that, in spite of possible menaces, can also lead to novel drug therapies (Desai et al. 2007). In SCD, erythrocyte membrane alterations might also change drug permeability, in addition to known cation permeability perturbation, which results in erythrocyte dehydration (Brugnara 2018), favoring aggregation through the increase of deoxy-HbS concentration.

With regard to the selective binding to specific protein sites, knowledge of the aggregation mechanism is pivotal to define suitable targets to block aggregation and/or promote oligomer and fiber disassembly. For proteinopathies, the most common targets are hydrophobic regions such as the hydrophobic host pocket in deoxy-HbS and the NAC or the N-term region in α-syn because these are involved in the main contacts found in the respective fibers. However, structural changes affecting these regions, and thus, aggregation, can, in principle, also be driven by protein–drug interactions involving protein segments far-off from the main protein aggregation binding contacts.

A link between many protein aggregation inhibitors, including amino acids and peptides, is the existence of aromatic rings, which allow the formation of non-covalent stacked structures with aromatic amino acids and the interaction with aliphatic hydrophobic side chains. In addition, hydroxyl groups can interact with polar side chains and have been found to play a role in the aggregation inhibitory potency of small aromatic molecules (Ross and Subramanian 1977; Meng et al. 2009; Caruana et al. 2011). In this sense and considering the large collection of polyphenols that showed aggregation inhibitory activity in vitro for different proteins, although with reported diverse mechanisms, it could seem that finding a drug for therapeutic purposes would not be difficult. Nonetheless, although polyphenols occupy a relevant position among protein aggregation inhibitors, in addition to their long reported role as protective agents in NDs through various mechanisms (Freyssin et al. 2018; Henríquez et al. 2020), their general protective role remains elusive (Colizzi 2019). In particular, concerning their reputation as aggregation inhibitors, a more fundamental knowledge about the action mechanism of polyphenols and aromatic rings and hydroxyl groups, in general, seems important, namely, (i) their impact in protein solvation, including water density fluctuations and structural transformations, (ii) whether direct (i.e. protein-polyphenol) and/or indirect (i.e. polyphenol induced alterations of water–water and protein–water interactions) aggregation inhibition mechanisms are involved, (iii) their selectivity toward hydrophobic and hydrophilic domains, or (iv) their influence on the structure of the monomeric proteins. This could allow unraveling the fundamental intermolecular interactions through which some small molecules induce the formation of stable (not-toxic) oligomers, precluding oligomerization and even disrupt mature fibrils (Zhu et al. 2004; Hong et al. 2008; Bieschke et al. 2010; Lorenzen et al. 2014), whereas others may inhibit nucleation by stabilizing the monomer, although not disrupting preformed fibrils.

We stress that, the fact that some polyphenols were found to inhibit aggregation of distinct proteins (Henríquez et al. 2020), behaving almost as universal-aggregation-inhibitors, could mean these are not effective therapeutic drugs because of low selectivity, requiring concentrations too high to hamper key contacts. For instance, there is a common view that a drug, either a small molecule or a peptide-based drug (Armiento et al. 2020), that can shield specific domains of the NAC (Bertoncini et al. 2005; Dedmon et al. 2005) is of potential therapeutic interest in PD. The concentration necessary to block some pivotal sub-domain of the NAC, however, may not be therapeutically relevant because of a selectivity deficit. Indeed, looking at the hydrophobic–hydrophilic amino acid sequence in α-syn (Figure 10(c)), it is difficult to argue on the putative preferential binding of a polyphenol toward a specific domain, let alone its influence on the structure, solubility, and aggregation propensity of the monomer.

A similar rationale applies to peptide-based drugs. While the importance of aromatic rings was long reported in the context of deoxy-HbS aggregation, concentrations too high for therapeutic purposes were required to observe significant solubility increases (Olubiyi et al. 2019). This suggests that, despite peptides carrying aromatic rings can compete more effectively than small molecules with proteins for specific extended amphiphilic surfaces involved in fiber contacts, they still lack enough selectivity. In addition, some peptides may also aggregate, a characteristic possibly shared with some (low solubility) polyphenols.

Thus, although the rational design of aggregation inhibitors continues to hold the promise of developing effective therapeutics for several proteinopathies, the fact that many small molecules in drug libraries are soluble polyphenols or other amphiphilic molecules with aromatic rings (and OH groups), some of which have been considered PAINS, can taint drug screening assays (Baell and Nissink 2018), unless the aggregation inhibition mechanisms of some of these molecules, at a molecular-level, is unraveled. In this sense, it seems particularly promising to investigate the molecular origin of the seeming “universal” aggregation inhibitory activity of some small molecule drugs, either to improve drug design strategies based on these molecules or to rule them out as promising leads possibly for reasons that put them into the category of PAINS.

Conclusions

The failure of many potential drugs for NDs, especially Alzheimer’s disease, propelled the discussion (Makin 2018) on whether the amyloid hypothesis is, in fact, correct, and protein aggregation is the primary cause of NDs. However, unlike NDs, the molecular origin of SCD is well-known and there has also been limited progress in finding effective aggregation inhibitors. This naturally raises questions concerning the efficacy of this approach in SCD, NDs, and other proteinopathies. However, given the large number of diseases that involve some form of protein aggregate and the economic burden associated with future stem cell and gene therapies, the development of small molecule or peptide-based drugs remains a worth-exploring therapeutic route.

Herein, we provided an overview of the drugs developed to treat two completely different proteinopathies, namely, PD and SCD, aiming to establish some links associated with the importance of hydrophobic interactions in protein aggregation. The small and large length-scale regimes of the hydrophobic effect were discussed, along with the apparent relevance of aromatic rings and hydroxyl groups to aggregation inhibition. Although the ubiquity of polyphenols as potential small molecule drugs concerning a multitude of diseases, ranging from cardiovascular diseases to cancer and NDs, stimulates their study, their mechanisms as protein aggregation inhibitors remain poorly understood. As a consequence, the design of improved drugs that could dodge intrinsic drawbacks of polyphenols that showed aggregation inhibitory activity in vitro even at nontoxic concentrations, including low bioavailability and selectivity, has proven difficult.

Significant advances might emerge in a near future, nevertheless, linked with improved machine learning-assisted drug design methodologies (Vamathevan et al. 2019) and a more fundamental understanding of the action mechanism of potential drugs, polyphenols and others, concerning their influence on the impediment of intertwined hydrophobic and electrostatic interactions and protein solvation. In addition, the design of cyclic peptides represents a promising route toward the development of alternative drugs with enhanced selectivity while avoiding some of the limitations of linear peptides.

Disclosure statement

The authors report no declarations of interest.

Additional information

Funding

GM acknowledge financial support from Fundação para a Ciência e a Tecnologia of Portugal for a PhD scholarship (2021.05348.BD) and NG acknowledge financial support from Fundação para a Ciência e a Tecnologia of Portugal and (CEEC/2018). Work supported by UIDB/04046/2020 and UIDP/04046/2020 center grants from FCT, Portugal (to BioISI).