Hydrogels are cross-linked by three-dimensional networks of modified molecules that hold a large amount of water, which were first reported by Wichterle and Lím (1960). By definition, water must account for at least 10% of the total weight (or volume) of hydrogels, which is very similar to natural tissue. Due to the high water content, favorable structural features, and biocompatibility, hydrogels have great potential in biomedical applications such as drug delivery, tissue engineering, sensing, and cell culture scaffolds (Ahmed, 2015; Ghosal et al., 2018; Zou et al., 2020). Based on the different gelation mechanisms, hydrogels can be classified into chemical hydrogels and physical hydrogels (Mahinroosta et al., 2018). The chemical hydrogels are cross-linked via covalent bonds, resulting in high mechanical strength, structural stability, shape memory, and irreversible properties (Liyan et al., 2018). The irreversible properties deprive them of self-healing and impair their injectability (Tu et al., 2019). More importantly, those limit cell proliferation and migration and hence limited application in three-dimensional cell culture (Wang et al., 2019, 2020). Physical hydrogels, also known as supramolecular hydrogels, could be produced by a molecular self-assembly process (Li et al., 2020; Xian and Webber, 2020; Li et al., 2021b). Molecular self-assembly relies on noncovalent interactions, such as hydrogen bonding, hydrophobic interactions, aromatic π–π stacking interactions, and electrostatic interactions, which are weak and reversible (Whitesides and Boncheva, 2002; Webber and Pashuck, 2021). The dynamic reversible nature of these interactions endows them with self-healing and shear thinning properties; hence, they are particularly suitable for biomedical applications, as they are more flexible and can be easily injected (Hu X. et al., 2020; Gupta et al., 2020; Vazquez-Gonzalez and Willner, 2020).

The physical hydrogelators include natural polymers, such as polysaccharides (Dai et al., 2019; Chen et al., 2020; Lei et al., 2020), proteins (Wlodarczyk-Biegun et al., 2016; Fisher et al., 2017; Singh et al., 2017; Zhao et al., 2018), and polynucleotides (RNA and DNA) (Korah et al., 2018; Li et al., 2019b), and synthetic materials, such as poly (lactic-co-glycolic acid (PLGA) (Ko and Kwon, 2020), polyglycolic acid (PGA) (Prabhu et al., 2020), and polylactic acid (PLA) (Basu et al., 2016). These various polymers establish complex 3D networks of intermolecular interactions between building blocks to form macroscopic objects. Besides polymers, low-molecular-weight gelators (LMWGs) (generally <2000 Da) are of great interest because the associated gelators have intrinsic self-assembly ability to form different morphologies of fibers, rods, ribbons, and nanotubes, which could be used to build attractive tools for various biomedical applications on account of their high biocompatibility and low toxicity (Nolan et al., 2017; Ramin et al., 2017; Piras et al., 2021). Among the LMWGs, self-assembly peptide-based hydrogels are well known and have been the subject of many studies because the peptides are the most attractive building blocks, which can be readily controlled by changing the amino acid sequence of the peptides and by modifying the side chains of the peptides (Frick et al., 2017; Fu et al., 2021; Zhang et al., 2021). In particular, many small amino acid sequences responsible for the biological assembly proteins such as the amyloid-beta polypeptide (Aβ42) (Qiu et al., 2015), tau proteins (Pîr et al., 2017), and islet amyloid polypeptide (IAPP) (Fawver et al., 2014) indicated in Alzheimer’s disease, have been identified; hence, researchers can design self-assembly peptides through a biomimetic approach. The shortest self-assembling motif of peptides was dipeptide, for example, L-Phe-L-Phe, a dipeptide from Aβ42, which is also the most widely studied scaffold of supramolecular hydrogelators (Diaferia et al., 2019b). However, despite the easy synthesis and decoration of dipeptide, the prediction and design of dipeptide-based hydrogels are still challenging due to the complex nature of the molecular structure and hydrogel behavior (Fichman and Gazit, 2014; Frederix et al., 2015; Fuentes-Caparrós et al., 2019). Researchers have tried many strategies in dipeptide design to resolve this problem, such as the introduction of a combinational approach, which generated a structurally diverse hydrogel library with more than 2,000 peptides (Li et al., 2019a).

Recently, self-assembly hydrogels based on the longer (Levin et al., 2020; Mondal et al., 2020) or short peptides (Fleming and Ulijn, 2014; Tao et al., 2016; Guyon et al., 2018) and their related gel-based biomedical applications (Fukunaga et al., 2019; Ni and Zhuo, 2019) have been reviewed but in a very limited manner. In this review, we focus on the dipeptide derived from common amino acids that form hydrogels and the possible relationship between the structures and hydrogel behavior. We hope that with a deep understanding of the relationship between structures and properties of dipeptides, the dipeptide hydrogels would have an effective application in preparing minimal biocompatible materials.

To form a hydrogel, the dipeptide hydrogelator first self-assembled into some kind of one-dimensional (1D) supramolecular structures, like nanofibers and twisted nanoribbons, and then into three-dimensional (3D) networks to ensnare a great number of water molecules inside (Du et al., 2015). This demands that the structure of the dipeptide hydrogelator should greatly facilitate the formation of one-dimensional assemblies. The classic dipeptide hydrogelator is divided into three parts, as shown in Figure 1. The two same or different amino acids are linked via an amide bond to form the main chain of dipeptide hydrogelator, which can provide intermolecular hydrogen bonding in the self-assembling process. The other two parts are also important and are recognized as the side chain, which helps the gelator to self-assemble into associated 3D networks. As we know, hydrogel is one of the most typical amphiphilic materials in which solvent molecules are entrapped within the 3D networks under suitable conditions (Bahram et al., 2016). Unlike in covalently connected polymers (polypeptide or proteins), the structure of dipeptide hydrogelator is relatively simple and allows one to tune the formation easily and subtle changes can be applied to the structure that may lead to various hydrogel behaviors: gel or not gel.

The amino acids have different side chains for various hydrophilic and hydrophobic properties. There are the 20 different most common amino acids in nature and each of them has specific chemical characteristics, which have a unique role in protein structure and function. Based on the characteristics of the side chain when it is in contact with water, amino acids can be classified into three categories: hydrophobic (low propensity to be in contact with water), polar, and charged (energetically favorable contacts with water). Due to the simple synthesis of dipeptides, they have significant advantages in regulating physicochemical behavior by peptide sequence design (Bak et al., 2015). Even if there is only one amino acid difference, the self-assembled nanostructures and the resulting hydrogels can be significantly changed (Habibi et al., 2016; Zhou et al., 2019).

The side chains of the N-terminus and C-terminus generally provide an auxiliary function for a better equilibrium of solvability and self-assembly capacity, such as enhancing the hydrophobicity and the intermolecular forces. It has been proven that an efficient way to construct a supramolecular hydrogelator is to attach a π-conjugated group to a short peptide (Adams and Topham, 2010). For example, the Fmoc-Phe-Phe (Fmoc-FF) dipeptide, derived from the β-amyloid peptide, is mediated by a potent combination of hydrophobic, π–π, and hydrogen bonding interactions that result in the formation of hydrogel (Ryan and Nilsson, 2012; Rajbhandary and Nilsson, 2017).

The strategies for preparing the new generation of dipeptide hydrogels can be divided into four categories based on the modification site of dipeptide: 1) N-terminal modified dipeptide; 2) C-terminal modified dipeptide; 3) uncapped dipeptide; 4) cyclic dipeptide. This review will focus on the linear dipeptide (1–3) and other recent reviews about cyclic dipeptide can be found in articles by Yu et al. (2020) and Balachandra et al. (2021).

To form an efficient dipeptide hydrogel, it is possible to modify a large aromatic group on the N-terminal of the dipeptide hydrogelator (Fichman and Gazit, 2014). Aromatic moieties can facilitate the self-assembly of peptides and stability of conformations and functions due to the favorite contribution of the aromatic group, such as π–π stacking and hydrophobic interactions. The aromatic group includes fluorenylmethoxycarbonyl (Fmoc), naphthalene (Nap) derivatives, phenothiazine (PTZ), carboxybenzyl (Cbz), azobenzene (Azo), and pyrene (Pyr). The literature has reported that unmodified dipeptides without aromatic capping do not form hydrogels, whereas several N-terminal modified peptides have been used as efficient hydrogelators; the detailed contents are discussed as follows.

The C-terminal carboxylic acid moieties of dipeptides could usually be modified to perturb the hydrophobicity and hydrogen bond capacity of the C-termius, which showed significant effects on the self-assembling propensity, such as the hydrogelation ability and morphology of the self-assembled nanostructures.

Tuttle and Ulijn et al. have provided a combined experimental/computational method to evaluate the hydrogelation properties of various C-terminus modifications (amide vs. ester) of Fmoc-dipeptide (Sasselli et al., 2016). The approach proved that amide modification was more important than ester modification because the Fmoc-Thr-Phe-NH₂ could form more stable hydrogels than Fmoc-Thr-Phe-OMe, possibly due to the extra hydrogen bonds that the amide group forms upon self-assembly. Ikeda et al. have reported C-terminal hydrazide-modified dipeptide, Cbz-Phe-Phe-NHNH₂, which showed limited aqueous solubility and could not form a hydrogel, but its carbohydrate derivatives could form a hydrogel (Tsuzuki et al., 2017) (Figure 9). They have found out that the disaccharide structures (epimer or glycosidic-bond geometry) have a significant effect on the formation ability of the hydrogel and the morphology of the self-assembled structures. The hydrazide group C-terminal modified p-nitro-phenylmethoxycarbonyl (NPmoc)-based dipeptide NPmoc-Phe-Ala-NHNH₂ could form a hydrogel, while the others like NPmoc-Ala-Phe-NHNH₂, NPmoc-Phe-Phe-NHNH_2, and NPmoc-Ala-Ala-NHNH₂ failed to form hydrogel at the same conditions (Ohtomi et al., 2020). Their results also indicated that adjusting the side chain phenyl group position in the dipeptide hydrazides could vary the dimensions of the self-assembled nanostructures from 1D (fiber and tubular) to 2D (sheet).

Coutsolelos et al. have synthesized several aliphatic dipeptides bearing various protecting groups and found out that Fmoc-Ile-Ile-TPP (TPP: tetraphenyl porphyrin) formed a hydrogel in HFIP–water (2:8) solvent mixture, whereas TPP-Ile-Ile-OMe and Boc-Ile-Ile-TPP failed to form hydrogel due to the spherical assemblies in solvents (Figure 10) (Nikoloudakis et al., 2019).

Although uncapped dipeptides are very attractive building blocks due to their chemical simplicity, they usually fail to form a stable hydrogel, e.g., the Phe-Phe hydrogels were reported to be metastable (Conte et al., 2016; Kurbasic et al., 2019; Kralj et al., 2020), unless the Phe benzene ring of the dipeptide was further modified by adding a p-nitro (Kurbasic et al., 2019) or p-iondin (Kralj et al., 2020) substitution. For the uncapped dipeptide composed of natural amino acids, Ventura and coworkers have reported that Ile-Phe dipeptide can form a stable hydrogel at pH 5.8 (de Groot et al., 2007); later, Marchesan et al. found out that Leu-Phe dipeptide could form a stable hydrogel in phosphate buffer (Bellotto et al., 2020). However, besides these two dipeptides, no other dipeptide hydrogelator composed of natural amino acids has been reported to form a stable hydrogel.

α,β-Dehydrophenylalanine (ΔPhe), an unsaturated analog of Phe, was found to induce conformational constraints in the peptide backbone and to usually yield a hydrogel. Chauhan et al. have reported that uncapped dipeptide Phe-ΔPhe could self-assemble into a stable hydrogel at minimal gelation concentrations of 0.2 wt% in a buffer solution at pH 7.0 (Panda et al., 2008). Yadav and Thota have reported another ΔPhe-containing dipeptide, Leu-ΔPhe, which could form a highly stable and mechanically strong hydrogel under mild physiological aqueous conditions (Thota et al., 2016). These hydrogels from ΔPhe-containing dipeptide were found to be nontoxic and were used to entrap several hydrophobic and hydrophilic drug molecules and release them in a controlled manner (Figure 11).

The chirality of the composed amino acids affects the hydrogelation ability of the uncapped dipeptide. Marchesan et al. chose diphenylalanine as a model compound and studied the self-organization of heterochiral D-Phe-L-Phe and its halogenated derivatives (Figure 12, Kralj et al., 2020). Their studies showed that D-Phe-L-Phe self-assembled into nanofibrils and resulted in a transparent hydrogel. Besides, they also found that halogenation had significant effects on the self-assembling process. For example, fluorination induced the analogous packing to nanotube formation, and iodination was the most effective strategy to augment the stability of the hydrogel.

Multicomponent self-assembly is a hot topic at present, which generally has advantages over a single component system (Li et al., 2021a). Multicomponent self-assembly of two or more different building blocks into one ordered nanostructure can promote the formation of a wider and more complex architecture, provide the tunable mechanical properties, enhance stability, and provide spatiotemporal control of self-assembly (Raeburn and Adams, 2015; Halperin-Sternfeld et al., 2017; Makam and Gazit, 2018). Recently, multicomponent hydrogels containing dipeptides have also attracted interest due to the innovative scaffolds from multicomponent self-assembly. Some studies have reported that the multicomponent self-assembly containing dipeptides has significant effects on the mechanical properties of those formed hydrogels. For example, Tendler and coworkers have reported that the mechanical properties of hydrogels formed by two peptides Phe-Phe and di-D-2-napthylalanine could be fine-tuned via changing their relative concentration (Sedman et al., 2013). Abramovich et al. have analyzed the gelation kinetics of mixed Fmoc-Phe-Phe/pentafluorinated Fmoc-Phe (Figure 13A) and discovered that the mechanical properties were improved (Halperin-Sternfeld et al., 2017). In the same way, the mix of Fmoc-Phe-Phe and Fmoc-Arg can form new hydrogels, which demonstrated high mechanical strength with a storage modulus of up to 29 kPa in combination with the bone mineral hydroxyapatite (Ghosh et al., 2017).

Fmoc-Phe-Phe and hexapeptide-based multicomponent hydrogels were studied by the Accardo group, and they found out that the combination of Fmoc-Phe-Phe with (Phe-Tyf)₃ hexapeptide or its PEGylated analog (Figure 13B) at different volumetric ratios (2/1 or 1/1 v/v) could provide hydrogels with tunable mechanical properties (Diaferia et al., 2019a). For instance, the PEGylated gel showed a decrease of the gel rigidity and slowing down of the gel formation kinetics compared to the un-PEGylated gel, and thus the mixed hydrogel Fmoc-Phe-Phe/PEG₈-(Phe-Tyf)₃ (1/1) have minor rigidity compared to Fmoc-Phe-Phe/PEG₈-(Phe-Tyf)₃ (2/1).

Ulijn and coworkers examined four different combinations of Fmoc- and pyrene (Py)-modified compounds (Fleming et al., 2014). The results showed that the combinations of two structurally different peptides (Py-Tyr-Leu/Fmoc-Ser and Fmoc-Tyr-Leu/Py-Ser) resulted in orthogonal coassembly, while the structurally similar peptides (Py-Tyr-Leu/Fmoc-Tyr-Leu and Py-Ser/Fmoc-Ser) followed the cooperative coassembly process (Figures 14A,B,D). They discovered that Pyr-Ser could perturb the intermolecular H-bonding and fiber formation associated with Fmoc-Tyr-Leu and Py-Tyr-Leu (Figures 14C,E), while Fmoc-Ser showed a disruptive effect on the self-assembly structure of Fmoc-Tyr-Leu (Figure 14E).

Yan et al. utilized the combination of dipeptide Fmoc-Phe-Phe and fullerene to improve the mechanical properties of the hydrogel for a better injectable formulation for biomedical applications (Zhang et al., 2018). Due to the coassembly of dipeptide and fullerene, the aggregation of fullerene in the hydrogel was largely inhibited via the noncovalent interactions between the dipeptide and the fullerene, and at the same time, fullerene enhanced the mechanical properties of the dipeptide hydrogel. The incorporation of the fullerene profoundly improved the photodynamic therapy (PDT) efficiency compared to the untreated fullerene, and the dipeptide-fullerene hydrogels could effectively inhibit multiantibiotic-resistant Staphylococcus aureus both in vitro and in vivo.

Self-assembly dipeptide hydrogel is the result of coordinated noncovalent interactions among molecular, environmental, and kinetic considerations. Although some successful dipeptide hydrogels have been reported, it is still difficult to predict and control the results of molecular assembly, which are derived by various factors that are external (solvent, temperature, light, etc.) and internal (molecular geometry, hydrophobicity, electronics, etc.). Nonetheless, through the subtle design of dipeptide molecules and the modification of the N- and C-terminus, dipeptide hydrogels can have great potential applications in preparing minimal biocompatible materials.