Synthesizing proteins is crucial for living cells, as these biomacromolecules take a significant role in almost all cellular behaviors. They have complicated quaternary structures and can function as structural supports of organisms or various bio-catalysts. Therefore, producing proteins on-demand in an independent and efficient way is fundamental for a biomimetic artificial cell. A cell-free protein synthesis (CFPS) system (Nirenberg and Matthaei, 1961) is an in vitro system, which enables transcription and translation processes outside cells. It simplifies the complex interactions in a living cell and only consists of the necessary elements for protein synthesis, such as gene templates, related enzymes, amino acids, salts, energy substrates, and cofactors. The principle behind CFPS is the biological central dogma, which refers to the transmission of genetic information from DNA to message RNA, and finally toward proteins. Due to the openness, simplicity, and operability, the CFPS system can provide an ideal platform for protein synthesis in artificial cells (Shin and Noireaux, 2012; Wu and Tan, 2014). Moreover, protein synthesis using recombinant elements (PURE) (Shimizu et al., 2001; Lavickova and Maerkl, 2019) system could completely customize the artificial cells, as its components are entirely purified and remain minimal but necessary for protein synthesis. For example, by adjusting substrates and enzymes in the PURE system, the metabolic pathways were controlled, and the detailed roles of ATPs were revealed in an ATP-producing artificial cell (Berhanu et al., 2019).

Although successfully producing model proteins like fluorescent proteins (Shin and Noireaux, 2012; Caschera and Noireaux, 2014), CFPS systems should be furtherly expanded to support the synthesis of complex and specific proteins. For example, membrane proteins are common proteins that perform crucial functions for cell survival (Almén et al., 2009). They could serve as enzymes, signal receptors, and transport machines. Therefore, the synthesis of membrane proteins has connected the boundary and the interior, thus facilitating the construction of artificial cells as an organic whole. Cell-free systems could provide an attractive platform to produce membrane proteins by adding surfactants or liposomes (Lu, 2017). These methods might still function in a membrane-bounded artificial cell. Several membrane proteins have been successfully synthesized and functioned in artificial cells, such as bacteriorhodopsins (Berhanu et al., 2019), cell division proteins FtsA (Furusato et al., 2018; Godino et al., 2020), and E. coli Min proteins (Godino et al., 2019). However, the correct assembling of de novo synthesized membrane proteins remains inefficient in artificial cells, partly due to the unclear integration mechanism of proteins into bio-membranes.

Another challenge is the post-translation modification of proteins in vitro. Most proteins require further modification after the translation process, such as glycosylation and phosphorylation, to become mature and undertake advanced physiological functions (Wang Y.-C. et al., 2014). However, these might be difficult or inefficient in vitro, as the explicit modification processes are often complicated, involving multiple metabolites (Kightlinger et al., 2019). Some countermeasures are the fabrication of cell extracts in CFPS from eukaryotic cells and the supplement of disulfide bond-forming enzymes or chaperones (Goerke and Swartz, 2008; Lu, 2017). Up to now, results have been achieved to introduce the complete enzyme systems or pathways for post-translation modifications in the cell-free solution (Kightlinger et al., 2019). Nevertheless, the ability of artificial cells to support modification processes needs to be improved, as they would be crucial in the construction of diverse metabolic pathways and other advanced cellular functions.

Intracellular homeostasis is important for normal metabolic activities. A significant difference between cells and solutions is the crowded environment inside the former. About 30% of the volume is occupied by macromolecules such as proteins, nucleic acids, and polysaccharides, whose concentration can reach 300–400 g/L (Zhu et al., 2020), and even 500 g/L in some compartments inside cells (Minton, 2015). The high concentration of macromolecules causes the so-called “macromolecular crowding” and brings significant effects to the interactions between participating biomolecules and the final metabolism (Rivas and Minton, 2016). First, the crowding environment can cause the excluded volume effect, which reduces the diffusion coefficients of both small and large molecules (Tan et al., 2013; Smith et al., 2017). Second, it also increases the equilibrium constants associated with macromolecules, probably due to the decrease in the total free energy of the solution after binding to molecules (Ellis, 2001). Moreover, the crowding effect can enhance the association of proteins, especially facilitating the folding or aggregation of proteins (Ellis, 2001; Bai et al., 2017). As bioreactions could be affected by both the diffusion process and molecule activities, the crowding effect on distinct bioreactions could be complicated with the increased concentration of macromolecules. In this case, creating a similar physiological medium inside an artificial cell deserves great attention.

Although the direct encapsulation could form a restricted interior, the concentration of macromolecules is usually not able to reach as high as in living cells (Minton, 2006). An intuitive strategy is to add inert crowding agents, such as poly (ethylene glycol) (PEG), dextran, and ficoll (van den Berg et al., 2017). These generally drift away from the major metabolism and might perform differently according to their species and concentrations (Ge et al., 2011). Therefore, this method might be limited to some extent. Considering the mismatching of diluted cell-free system with the large size of compartments, it would also be applicable to condense the cell-free reaction mixtures (Fujiwara and Nomura, 2013) and adjust the size of artificial cells through osmotic pressure regulation to create a more crowding internal environment. For example, monodisperse liposomes could shrink three orders of magnitudes to form crowded artificial cells (Deng et al., 2018). The interior solution finally reached about 272 g/L, which was close to the living cell.

Reconstitution of the crowded environment in an artificial cell is crucial in a broader sense. It causes the cytoplasm to be more similar to gels rather than solutions, and even leads to the lipid-lipid phase separation (LLPS) to get membrane-free compartments (Ma et al., 2020). These intracellular structures are involved in diverse metabolism in both eukaryotes and prokaryotes (Wei et al., 2020), and have been proved to benefit the protein synthesis (Sokolova et al., 2013; Tang et al., 2015; Deng et al., 2018). Condensates inside artificial cells were successfully triggered through LLPS to conduct metabolic behaviors, such as enzyme catalysis (Deshpande et al., 2019a; Love et al., 2020), and study the potential coacervate-membrane interactions (Last et al., 2020). In a complicated metabolic network, these membrane-less organelles would ensure orderly biochemical reactions and provide valuable insights toward the real metabolism inside cells (Deshpande and Dekker, 2021). In this case, the crowded internal homeostasis and the local confinement would help construct an efficient and robust metabolic network inside an artificial cell.

The encapsulation of one single enzymatic reaction would be relatively simple in the artificial cell. However, the complex bioreaction cascades or networks seem challenging because of the complicated processes and ambiguous mechanisms. Here, we propose that the recycling processes, which mainly refer to the circulation of metabolites and enzymes, should be carefully considered to construct complicated metabolism. In this case, more steady output and complex metabolic operation could be realized.

A robust and sustainable reaction network relies on stable conditions, which could be simplified as the continuous supply of substrates and elimination of products. This was primarily raised to prolong the life of cell-free protein synthesis, mainly through the selective permeability (Spirin et al., 1988) or anchor of the pore proteins onto membranes (Noireaux and Libchaber, 2004). After that, nanoliter-scale reactors, which could constantly exchange all molecules, were constructed to carry out diverse regulatory mechanisms (Niederholtmeyer et al., 2013). Meanwhile, the auto-regulated protein synthesis was achieved based on the on-chip DNA compartments (Karzbrun et al., 2014). These results proved the necessity of steady-state conditions for complex and continuous metabolic regulation. Recently, hydrogels with porous structures were applied to encapsulate the PURE system and support genetic oscillators through the continuous exchange of substrates and byproducts (Zhou et al., 2018; Lai et al., 2019). The circulation of materials could provide an opportunity for the construction and regulation of complicated metabolic networks.

Another crucial factor is the regeneration and digestion of biomolecules, such as nucleic acids and enzymes. As these biomolecules could lose activity during the reaction process, it is important for sustainable artificial cells to break them down and synthesize new substitutes. However, the de novo synthesis of original components is not a trivial task. On the one hand, most complex biomolecules are still hardly produced for the low efficiency or complicated assembly in vitro. On the other hand, as some functional enzymes prefer to function on membranes, the correct location is also essential. Up to now, significant progress has been made in reconstituting several vital biomolecules evolved in metabolism, including DNAs (van Nies et al., 2018; Libicher et al., 2020), ribosomes (Jewett et al., 2013; Shimojo et al., 2020), and tRNAs (Hibi et al., 2020). Recently, a partial self-regeneration artificial cell system capable of regenerating essential components was also constructed successfully (Lavickova et al., 2020). These results were mainly based on the minimal reaction demand. The next step would be substrate recycling in more complicated metabolic networks.

As for the digestion part, hydrolytic enzymes are commonly used in living cells for the digestion of biomolecules. However, their optimal pH is often at 4.5–5. To provide a proper degradation environment and prevent interference with other bioreactions, the encapsulation of hydrolytic enzymes as a lysosome-like module seems applicable. Actually, lysosomes inside eukaryotic cells consist of many hydrolytic enzymes, and the guidance of inactive biomolecules toward digestion is complicated and still under exploration (Bonam et al., 2019). Carreira et al. (2017) developed a type of lysosome-mimicking vesicles and discovered that the accumulation of sphingosines could cause the permeability and other biophysical changes of the membrane, which might facilitate the digestive function of lysosomes. The degradation of enzymes or other biomolecules is generally a necessary part of metabolism. Conclusively, more efforts should be made to construct an efficient recycling system inside an artificial cell or between artificial cells and the environment.

The substrate-level phosphorylation refers to the direct coupling of energy-rich compounds with ATP production. Phosphate groups are extracted from the substrates and transferred to ADP or GDP to produce ATP or GTP. Because of the simplicity and easy-operating, this is widely applied in cell-free systems. Traditional high-energy substrates include phosphoenolpyruvate, creatine phosphate, and acetyl phosphate, which could support protein synthesis at the level of g/L (Des Soye et al., 2018). However, the phosphate compounds could hamper the synthesis in a batch reaction, for their high costs, the accumulation of inorganic phosphate, and the pH change (Jewett and Swartz, 2004; Kim et al., 2007). In this case, glycolytic intermediates such as pyruvate (Jewett and Swartz, 2004), maltodextrin (Kim et al., 2011), and maltose (Caschera and Noireaux, 2014) were introduced for the recycling of inorganic phosphate, thus facilitating the ATP regeneration and improving the homeostasis of the cell-free system. Nonetheless, they might suffer from the inefficient utilization of the ATPs and other unexpected byproducts, as additional metabolic pathways were generally included (Caschera and Noireaux, 2014).

Another approach is to build a material exchange system. Through the continuous supply of energy and removal of harmful byproducts, artificial cells could reach a longer lifespan. For example, dialysis membranes were used to separate the cell-free system from the feeding buffer (Kim and Choi, 1996). Therefore, small molecules such as the energy substrates could diffuse across the membrane and support sustainable protein synthesis. Furthermore, pore-forming proteins were also anchored onto the membrane to improve the mass transfer (Noireaux and Libchaber, 2004). Recently, artificial cells based on porous hydrogels were constructed to conduct long-lived protein synthesis (Zhou et al., 2018; Lai et al., 2019). Although the interaction with the environment was a primary characteristic of a living system, the continuous supply of energy substrates might be limited in practice by the high costs and the specific feeding system.

It is worthwhile to construct a self-regeneration system for energy in an artificial cell from the perspective of reconstructing life. In nature, living cells can produce ATPs efficiently through the oxidation-phosphorylation and photo-phosphorylation processes, which can be summarized in two steps (Jeong et al., 2020). First, a proton gradient is established across the membrane boundary. In the oxidation-phosphorylation process, the driving force can be attributed to the oxidation of energy-rich molecules, in which electrons are transported through the electron transport chains (ETCs) (Nath and Villadsen, 2015). While in the photo-phosphorylation, it is driven by light with the help of photosynthetic pigments. Second, the proton gradient drives the ATP synthase (Nesci et al., 2015) to produce ATPs. The recovery of these two processes inside an artificial cell might lead to a self-sufficient supply of energy.

The construction of the ATP synthesis system can date back to 1974, when the purple membrane from Halobacferium halobium was packaged into phospholipid vesicles to support light-driven ATP synthesis (Racker and Stoeckenius, 1974). After that, the reconstitution of purified bacteriorhodopsin (bR) and ATP synthase in proteoliposomes enabled the formation of the proton gradient and ATP synthase (Sone et al., 1977). This indicated the possibility of ATP synthesis from completely synthetic components. In the next 30 years, the efficiency and stability of the ATP synthesis system were improved through the reconstitution of ATP synthases from bacteria (Richard et al., 1995; Choi et al., 2006), plants (Feng et al., 2016; Li et al., 2019) or mitochondria (Matuschka et al., 1995), and different types of proton gradient generators (Li et al., 2019; Tarryn et al., 2020; Xu et al., 2021). However, the membrane materials were generally based on lipids, probably due to their excellent compatibility with membrane proteins. Therefore, the various lipid compositions were studied to test their effects, and the protein densities were found to play a great role in ATP synthesis (Nilsson et al., 2016). In this case, the ATP synthesis system, or namely the artificial organelle for energy generation, was established in an efficient way.

These pre-prepared modules could be integrated into vesicles to drive other bioreactions. In 2018, a switchable ATP synthesis system, which could respond to light with different wavelengths, was coupled with actin polymerization and finally led to membrane growth (Lee et al., 2018). Then, in the next year, a similar photosynthetic organelle composed of bR proteins and ATP synthases was introduced into the recombinant cell-free system to support the complex transcription and translation processes (Berhanu et al., 2019). This fascinating result proved the compatibility of self-sufficient ATP synthesis with protein production. Recently, a chloroplast mimic with natural and synthetic parts was constructed to conduct a complete metabolic cycle for continuous CO₂ fixing (Tarryn et al., 2020). Although significant progress has been achieved, these synthetic energy modules might still suffer from low efficiency. Most of them were only utilized in simple reaction cascades. A possible solution might be the loading of more functional proteins. For instance, the vast surfaces of mitochondria and chloroplasts in nature have enabled the complex metabolic networks inside cells. Therefore, the construction of membrane structures with a considerable surface area could facilitate the energy supplement in logic and finally help construct an energetically independent artificial cell.

Growth represents the volume increase and mainly happens prior to the division. The metabolism should also get enhanced in this process. Generally, artificial cells could grow in three forms: absorbing materials from the environment, fusing with each other, and creating the materials from inside.

Membrane vesicles composed of fatty acids could absorb alkaline micelles spontaneously from outside and were first studied to imitate the cell growing process (Chen and Szostak, 2004; Stano and Luisi, 2010). This model might illustrate the state of primitive cells, but was probably not suitable for artificial cells with complex functions, as fatty acids were not compatible with some biological reactions and limited as a membrane material (Schwille et al., 2018). Afterward, phospholipid-based vesicles were successfully constructed to grow by fusing with each other. Giant unilamellar vesicles (GUVs) could gradually grow when supplied with small unilamellar vesicles (SUVs). The processes were mediated by the membrane tension through a transmembrane osmotic pressure (Deshpande et al., 2019b). To avoid the disordered fusion, the lipid membranes could be modified by specific molecules so that the fusion process could be initiated in the desired manner. For example, by using negatively charged lipids, the fusion of vesicles could be achieved through electrostatic repulsion (Terasawa et al., 2012). The growing process could also be light-controlled through the use of photo-sensitive surfactants (Suzuki et al., 2017). Besides, the protein-lipid interaction was proved promising for the induction of membrane tethering and fusion (Cai et al., 2019). Some polymer-based vesicles could grow upon integrating stimuli-responsive nanoparticles, which contained the membrane materials (Schwille et al., 2018). Compared with the absorption mechanism, the fusion process could supply other necessary substrates to meet the metabolic burden brought by the volume increase, thus might be popular in future studies. These two strategies both rely on the acquisition of the homologous membrane materials from the environment and reflect a primary growing form. A more lucrative way is that artificial cells synthesize their own membrane components and integrate them on the preexisting membrane. The synthesis of phospholipids has been achieved by purified enzymes in vitro (Schmidli et al., 1991), which provided a solid precondition. Recently, membrane growing was realized through the internal biosynthesis of lipids (Exterkate et al., 2018; Blanken et al., 2020). These excellent works would greatly improve the construction of an autonomous and self-reproducible artificial cell. Except for the volume increase, the morphological change was another growing form, such as the axon outgrowth. Artificial cells growing into asymmetric forms might help us to understand cell differentiation and deserve more exploring.

As the next step of growth, division is almost inevitable for most cells. In this process, the cell membrane would deform to decrease its volume-to-surface ratio and become the dumbbell shape connected by a narrow membrane neck. Meanwhile, the interior is distributed into two different compartments. After the collapse of the membrane neck, the interior is separated, and the cell eventually divides into two daughter cells. Inspired by this, the division of an artificial cell could be resolved into the division of compartments and the allocation of crucial components such as genetic materials and enzymes. In the following text, these two aspects will be discussed in details.

GUVs are broadly used to study the division process, as they possess great fluidity and are most similar to natural membrane systems. By constructing fissionable GUVs, several models of membrane division induced by physical mechanism, chemical synthesis, or fundamental biological processes had been established (Caspi and Dekker, 2014). For example, some division machinery based on structure proteins, such as actomyosin and FtsZ, were found necessary in the division cycle. After the reconstitution of these proteins on GUVs, the liposome division could be observed in some cases (Osawa and Erickson, 2013). Besides, the cell-free synthesized bacterial division proto-rings could help to constrict liposomes and generate budding vesicles (Godino et al., 2020). This model relied on the specific splitting proteins anchored on lipid membranes. However, membrane fission could also be induced regardless of the protein structure. The steric pressure among the closely arranged membrane proteins could increase the membrane curvature and finally lead to a division. As a result, even green fluorescent proteins bounded on the membrane could drive fission efficiently (Snead et al., 2017). Furthermore, even low-density membrane proteins were feasible to induce the vesicle division by the precise control of the spontaneous membrane curvature (Steinkuhler et al., 2020). These efforts greatly decrease the complexity of the division process in vitro. However, they often resulted in unpredictable division results both of the volume and the components. As a result, microfluidics was applied to cut liposomes mechanically, thus precisely controlling the generation of daughter cells (Deshpande et al., 2018). This strategy could produce homogeneous offspring but be limited by the size of the vesicles and the specific wedge-shaped splitter. Recently, the spatiotemporal control of division was realized through the suitable osmolarity change based on the phase-separated GUVs. In this case, only the target GUVs would divide, while the others were not affected (Dreher et al., 2021). Based on these current works, one typical issue is that the allocation of the volume and components seems pretty random during the membrane fission. Considering the complex metabolism encapsulated inside vesicles, the precise and average division of the artificial mother cell would be pretty difficult. Some breakthroughs were achieved by the coupling of compartments and gene materials. Through the interaction between the negatively charged DNA and the cationic membrane, the division of the compartments could be linked with the amplification of DNA (Kurihara et al., 2011). As a result, the latter could significantly induce the growth and division of vesicles. On the other hand, RNA as another gene information carrier was encapsulated inside fatty acid vesicles and successfully distributed to the daughter cells (Zhu and Szostak, 2009). Nevertheless, these models could not illustrate the precisely average distribution of maternal genetic information. This is a crucial question, as the volume and other components might be separated mechanically through microfluidics, but the information materials should be replicated and distributed accurately.

Although the growth and division of artificial cells have distinctive processes, they all experience the change in volume or morphology. Especially, the asymmetric division, in which the offspring can inherit different components, should not be neglected, as it is inevitable in cell differentiation and various life course (Andes-Koback and Keating, 2011). In reference to the natural processes, the accurate distribution of gene materials and the precise control of division, both average and asymmetric, would be the focus of future works.

Generally, the communication process requires three elements, the sender, the receiver, and the signal-transmitting information. The senders catch the signals released by the receivers and make responses accordingly. The signals could be physically, chemically, or biologically originated. The response could be the production or release of target molecules, cell motility, or just the simple changes in morphology. The simplest model of the artificial communication process is that an artificial cell receives signals from the environment and takes specific responses. For example, semipermeable compartments could allow the exchange of reaction substrates, thus facilitating protein synthesis (Zhou et al., 2018; Lai et al., 2019). The intelligent release of agents was realized based on the induction of various environmental stimuli, including pH, osmotic pressure, light, ultrasound, and temperature (Fomina et al., 2010; Wang Y. C. et al., 2014; Basuki et al., 2017). However, the signal sources are not rigidly limited, which can also be artificially supplied. A more complex communication model requires specific signal senders. In this case, scientists aimed to construct another group of artificial sender cells, which released signals to the environment (Adamala et al., 2017; Joesaar et al., 2019). Except for all syntheses, natural cells could also be used as senders or receivers to interfere with artificial cells (Elani, 2021). Furthermore, based on the sender-receiver model, feedbacks could be introduced rather than one-way communication. For example, while the senders were the receivers themselves, they could regulate bioreactions based on their own products, which showed interesting collective behaviors as a community (Gines et al., 2017). Besides, the bidirectional communication was realized between artificial cells based on DNA strand-displacement circuits (Joesaar et al., 2019). Nevertheless, the typical communication types are not so rich. The construction of multi-directional communication still needs further studies, as well as the precise control of information processing.

Signals hold an essential place in an artificial communication system, for they relate to the design of artificial cells and determine the transmission type. As discussed in Section 2.4.1, signals could be various. The intuitive method of receiving external signals utilizes stimuli-sensitive materials to construct receivers (Wang Y.-C. et al., 2014). These materials could change their morphology, swell or shrink, or modify their pore structures to regulate the transmission of signal molecules (Yavuz et al., 2009; Yan et al., 2013; Wang et al., 2020). More generally, it is applicable to introduce controlling modules from gene level, such as light-activated T7 promoter (Booth et al., 2016), various biosensors for small molecules, or specific ions (Jung et al., 2020; Thavarajah et al., 2020), to regulate the gene expression. This strategy is not limited by the sensitive materials. After the integration of different sensing modules inside one compartment, more complicated communication abilities would be practicable in artificial cells.

Another remarkable point is the transmission style of signals if information molecules are applied. Artificial cells could release or receive signals just through the diffusion of chemical molecules. Compared with the macromolecules, smaller molecules like CO₂ (Yan et al., 2013), glucose (Arya et al., 2016), N-(3-oxo-hexanoyl)-L-homoserine lactone (3OC6HSL) (Lentini et al., 2017), can pass through the semipermeable membranes more easily, thus are broadly used as signals. Large molecules are also applicable, but they generally need large pore sizes or just membrane-free compartments. For example, DNA segments could diffuse into proteinosomes with porous structures to initiate the protein synthesis (Joesaar et al., 2019). Artificial cells based on clay hydrogels possessed large pores and even communicated through diffusive protein signals (Niederholtmeyer et al., 2018). These porous materials were more tolerant for molecular transfer than the traditional liposomes. Nonetheless, it would be more difficult to modify their surface with functional biological macromolecules compared with liposomes (Rideau et al., 2018). The impermeable molecules could go through the lipid membrane by the reconstitution of membrane pores, such as α-hemolysin (Adamala et al., 2017; Hilburger et al., 2019). These transmission processes were mostly driven by passive molecular motion and almost uncontrollable. They might be inefficient and interfere with each other in a complicated system with multi signals. To solve this problem, communications based on specific interactions have provided a feasible idea. For instance, the antibody binding could selectively target the specific bacterial and trigger a response when supplied in two groups of bacterial (Fernandes et al., 2010). Similar interaction includes the avidin-biotin complexation, which was realized in a synthetic vesicle-to-vesicle communication system (Ding et al., 2019). These specific interactions mostly rely on the biomolecules anchoring on the membrane, and thus the bio-compatibility and modifiability of membrane materials are extremely important.

Communication is a kind of group behavior, which involves a large number of individual artificial cells. Based on the sender-signal-receiver models mentioned above, various forms of communications could be established, whether they exist in reality or not. Quorum sensing (QS) is pretty common in bacterial populations, which refers to the regulation of gene expression in response to the concentration of autoinducers produced by bacteria themselves (Miller and Bassler, 2001). Quorum sensing is usually determined by the cell-population density. For instance, fluorescence could accumulate in artificial cells only at high densities by using T3 RNA polymerases as QS signals (Niederholtmeyer et al., 2018). The signal molecules could also spread and form spatial distribution to induce different activities like pulses and waves in the artificial cell groups (Tayar et al., 2015). In contrast to this heterogeneous phenomena, synchrony and pattern formation were also realized through the nonlinear genetic oscillators coupling on a chip of artificial cells (Tayar et al., 2017). However, this collective behavior was still hard to be reproduced in vitro, probably due to the low robustness of cell-free oscillators. Another exciting application of the synthetic communication process is to mimic the natural differentiation of cell groups, which is the basis of development for advanced species. For example, by implementing different types of feedback gene circuits, artificial cells could exhibit differentiated gene expression during the signaling process (Dupin and Simmel, 2019). In a word, the studies of synthetic communication systems would not only improve our cognitions of collective cellular behaviors but facilitate the construction of a dynamically connected artificial cell community.

Evolution is a typical and crucial characteristic of life (Szostak et al., 2001; Schwille et al., 2018). Here we put it in this section as it reflects the communication between artificial cells and the environment. The artificial cells can react to the environmental pressure and take a response to change themselves, which could be inherited by their future generations. Therefore, the evolution process is largely related to the information coupled with the fitness of artificial cells. A typical model was that RNAs with special structures coupled with the interior vesicle wall could facilitate the stability of vesicle and improve survival (Szostak et al., 2001). The complementary pairing between short oligonucleotides and replicase ribozymes was also studied as an evolution system model (Engelhart et al., 2016). The oligonucleotides could combine with the ribozymes and inhibit vesicle growth. In addition to the typical response-receive model, the tight connection between the response and the survival of artificial cells is also required to provide the evolution engine. Since the proliferation of artificial cells has not been stably achieved, the construction of artificial evolution systems still has a long way to go. However, it should be considered necessary to build a self-sustaining artificial cell.