Chemical–biological approaches for the direct regulation of cell–cell aggregation

Cell–cell aggregation is one of the most well‐known modes of intercellular communication. The aggregation also plays a vital role in the formation of multicellularity, thus manipulating the growth and development of organisms. In the past decades, cell–cell aggregation‐related bioprocesses and molecular mechanisms have attracted enormous interest from scientists in biology, and bioengineering. People have developed a series of strategies to artificially regulate cell–cell aggregation through chemical–biological approaches. To date, not only the chemical reagents such as coordination compounds and polymers but also the biomacromolecules such as proteins and nucleic acids, are employed as the “cell glue” to achieve the control of the cell aggregation. So it is meaningful to review the recent advances of the chemical–biological approaches in cell–cell aggregation manipulation. In this review, we discuss the mechanisms and features of recently developed strategies to control cell–cell aggregation. We introduce molecules and designs relying on chemical reactions and biological conjugations respectively, and talk about their advantages and suitable applications. A perspective on the challenges in future applications in cell manipulation and cell‐based therapy is also proposed. We expect this review could inspire innovative work on manipulating cell–cell aggregation and further modulate cell–cell interactions in the research of bio/chemical fields.


INTRODUCTION
Cell-cell aggregation, which biologically appears in cell recognition and cell communications, has been widely known as an essential process of cell-cell interactions, formation of multicellular organisms, and integration of biofunctions. [1,2] Naturally, aggregation between cells is achieved through the interactions of particular molecules anchoring on cell surfaces. Through cell-cell aggregation, basic cell behaviors, such as cell activation, apoptosis, migration, proliferation, and differentiation, are elaborately controlled to ensure the regulatory status of the physiological processes including the regulation of tissue homeostasis, the immune response, and so on. [3] For example, recent researches have reported that the circular tumor cells would perform accelerated metabolism once they have formed into aggregated clusters. [4][5][6][7] Accordingly, these intercellular reactions should be highly specific and tightly regulated. Accumulated evidence has shown that many diseases such as metabolic disorders, autoimmune disruption, and even cancers are related to dysfunctional cell-cell aggregation. [8,9] Therefore, precise cell-cell aggregation manipulation would help push forward the study of various cellular processes in the molecular mechanism. Artificial control of cell-cell aggregation has shown great potential not only in cell biology study, but also in many research fields including tissue engineering, organ reproduction, and cell-based diagnosis and treatment. [10][11][12][13][14] Nowadays, numerous strategies have been developed to achieve programmed cell-cell aggregations, thus promoting people to better understand the biological processes in living organisms and to deal with different diseases. [15,16] In natural biosystems, the intercellular interactions are usually modulated by cell-surface molecules including nonspecific aggregation molecules and specific receptor/ligand. Although different cells may be close to each other randomly during migration, such kind of proximity may not initiate any communications between them, which means that regulated cell-cell aggregation may only occur between a few specific cells. To expand the dimension of cell-cell aggregation, scientists attempted to artificially modify the cell surfaces by molecules that may interact with each other for connection or crosslink. [17] Originally, cellular engineering through genetic manipulation has been exploited to modulate the expression of specific surface receptors to improve the junction of the cells. However, the expression of endogenetic proteins is not easy to fine-tune, and the overexpression may cause undesired cell phenotypes. In addition, when this method is translated from the lab to the clinic, the obstacles including low survival and poor tissue specificity are also difficult to deal with. [18,19] Due to the development of biochemistry and molecular biology, now the novel "cell glue" molecules, including chemical molecules and biomolecules, can be artificially modified on the surface of target cells, thus achieving regulated cell aggregation. In general, the major challenges of a suitable "glue" design include (1) freely modified on cell membrane without affecting the cell function and activation; (2) stably anchored on cell membrane without continuously internalized, degraded and replaced; (3) efficiently reacted with each other to build the crosslink under cellular conditions. To date, the biomolecules-based "glue" such as complementary DNA strands, antibodies, aptamers, and acceptorligand systems (e.g., biotin-streptavidin) have demonstrated higher biocompatibility, specificity, and programmability. On the other hand, the nonbiomolecules such as polymers and host-guest molecular pairs have showed better affinity and lower cost. The development of new approaches will provide exciting tools for people who are interested in the scientific fields of biology, pharmacology, and materialogy. Therefore, it is meaningful to summarize the recent design strategies people have introduced for manipulating cell aggregation. In this review, the reported strategies for cell aggregation, including both chemical methods and biological methods, will be introduced, corresponding reaction mechanisms will be detailed, their advantages and limitations will be discussed. We are hoping that this review can give guidance to scientists in biology and chemical biology when they are trying to pick up the most suitable "glue" for study in cellrelated research, as well as inspire scientists to provide more promising "glue" strategies to promote the application of these encouraging tools.

NONBIOLOGICAL MOLECULES-INDUCED CELL AGGREGATION
To achieve precise control of cell assembly, the most superior step to be dealt with is to anchor the "glue molecules" on the cell surface stably. As we know, the cell membrane has a heterogeneous, complex, and dynamic biological structure consisting of many lipids, proteins, and carbohydrates, which contains specific functional groups such as primary amines, thiols, and diols. These functional groups may involve in versatile chemical conjugation reactions. On this account, many chemical reagents have been designed for specific modification of target cell surface. Once a surface modifica-tion is achieved, artificially manipulated cell-cell aggregation can be achieved through different chemical reaction routes including covalent connection, host-guest interaction, polymerization, and so on. For chemical reaction-induced cell aggregation, a fundamental requirement is that the reaction should not interfere with the cell activity and its physiological function. In 2000, Bertozzi and coworkers introduced the "Staudinger ligation" reaction, proposing a new term of bioorthogonal reaction. [20] To date, the well-developed bioorthogonal chemistry methodology has provided several chemical reaction approaches such as Cu-catalytic azidealkyne click (CuAAC), Staudinger ligation, cross-metathesis, and else. [21] These bioorthogonal reaction methods not only can help to modify the desired compounds on the surface of target cells but also make the linkage between cells possible. [22,23] As an improvment, scientists can also break the established linkage through photoisomerization, introduction of competitive molecules or other designed methods to realize reversible cell-cell aggregation. Reversible programming of cell-cell aggregation has the following advantages: (1) It can realize the reuse of surface-altered tool-cells; (2) it can better explore the influence temporally on dynamic cellcell aggregation and intercellular signal transduction; (3) it also provides the possibility for some extended applications, such as cell sorting, cell patterning and tissue capture/release biotechnologies. In this chapter, we will introduce those pure chemical molecules and strategies people employed, according to the reaction mechanisms used for inducing cell-cell aggregation.

Covalent chemical reaction
Generally, cell-cell connections can be actualized through the direct reaction between the molecules modified on the surfaces, so the covalent linkages are usually the first choice due to the stable chemical bonds and diverse designs. Click chemistry is a typical reaction that can provide stable covalent bonds with high reaction efficiency under gentle reaction conditions. Since the first proposition by Sharpless in 2001, [24] click chemistry has been widely used in drug development, biomedical material construction, and other fields. Taking its advantages of high yield, inoffensive by-products, simple reaction conditions, and fast reaction rate, many strategies for manipulating cell aggregation based on click chemistry methods have been developed. Yousaf group [25] realized the cell assembly via oximebased click chemistry ( Figure 1A). They constructed lipid vesicles modified with ketone and oxyamine groups, respectively. Then the functional groups were delivered to membranes of different cell populations by liposome fusion process and subsequently reacted with each other via oxime ligation to enable the modified cells to connect with each other. Using this method, researchers constructed small threedimensional (3D) spherical cell assemblies and even large and dense 3D multilayered tissue-like structures from bottom to up. In this strategy, the bioprocess-mimicking lipid fusion helped the functional groups to insert into target cells without disrupting the normal cell physiological process. Moreover, the generated covalent oxime bond was endowed with large bond energy, thus ensuring the high stability of the cell aggregates. As an improvement, the same group has embedded a F I G U R E 1 Covalent chemical reaction for cell aggregation. (A) General schematic of bio-orthogonal ketone and oxyamine molecules for subsequent chemoselective oxime ligation to realize cell-cell aggregation. Reproduced with permission: Copyright 2011, American Chemical Society. [25] (B) Schematic describes the molecular level control of tissue assembly and disassembly via a photo-switchable cell surface engineering approach. Reproduced with permission: Copyright 2014, Nature. [26] (C) Scheme shows dynamic liposome-liposome fusion and liposome-cell fusion for tailoring cell surfaces. Reproduced with permission: Copyright 2014, Wiley. [27] (D) Illustration of the cellular gluing method based on metabolic glycoengineering and double click chemistry. Reproduced with permission: Copyright 2015, Wiley. [28] (E) Illustration of the chemically detachable cellular glue system based on click chemistry linkers with degradable disulfide bonds. Reproduced with permission: Copyright 2016, American Chemical Society [29] photo-cleavable center between the lipid hydrophobic moiety and the oxyamine group ( Figure 1B). [26] Under mild UV radiation (365 nm, 10 mW/cm 2 , 5 min), the UV-controlled component could be cleaved, leading to the completed dispersal of the assembled cell aggregates. Then a remote, spatial and temporal control of cell interactions could be realized. Besides this photo-switchable method, Yousaf group [27] proposed an electrochemical reduction strategy to modulate the assembly and disassembly of cells. As shown in Figure 1C, instead of the ketone group, hydroquinone was delivered onto the cell surface via liposome fusion. After activating by chemical (CuSO 4 ⋅5H 2 O, 5 min) or electrochemical oxidation (−100 to 650 mV, 100 mV/s), hydroquinone converted to quinine and further reacted with oxyamine, thus connecting target cells via the oxime bond. The oxime would be cleaved under reductive potential (−100 mV, 10 s), causing a quick reverse to the original hydroquinone and oxyamine on the cell surfaces. In contrast to photo-controlled disassembly, this electrochemical method realized cyclic conversion between the cell assembly and disassembly with a more rapid reaction rate.
In addition to the ketone-oxylamine condensation reaction, azide-dibenzocyclootyne (DBCO) and trans-cyclooctene (TCO)-tetrazine (Tz) are also frequently-used covalent chemical linkers for cell aggregation because of their aqueous-phase reactions and copper-independent nontoxic catalyst. In recent years, Yun group [28] established cell-cell contacts via the double click chemistry method ( Figure 1D). First, the azide (N 3 ) groups were introduced onto the cell membrane through metabolic glycoengineering. Next, the N 3 groups on different cells would react with DBCO preattached with Tz or TCO, respectively. Finally, the obtained Tz-and TCO-modified cells were mixed together and connected with each other through the Tz-TCO click chemistry reaction within 10 min. In addition to the advantage of efficient linkage, the formed Jurkat T-NIH3T3 cell pair showed good mobility, high stability, and strong linkage at flow rates up to 60 ml/min corresponding to shear stress over 20 dyn/cm 2 , which is higher than typical vessel-wall shear stress in veins and arteries. Moreover, they demonstrated that after being intravenously injected into live mice, the Jurkat T-Jurkat T cell pairs could be observed in circulation and in the lung F I G U R E 2 Typical host-guest interaction for cell aggregation. (A) Structures of hosts and guests for vesicles aggregation and corresponding lightresponsive mechanism. Reproduced with permission: Copyright 2010, Wiley. [31] (B) Reversible manipulation of cell assembly and disassembly by lightresponsive host/guest pair. Reproduced with permission: Copyright 2016, Nature. [32] (C) Schematic representation of the supramolecular functionalization of cell surfaces via targeting of the membrane-receptor CXCR4 (green). Reproduced with permission: Copyright 2017, Nature [34] tissues, suggesting the potential of a novel cell delivery strategy by attaching a cargo cell to a mobile carrier cell. To extend this method, Yun group [29] synthesized DCBO-SS-Tz/TCO which has an inserted disulfide bond (S-S) between DCBO and Tz/TCO ( Figure 1E). Based on the degradability of the S-S bond caused by glutathione (GSH), controlled disassembly of cell pairs could be realized with the addition of 5 μM GSH in only 10 min. They also showed that the introduction of the S-S bond would not decrease the strength of the Tz-TCO connection. The great efficiency of both the assembly and disassembly of the cell aggregates ensures the application of the new method in tissue engineering and cell biology fields.

Host-guest interaction
Host-guest interaction is a well-noticed reaction in many research fields due to its simple reaction conditions, good reaction specificity, and high reaction efficiency. As one of the star host molecules, cyclodextrin (CD), which has been widely used in many biological applications such as drug delivery and biomedical materials synthesis, [30] has attracted great interest from people in cell aggregation manipulation. Ravoo group [31] prepared a unilamellar CD (α-CD or β-CD) bilayer vesicle as a versatile model system for the recognition, adhesion, and fusion of biological cell membranes (Fig-ure 2A). Two vesicles made from α-CD and β-CD, respectively, are linked with the addition of azobenzene (azo) because azo can react with the α-CD and β-CD through the host-guest interaction. Similarly, tert-butylbenzene is also a specific guest for β-CD, which can connect with two vesicles containing β-CD, thus regulating cell-cell contact via host-guest interaction. Qu group [32] introduced a strategy to achieve a fast and convenient reversible cycle of cell assembly and disassembly ( Figure 2B). They first modified the cell membrane with azide group through metabolic glycoengineering, then the host molecule of alkynyl-PEG-β-CD (PEG = polyethylene glycol) was installed successfully on the cell surface utilizing CuAAC. A series of azobenzene derivates were employed as the guest molecules. The trans-azobenzene could transform into cis-azobenzene under UV light irradiation and go back to trans-configuration with visible-light irradiation, while the cis-azobenzene might not react with β-CD through host-guest reaction because of the steric hindrance. Such a design provided spatiotemporal control for cell engineering. For example, the introduction of fluorescent-labeled azo achieved controllable cell imaging and the introduction of a homobifunctional guest molecule (azo-PEG-azo) made reversible cell assembly successful. Through this method, they have built up a cell cluster between peripheral blood mononuclear cells (PBMCs) and targeted cancer cells, and demonstrated that the programmable cell aggregation might induce the apoptosis of cancer cells, thus providing a new cell-based therapy for different diseases. In recent years, the same group improved their method mentioned above. They utilized lipid-modified β-CD (lipid-PEG-β-CD) to anchor cell membrane via hydrophobic insertion, thus avoiding the complicated process of β-CD introduction previously used. [33] They showed an easier way to achieve reversible cell aggregation manipulation. Different from click chemistry, host-guest interaction can induce cell-cell connection through a noncovalent bond, which is beneficial to provide reverse assembly due to the low bond energy. In addition to azobenzene, adamantane (Ad) is another accepted guest for cyclodextrin. Rood et al. introduced Ad to cell surface utilizing specific binding between chemokine receptor 4 (CXCR4) and Ac-TZ14011 peptide ( Figure 2C). [34] The polymer containing multiple β-CD molecules was used as a linker to hold multiple Ad molecules together. It is reported that Ad-β-CD has a higher association constant than azo-β-CD, [32] thus achieving a more stable cell-cell adhesion for further cell/tissue engineering applications. Moreover, they also demonstrated that the excessive β-CD between cells can further be functionalized with fluorophores or therapeutic agents for signal label or drug delivery.

Polymeric technology
Several works have linked polymeric technology with the applications of living cell encapsulation, [35] cell surface receptor aggregation, [36] modulation of cell-surface properties, [37] and so on. However, the applications of polymers in direct manipulation of cell-cell aggregation were barely reported. The investigation of polymerization reaction on the living cell surface is still in an early stage because of the difficulty to maintain cell viability under the polymerization reaction conditions due to the use of cytotoxic transition-metal catalysts, organic solvents, and the production of radical species during the reaction. The good news is that a few attempts have been introduced to regulate cell aggregation by presynthesized polymers. The Yusa group [38] has successfully introduced a typical method to utilize the phase-separation property of polymers in lower critical solution temperature (LCST) to realize cell aggregation. Poly(N-isopropylacrylamide), an ideal thermosensitive material extensively applied in biomedical systems, was employed because of its near-body-temperature LCST between 32 • C and 35 • C. The methacryloyl group was first delivered to sialic residues on the cell surface by metabolic glycoengineering and further reacted with terminal thiol-modified poly(Nisopropylacrylamide) (PNIPAM-SH) by thiol-ene reaction under 365 nm UV irradiation, thus anchoring the PNIPAM on the cell membrane. At a temperature lower than LCST (25 • C), the polymers showed hydrophilicity because the hydrophilic functional groups of the polymer chains could form hydrogen bonds with water molecules. Therefore, the PNIPAM might maintain in swelling state and keep the cells leave apart. If the temperature rose to 37 • C, which was higher than its LCST, enhanced hydrophobic interaction could make PNIPAM dehydrate and shrink, thus inducing cellcell aggregation. Di(ethylene glycol) methyl ether methacrylate (DEGMA) is another promising thermosensitive polymer candidate. Pasparakis and coworkers [39] prepared the copoly-mer of DEGMA and N-hydroxysuccinimide methacrylate (NHS-MA), which could be modified on the surface of cells through reacting with amino groups of the membrane proteins ( Figure 3A). Similar as described previously, if the temperature is above LCST, phase transition-triggered cell aggregation could be observed. Besides, Pasparakis and coworkers prepared another copolymer of water-soluble Nvinylpyrrolidone (NVP) and 3-(acrylamido) phenylboronic acid (APBA) as target moiety, which could react with the cis-diol group on sialic acid to form boronate ester bonds. As a "glue molecule," the polymer enabled cells to adhere in just several minutes to form a spherical cell nanostructure. Moreover, the reversibility of boronate-ester bond could make cell spheroids disassemble by adding 0.01 mM glucose. One challenge of the polymer methods is the requirement for high concentrations of polymers, which may cause toxicity to cells and organisms. So, scientists developed a new copolymer that could accelerate the cell aggregation by two distinct mechanisms (diol-boronate ester-induced intercellular crosslinking and polymer-polymer hydrophobic interactions above LCST). [40] The LCST of the copolymer was between 32 • C and 34 • C, which is suitable for cell culture. As a result, only 25 μg/ml new copolymer was needed for cell aggregation control, while the minimum concentrations of the previous two copolymers were 200 μg/ml. Because of the negatively charged nature of the cell membrane, electrostatic adsorption between positively charged polymer and the cell membrane is a supplementary strategy to trigger cell aggregation. A polyethyleneimine (PEI) backbone conjugated with hydrazide groups (PEI-hy) can be an efficient "glue molecule" among cells ( Figure 3B). [41] Aldehyde residue, which is introduced onto cell surface by periodate oxidation, can react with the hydrazide group of PEI-hy and consequently mediate cell-cell connection, while neutral hydrazide cannot trigger such aggregation. Such a result indicates that the positively charged polymer may concentrate the linkers onto the negatively charged cell membrane to improve efficient covalent linkage. However, people have realized that the electrostatic force alone is not strong enough to trigger cell aggregation. Electrostatic adsorption is usually used as an auxiliary force to combine with other force to accelerate cell aggregation and improve the stability of the aggregates. For example, Mo et al. introduced an oleyl-PEG conjugated 16 arms-polypropylenimine hexadecaamine (DAB) dendrimer, which was consist of positively charged dendrimeric linker and hydrophobic cell membrane-binding moieties to rapidly stabilize cell-cell contacts. [42] They demonstrated that it was a quick and simple way to induce cell aggregation with 1 min centrifugation at 40 rcf.
PEG is another famous polymer that has been widely applied in many biological systems. Recently, Hawker's group has introduced a controlled radical polymerization (CAR) strategy to link polymers on cell surface. With PEG as the example molecule, they showed that the functionalized PEG could interact with tannic acid (TA) via hydrogen bond to perform TA-triggered aggregation of PEG-modified yeast cells. [35] More generally, PEG usually works as an important linker between anchoring molecules and functional molecules such as lipid, aptamer, antibody, and bioorthogonal group. [33,43,44] Because the 20-nm-long glycocalyx on the cell surface can cause nonspecific and unavoidable steric hindrance between the cell-surface molecules and the F I G U R E 3 Polymer-induced cell aggregation. (A) Illustration of the macromolecular cell surface modification concept with copolymers of P1 and P2. P1 induces cell aggregation through intercellular diol-boronate ester formation that can be reversed by the addition of diol-rich compounds such as glucose; P2 promotes cell aggregation by covalent anchoring on the cell membrane and subsequent formation of cell aggregates via thermoresponsive coil-to-globule phase transition of the polymer above the LCST. Reproduced with permission: Copyright 2015, Royal Society of Chemistry. [39] (B) Structure of (a) positively charged PEI-hydrazide and (b) neutral hydrazide. Electrostatic adsorption assists cell aggregation. Reproduced with permission: Copyright 2007, Elsevier [41] exogenous modifying molecules, PEG is an ideal candidate to extend the functional groups from the cell surface without affecting their conformations, thus making the cell-cell interaction go smoothly. [45]

Coordination interaction
Ionic coordination interaction is common in the construction of nanomaterials such as hydrogels or collagen nanoassembly; now it has also been reported as a technique for inducing cell aggregation. The Caggiano group [46] successfully realized rapid cell aggregation via coordination interaction between Fe 3+ ion and maltol ( Figure 4). In their strategy, maltol was introduced by maltol-derived hydrazide, which could react with the nonnative aldehydes modifying on the cell surface by periodate mild oxidation of sialic acid residues. Subsequent addition of Fe 3+ might induce nonspecific cell aggregation by gentle agitation in 10 min since one Fe 3+ ion can coordinate with three maltol molecules to form a crosslink. However, a barrier of cell aggregation using such technology is that iron-chelating proteins such as transferrin in the serum can also coordinate with Fe 3+ ion, [47] limiting their application in vivo.

F I G U R E 4
Schematic representation of chelate-mediated cell aggregation. Reproduced with permission: Copyright 2013, Royal Society of Chemistry [46] 3

BIOMOLECULES-INDUCED CELLS ASSEMBLY
In the pure chemical methods, many chemical reagents are used for both modification and interconnection, which may cause negative effects on cell viability and proliferation. Moreover, chemical reactions usually lack the specificity to distinguish different cell lines in multiple cell assembly systems. In this regard, many biomolecules have been exploited as the new generation of "glue molecules." Compared to chemical molecules, biomolecules have superior performance in biological applications due to their biological compatibility and specificity. To date, numerous biomolecules have been introduced for cell engineering applications, such as biotin-streptavidin, antigen-antibody, and DNA molecules.

Biotin-streptavidin system
The biotin-streptavidin system is the most widely applied biomolecule pair for generating stable linkage. Streptavidin and biotin have a firmly specific combination with an extremely low dissociation constant (K d ) about 1.3 × 10 −15 M. [48] Streptavidin is a kind of protein with a total molecular weight (Mw) of 60 kD and consists of four identical subunits. Each subunit can especially interact with one biotin, a kind of vitamin molecule with a Mw of 244. That means, one streptavidin can contact with four biotins simultaneously.
Considering that the biotin can be easily attached to various substances such as nucleic acid, protein, organic molecules, and so on, a stable network constructed from the multiple linkages between streptavidin and biotinylated substrates can be fabricated conveniently. These advantages make the biotin-streptavidin system widely applied in many fields including chemistry, biology, and even engineering. No doubt, such a system also attracted intense attention from people studying cell-cell interaction due to its specific affinity and good biocompatibility. Similar to the functional molecule maltol mentioned above, biotin could be introduced by biotin hydrazide, which can react with aldehyde modifying on cell surface by periodate oxidation of sialic acid residues ( Figure 5A). [49] Subsequently, the cell aggregation can be induced through the addition of streptavidin. To further simplify the operation process, scientists developed another approach for one-step biotinylation of the cell surface through the endogenous amine group, without addition of the aldehyde. In their strategy, N-hydroxyl-succinimide biotin (sulfo-NHS-LC-biotin) was employed to directly link the biotin with the amine group on the cell membrane within 10 min. Using this method, Sakai group [48] successfully achieved to adhere biotinylated cells into multicellular spheroids in a 6-well plate by shaking at 60 rpm for only 3 min ( Figure 5B). The strong binding affinity of the biotin-streptavidin system would not only make cell aggregation in a fast speed, but also make homogeneous aggregation of different types of cells possible. As an improvement of compacted and multilayered tissue-like structures formed by cell-cell connection, 3D tubular structure induced by biotin-streptavidin was also constructed using stress-induced rolling membrane (SIRM) technique ( Figure 5C). [50] The proposed 3D cell aggregate structure can mimic tubular structures in vivo such as blood vessels. This biotin-streptavidin strategy for cell aggregation has also been applied in cellular therapy. For example, the Wang group demonstrated that both of natural killer (NK) cell and Jurkat cell (T-lymphoma cell) could be biotinylated and then connect with each other using the biotin-streptavidin linkage ( Figure 5D). [51] NK cell would induce the apoptosis of Jurkat cell more effectively through proactively shortening distance, thus achieving targeted cell killing with outstanding specificity and efficiency. Moreover, scientists have indicated that such a cell-cell communication can be regulated by light manipulation in a noninvasive and remote-control manner. To achieve this, polythiophene derivative (PTP), a small molecule that can generate reactive oxygen species (ROS) under light irradiation, was selectively internalized into NK cells. It could be observed that light-induced PTP activation increased the mortality of NK cells in the biotin-streptavidininduced cell aggregation system, thus reducing the damage of NK cells to Jurkat cells in a light-dose-dependent manner.

Specific binding of protein-protein or protein-peptide
The biotin-streptavidin system has been well characterized and widely used in the biological study; however, it still has some deficiencies. For example, streptavidin is derived from bacteria and is a potent antigen that may induce immune response in mammalian cells. Moreover, the biotinstreptavidin association is too strong to be dissociated, which limits its application in flexible control of cell assembly.
Some proteins have been proved to dominate cell aggregation naturally, but this in vivo approach is still not universally, specifically, and artificially controlled because the mechanisms remain unclear. Nevertheless, these natural pathways remind people that protein is a class of important biomolecules that should not be ignored in the application of cell-cell connection. To date, some proteins have been used to specifically target at the biomarkers on the cell membrane and thus to purposefully regulate the aggregation of particular cell populations.
Antigen-antibody interaction is commonly used in various biological methods such as western blotting, enzymelinked immunosorbent assay (ELISA), and so on. Scientists have introduced various antibodies (including single-chain antibody fragment, [52] nanobody, [53] etc.) and attached them with prepared nanoparticles to recognize and bind with specific targets on cell surface. The high specificity and affinity of the antigen-antibody union provide the possibility of establishing cell-cell interaction systems through such a method.
Bispecific antibody is a kind of antibody constructed by the fusion of two different antibodies, so that one bispecific antibody molecule can usually recognize and bind with two different targets, making it a potential "glue" to link different cells together. For example, CD19-CD3 bispecific antibody-Blincyto-has been used to engage cytotoxic T cells for redirected lysis of tumor cells. This strategy has been confirmed to be effective in clinic and just be approved by FDA for immunotherapy. [54] So far, many methods have been developed to connect two kinds of antibodies for further inducing cell aggregation, such as chemical crosslinking method, [44] recombinant antibody engineering, [12,55,56] F I G U R E 5 Biotin-streptavidin system used for cell modification and aggregation. (A) Reaction scheme for the biotinylation of cells via periodate oxidation. Incubation of cells with sodium periodate causes the oxidation of the vicinal diol of cell surface sialic acid residues. The resulting aldehyde groups can be used to selectively ligate biotin hydrazide to the cell surface via the formation of a hydrazone bond. Reproduced with permission: Copyright 2003, Wiley. [49] (B) Biotin-streptavidin-induced rapid formation of multicellular heterospheroid. Reproduced with permission: Copyright 2011, Elsevier. [48] (C) Schematic diagram of cell surface modification and stepwise formation of multicellular structures. Reproduced with permission: Copyright 2013, Wiley. [50] (D) Multicellular assembly of immune cells (NK-92MI) and cancer cells (Jurkat) by the chemical bottom-up approach. Reproduced with permission: Copyright 2014, Wiley [51] DNA-mediated antibodies assembly method, [57,58] and further developed multispecific antibodies. [57,59,60] Antibodies can also be directly anchored on the cell surface through lipid insertion, which is more convenient for operation comparing to the conventional bioengineering method. The Wagner group [61] has prepared chemically self-assembled nanorings (CSANs), which contain both single-chain antibody (scFv) and lipid-linking group, to recognize cell-surface receptors and thus direct cell assembly ( Figure 6A). In their design, two dihydrofolate reductases (DHFR) connecting with each other by a linker peptide can rapidly self-assemble into CSAN. This CSAN can bind tightly with lipid-linked methotrexate (bisMTX), a dimer of DHFR inhibitor. Through the recombinantly fusing scFv with DHFR, a scFv-CSAN-bisMTX-lipid construct can be obtained finally and such a structure can insert into the cell membrane. Utilizing this method, they modified the tool cells with the scFv of epithelial cell adhesion molecule (anti-EpCAM), and the modified cells could connect with EpCAM positive cells, such as MCF-7 cells, forming stable cell assembly. As an improvement, the same group [62] reported a method that mixed two kinds of fusing scFv-DHFR subunits with different targeting molecules to form a bispecific CSAN. Furthermore, they also demonstrated that the binding between bisMTX and DHFR can be broken by the addition of excess trimethoprim, a competitive inhibitor of DHFR. This is an advantage that the trimethoprim is an FDAapproved antibiotic that is low-toxicity in vivo and can be widely distributed in the whole organism, making the regulation of cell assembly and disassembly in vivo being probably. Recent studies have shown that clustered distribution of T cell receptor ligands on antigen-presenting cells (APCs) affect T cell activation in immune therapy. [63,64] A shorter distance between antibodies and cell membrane may provide better results, so people tried to find the approach to fine-tune cellcell distance for immune therapy. The Fan group [14] prepared a cholesterol-DNA-biotin linker to anchor peptide-major histocompatibility complex (pMHC) and CD28 antibody on the red blood cell (RBC) membrane as the APC. Taking advantage of the length-controllability of DNA, they found that a shorter distance between the clustered pMHCs and the cell membrane was benefit for more effective T cell activation.
Thanks to the development of genetic engineering technology, people working on cell biology and molecular engineering have powerful tools to express proteins directly on cell surface now. As an example, the Notch protein has been recognized as a target. [65][66][67] The Notch protein is a transmembrane receptor protein containing three parts (extracellular ligand-binding module, intracellular transcriptional module, and central regulatory module). However, the interaction between the exposed wild-type Notch and its ligand will initiate intramembrane proteolysis (i.e., cleavage of the receptor), [68,69] so the Notch protein and its ligand cannot be applied for cell aggregation directly. The Lim group [70] discovered that the intracellular domain and extracellular domain of Notch can be replaced by heterologous F I G U R E 6 Specific binding of protein-protein or protein-peptide in cell assembly/disassembly. (A) Working principle of the introduction of single-chain antibody (scFv) by the interaction between dihydrofolate reductases (DHFR) and DHFR inhibitor methotrexate (bisMTX) as well as subsequent cell-cell assembly and disassembly. Reproduced with permission: Copyright 2014, Wiley. [123] (B) Structure and working principle of synthetic Notch receptors (synNotch) (a) and the assembly of sender and receiver cells via synNotch receptors-ligands interaction (b). Reproduced with permission: Copyright 2016, Elsevier. [70] (C) Design and characterization of a dual cell surface receptor and reporter system. Reproduced with permission: Copyright 2015, American Chemical Society. [73] (D) Schematic illustration of cell-cell interaction triggered by blue light switchable protein pairs. (a) CRY2/CIBN [75] and (b) nMag/pMag, iLID/Nano. [77] Reproduced with permission: Copyright 2019, Wiley, [75] Copyright 2020, American Chemical Society [77] amino acids sequences ( Figure 6B). Using lentivirus vectors to express the designed Notch protein in host cells, they successfully prepared synthetic Notch (synNotch) receptors and observed regulated cell-cell communication and signal transduction. As an example, a receiver cell with a syn-Notch receptor, whose extracellular domain is scFv for CD19 (anti-CD19), can be activated by connecting with CD19expressing sender cell and then trigger intracellular green fluorescent protein (GFP) expressing. Moreover, multiple syn-Notch proteins can be constructed on one receiver cell, so two or more reporter-ligand pairs can be employed to synergistically mediate the cell-cell connection. People showed that with sophisticated design, respective signaling pathways could be modulated orthogonally. Therefore, cascades of cell-cell signaling pathways could be constructed artificially via multiple synNotch proteins engineering. In the recent work of the same group, they have achieved the programme of the multidomain cell-cell aggregation using synNotch-based platform. [17] In their method, the recognition of the sender cell to the receiver cell can induce the change of cadherin, which can subsequently influence the cell-cell aggregation. This synNotch toolkit has now attracted intense attention of research in cell biology and bioengineering owing to its powerful performance.
Similar to proteins, oligopeptides can also play as specific recognition segments. The Luo group [71] conjugated RGD, a peptide that was demonstrated to play as an epitope of extracellular matrix proteins such as fibronectin, [72] with the polyamindoamine (PAMAM) dendrimer, making RGD present a clustered state. Such an RGD cluster can be utilized as the "glue molecule" for regulate cell aggregation. The Yousaf group [73] has presented a method to initiate cell-cell interaction using RGD ( Figure 6C). They first prepared a bilayer membrane structure by combining calceintethered alkyl chain (calcein lipid) with other lipids (POPC and DOTAP). Then the calcein would be introduced into the target cell membrane by liposome fusion. Next, calcein could react with dabcyl hydrazide, showing a state of fluorescence quenching caused by the linkage. Finally, a ligand exchange reaction would be activated when an oxyamine-tethered new ligand was introduced in the system. The oxyamine group could react with the calcein and replace the dabcyl hydrazide group, thus linking the new ligand on the cell surface. Besides RGD, other oligopeptides were also successfully used for cell aggregation. For example, the adhesive peptide IKVAV was conjugated to PAMAM to act as the "glue" for cell aggregation, too. [74] People reported that IKVAV-PAMAM had an enhanced service capability than RGD-PAMAM, probably because the modification of IKVAV provided a dendrimer scaffold with greater hydrophilicity. Moreover, people also demonstrated that this multivalent adhesive conjugate could enhance cell proliferation and expression compared to the cells treated with monovalent ligands. It means the peptidebased "cell glue" is a potential tool for the construction of multicellular structures with high efficiency.
Similar to those nonbiological methods, one challenge maintaining in the application of the proteins or peptides is also to make the aggregation reversible and programmable. Recently, several light switchable protein pairs were developed in cell-cell interaction. The Wegner group successfully established cell-cell connection via selected protein pairs such as CRY2 (cryptochrome 2) and its interaction protein CIBN (N-terminal of Cry-interacting basic helixloop-helix protein 1) ( Figure 6D). [75] It took just a few minutes for CRY2-CIBN to assemble under irradiation of 480 nm light and disassemble in dark. [76] When CRY2 and CIBN were expressed in two groups of MDA-MB-231 cells (MDA), respectively; the CRY2-MDA and the CIBN-MDA could assemble rapidly once they found each other, forming observed cell clusters or multilayered 3D structures. Further, they demonstrated that such a method can trigger cell assembly in a noninvasive (blue light-mediated), repetitive, and reversible way. As an improvement, the Wegner group introduced the application of other light-responsive protein pairs including iLID and its interaction protein Nano (iLID/Nano), nMag and its interaction protein pMag (nMag/pMag), as well as nMagHigh and its interaction protein pMagHigh (nMagHigh/pMagHigh) ( Figure 6D). [77] Considering these protein pairs had a decreasing dark reversion rate in turn, the researchers evaluated the effect of dynamics on the sizes and shapes of the cell clusters that were constructed via the formation of these protein pairs. Using pulsed illumination with different frequencies to change the dynamics, they indicated that compact and round cell aggregates would be obtained under thermodynamic control, while branched and loose cell aggregates would be obtained under kinetic control. Moreover, self-sorting of the cell clusters with different sizes and shapes could be realized. Once they mixed four different cell types together, two separated cell assemblies, such as iLID-/Nano-MDA cells and nMag-/pMag-MDA cells, could be obtained.

DNA
Proteins have provided promising tools in biology and cell engineering. However, there are still some challenges for the accurate manipulation of cell aggregation. For example, the screening and synthesis of antibodies and peptides, particu-larly the bispecific antibodies, is time consuming and high cost. What's more, the prepared exogenous antibodies tend to show strong immunogenicity, thus limiting their applications in living body. For these reasons, more suitable biomaterials for cell assemble are still in urgent need. DNA was previously known as biomolecule that was used to store genetic information. James Watson and Francis Crick first suggested the highly stable and specific doublehelix model of double-stranded DNA more than six decades ago. [78] Nowadays, this classic structure has widely considered to be the basic character of the complicated DNA nanostructures. In the 1980s, Seeman's pioneering work [79] in the area of DNA nanotechnology assigned DNA as biomaterial usage. Since then, DNA has been recognized as a kind of material similar to polymers. Thanks to the development of DNA nanotechnology, highly specific self-recognition and programmability of DNA molecules have been successfully used for creating considerable DNA nanostructures with predesigned structures and DNA nanomachines with customized functionality. There are numerous styles of DNA nanostructures such as tetrahedron, [80] cubic, [81] hairpin-like molecular beacon, [82] nanotube, [83] nanoflower, [84] hydrogel, [85] and DNA-nanoparticle superstructures [86] that have been synthesized successfully. These DNA-based structures have been widely used in different fields, including nanomaterial synthesis, molecular motor, biosensing, bioimaging, drug delivery, and so on. [87][88][89][90][91][92] To date, many studies have been carried out to establish cell-cell connection systems based on the special DNA nanostructures. It is worth to mention that although the exogenous DNA molecules have been noticed to initiate the immunoresponse through cGAS-STING pathway, the DNA strands applied in DNA nanotechnology now are still too short (∼100 bp) to show obvious immunoresponse. [93] When people try to use DNA as the "glue molecule" for cell-cell aggregation, the first challenge is the immobilization of DNA molecules on cell surface. Currently, the commonly used methods include the introduction of DNA aptamers and the hydrophobic modification of DNA strands. Noticeably, the aptamer is a kind of special DNA molecule, which has a certain sequence and fold into specific spatial structure that can specifically recognize a target varying from small molecules, macromolecules to even entire cells. [94][95][96] Once the target is selected, the aptamer with high recognition specificity and binding affinity to the target can be screened based on the technology of Systematic Evolution of Ligands by Exponential Enrichment (SELEX). [97] In addition to the specificity and affinity required for diagnostic [98] and therapeutic applications, [99] aptamers show more attractive properties compared to antibodies. They are usually smaller and more stable, resulting in better tissue penetration ability. The synthesis of DNA molecules is commercial now with a lower cost. Moreover, the nucleic acids can be readily adapted for modifications to meet different needs. Taking these advantages, the Chang group [100] designed a dimerized five-point-star DNA nanostructure to link two tetravalent aptamers TE02 and LD201t1, which can respectively target Jurkat cell and Ramos cell ( Figure 7A). They utilized the designed nanostructure to achieve the interaction between two cells and proved that the rigid and multivalent aptamers could offer more robust binding than the flexible and monovalent one. Such a DNA nanostructure needs to  [45] (C) Schematic illustration of cell-cell attachment through DNA hybridization between complementary ssDNA-PEG-lipids incorporated into the outer cell membranes. Reproduced with permission: Copyright 2013, Elsevier. [102] (D) Stepwise anchoring of fatty acid (FA)-modified ssDNA on cell membranes and subsequent cell assembly via DNA hybridization. Reproduced with permission: Copyright 2014, American Chemical Society [107] search and anchor two different cells simultaneously. Therefore, if the number of targets on the cell surface is not sufficient, the decreased binding efficiency might sharply limit the utility for triggering cell-cell interaction. The Tan group [45] reported another way to induce the cell-cell aggregation with high efficiency ( Figure 7B). They first synthesized a complex of diacyl lipid-PEG-aptamer. Then, the lipid tail of the complex could firmly insert into the cell membrane through its hydrophobicity, so that the aptamer TD05 could be directly anchored on the surface of cytomegalovirus (CMV)-specific CD8 + cytotoxic T lymphocyte (CTL). The aptamer-modified CTL might work as an immune effector cell to target Ramos cell because the TD05 could specifically recognize the immune globulin heavy mu chain on the surface of Ramos. Finally, they demonstrated that such cellcell aggregation might trigger Ramos cell killing, thus providing a redirected cell killing method mediated by aptamers. Similarly, scientists have also proved that the aptamers could be introduced to T cell surface by metabolic glycoengineering and click chemistry. [101] This nonviral aptamer-replacing CD19 chimeric antigen receptor (CAR) T (CAR-T) strategy might be further used in immunotherapy in mouse tumor models.
While aptamers with high specificity have advantageous performance in recognizing target cells for cell assembly, they still have to be screened through a complex and rigorous process. Directly anchoring different single-stranded DNA (ssDNA) oligonucleotides on different cell surfaces might be a more straightforward strategy. As mentioned above, ssDNA can be easily modified with other functional groups and molecules, so inserting lipid into cell membrane via the hydrophobic interaction is one of the most commonly used methods to anchor ssDNA on cell surface. Teramura and coworkers [102] synthesized a complex of ssDNA-PEG-lipid ( Figure 7C). Through the hydrophobic lipid, ssDNA could anchor on the cell surface. When cells modified with complementary DNAs met each other, cell-cell attachment would be built. This method is available for not only homogeneous but also heterogeneous cell-cell interactions. [103] People have proved that the more ssDNA-PEG-lipid are inserted into the cell membranes, the larger interface between two cells will participate to trigger the stronger the cell-cell interaction. For E-cadherin-expressing cells, the cell-in-cell invasion process (the normal cell MCF-10A was internalized into the cancer cell MCF-7) was observed especially when ssDNA was in a high ratio.
In fact, the binding strength via hydrophobicity of lipid is not very robust, because cell membrane fluidity and continual refreshment may cause lipid detachment from the membrane or cellular internalization. [104] Thus, improving the stability of lipid insertion is necessary. People tried to use lipids with different lengths as the anchors, and it has been proven that the C 16 dialkylphosphoglyceridemodified ssDNA has a more stable anchor ability on cell membrane than C 18 dialkylphosphoglyceride-modified ssDNA. [105] Another common feature is that the dialkylmodified ssDNA has a more stable anchor ability than monoalkyl-modified ssDNA. In recent years, the Gartner group [106] have made a great progress in improving anchoring stability of monoalkyl fatty acid-modified ssDNA (Figure 7D). In their strategy, two complementary ssDNA linked with fatty acid were added step by step. The first strand would anchor on the lipid bilayer but remained in relatively rapid equilibrium with the medium as well as the second coanchor strand. Because of the membrane fluidity, the two anchored ssDNA strands could meet and hybridize with each other, resulting in overall improvement of the hydrophobicity and the binding strength. In their design, the coanchor strand was shorter than the first anchor strand, thus the longer anchor strand could further hybridize with the other complementary anchor strand anchored on the other cell to construct cellcell aggregation. With such a DNA-programmed assembly of cells (DPAC) method, a bottom-up cell assembly strategy was developed to build a 3D tissue by the same group. [107] In their design, the ssDNA was patterned onto a glass slide through covalent linkage. Cells modified with complementary ssDNA were introduced through a flow cell over the modified glass slide with programmable cell concentration and flow rate. Through DNA hybridization, the DNA-modified cells would then be anchored on the substrate surface. After repeating such flow cell incubation process, 3D microtissue structure was formed through the aggregation of cells layer by layer via DNA linkage. Further, by merging the DPAC method with microfluidic and 3D printing technologies, they demonstrated that this strategy performed very well in the synthesis of cell aggregation structures.
A challenge of ssDNA-based cell aggregation is building cell clusters with asymmetric or directional arrangements, which is limited by the diffusivity of ssDNA on fluidic cell membranes. To improve the anchoring stability on the cell membrane, DNA frameworks have been exploited as anchoring modules for cell modification. Taking advantage of their rigid construction and larger size, 3D DNA nanostructures exhibit much slower endocytic kinetics once anchoring on the cell surface compared to ssDNA. Thus, they can stay on the cell membrane for a longer time for further reactions. Tan group [108] has presented a strategy using DNA tetrahedron structure as the "glue" of cell aggregation ( Figure 8A). One of the four vertices of the DNA tetrahedron was applied to form the connection between cells, the remaining three vertices were all used to immobilize on the cell surface through cholesterol modification. It could be observed that DNA tetrahedron with three cholesterols shows a significantly improved stability compared to DNA tetrahedron with only one cholesterol. The DNA tetrahedron with three cholesterols also perform better in membrane anchoring than ssDNA with three cholesterols modification, demonstrating that tetrahedral scaffold was an important factor for enhancing the membrane-anchoring stability. Through the cell-cell connection mediated by DNA tetrahedron, the signal response between Raji B cell and A549 cell under lipopolysaccharide (LPS) stimulation could be detected. Recently, they designed an improved DNA tetrahedron structure, which was first blocked by a special aptamer sequence so that the cell-cell interaction could not be initiated. In the presence of target molecules such as ATP, the block sequence would be removed by allosteric modulation. Therefore, the exposed DNA tetrahedron arms could induce the cell-cell interaction. [109] Our group introduced another improved DNA tetrahedron linker for programmed cell-cell aggregation. [110] Metastable DNA hairpins were linked to the vertices of DNA tetrahedron, and the cell-cell aggregation could be triggered once a trigger strand was added into the system to initiate hybridized chain reaction (HCR), which could link the DNA tetrahedrons on cell surface together. The results demonstrated that increasing the number of the modified hairpins would induce larger cell assembly.
Although anchoring DNA on cell surface through hydrophobic insertion is easy to operate, a covalent connection between the DNA oligonucleotides and the cell membrane is always desired. On the one hand, the covalent chemical bonds are more robust. On the other hand, the covalent method may provide the more specific link between DNA and cells. To date, DNA modification by various functional groups has been commercialized, so that the metabolic glycoengineering and click chemistry methods can be applied to introduce ssDNA on cell membrane with a simple reaction process. The Bertozzi group [111] has proved that the control of cell-cell contact could be highly dependent on DNA sequence complexity, density, and total cell concentration, so a programmable cell assembly can produce desirable microtissue step by step. A similar mechanism has also been proved by the Song group, who used the DNA-mediated cell-cell connection to promote stemness maintenance and expansion of stem cells for the first time. [112] Generally, disassembly of DNA-dependent cell aggregates can be accomplished by adding DNase or increasing temperature over the melting temperature of the DNA duplex. [111] With the development of DNA technology, more and more special characters of DNA have also been applied to control the aggregation of cells. Hou et al. reported a cytosinerich DNA triplex platform that could regulate the assembly of cells by controlling pH. [113] In recent years, Lu group has engineered DNAzymes on the cell surface ( Figure 8B). [114] DNAzymes are special DNA sequences that can perform catalytic function toward DNA/RNA cleavage using specific metal ions as cofactors. So, they built a platform that could regulate dynamic cell behaviors using different metal ions.
The powerful amplification capability of DNA also makes it promising material for cell aggregation applications. Polymerase chain reaction (PCR) is the most famous DNA amplification method, but it needs extreme temperature conditions. [115] Due to the development of molecular biotechnology, DNA molecules can be amplified in cellular conditions and in situ now. By introducing HCR amplification technique, Wang group [116] achieved the branch-like amplification of DNA on the cell surface ( Figure 8C). In their strategy, DNA initiator was first introduced by metabolic glycoengineering and click chemistry. Using the special design, the branched DNA structure produced by HCR contained many extended unhybridized ssDNA parts, whose sequences were complimentary to those of the ssDNA parts on other cells. This polyvalent DNA hybridization pathway would trigger a more effective and robust cell-cell connection which could even be observed over 20 days during the culture. What's more, one important point in this method is that only the  [108] (B) DNAzyme controlled cell-cell interaction and two-factor disassembly control of cell assemblies by DNAzymes. Reproduced with permission: Copyright 2021, American Chemical Society. [114] (C) Schematic illustration of in situ formation of polyvalent DNA polymers on the cell surface. Reproduced with permission: Copyright 2018, Wiley. [116] (D) Scheme of DNA origami nanostructure-based organization of cell origami clusters. Reproduced with permission: Copyright 2020, American Chemical Society [119] initiator DNA strand can be settled on the cell surface. The reduced occupation of the cell membrane will minimize the effects of cell surface modifications on cells. In most recent years, the same group has developed their strategy to construct a branched structure containing multiple aptamers on the cell surface. Such a structure could perform as a polyvalent antibody mimic (PAM) to initiate the aggregation of different cells. [117] The improvement of DNA nanotechnology has provided more and more promising DNA nanostructures as tools for manipulating cell assembly. For example, the structures constructed by DNA origami technology have been shown to be powerful tools to engineer cell surface functions. [118] Fan group [119] introduced a DNA origami nanostructure (DON) to organize the aggregation between homo-/hetero-type cells ( Figure 8D). The structure, which was termed as Janus DON, contained four tube-like subunits and was modified with numerous ssDNA strands on the surface. The target cells were modified with the complementary ssDNA, which were premodified with thiol groups, through the crosslink between the thiol and the amino groups on the cell membrane. By constructing different types of DONs, people demonstrated that different types of intercellular communications could be manipulated by DONs. Lately, new designs such as DNA hairpin motif have also been introduced to manipulate cell-cell interactions. [120]

CONCLUSION AND OUTLOOK
The development of programmable cell-cell aggregation provides not only a useful method for investigating the mecha-nism of cell interaction in the cell biology field but also a promising tool for building artificial tissues in the bioengineering field. During the past decades, numerous chemical or biological pathways have been presented to achieve the goal of precisely controlling cell-cell aggregation, the characteristics of which were summarized in Table 1. Immobilizing the "glue molecules" on the cell membrane is the first vital process. Although the anchoring via hydrophobic tails insertion is a convenient way, the stable immobilization through covalent bonds may always provide a more robust system. Bioorthogonal chemistry strategies make the linkage between the "glue molecules" and cell surfaces being efficient with the help of bioengineering techniques. Moreover, the bioorthogonal strategy can also help people to overcome the other important barrier in cell-cell aggregation programming, that is, how to manage the adhesion between the "glue molecules." Besides these pure chemical methods, biomolecules have been also applied to build stable linkage between cells to physiologically mimic what happened in vivo.
In this review, we summarize the recent advances in the molecule engineering of cell assembly and mainly discuss the mechanism of chemical and biological methods. As we know, the study of human tissues and diseases is often hindered by difficulty with the construction of in vivo models which maintain differences to the in vitro ones. [121,122] Studies in cell assembly can effectively deal with this problem via building materials for in vivo tissue repair or for constructing realistic in vitro tissue models. Beyond this, the selective assembly of multifarious cell types, such as tumor cells and immune cells, can make vast differences in immunotherapy. The tumor cells can be directionally killed by immune cells via accurately Easy (DNA is readily adapted for various modifications.) The interaction between the DNA aptamers and the target proteins may influnce their regular physiological functions. [45,[100][101][102][103][105][106][107][108][109][110][111][112][113][114][116][117][118][119][120] controlling cell-cell interaction. What's more, molecular engineering potentially improves our understanding of the signaling mechanisms of cellular behavior and the regulation criteria for programing cells toward desired outcomes. There still remains multiple challenges to be addressed to improve the controllable assembly between multiple populations of cells. (i) It is necessary to systematically study and reduce any negative impacts of cell-surface-loaded materials and underlying processes on cellular function. (ii) Some of the reported cell-cell aggregation processes have to maintain at a relatively low temperature to slow down the fluidity of cell membrane to avoid the endocytosis and loss. So, more stable surface immobilization methods are still necessary.
(iii) Nongenetic cell engineering technologies are still in their infancy and none have been translated into the clinic to date, and studying the advanced technology in vivo and in clinic is significant. Obviously, these demands probably cannot be fixed by only the chemists or biologists themselves. The cooperation between experts in different fields, such as bioengineering, molecular biology, cell biology, chemical synthesis, nanotechnology, and so on, becomes more and more important to achieve a big goal to improve the programmed cell-cell aggregation as a practical tool. In the future, precise engineering of cell surface will make it possible to manipulate specific and controllable cell aggregation between different cell types as expected with satisfy efficiency and applicability, which has immense potential in tissue engineering, organ reconstruction, cell-based diagnostics, and therapeutics.

A C K N O W L E D G M E N T S
This work was supported by the National Natural Science Foundation of China (No. 22074068, 591859123 and 21874075) and "the Fundamental Research Funds for the Central Universities," Nankai University (63211050).

C O N F L I C T O F I N T E R E S T
There are no conflicts to declare.

D ATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available from the corresponding author(s) upon reasonable request.