Hybrid BioMicromotors

One of the currently most exciting engineering challenges is to create miniaturized functional vehicles that can carry out complex tasks at a small scale that would be otherwise impractical, inefficient, or outright impossible by conventional means. These vehicles are termed nano- and micromotors or -robots and should be distinguished from even smaller molecular machines for energy, computing, or other applications on the one side and static microelectromechanical systems (MEMS) on the other side of this size scale. Rather than being electronic devices on a chip, micromotors are able to move freely through a liquid medium while being steered or directed externally or by intrinsic design, which can be achieved by various mechanisms, most importantly catalytic reactions, magnetic fields, or ultrasonic waves.^1–3 There is a variety of sensing, actuating, or pickup-and-delivery applications that scientists are currently aiming for, with local drug targeting for cancer treatment being one of the most prominent examples.^4,5 For such an application, a micromotor needs to be able to move, i.e., to swim, freely in three dimensions efficiently with a reliable mechanism to be controlled and directed. The physics of swimming is hereby dominated by viscous drag forces, which is a direct consequence of the small size scale that leads to a low Reynolds number, a problem that was discussed extensively by physicists in the field and will be shortly covered in this review as well.^2,6,7 This kind of swimming has challenged engineers as it is not commonly experienced in everyday life, but can nonetheless be observed in nature for motile microorganisms like sperm or certain bacteria. Naturally, these microorganisms served as inspiration from the very beginning to create artificial micromotors, as they were able to tackle the challenges that an active, self-sufficient microswimmer vehicle has to face.⁸ With biomimetic approaches, researchers were able to imitate the flagella-based motion strategy of sperm and Escherichia coli bacteria by reproducing their respective flagellum shape and actuating it with magnetic fields.^9,10 Unfortunately, the convenient biochemical on-board power supply of these microorganisms could not be reproduced at this size scale to date. As mentioned earlier, external power sources like magnetic fields, ultrasonic waves, light, or chemical fuels in the liquid medium have been employed to solve this problem.^1,2,11 A fundamentally different solution was proposed to deal with the problem of motor miniaturization, and further extended toward other functionalities of micromotors ever since: The so-called biohybrid approach directly employs living microorganisms to be a main component or modified base of a functional microswimmer.^12,13 Often, the microorganisms thereby served as motor units for artificial devices, but this role was extended and modified in recent years toward other functionalities that take advantage of the biological capabilities of these organisms considering their means of interacting with other cells and living matter, specifically for applications inside the human body like drug delivery or fertilization.^14,15 It is a distinctive advantage of microorganisms that they naturally integrate motility and various biological functions in a conveniently miniaturized package, coupled with autonomous sensing and decision-making capabilities. They are able to adapt and thrive in complex in vivo environments and capable of self-repair and self-assembly upon interaction with their surroundings. In that sense, self-sufficient microorganisms naturally function very similar to what we envision for artificially created microrobots: They harvest chemical energy from their surroundings to power molecular motor proteins that serve as actuators, they employ ion channels and microtubular networks to act as intracellular wiring, they rely on RNA or DNA as memory for control algorithms, and they feature an array of various membrane proteins to sense and evaluate their surroundings. All these abilities act together to allow microbes to thrive and pursue their goal and function. In principle, these abilities also qualify them as biological microrobots for novel operations like theranostics, the combination of diagnosis and therapy, if we are able to impose such functions artificially, for example, by functionalization with therapeutics. Furthermore, artificial extensions may be used as handles for external control and supervision mechanisms or to enhance the microbe's performance to guide and tailor its functions for specific applications. In fact, the biohybrid approach can be conceived in a dualistic way, with respect to the three basic ingredients of an in vivo microrobot, which are motility, control, and functionality. Figure 1 illustrates how these three ingredients can be either realized biologically, i.e., by the microorganism, or artificially, i.e., by the synthetic component. For example, a hybrid biomicromotor based on a sperm cell can be driven by the flagellum of the sperm or by an attached artificial helical flagellum.^14,15 It can orient itself autonomously via biological interactions with its surroundings and other cells, or be controlled and supervised externally via artificial sensors and actuators. Finally, it can carry out a biological function, like its inherent ability to fertilize an egg cell, or an artificially imposed function, like the delivery of synthetic drugs or DNA vectors. A biohybrid device may deploy any feasible combination of such biological and artificial components in order to carry out a specific application. This review gives an overview of such hybrid biomicromotors and evaluates the strengths and weaknesses of current systems, as well as the future potential of this still novel approach.

The fundamental differences of motion at the micro- compared to the macroscale have been explained and made public to generations of scientists by Edward Mills Purcell's famous lecture “Life at low Reynolds number” in 1977.¹⁶ Ever since, physicists have discussed and worked on the so-called “Scallop Theorem” which describes the necessity of non-reciprocal motion due to the dominance of viscous over inertial forces at the microscale in order to move forward.^2,6–8,16 We will not explain these considerations in detail here, but rather focus on the different strategies to realize microscale motion of synthetic devices and how they were conceived.

To answer the question of how things are moved efficiently at the microscale, one should first take a look at how nature is doing it. Submicron-sized particles, ions, and small molecules are transported over short distances mainly by diffusion.^17–19 This can be directed by specified channels within cell membranes for intercellular transport, but in general is not targeted at a specific point. A more targeted transport on that small size scale is achieved by cargo-loaded vesicles riding on intracellular tracks, so-called microtubules.^20–22 The vesicles are thereby actively pushed and pulled by motor proteins, for example, kinesin, along the microtubules.^23,24 This type of motion is similar to the myosin-based force generation in muscle cells and is a basic example of nature's molecular machines that can generate kinetic energy from chemical reactions.^25,26 These motor units work nicely as small transporters on predefined tracks, but they need to be employed within a higher order geometry in order to power freely moving, self-sufficient, microscale vehicles. Such geometries in nature are, for example, pseudopodia, cilia, and flagella that move cellular organisms like macrophages, bacteria, protozoa, and sperm cells by different mechanisms. While macrophages use pseudopodia to get a hold on their surroundings and migrate via amoeboid motion, ciliated or flagellated microorganisms can swim freely in liquid medium.^27,28 A non-reciprocal beating pattern is thereby necessary that can be achieved in different ways, for example, metachronal waves of multiple cilia, or helical waves of flagella bundles, or the overhanging stroke of a sperm flagellum.^29–33 

These solutions inspired scientists to design artificial flagella to imitate nature's microswimmers and obtain free swimming microrobots.^9,34–36 Unfortunately, despite considerable advances in the field of molecular machines which were recently awarded with the Nobel Prize in Chemistry,³⁷ the molecular motor units that drive the biological flagella could not be reproduced similarly yet, which made it necessary to think of alternative ways to provide the necessary energy for the synthetic motors. Our capability to miniaturize our machinery has not yet reached the complexity of nature's molecular devices which is why we have to rely on external power sources to drive biomimetic microrobots. As these microrobots should be untethered to freely operate in complex environments, permeating energy fields like light, ultrasonic waves, and magnetic fields have been the sources of choice so far.^1,4,11 In contrast to the idea of alternatively powered biomimetic swimmer geometries, the approach to use chemical substances in the surrounding medium as an energy source has led to different kinds of microswimmers that are based on motion principles that are not directly inspired by nature in terms of the swimmer shape, but in terms of chemical to mechanical energy conversion. Such micromotors are, for example, electrophoretic Janus swimmers³⁸ as well as bubble-propelled micromotors³⁹ that both rely on the catalytic decomposition of hydrogen peroxide as a chemical fuel.^1,40 However, although these motors are self-sufficient swimmers in their respective medium, they cannot be considered completely free as the fuel component must be added first to the medium (and should not run out), which is especially problematic considering applications inside living bodies. An even bigger problem in that case is the biocompatibility of the fuel which is, in the case of hydrogen peroxide, not given. Consequently, people are now searching for alternative fuels that are biocompatible or ideally already present at the intended application site, as well as suitable catalysts for efficient reactions to generate kinetic energy.^41–44 This however leads back again to nature, namely, enzymatic reactions that convert energy from in vivo biochemicals like, for example, from glucose^45,46 or urea,^47–49 which brings about the central topic of this review: Instead of imitating nature, biological entities might as well just be incorporated into or enhanced by synthetic devices in a biohybrid approach. This was at first done to make use of the convenient on-board power supply of motile microorganisms, but the concept was further extended in recent years to also or exclusively benefit from other advantages and functionalities of living organisms. These will also be covered in the main part of the review. In both cases, we will focus on biohybrid devices that are able to swim freely in three dimensions and are either autonomous or supervised and controlled externally by non-invasive methods to operate in a robotic manner. This is especially important for biomedical in vivo applications, which will be discussed in detail, as against microrobotics for MEMS technology,^50–53 microassembly,^54–56 or environmental applications,^57–59 which are not subject of this review.

Unicellular microorganisms use cilia or flagella to propel in fluids at the microscale. These elastic paddles exhibit a specific beating pattern to overcome the scallop theorem.¹⁶ Although there are thousands of such individual swimming microorganisms, the biomolecular composition and propulsion mechanism of their motor units are surprisingly similar. There is however a sharp distinction between bacterial and eukaryotic flagella (see Fig. 2). The first of these two big groups, bacteria flagella, are connected to the main body of the respective bacterium by a literally rotating molecular joint.^32,60–62 Upon rotation, the flagella form a helical shape due to their elasticity which is subject to the anisotropic drag of the fluidic surrounding and leads to a corkscrew-like motion of the bacterium.^16,62–64 Usually, a bacterium has multiple flagella which are configured either lophotrichous, i.e., on one side of the bacterium's body, amphitrichous, i.e., on opposite sides of the body, or peritrichous, i.e., all over the bacterium's surface.⁶⁵ These flagella may rotate individually or form bundles of several flagella that intertwine and rotate together.^62,65 If a bacterium has only one flagellum, the configuration is called monotrichous. Most bacteria swim in the so-called run-and-tumble pattern, which describes unidirectional motion periods (i.e., run) that are followed by a stationary state (i.e., tumble) where the bacterium stops to reorient.⁶³ For well-known peritrichous bacteria like Escherichia coli (E. coli), Serratia marcescens (S. marcescens), and Salmonella Typhimurium (S. typhimurium), it was observed that the propelling flagella bundles untangle during this reorientation to allow single rotating flagella to alter the direction of the bacteria's movement.^62,63 These three are among the most frequently employed species for bacteria-powered hybrid microswimmers because of their well-known behavior, easy culture conditions, relatively high swimming speeds of roughly 40 μm/s (20 bodylengths per second), chemotaxis capabilities, and genetic alteration possibilities. Other employed bacteria are listed in Table I. A different class of bacteria that proved to be very interesting for hybrid biomicromotors are magnetotactic bacteria (MTB), also listed in Table I. They feature different flagella configurations, e.g., Magnetospirillum gryphiswaldense (M. gryphiswaldense) with two individual amphitrichous flagella, and especially Magnetococcus marinus strains (e.g., MC-1 and MO-1) reach very high speeds of roughly 200 μm/s (100 bodylengths per second), but their most important advantage is the magnetosomes inside their bodies.^66,67 These are naturally synthesized magnetite (Fe₃O₄) nanocrystals that are arranged as a superparamagnetic chain within the bacterium that naturally aligns its magnetic dipole moments to the earth's magnetic field.^68,69 This serves as a means for orientation for the bacterium but can be exploited with stronger artificial external magnetic fields to forcefully align and control the magnetotactic bacterium and its motion.

The second big group of flagellated microorganisms are eukaryotic swimmers. Eukaryotic flagella are different from their bacterial counterparts as they are not rotating, but beating in a whiplash fashion.^16,70,71 Although the power stroke of the molecular motor unit is actually reciprocal, the elasticity of the flagellum leads to an overhanging, damped wave-like beating pattern that allows non-reciprocal motion in low Reynolds number fluids.^16,72 Most prominent examples are mammalian sperm cells that are monoflagellated and have various sizes and swimming speeds depending on the species (see Table I).^73–75 Another eukaryote that was employed for hybrid microswimmers is Chlamydomonas Reinhardtii (Chlamydomonas), which is a microalga with two lophotrichous flagella.⁷⁶ Trypanosomes are eukaryotic parasites that have their flagellum tightly bound to the cell membrane over almost all of its length, which makes the whole body an undulating swimmer that moves with the tail tip in front.^77–79 Their distinctive ability to adapt their surface to make them invisible to the immune system in a mammalian host's blood makes them interesting candidates for in vivo hybrid biomicromotors, but the highly dangerous pathogenicity of common trypanosome species prevented the exploration of this possibility so far.⁸⁰ Microorganisms that employ cilia instead of flagella were also barely used in hybrid approaches. Cilia are structurally similar to flagella, but generally shorter, and are usually distributed all over the surface of the respective microbe, which is called holotrichous arrangement.^71,76,81,82 The non-reciprocal wave pattern that is necessary for propulsion thereby does not arise from the beating pattern of individual cilia, but from their concerted beating as a metachronal wave, i.e., like a Mexican wave performed by a sports stadium crowd.^83–86 One example that was employed in a hybrid approach is Tetrahymena pyriformis (Tetrahymena), a eukaryotic protozoon that is bigger than most other employed microorganisms, but also significantly faster, with a swimming speed of roughly 500 μm/s (10 bodylengths per second).⁸⁷ 

There are other unicellular organisms that move without any cilia or flagella. In principle, most cells can migrate by pushing and pulling their bodies by membrane protrusions formed by actin-filaments, so-called pseudopodia.^88,89 This crawling motion is called amoeboid movement.^90,91 It is not suitable to propel free-swimming micromotors because it is very slow (<1 μm/s) and needs surfaces to crawl along. However, some suspension cells that feature this motion mechanism are nonetheless interesting for hybrid biomicromotor applications. These are mainly blood cells like macrophages and T cells that can be employed to infiltrate solid tumors. This will be discussed further below. Also, red blood cells (RBCs) were employed as hybrid drug carriers.⁹² They were propelled artificially, by ultrasound, as they naturally only move passively in the flow of blood. Such hybrid biomicromotors that rely on artificial instead of cellular propulsion are not as common as cell-propelled swimmers and will be discussed individually further below as well.

A final type of cellular motion is the compression of muscle cells (myocytes). Of course, muscle cells do not swim in nature, as they are stationary, but can be employed in a biohybrid approach to move synthetic devices and make them swim. Heart muscle cells of rats (cardiomyocytes) were employed to actuate microtools, but also micromotors like artificial flagella.^93,94 For larger devices, skeletal muscle cells and insect dorsal vessel tissue were used, but these are not the focus of this review.^95,96

Apart from propulsion, another important aspect of microswimming is taxis. Taxis is the ability to respond to an external stimulus with a change in motion direction.¹⁰¹ For motile microorganisms, this is necessary to find food and purpose.^102,103 The most common form of taxis is chemotaxis, i.e., the response to a chemical cue that is present in the surrounding medium.^104,105 Practically, all motile microorganisms are capable of some form of chemotactic response, and consequently, many cell-driven hybrid biomicromotors were designed to rely on these capabilities as well, as will be discussed below. Aerotaxis^106,107 and pH-taxis^108,109 are distinctive forms of chemotaxis that describe the response to oxygen concentration or pH of the liquid medium, respectively. Physical instead of chemical stimuli may also induce tactic behavior in certain microbes, for example, light (phototaxis),^110–112 temperature (thermotaxis),^113,114 electric fields (galvanotaxis),¹¹⁵ or magnetic fields (magnetotaxis).^116–119 Response to mechanical stimuli like liquid flow or to the contact with solid objects can be described as rheotaxis^120–122 and thigmotaxis,^123,124 respectively. The response in all of these cases can be positive or negative, i.e., the microswimmer may move towards or away from the respective stimulus. An important point here is the difference between klinotactic response and tropism. For example, considering chemotaxis, most microbes have only one sensory capability, i.e., they can sense whether the concentration of a certain chemical increased or decreased at their current location, compared to their previous one, and then decide to alter their movement direction or not.^125,126 They can change their direction klinotactically, but they cannot decide on a particular new direction based on their sensing. By contrast, tropism requires more complex sensory organs to go directly up a perceived gradient.¹²⁷ For example, our eyes allow us to walk directly toward a light source that we see, i.e., phototropically. A phototactic bacterium however would not see the light source, it would have to stop to sense the light intensity in its immediate surroundings, and then, it would move again and stop to sense again until it reaches the point with the highest relative intensity after several directional changes. With the biohybrid approach, we can impose more effective means of taxis artificially, for example, magnetotactic behavior via synthetic magnetic attachments, and we can impose tropism by altering the magnetic field in response to the visual feedback that we receive and judge by ourselves. The microorganisms however still offer the advantage of combining multiple forms of taxis simultaneously, as was pointed out and employed in several hybrid biomicromotor approaches that will be discussed in the main section of this review.^101,128

In the earliest free-swimming cell-driven hybrid biomicromotors, the artificial component was an inert particle, e.g., a spherical polystyrene (PS) microbead, to model synthetic cargo, for example, imitating encapsulated drugs for targeted tumor therapy, or polymeric parts for microassembly in general. Later, liposomes or Fe₃O₄ nanoparticles (NPs) were loaded to exert specific functions like actual drug encapsulation or magnetic controllability, as will be discussed below. These were fabricated by already established chemical protocols. Relatively novel fabrication techniques were employed to create synthetic components specifically tailored to fit to the respective microorganism, for example, polymer-metal composite microtubes and microhelices to be mechanically attached to sperm cells, as depicted in Fig. 1. These were fabricated by strain-engineered rolled-up nanotechnology^149,150 and 3D laser lithography,^151,152 respectively.

Following the schematic depicted in Fig. 1, hybrid biomicromotors can be distinguished by nature of motion, control, and functionality. These three features offer two possibilities each, i.e., either biological or artificial, and can be combined arbitrarily, which means that if every combination would be one category of hybrid biomicromotor, this would lead to a multitude of categories that would be clearly defined, but not overseeable. Considering the current state-of-the art of such biohybrid microdevices, this is not necessary. The vast majority of hybrid biomicromotors so far relied on cellular propulsion and showed transportation of synthetic microcargo with the intent of targeted drug delivery as potential application, i.e., an artificially imposed functionality. Consequently, in Sec. III A, we will focus on such hybrid biomicromotors that employ microorganisms mainly as propulsion units, i.e., for biological motility, and discuss how both biological and artificial control mechanisms were developed to enhance their functionality with respect to synthetic cargo, for example, by taking advantage of microbial or artificially imposed taxis. In the second part, hybrid biomicromotors that employ microorganisms not for motility, but for other biological functionalities, mainly based on particular cellular interactions, will be discussed. These advanced cellular functions may in some cases also be present in hybrid biomicromotors that rely on motile cells; however, we will discuss them in detail in the second part to stress their importance, illustrating the development of microbial biorobots from mere microtransporters to motile and controllable biological sensors and actuators for in vivo applications. This development is schematically depicted in Fig. 3, where different cellular functions are shown that will be discussed in detail in Sec. III B.

The incorporation of microorganisms to serve as motor units for miniaturized synthetic vehicles represents the most common and immediately plausible application of the biohybrid approach. As mentioned earlier, many of the biohybrid micromotors that use the microbial component for other purposes also take advantage of the convenient on-board power supply of such microorganisms. In either case, different bacterial and eukaryotic cells are employed in different ways depending on the design and purpose of the specific microswimmer. Formally, we divide such cell-driven swimmers in three types that simply describe the size relationship between the cellular carrier and synthetic cargo: type I, where the synthetic part is significantly larger than the employed microorganisms; type II, where synthetic and biological parts are about the same size; and type III, where the synthetic components are orders of magnitudes smaller than the microorganism (see Fig. 4). We choose this simple classification because it describes three different ways of how cargo can be transported in general, i.e., collectively, individually, or in small compartments. At this stage, it is difficult to judge which of these strategies is the most efficient because in most studies model cargo particles were used or studies on cargo release and efficiency were not compared to other delivery systems. Given the fact that for most microcarrier applications quite a large amount of individual swimmers will be necessary to achieve efficient and reliable cargo deliveries, it might ultimately be irrelevant if it is a type I, II, or III delivery system once a stable performance is achieved. However, in the development of these three swimmer types, different challenges need to be tackled, for example, considering different bonding and release mechanisms between biological and synthetic components, and these will be discussed in detail in the following parts.

We want to focus specifically on hybrid biomicromotors that can move freely in a 3D environment and perform tasks self-sufficiently, or under untethered external control in a robotic manner, especially considering biomedical applications inside the human body. The first works to move synthetic matter with the help of whole living microorganisms were published by Berg et al. in 2004 and Uyeda et al. in 2005. The former attached multiple flagellated Serratia marcescens bacteria (S. marcescens) onto the surfaces of polystyrene (PS) beads and pieces of polydimethylsiloxane (PDMS) to obtain the so-called auto-mobile beads and chips for microfluidic flow manipulation,¹⁵⁶ whereas the latter guided gliding, Mycoplasma mobile bacteria (M. mobile) along the tracks in a microchip environment to push streptavidin-coated PS microbeads and thus acting as microtransporters.¹⁵⁷ Following our categorization, the auto-mobile PS bead by Bert et al. 2004 was the first type I cell-driven hybrid biomicromotor, whereas Behkam and Sitti were the first to achieve a certain level of control over such a motor's movement and proposed its use as a robotic microswimmer in 2006.¹⁵³ In parallel, the first type II hybrid biomicromotors, following that of Uyeda et al., but with free-swimming cells, were developed by Whitesides et al. in 2005 and Martel et al. in 2006, the former combining a PS bead with a Chlamydomonas reinhardtii alga (Chlamydomonas),¹⁴³ and the latter combining melamine resin beads with different magnetotactic bacteria (MTB), namely, Magnetospirillum gryphiswaldense (M. gryphiswaldense) and Magnetococcus marinus (strain MC-1).¹⁴⁰ Martel et al. were also among the first to propose type III hybrid biomicromotors based on MTB combined with nanoparticles,¹⁵⁸ although this concept took a little bit longer to be realized¹⁵⁵ because of the more complicated mechanism to achieve a reliable and quantifiable bonding between the bacterium and synthetic particles. A special case of type III is given by the possibility to integrate small substances directly into the motile microorganism. Such a modified microorganism can be characterized as a hybrid biomicromotor as well, although the artificially placed substances might be of biological origin. In a way, this may also apply to genetically modified bacteria that have been proposed as delivery vehicles already by Schaffner in 1980.¹⁵⁹ This concept is called bactofection and will be discussed separately further below.

Type I cell-driven hybrid biomicromotors are characterized by a large piece of synthetic cargo (Ø ≈ 10 μm) motorized by several small flagellated bacteria (Ø ≈ 1 μm). The auto-mobile bead in Berg et al. (2004) was the prototype of such a microswimmer, featuring a PS bead of 10 μm diameter and several tens of S. marcescens bacteria randomly attached to the hydrophobic surfaces.¹⁵⁶ This first swimmer reached a velocity of ca. 5 μm/s, which was only 10% of one single free bacterium. This poor performance was a result of the random distribution of bacteria on the microsphere and no control on the directionality of the motion. Bekham and Sitti reduced the number of attached bacteria to around 10 to obtain more favorable bacteria distributions on the bead and consequently observed an increase in velocity to roughly 15 μm/s. More importantly, they presented a means to control the micromotor motion. They showed that the introduction of copper ions (Cu²⁺) to the liquid medium will stop the micromotors, while the addition of ethylenediaminetetraacetic acid (EDTA) will restore their motion.¹⁵³ This kind of on/off-control was also demonstrated by Kim et al. for S. marcescens-driven systems, but with UV light as the switch actuator instead of chemicals.¹⁶⁰ Systems by Kim et al. were large triangular and rectangular sheets of a SU-8 photoresist with lengths up to 400 μm, as well as smaller Silicon (Si) shapes around 5 μm diameter.^160,161 They also confirmed that smaller structures are faster (up to 15 μm/s) because of the more favorable distribution of attached bacteria, but probably also because of the different size and shape of the synthetic cargo. They chose these materials because they envisioned micromanipulation and assembly applications of these micromotors in the MEMS field. Behkam and Sitti however were the first to propose the biomedical applications of these kinds of micromotors in disease diagnosis and targeted drug delivery,¹⁵³ which are the focus of this review. For this, the performance of the motors must be greatly improved, which was attempted by patterning the attachment of bacteria onto the cargo particle to obtain a directed, concerted propulsion. Behkam and Sitti achieved this by lateral plasma treatment to promote hydrophilic bacteria adhesion on roughly half of the PS bead and thus obtain bacteria-propelled Janus particles with clear directional bias and enhanced motility.¹⁶² They observed velocities of up to 50 μm/s, again more successfully with a low number of bacteria per bead [less than 10, see Fig. 5(a)]. The control over the adhesion of bacteria was also achieved by Park et al. with different materials, namely, bovine serum albumin (BSA)-patterned SU-8 microcubes¹⁶³ and poly-L-lysine (PLL)-coated polyethylene glycol (PEG) beads [see Fig. 5(b)].¹⁶⁴ In the latter case, they also changed the type of bacteria to S. typhimurium. Although being inferior to S. marcescens in terms of speed, they claimed that S. typhimurium is the better choice because it can be easily genetically modified to exhibit fluorescence as well as attenuated pathogenicity and because of its tumor-targeting ability.¹⁶⁴ Some kind of response of the motor bacteria to the in vivo environment is a very important factor to control the micromotors' movement inside the body in a collective, application-oriented way. Behkam et al. and Sitti et al. already stressed this in their studies on chemotaxis of S. marcescens attached to PS beads,^165,166 but Park et al. were the first to show this functionality in vivo with their so-called Bacteriobot system (PS bead + S. typhimurium) with tumor-targeting behavior inside mice, namely, the tendency of S. typhimurium to accumulate in the hypoxic regions of a solid tumor—a type of aerotactic behavior that we call “tumortaxis” in Table II to stress its application-related relevance [see Fig. 5(c)].¹⁶⁷ In this remarkable work, they also employed a different kind of bonding between bacteria and microbeads based on the linking of biotin and streptavidin. This covalent bonding provides a more directed and reliable attachment compared to the previous systems that were based on hydrophilic/hydrophobic adhesion. It was done before by Fukuda et al. with Vibrio alginolyticus bacteria (V. alginolyticus) attached to a liposome with a raft domain.¹⁶⁸ This system is interesting as the liposome cargo was specifically chosen because it is a potent drug carrier for targeted delivery applications, unlike the model structures made of PS or SU-8 that were used earlier. In recent years, there was a tendency to employ more biologically active materials as cargo, like bacterial cellulose¹³³ and alginate.¹⁶⁹ In 2016, Park et al. finally published actually drug-loaded Bacteriobots made of liposomes with paclitaxel (PTX)¹⁷⁰ and hyaluronic acid (HA) beads with docetaxel (DTX),¹⁷¹ respectively, and studied their drug release kinetics, tumor targeting capabilities, and tumor cytotoxicity [see Fig. 5(d)]. These studies mark the current state-of-the-art of type I hybrid biomicromotors as they link the cargo-carrier concept with the collective control by the given biochemistry in the liquid environment and drug loading for specific applicability. Unfortunately, drug delivery and release was not shown in an in vivo model yet, as there were open questions considering the necessary number of microrobots, their motility in actual blood flow, and their immunological properties and long-term toxicity considering unwanted tissue interactions. Also, chemical control alone might not be enough to achieve efficient targeted therapy, although there are promising recent studies on pH-taxis¹⁷² and chemotaxis^173,174 to achieve this [see Fig. 5(e)]. In an earlier study, Kim et al. explored the use of electric fields and UV-light to control the motion of their SU-8-based flat microswimmers propelled by S. marcescens.^175,176 However, these methods are not suitable for in vivo applications, as they do not have the necessary tissue penetration capability for external control, and their biocompatibility remains questionable. These microswimmers were indeed proposed for in vitro micromanipulation and assembly tasks, integrated in MEMS technology, which is not the subject of this review. For biocompatible external control, Sitti et al. investigated the use of a superparamagnetic cargo bead to enable magnetic steering of their hybrid biomicromotor.¹⁷⁷ They claimed that this might provide a means to guide multiple microdevices to the region of interest, where they can then respond to local biochemical signals for targeted delivery. Another concept would be to guide a swarm of micromotors with one magnetically steered swimmer that gradually releases a chemoattractant substance that the others can follow [see Fig. 5(f)]. Either way, multiple control mechanisms for a collective of microdevices are an important step toward efficient and safe in vivo drug delivery applications, which is the main goal for type I microswimmers that are summarized in Table II.

Type II cell-driven hybrid biomicromotors are characterized by an individual motile microorganism (Ø ≈ 1–10 μm) that carries and propels a synthetic cargo particle of a similar size. Contrary to type I micromotors, there is no one standard system that was gradually improved over the years, but several different approaches right from the beginning, employing different bacteria with different control mechanisms. Individual control of these systems was sought after already in the first few works on this subject. Uyeda et al. investigated M. mobile attached to PS beads (both with Ø ≈ 0.5 μm) by biotin-streptavidin interactions and guided them along the engineered tracks in a microchip environment.¹⁵⁷ Whitesides et al. worked using Chlamydomonas, a eukaryotic alga of roughly 10 μm diameter with two flagella, to pick up, transport, and drop off slightly smaller PS beads individually, guided by visible light due to their natural phototaxis ability [see Fig. 6(a)].¹⁴³ Martel et al. employed M. gryphiswaldense, a magnetotactic bacterium with two flagella at its polar ends, to adhere to and propel a slightly larger melamine resin bead.¹⁴⁰ Magnetotactic bacteria like M. gryphiswaldense feature a chain of iron oxide nanoparticles, so-called magnetosomes, within their cell body which allows them to align to the Earth's magnetic field but also makes them susceptible to man-made magnetic fields to control their direction of motion. This useful functionality distinguishes magnetotactic bacteria-based hybrid biomicromotors from other magnetically manipulated hybrid systems that need to include synthetic magnetic materials. Consequently, it was proposed to transfer bacterial magnetosomes to other microorganisms that might be better suited for a given application of a hybrid biomicromotor. This idea of biohybrid integration also describes a departure from the carrier-cargo paradigm of the three introduced type II systems that exhibit only loose interaction between the microorganism and synthetic particles, also clearly indicated by their proposed applications as microtransporters, so-called microoxen,¹⁴³ or micromanipulators, respectively, toward autonomous microrobots. This is perfectly illustrated by the work of Takeuchi et al. in which they attached the flagella of Chlamydomonas to PS beads and liposomes by biotin-streptavidin bonding and controlled their movement via addition of adenosine triphosphate (ATP) to the liquid medium [see Fig. 6(b)].¹⁸⁶ A particularly in vivo application-related strategy to functionally integrate synthetic and biological components was proposed by our Institute in 2013: Bovine sperm cells were coupled to specifically tailored Ti/Fe-microtubes by mechanically trapping individual cells inside them.¹⁸⁷ A sperm cell that became trapped in that way by swimming into a microtube continued to move and thus propelled the tube together with itself [see Fig. 6(c)]. The synthetic tube provided the means to magnetically control the swimming direction of the sperm, but the function and purpose of the sperm to fertilize were retained. This is, to the best of our knowledge, the first hybrid biomicromotor where a microorganism is not solely abused to push around synthetic cargo, but actually supported by the synthetic component to carry out its original intention. This concept was further explored in the following efforts to improve control¹⁵⁴ and functionality¹⁸⁸ by further optimizing geometry and physical properties of the microtubes, for example, by changing the tube material to a polymer-metal-composite that was still magnetic, but also thermoresponsive, containing poly(N-isopropylacrylamide) (PNIPAM), to control sperm capture and release via temperature [see Fig. 6(c)].¹⁸⁸ Similar to type I hybrid biomicromotors, drug delivery and tumor targeting applications are also possible for these microswimmers, as was shown in a recent study on possible treatment of gynecological cancers by the so-called sperm-driven tetrapods.¹⁸⁹ Stanton et al. also published a study on the drug-carrying capability of PS beads half-coated with different metals to promote guided adhesion of up to three E. coli bacteria [see Fig. 6(d)].¹⁹⁰ Similar to the sperm-driven microswimmers, magnetic guidance was enabled by an iron coating of the synthetic component and in vitro cytotoxicity tests were conducted to ensure the viability of the biological motor unit. The tubular geometry of mentioned sperm-driven microswimmers was also adapted to E. coli to make use of enhanced magnetic guiding capability with polypyrrole (PPY)-Nickel (Ni)-polydopamine (PDA) composite tubes.¹⁹¹ Furthermore, these E. coli-driven micromotors offered a literal “kill” trigger for their motion when functionalized with urease: It was shown that once urea was added to the micromotors' surroundings, they stopped moving as the E. coli were literally killed by ammonia which was produced during the enzymatic decomposition of urea in the medium by urease that was functionalized onto the micromotors. The researchers claimed that this provides a biocompatible way to stop the micromotors' motion as urea can be naturally present inside the body. Furthermore, the bacteria-killing effect of produced ammonia may prevent infections after the proposed application of drug delivery by the E. coli-driven microswimmers is accomplished. In vivo tests in animal models are yet to be realized for any of these type II hybrid biomicromotors that are summarized in Table III. Therefore, it is absolutely necessary to visualize them in deep-tissue conditions, preferably in real-time. Magnetic Resonance Imaging (MRI) was proposed by Martel et al. and others to achieve this for such magnetically active hybrid microdevices. This, as well as other possibilities of in vivo tracking, will be discussed further below.

Type III cell-driven hybrid biomicromotors are characterized by an individual motile microorganism (Ø ≈ 1–10 μm) that is functionalized with significantly smaller synthetic cargo particles (Ø < 500 nm). For these kinds of hybrid biomicromotors, a highly autonomous microorganism is desirable, that is ideally susceptible to external control already by nature. For example, magnetotactic bacteria are a reasonable choice as they are potent microswimmers that can be controlled by an external magnetic field without addition of any synthetic component. Martel et al., who worked with M. gryphiswaldense to carry microbeads of similar size (see above), switched to the faster Magnetococcus marinus strain MC-1 and proposed that functionalization of these bacteria with drug-loaded polymeric nanoparticles (NPs) of roughly 150 nm diameter via antibody-mediated covalent binding might provide more effective agents for in vivo targeted therapy.¹⁵⁸ They elaborated their concept with calculations on magnetic controllability and swimming speed, studies on lifetime and potential cytotoxicity, and the ability of MC-1 to target and penetrate solid tumors via aerotactic behavior, especially considering breast cancer detection and diagnosis with follow-up drug delivery.^138,158 The first of these functionalized MC-1 microswimmers was published in 2014 and featured around 70 liposomes of ca. 170 nm diameter attached via carboiimide cross-linking chemistry.¹⁵⁵ A microswimmer speed of ca. 80 μm/s was measured during the first 30 min of application. Biocompatibility was shown in cytotoxicity assays with various cell lines, which also indicated cellular uptake of the liposome cargo. A recent follow-up study of the same system featured in vivo experiments in mice and proved the magnetic control of tens of millions of microswimmers and subsequent aerotactic tumor penetration with fluorescent antibodies.¹²⁸ This can be considered state-of-the-art for type III hybrid biomicromotors, analogous to Park et al.'s work on type I microswimmers,^170,171 with the restriction that no tumor cytotoxic effect was shown, but better external controllability was achieved due to the magnetic field guidance [see Fig. 7(a)]. An even faster Type III microbial swimmer was reported by Kim et al., based on the ciliated protozoon Tetrahymena Pyriformis (Tetrahymena).^87,195 Here, the synthetic components, Fe₃O₄ nanoparticles of ca. 50 nm diameter, were not functionalized onto the surface, but completely internalized by the microorganism by literal ingestion. Tetrahymena's cilia propulsion resulted in velocities as high as 800 μm/s, but unfortunately the microorganism is not suitable for in vivo applications and, apart from magnetic steering and closed-loop control of the protozoon's motion similar to magnetotactic bacteria, no function or application was shown. Another type III concept was introduced by Vanderlick et al. which relied on sperm cells as carriers for liposomes of 200 nm diameter.¹⁹⁶ Unlike the first hybrid biomicromotor based on a sperm cell introduced by our group, which was introduced as a type II micromotor,¹⁸⁷ this one employed human and murine sperm instead of bovine. Similar to the aforementioned work, however, the sperm cell stuck to its original task of fertilization. The attached liposomes were delivered into the egg cell (oocyte) in the process, which was verified by fluorescence [see Fig. 7(b)]. Non-invasive drug administration and gene delivery to alter the genome of an organism before embryo development, for example, to create transgenic mice, was proposed as a possible application. The average velocity of these sperm-driven microswimmers was not mentioned, but it was stated that sperm viability, membrane integrity, and motility patterns remained normal after cargo attachment, which would mean a velocity in the range of 100 μm/s and a lifetime of several hours. No external magnetic or chemotactic control was shown, as this approach relies on the sperm cells' natural abilities to find the oocyte (which is known to involve local chemotaxis and rheotaxis).¹⁹⁷ As mentioned before, staying close to the sperm cell's natural mission and environment can also be used for targeting and drug therapy of gynecologic tumors¹⁸⁹ or assisting fertilization itself.¹⁸⁷ All these systems can be classified as hybrid drug, DNA, or gene delivery agents, and it is important to stress that, contrary to purely synthetic delivery systems, they take advantage of the sperm cell's ability to fuse with another cell—a biological interaction that man-made drug delivery agents cannot emulate easily, as will be discussed further below. A third microorganism that was used for type III hybrid biomicromotors was E. coli, which is thus the only biological microswimmer that was employed for all three types of cell-driven devices (see Tables II, III, and IV). Behkam et al. attached PS nanoparticles of two different sizes (109 and 390 nm) by biotin-streptavidin interaction to E. coli and reported an average velocity of roughly 20 μm/s,¹⁹⁸ which is comparable to the type II system reported by Gracias et al.¹⁹² and significantly better than their own type I E. coli-driven microswimmer.¹⁷⁸ In a subsequent study, they proposed an application different from drug delivery as they showed the possibility to sort similar-sized nanoparticles by surface chemistry with this system.¹⁹⁹ E. coli was shown to bind preferably either specifically to streptavidin-coated particles or unspecifically to positively charged particles, but showed no binding to uncoated and uncharged particles, respectively. This was exploited to separate mixed particles in a microfluidic chamber by selective binding to E. coli and subsequent separation by E. coli 's chemotactic response to swim up a casamino acid gradient in the chamber [see Fig. 7(c)]. Actually, Chen et al. already published a short study in 2005, where they showed unspecific adsorption of Fe₃O₄ nanoparticles of 100 nm diameter onto E. coli bacteria and their chemotactic response to glucose.²⁰⁰ This was one of the first studies where the concept of a bacterial mini robot transporting nano-objects was specified. A recent study on E. coli 's particle binding capacities was conducted by Vanderlick et al. with differently sized liposomes.²⁰¹ The employed gangliosides, a glycolipid protein, covalently bind E. coli to liposomes of 100–200 nm, 1–2 μm, and 10–20 μm diameter. Following the definition in this review, these are actually types III, II, and I cell-driven microswimmers [see Fig. 7(d)]. For these, velocities of roughly 30, 15, and 2 μm/s were reported, which are 100%, 50%, and 7% of the speed of one single E. coli bacterium, respectively. This serves well to illustrate what we found also comparing the other hybrid biomicromotors: Type III microswimmers are generally faster than type II microswimmers because the smaller cargo has less influence on the rheology of the swimmer, i.e., the original performance of the microorganism can be retained. Both are faster than type I microswimmers because one microorganism is faster than a collective of several swimmers when these are not perfectly aligned on one end of the synthetic cargo particle to swim in parallel. However, individual free swimming is not what will be observed in most microswimmer applications, because a collective of thousands if not millions of individual microrobots is necessary to make, for instance, efficient drug delivery feasible. In this sense, considering the pathogenicity of most bacteria strains that were employed so far, it might be better to maximize the synthetic cargo volume and minimize the amount of biological carriers even if it would mean poorer individual swimming performance. Nonetheless, what really distinguishes type III devices from the other hybrid biomicromotors is their potential to interact with the target tissue on a biological level, from cell to cell. As indicated by the aforementioned works, this may lead to solid tumor penetration, fertilization, or cell fusion and cellular uptake in general. An important additional functionality in this respect is the fact that the synthetic cargo does not necessarily need to be attached to the outside of the carrier microorganism, but can be inside the cell and thus protected from the environment. This and further functionalities of hybrid biomicromotors will be discussed in Sec. III B.

There are other types of cells that can be employed as motors for microdevices, not based on flagellar propulsion. The most prominent motors in our body are muscle cells. These cells do not propel themselves, but rather act on a higher-order structure to move it via contraction and subsequent relaxation. Skeletal muscle cells, heart muscle cells (cardiomyocytes), and insect dorsal vessel tissues have been employed to actuate the so-called soft-robotic devices, i.e., flexible or compliant tools for mechanical applications, for example, pumping,^202–204 bending,^95,205–209 and walking/crawling machines.^{93,96,210–213} Also, a biogenerator based on the energy conversion of piezoelectric nanofibers bent by contractile cardiomyocytes was reported.²¹⁴ As before, we want to focus on free-swimming, untethered hybrid biomicromotors. A very illustrative example was given by Saif et al. which is directly related to previously discussed flagellar swimmers: They designed a synthetic, biomimetic sperm flagellum that was actuated by attached rat cardiomyocytes.⁹⁴ They showed the preferential adhesion of the cells to fibronectin-functionalized regions on the PDMS flagellum and reported swimming speeds of roughly 10 and 80 μm/s for two different geometries. However, these velocities are relatively low considering the flagellum length of ca. 2000 μm. In general, the reported cardiomyocyte swimmers are several orders of magnitude larger than the previously discussed hybrid biomicromotors, because one muscle cell alone is far bigger than most motile microorganisms (see Table I).²¹⁵ This also applies to recent works on artificial legged-,²¹⁶ ray-,²¹⁷ and dolphin-²¹⁸ swimmers which all feature a synthetic but biomimetic architecture actuated by a seeded array of muscle cells.

Microorganisms offer many interesting abilities that go far beyond the function as convenient motorization units in hybrid biomicromotors. Some of them, like the capability to respond to their chemical environment, as well as cell-cell interactions, have been already discussed in the previous paragraphs. In the following, we want to discuss hybrid biomicromotors that do not rely on microbial motility, but on one or more different biological abilities of a microorganism that allow advanced functionalities. In some of these cases, motility was provided by an artificial component, as introduced in Fig. 1. Where appropriate, we will also repeat previously discussed cell-driven hybrid biomicromotors in which the microorganism contributed a unique cellular function along with its motility. Figure 8 depicts a possible categorization of such cellular functions, which was already outlined in Fig. 3 where we showed the continuous development of hybrid biomicromotors toward multi-sensitive, multi-functional microdevices. We will discuss how the properties of a hybrid micromotor can not only be altered by artificial extensions, but also by tailoring the abilities of the microorganism via genetic engineering. Furthermore, we will illustrate the advantages of employing microorganism's abilities to sense and respond to its biological environment. In this respect, we already mentioned chemotaxis and tumortaxis that may guide cell-driven hybrid microswimmers, but we will now further elaborate on biocompatibility and in vivo performance in general and also discuss biological interactions that are beneficial for specific processes like uptake and release of synthetic drugs. This finally leads to intercellular communication mechanisms via membrane interactions, cellular uptake, and cell fusion, which can hardly be emulated artificially at all, but nonetheless supported or utilized by synthetic components in a biohybrid approach.

Szalay et al. showed how E. coli, Vibrio cholerae (V. cholerae), S. typhimurium, and L. monocytogenes bacteria, as well as vaccinia virus, can be genetically engineered to express luciferase-catalyzed luminescence or green fluorescent protein (GFP) activity.²²² Such modification of cellular organisms to express certain genes and promote properties that would naturally not occur in the given organism has been accomplished with many different biological systems, including fully grown animals.^{219,223–225} What is exceptional about the mentioned work is that microorganisms like E. coli and S. typhimurium were modified, which are common carrier bacteria in hybrid biomicromotors. The tumor targeting ability of these bacteria was shown in mice, as well as their ability to survive and replicate within the tumor. This could be visualized in vivo and in real time by the genetically induced fluorescence of the bacteria [see Fig. 9(a)]. Consequently, the potential for biorobotic tumor metastase detection and tumor-specific gene therapy was shown with genetically engineered bacteria.²²² In a similar study, Choy et al. injected loaded plasmids into E. coli to express bacterial luciferase and showed real-time monitoring of bacterial migration in tumors and non-invasive tumor imaging also in mice [see Fig. 9(b)].²²⁶ These two examples illustrate how genetic engineering may serve to specifically modify hybrid biomicromotors by tailoring the microorganism's properties and adding new functions, for example, fluorescence, to the system. Park et al. made use of this concept when they employed S. typhimurium for their type I hybrid biomicromotors that were genetically modified to be attenuated (i.e., less pathogenic) and to express GFP.¹⁶⁴ Another possibility might be to improve a carrier microorganism's swimming speed or chemotaxis abilities by genetic engineering. We consider such cell-specific biological modifications a means to minimize physical changes like the addition of more synthetic components to achieve desired functions where possible, because the biocompatibility of the whole biohybrid system might be compromised by more and more synthetic materials, as well as the swimming performance of the microorganism, if it is loaded or functionalized with many different synthetic particles. Consequently, the possibility to genetically modify biological entities is an additional advantage of hybrid biomicromotors that employ such tailored microorganisms.

Next to their swimming ability, biocompatibility is indeed the most important argument for employing cellular organisms as microrobots. This might sound counterintuitive, as most hybrid biomicromotors reported so far were based on bacteria that are actually pathogenic which is, of course, not favorable for the host organism, i.e., the patient. However, we need to rethink the concept of biocompatibility when considering in vivo microswimmers. In general, the biocompatibility of a synthetic microdevice is assessed by incubating it together with the target cells or tissue to investigate the material's possible toxicity which might disturb the cell's metabolism and proliferation or even cause cell death. But biocompatibility can not only be described as the absence of cell toxicity, it needs to take into account the reaction of the whole body and immune system to the infiltrating microdevice. Especially, synthetic micromotors that enter the bloodstream will be attacked by immune cells as they are recognized to be of foreign origin and disposed of or isolated even if they are not releasing harmful substances or causing inflammations. This microscale response might not be a big problem for static macroscale implants like artificial hip joints, but tiny micromotors are already useless when they are stuck and fixed at some nook or twist inside the body. Microbes like bacteria, by contrast, are adapted to this environment and have developed abilities to thrive or at least to hide their presence from the immune system inside a foreign organism. This advantage can be exploited in biohybrid devices even with immotile microorganisms—a feature that may be called biocamouflage. A good example was given by Wang et al. who employed red blood cells (erythrocytes) as microrobots.⁹² Considering biocompatibility, erythrocytes are definitely among the best solutions for in vivo microcarriers as they can be obtained directly from the patient and thus will not provoke any negative effects when (re-)administered to the same patient as they are not recognized as foreign objects. However, erythrocytes are not motile on their own, as they are usually dragged along passively by the flow of blood. Wang et al. proposed to propel the cells with externally applied ultrasound pulses and reported velocities of up to 30 μm/s in phosphate buffered saline solution (PBS) and 10 μm/s in blood. Furthermore, they loaded the cells with Fe₃O₄ nanoparticles with a method called hypotonic dilution encapsulation, where a sequence of hypotonic and isotonic treatments leads to diffusion of the particles into the erythrocytes. The synthetic particles allowed magnetic steering of the resulting hybrid biomicromotors [see Fig. 10(a)]. The addition of further functional synthetic components to employ the controllable microrobot for specific biomedical applications like biosensing and drug delivery was proposed.⁹² Considering the already discussed cell-driven hybrid biomicromotors, it is not yet clear which microswimmer cells are the best at working around the obstacles in vivo. It is most likely dependent on the application and specific in vivo environment. For example, sperm-driven micromotors obviously excel when applied in the female reproductive tract for assisted fertilization,¹⁸⁷ prenatal gene delivery,¹⁹⁶ or targeted therapy of gynecologic cancers,¹⁸⁹ but are probably not the best solution for injection into the blood stream. Also, the biocamouflage effect of hybrid microbial swimmers is presumably compromised by the synthetic component, which means that type III would be preferable to type I cell-driven microswimmers in that respect, because there the synthetic component is less prominent. Ideally, the synthetic component is entirely encapsulated by the microorganism, like in example of Wang et al. In the case of synthetic drug cargo for targeted delivery, thereby, the drug would be also protected by the carrier cell's membrane to avoid unintentional dissipation,²²⁷ similar to liposomes that were used as synthetic drug carriers for type I microswimmers.^170,171 However, the carrier cell's viability and functionality must remain intact, and a proper cargo release mechanism needs to be established in these cases. Interesting microbial carriers to act as such advanced drug delivery systems are also other suspension cells like T cells and macrophages. They were proposed as therapeutic delivery agents already in 2000 by Naylor et al. because of their ability to naturally accumulate in hypoxic tumor regions and being recruited as part of the endogenous immune system.²²⁸ This so-called tumortropic migration behavior (that we termed “tumortaxis” in Table II to indicate its similarity to aerotaxis and chemotaxis), along with their endogenous nature (i.e., cells taken directly from the respective patient) can be combined with therapeutic cargo and further synthetic extensions to build a biohybrid micromotor that exhibits biocamouflage and acts as a “Trojan horse” to infiltrate tumor tissue, as shown by Bashir et al. in 2007.²²⁰ They reported that gold nanoshells of ca. 150 diameter could be taken up by macrophages conveniently via phagocytosis. They exposed the biohybrid devices to near infra-red (NIR) light to release the nanoshells by photoinduced ablation of the macrophages and reported subsequent photoinduced cell death of infiltrated breast tumor tissue [see Fig. 10(b)]. Hirschberg et al. presented a similar approach of silica/gold core/shell NPs taken up by macrophages to penetrate the Blood-Brain-Barrier (BBB) and treat glioma (brain tumor tissue) by NIR-induced photothermal ablation.²²⁹ Drezek et al. employed T cells filled with 45 nm diameter gold colloids and showed tumortropic migration and nanoparticle distribution in vivo with mice.²³⁰ Mouse model experiments were also conducted by Chu et al. with macrophages and gold nanorods of ca. 7 nm diameter and 25 nm length.²³¹ They proved that the photothermal treatment approach can inhibit tumor growth and recurrence rate significantly in vivo. Another approach employing macrophages as “Trojan horses,” but with different artificial functionality, was proposed by Troyer et al.: Instead of photothermal treatment, they relied on alternating magnetic fields that induced hyperthermia on paramagnetic Fe/Fe₃O₄ core/shell NPs (ca. 30 nm diameter) loaded into macrophages.²³² Also in a mouse model, they showed tumortropic migration of the biohybrid system and subsequent tumor growth inhibition of pancreatic cancer by magnetically induced hyperthermia. Finally, chemical drugs like doxorubicin (DOX), viral vectors, or interferons and other cytokines can also be loaded into macrophages^233–235 and T cells.^236–238 A synthetic container for these therapeutics, which can be taken up by phagocytosis and ensures controlled release and distribution upon tumor infiltration, is however still desirable, as was shown by Chiu et al. who filled macrophages with perfluoropentane composite bubbles and DOX-loaded polymer vesicles.²³⁹ The echogenic composite particles thereby served two purposes: First, the tumortropic infiltration of hypoxic tumor regions could be supervised by ultrasound reflection, which is a non-invasive in vivo imaging technique superior to fluorescence visualization in terms of deep tissue penetration. Second, focused ultrasound was used to collapse the bubbles, which induced structural disruption of neighboring polymer vesicles and thus triggered DOX release. However, contrary to previously presented drug-loaded type III (or other) cell-driven swimmers, all these leukocyte-based systems can only move by cell migration, i.e., with a velocity of usually less than 1 μm/s. This compromises their efficiency if they are not directly injected onto the tumor, which is not always possible. Detailed reviews of such systems that are not actually microswimmers were published in recent years.^240–242 To overcome this weakness, Park et al. proposed to let macrophages internalize poly-lactic-co-glycolic acid (PLGA) particles of 300 nm diameter that contained the drug DTX, but also Fe₃O₄ NPs (Ø ≈ 10 nm) to enable magnetic manipulation.²⁴³ They showed that aggregates of such biohybrid systems can be moved toward a tumor spheroid in a microfluidic channel with a velocity of up to 50 μm/s by an external magnetic field gradient, and then infiltrate the tumor tissue by tumor-triggered macrophage recruitment [see Fig. 10(c)]. In a recent study, they refined this approach by employing macrophages filled with liposomes of roughly 150 nm diameter filled with Fe₃O₄ NPs and PTX instead of DTX.²⁴⁴ The dual-targeting by magnetic and chemotactic (tumortropic) stimuli, as well as chemotherapeutic activity toward breast and colorectal cancer, was assessed in an in vitro model. As discussed earlier, the biocamouflage effect of blood cells is induced by their endogenous nature that allows them to act as “Trojan horses” within the patient's immune system.²²⁰ Their tumortropic behavior towards hypoxic regions is however based on chemotactic cell-cell interactions that include, for example, active macrophage recruitment by the tumor.²⁴⁵ The utilization of such cell-to-cell communication and specific cellular membrane interactions is another important asset of microbial systems and marks a further extension of the biohybrid approach that will be discussed in detail in the following.

The ability of macrophages to infiltrate and communicate with tumor tissue that was described earlier can also be found in several bacteria, and bacterial tumor targeting was already mentioned, too.^167,222 The concept of manipulating bacteria to use them in tumor therapy actually already existed before their application as freely swimming micromotors and was termed bactofection, i.e., transfecting biochemical agents like genes or therapeutical proteins from bacteria to tumors or other cells.^{159,228,246,247} This involves cellular uptake of the bacterium by the target cell in most cases. Several reviews addressed these systems and also recent progress and challenges.^{13,240,246–250} We want to focus on works that combine the biological functions of microorganisms that allow intracellular gene delivery with microrobotic control mechanisms that enable efficient targeting and external supervision. One of the earliest works where such a bactofection microrobot was presented was published by Bashir et al. in 2007. They functionalized Listeria monocytogenes bacteria (L. monocytogenes) with PS nanoparticles that contained fluorescent or bioluminescent gene model drugs and showed how these hybrid biomicromotors were internalized by various tumor cell lines in vitro.²⁵¹ In vivo studies in mice confirmed the transcription and expression of successfully delivered GFP [see Fig. 11(a)]. It was stressed that this bacteria-mediated delivery of plasmid DNAs is superior to delivery just by loaded nanoparticles, because the bacteria were on the one hand able to penetrate and colonize solid organ tumors, and on the other hand able to escape from entrapment in intracellular vesicles to release their cargo. Similar to Park et al. work with macrophages that was discussed earlier,²⁴³ we did not classify this system as a type III cell-driven microswimmer because the swimming speed of L. monocytogenes is less than 1 μm/s. However, both works describe an important extension of the hybrid biomicromotor approach, i.e., the combination of cellular uptake mechanisms with the delivery of synthetic cargo. Efficient cell-driven microswimmers that feature these abilities were presented by Vanderlick et al.¹⁹⁶ and our group,¹⁸⁹ which are already discussed earlier. These hybrid biomicromotors relied on sperm cells and their capability of cell fusion. As already mentioned, the use of sperm cells in biohybrid devices was initially proposed by our institute to assist fertilization, so in that case no synthetic cargo was delivered, but the synthetic component served solely to allow external manipulation.¹⁸⁷ This approach was further extended in 2016, where the sperm cell was assisted in cell fusion and fertilization by also taking over its propulsion duty, i.e., to allow fertilization with immotile sperm cells by propelling them artificially.²²¹ This important step to combat male infertility was achieved by mechanically coupling polymer-metal composite microhelices of ca. 5 μm diameter and 25 μm length to individual bovine sperm cells and rotating them to propel and swim by an external magnetic field [see Fig. 11(b)]. Velocities of ca. 20 μm/s were reported for the coupled system and cytotoxicity studies were conducted to confirm the sperm cells' viability and functionality. This work is contrasting the previously discussed cell-driven microswimmers as it is reverting the initial idea of biological propulsion combined with an artificially imposed function, to an application that demands biological function combined with artificial micromotion. Nonetheless, it describes a motile biohybrid device with an advanced function that is tailored for a specific application, following the definition illustrated in Fig. 1. Another work where the microorganism represented the cargo and not the carrier was published by Suzuki et al.: They reported the nonspecific hydrophobic adhesion of multiple E. coli bacteria to one PS bead of ca. 5 μm diameter that was coated with zinc and platinum in a Janus configuration to propel itself by a well-known redox reaction with methanol or p-benzoquinone.²⁵² Here, the E. coli bacteria did not contribute to the system's chemical self-propulsion of ca. 0.15 μm/s, but magnetic guidance was enabled by an additional nickel coating. Although bacteria and synthetic beads were linked to form a biohybrid entity, it is questionable if this can be characterized as a hybrid biomicromotor in the frame of this review at the current stage, because the proposed application as the device for spatial cell manipulation and assembly would imply that any synthetic microtool that is used to pick and place cells could be considered a biohybrid device. By contrast, we want to point out that the previously discussed sperm-carrying micromotor of our institute aimed to perform a biological task, i.e., fertilization with an immotile sperm cell, by complementing the cell's deficiency with a synthetic component. In that case, the specific cellular interaction of the sperm cell was the center of the approach. Another case of cell manipulation was reported by Yang et al. who artificially emulated magnetotactic properties by grafting a Fe₃O₄-doped alginate hydrogel to yeast cells (Saccharomyces cerevisiae) and bacteria (Flavobacterium heparinum), respectively.²⁵³ They claimed that the enabled spatial manipulation by magnetic fields can be transferred also to other microorganisms to obtain mobile building blocks for micro-patterned cell arrays that may serve for highly specialized cell studies or tissue engineering. This may lead away from the review's focus on free-swimming, robotic hybrid biomicromotors, but serves as another example where the cell's natural function should remain untouched, while control and motility are provided artificially. A final example for a hybrid biomicromotor that relies on cell-cell interactions and also shows cell transport was proposed by Song et al.: They presented the guided transport of Staphylococcus aureus bacteria (S. aureus) by magnetotactic MO-1 bacteria. S. aureus are spherical bacteria with a diameter of roughly 1 μm and are thus slightly smaller than MO-1.²⁵⁴ Individual and up to four S. aureus were shown to bind to MO-1 which were functionalized with polyclonal antibodies [see Fig. 11(c)]. Swimming velocities of up to 200 μm/s were reported which is remarkable, as it is roughly the same speed that unloaded MO-1 exhibit. The system is a hybrid biomicromotor because the connection of microbial cargo and microbial carrier was brought about by synthetic means, i.e., functionalization with polyclonal antibodies, although it is of course the microorganisms' distinct membrane properties that make such an artificial bonding possible. The proposed application of this approach was pathogen separation and isolation, which also qualifies it as a biomedical microrobotic device. A summary of important hybrid biomicromotors that employ advanced cellular functions is given in Table V.

There are several key challenges regarding the performance of hybrid biomicromotors for in vivo applications with respect to motility, control, and functionality (see Fig. 1). Motility, i.e., microscale swimming, is arguably the most developed feature of many hybrid micromotor approaches. In in vitro experiments, many of the reviewed works reported swimming velocities that were well in the range of natural in vivo swimmers like bacteria or sperm, i.e., several tens of micrometers per second, which can be considered a reasonable reference point. Unfortunately, in vivo conditions are different from laboratory experiments. Some studies tried to mimic swimming in body fluids by emulating rheological properties and temperature conditions of the target medium; however, there are many more obstacles like suspension cells, extracellular matrix, and epithelial walls that may hinder a micromotor's motion in vivo.^2,255,256 In general, the elasticity of tissue-lined channels in the body, their geometrical complexity with numerous constrictions and twists, and the abundance of sticky proteins on virtually any in vivo surface may pose a big threat for any microswimmer. Precise control of the devices' motion is absolutely necessary. Chemotactic steering of cell-driven swimmers and magnetic control, either of magnetotactic bacteria or magnetic artificial components, were the favored means of guidance until now. The former because chemical clues are in many cases already present at the target location, for example, a hypoxic gradient in solid tumors for drug delivery (aerotaxis is a form of chemotaxis) or progesterone secreted by the oocyte to attract sperm cells for fertilization. The latter because magnetic fields can be applied externally, i.e., non-invasively, and are harmless to the body. Exploiting the biohybrid approach, both methods can also be combined for the so-called dual-targeting with long-range magnetic steering and short-range chemotactic/aerotactic targeting, as was shown by several groups (as discussed earlier). An important advantage of this concept is that it can be applied collectively on a large band of micromotors, which is necessary, for example, to achieve significant doses for drug delivery. This was shown in several in vivo studies with mice for various biohybrid systems (as discussed earlier). However, then the individual address of microswimmers is aggravated, which might be a problem considering the aforementioned difficulties of in vivo navigation. It is very likely that individual swimmers get scattered and stuck somewhere in the body where they do not belong. In these cases, a rescue mechanism needs to be established. Ideally, synthetic materials are biocompatible and also biodegradable, i.e., can be resorbed by the body fluids. Cellular components like bacteria could be cleared by the immune system, for example, phagocytized by macrophages. If this is not the case, inflammations might occur and postoperative treatment might be necessary to remove the foreign devices. This depends also on the site of application in the body. For example, micromotors that operate in the gastrointestinal tract are arguably easier to wash out or excrete than micromotors that traverse the blood-brain-barrier. The site of application also influences the number of micromotors that is necessary for a given task, as well as the requirements for biocompatibility. Little research has been done on these issues. Depending on the accessibility of the application site, the life-time of swimming microorganisms or the externally penetrating signal intensity to artificially actuate microswimmers may also be crucial factors that need to be addressed. In this respect, the most important challenge considering controllability and applicability is probably in vivo tracking and imaging of the operating microdevices, or some other means of feedback for supervision. There are several critical requirements. Spatial and temporal resolution is important to track micromotors in real time. Optical microscopy techniques are suitable for this, but they lack deep tissue penetration capability which is important for non-invasive visualization. Fluorescence signals from microdevices have been detected in in vivo experiments, but only from millimeters under the skin of mice, not from bulk tissue or larger blood vessel networks. Near infrared (NIR)-emitting sources might be able to overcome this weakness in the first (650–950 nm), second (1100–1350 nm), and third biological window (1600–1870 nm),^257–259 but real time tracking of swimming micromotors has not been shown yet. Another option for real time deep tissue penetration is magnetic resonance imaging (MRI).²⁶⁰ For example, large collectives of magnetotactic bacteria could be tracked this way;²⁶¹ however, the spatial resolution is not good enough to visualize individual microswimmers yet. A third option is acoustic signaling, for example, ultrasound. Ultrasound has been applied for decades to image live fetuses, although of course not with microscale resolution.²⁶² Studies on photoacoustic imaging however indicated the possibility of microscale in vivo imaging.^263–265 Echogenic components, i.e., materials that reflect ultrasonic actuation with a high signal-to-noise ratio, are crucial in order to achieve high resolution that enables individual real time tracking of micromotors. This was not achieved by any of the mentioned approaches in an in vivo experiment yet, but may be possible in the near future due to continuous and promising advances in the field. Also, a combination of these imaging techniques, as well as others, might be better suitable than one approach alone to face this challenge. Another aspect that can play a role in in vivo micromotor guidance and control is the autonomous decision-making ability of swimming microorganisms, for example, chemotactic response to the environment. For instance, sperm cells that are loaded with synthetic drugs to treat ovarian or fallopian tube cancer should be able to find their destination autonomously, since it is coincident with their natural mission, i.e., to reach the oocyte. In that sense, such a sperm-driven hybrid biomicromotor would be a true microrobot. However, it would be still desirable to be able to track the individual swimmers in order to retrieve failed devices. Also, depending on the amount of employed autonomous microrobots, their behavior might be different, depending on collective interactions that might lead to different “decisions.” This needs to be investigated carefully in suitable in vitro and in vivo models for a given application.

Apart from such technical challenges and requirements that need to be met before application, there is also the problem of approval. Micro- and nano-sized devices that operate inside the body are still far from being widely accepted in medicine, not only by the authorities that rightfully demand elaborate risk evaluations and long-term toxicity studies, but also by the potential patients. Especially, the abovementioned autonomously operating microrobots that employ living microorganisms will potentially cause more personal discomfort than a conventional, although bitter, pill.

The largest field of application for hybrid biomicromotors is cancer therapy, not only because cancer affects millions of people worldwide but also because it occurs in so many different forms in different parts of the body. Considering the discussed strengths and weaknesses of current micromotor approaches, it is clear that such different disease patterns demand highly specialized microdevices to achieve targeted therapy. Therapy in these cases mostly means drug delivery, i.e., the transportation of therapeutics to the diseased location (e.g., the tumor) with tailored dosage and drug release kinetics. Even before that, hybrid micromotors may serve for diagnosis, e.g., tumor localization, taking advantage of unique biosensing capabilities of various microorganism. A self-sufficiently operating hybrid microrobot may even be able to sense, diagnose, and treat diseased tissue autonomously, a concept that was labeled theranostics. This concept is in principle also applicable to other diseases, but cancer therapy is decidedly the main focus of hybrid biomicromotor research for biomedical applications at the moment. Apart from diagnosis and drug delivery, other therapeutic operations like microsurgery by drilling or localized thermal treatment were proposed. Also, other biomedical applications like assisted fertilization, gene transfection, and cell sorting for pathogen isolation or in situ tissue engineering have been explored. In all these cases, the biological sensing and interaction capabilities of living microorgansims are key to successful microscale manipulation in vivo. Artificial components serve to extend, enhance, or support these cellular functions, and provide means to externally control and supervise the hybrid device's task. Microscale swimming is thereby a decisive asset to ensure localized treatment, which serves to maximize efficiency while minimizing risks of side effects. Motility can be provided either biologically, i.e., by the microorganism, or artificially by an engineered carrier or synthetic attachments that respond to external stimuli like magnetic fields or ultrasound. We reviewed such hybrid biomicromotors to underline their potential for in vivo applications and tried to bring order into the variety of organisms and mechanisms that have been employed to pursue this exciting goal.

The authors would like to acknowledge Veronika Magdanz, Haifeng Xu, and Azaam Aziz for fruitful discussions and input. Financial support from the German science foundation (DFG) within the priority program SPP 1726 “Microswimmers–From Single Particle Motion to Collective Behavior,” as well as from the Leibniz Institute for Solid State and Materials Research Excellence Initiative (Project No. 00506024), is highly appreciated.