You are on page 1of 26

BE17CH10-Feinberg ARI 31 October 2015 15:30

ANNUAL
REVIEWS Further
Click here to view this article's
online features:
• Download figures as PPT slides
• Navigate linked references
• Download citations
• Explore related articles
Biological Soft Robotics
• Search keywords

Adam W. Feinberg
Department of Biomedical Engineering and Department of Materials Science and Engineering,
Carnegie Mellon University, Pittsburgh, Pennsylvania 15213; email: feinberg@andrew.cmu.edu
Annu. Rev. Biomed. Eng. 2015.17:243-265. Downloaded from www.annualreviews.org
Access provided by 38.25.16.177 on 10/29/22. For personal use only.

Annu. Rev. Biomed. Eng. 2015. 17:243–65 Keywords


The Annual Review of Biomedical Engineering is soft robotics, molecular motors, tissue engineering, bioinspired design,
online at bioeng.annualreviews.org
biomimetics, skeletal muscle, cardiac muscle
This article’s doi:
10.1146/annurev-bioeng-071114-040632 Abstract
Copyright  c 2015 by Annual Reviews. In nature, nanometer-scale molecular motors are used to generate force
All rights reserved
within cells for diverse processes from transcription and transport to muscle
contraction. This adaptability and scalability across wide temporal, spatial,
and force regimes have spurred the development of biological soft robotic
systems that seek to mimic and extend these capabilities. This review de-
scribes how molecular motors are hierarchically organized into larger-scale
structures in order to provide a basic understanding of how these systems
work in nature and the complexity and functionality we hope to replicate
in biological soft robotics. These span the subcellular scale to macroscale,
and this article focuses on the integration of biological components with
synthetic materials, coupled with bioinspired robotic design. Key examples
include nanoscale molecular motor-powered actuators, microscale bacteria-
controlled devices, and macroscale muscle-powered robots that grasp, walk,
and swim. Finally, the current challenges and future opportunities in the
field are addressed.

243
BE17CH10-Feinberg ARI 31 October 2015 15:30

Contents
1. INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 244
2. OVERVIEW OF BIOLOGICAL ACTUATORS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 246
2.1. Basics of Molecular Motors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 246
2.2. Hierarchical Organization and Control of Molecular Motors
in Cells and Tissues. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 247
3. BIOLOGICAL SOFT ROBOTS AT THE SUBCELLULAR
TO CELLULAR SCALE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 248
4. MUSCLE-POWERED ACTUATORS FROM THE
MICRO- TO MACROSCALE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 249
4.1. Devices Based on Two-Dimensional Cantilever and Film Bending . . . . . . . . . . . . 250
Annu. Rev. Biomed. Eng. 2015.17:243-265. Downloaded from www.annualreviews.org

4.2. Three-Dimensional Muscle Bundle–Based Constructs . . . . . . . . . . . . . . . . . . . . . . . . 251


5. BIOLOGICAL SOFT ROBOTS WITH COMPLEX FUNCTIONALITY
POWERED BY MUSCLE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 252
Access provided by 38.25.16.177 on 10/29/22. For personal use only.

5.1. Muscle-Powered Pumps . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 252


5.2. Muscle-Powered Grippers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 254
5.3. Robots That Walk or Crawl . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 254
5.4. Swimming Robots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 255
6. CURRENT AND LONG-TERM CHALLENGES
AND POTENTIAL SOLUTIONS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 255
6.1. Sourcing Molecular Motors and Cells for Building Devices . . . . . . . . . . . . . . . . . . . 255
6.2. Limitations of Biological Soft Robots Based on Muscle Cell Type . . . . . . . . . . . . . 257
6.3. Long-Term Maintenance of Biological Components In Vitro . . . . . . . . . . . . . . . . . 257
6.4. Energy Storage, Supply Systems, and Handling of Metabolic Waste. . . . . . . . . . . 258
6.5. Motor and Muscle Control Systems for Selective
and Coordinated Actuation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 259
7. CONCLUSIONS AND FUTURE DIRECTIONS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 259

1. INTRODUCTION
Nature has solved the problem of motility across a wide range of length scales from a few
nanometers to thousands of kilometers by using molecular motors as a functional building block.
Molecular motors are protein machines that convert chemical energy into work (1) and power a
diverse number of biological processes from DNA transcription (2), to bacteria swimming (3), to
muscle contraction (4). Coupled with a broad array of feedback and control mechanisms (5), these
molecular motors operate in scalable spatial, temporal, and force regimes (6, 7) that are difficult to
achieve using synthetic, engineered alternatives (8). The field of biological soft robotics seeks to
combine these molecular motor actuator systems with multiscale biological design principles
to engineer dynamic systems with functionality that exceeds that achieved using traditional
approaches. For example, striated skeletal muscle has superior power and force-to-weight
ratios and systems control compared to traditional mechanical actuators (9), including dynamic
modulation of force generation through spatial and/or temporal summation of innervated
motor units (10). Biology also provides design strategies to create complex actuator systems that
improve upon traditional robotics (11), such as muscular appendages like the elephant trunk and
the octopus arm (12), which have greater degrees of freedom. Understanding how these living

244 Feinberg
BE17CH10-Feinberg ARI 31 October 2015 15:30

systems work and how they can be utilized directly, and/or effectively mimicked, is the foundation
of biological soft robotics and necessary to achieve broad application in engineered systems.
Here biological soft robotics are explored from the standpoint of integrating a biological
component with a synthetic component to create a biohybrid system. The biological component
may be individual or ensembles of molecular motors or larger assemblies of molecular motors in
the context of the contractile cytoskeleton of muscle cells and tissues. The synthetic component is a
material fabricated to selectively interact with the biological component and provide a mechanical
framework against which to act. Generally, this is a polymer elastomer or hydrogel, but other
materials are also used. In addition, these biological soft robots take design principles directly
from living systems, such as bacterial flagella at the microscale (13) or the jellyfish body plan at the
centimeter scale (14) to achieve swimming at the appropriate Reynolds number. It is important
to understand that here the word soft refers to the use of living components (e.g., proteins, cells,
and tissues) that inherently have a low elastic modulus <100 kPa (kilopascal). In comparison,
Annu. Rev. Biomed. Eng. 2015.17:243-265. Downloaded from www.annualreviews.org

traditional robots are composed of polymer, metal, and composite materials with elastic moduli
that range from 1 MPa to 100 GPa. Soft, elastomeric materials are now being integrated into
Access provided by 38.25.16.177 on 10/29/22. For personal use only.

robotics systems to improve interaction with the environment and provide enhanced functionality
(15), but are not considered here (see Reference 16 for a recent review). It is also important to
incorporate bioinspired design, leveraging not just the molecular motors as an actuator system
but also how they are coupled into larger cell and tissue architectures (17). Note that, although
biomimetics and bioinspired design are active areas of research and development throughout
the field of robotics because of the advanced functionality these approaches enable (11, 18), this
article focuses on soft robots that incorporate living biological components. Specifically, using a
scalable actuator system built on molecular motors provides flexibility in configuration and force
generation that is not possible using synthetic alternatives.
In this review, biological soft robotics are covered from the standpoint of the nanoscale molec-
ular motors that power these systems and the way they are scaled and integrated with synthetic
components to create functional devices (Figure 1). First, biological actuators are discussed as
they operate in vivo, spanning from subcellular to cellular to tissue length scales. This includes
the hierarchical organization of molecular motors into larger-scale structures and the regulation
of actuation through integrated control systems. Next, examples of biological soft robotics across
length scales are explored, starting with nanoscale molecular machines and moving to larger mi-
croscale cellular systems and then on to macroscale multicellular tissues integrated into larger
devices. At the macroscale, the article focuses on biological soft robots that have achieved a level
of functionality similar to traditional robotics, including grasping, walking, and swimming. Fi-
nally, there is a discussion of the current and long-term challenges that face the field of biological
soft robotics and considers some of the potential solutions and future applications. The challenges
include (a) sourcing molecular motors and cells for building devices; (b) limitations of biological
soft robots based on muscle cell type; (c) long-term maintenance of biological components in
vitro; (d ) energy storage and supply systems and handling of metabolic waste; and (e) motor and
muscle control systems for selective and coordinated actuation. Recent reviews have covered re-
lated aspects of this field from the standpoint of the molecular motors used in multiscale actuator
applications (19) and the specific challenges of using skeletal muscle as an actuator in soft robotic
devices (20). This review touches on these topics but also discusses many of the challenges that
must be overcome to use these living components in engineered robotic systems. Furthermore,
it provides a detailed understanding of the current state of biological soft robotics, the unique
advantages of these systems, and the challenges that researchers are actively working to address
as innovation in this field continues to progress.

www.annualreviews.org • Biological Soft Robotics 245


BE17CH10-Feinberg ARI 31 October 2015 15:30

Sarcomere Biological soft robot

Actin–myosin Myocytes
2D muscle
motors

10 –9 10 –8 10 –7 10 –6 10 –5 10 –4 10 –3 10 –2
Multiscale molecular motor assembly (m)
Annu. Rev. Biomed. Eng. 2015.17:243-265. Downloaded from www.annualreviews.org

Figure 1
Schematic showing the multiscale molecular motor assembly used to engineer a biological soft robot.
Access provided by 38.25.16.177 on 10/29/22. For personal use only.

Actin–myosin molecular motors are scaled from single motors (∼5 nm) by assembly into larger-scale
structures that spatially organize these motors into high-density three-dimensional arrays termed sarcomeres
(∼2 µm). Myofibrils organize sarcomeres into linear, contractile cytoskeletal filaments that typically span the
length of cardiac myocytes (∼100 µm). The myocytes, as differentiated heart muscle cells, contain the
cellular machinery to regulate calcium influx into the cells when stimulated and to regulate contraction.
Using tissue-engineering techniques, the myocytes can be assembled into two-dimensional (2D) sheets of
muscle cells that couple together via gap junctions to form a highly aligned syncytium (∼1 mm). This 2D
tissue can be extended and integrated with engineered structures (e.g., polydimethylsiloxane films) to create
biological soft robots, shown here for a basic triangular, muscular thin film swimmer (77).

2. OVERVIEW OF BIOLOGICAL ACTUATORS


At the core of biological soft robotics is the ability to move in the environment, which requires
actuators capable of generating force. The primary way this is done in cells is using molecular
motors, tiny protein machines that are able to convert chemical energy in the form of adenosine
triphosphate (ATP) or guanosine triphosphate (GTP) into work (21, 22). Although there are many
molecular motors that perform diverse functions within the cell, it is the myosin motor, and its
ability to processively move along actin filaments, that enables formation of contractile cytoskeletal
elements, such as stress fibers and myofibrils. The assembly of cells that contain these contractile
cytoskeletal elements into muscle tissues, in turn, enables the formation of biological actuators at
the macroscale. In addition to the molecular motors, cells and tissues also contain metabolic path-
ways that provide critical support including (a) cellular respiration to generate more ATP, (b) pro-
tein synthesis and degradation to maintain structure, and (c) calcium handling to temporally regu-
late when motor function (contraction) occurs. This section provides a brief overview of molecular
motors and the way they are structured in muscle cells and tissues. More in depth discussion is
available in the cited references and in standard textbooks on cell biology and muscle physiology.

2.1. Basics of Molecular Motors


There are a wide range of molecular motors specialized for tasks throughout the cell. Generally,
these molecular motors move along a filament or other linear macromolecule in a processive
manner powered by ATP hydrolysis (1) and encompass processes that are critical and integral
to cell function (23). For example, RNA and DNA polymerases are nucleic acid motors that can
transcribe RNA from DNA (2) and form double-stranded DNA from single-stranded DNA (24),
respectively. Dynein and kinesin are molecular motors that move along microtubules and transport

246 Feinberg
BE17CH10-Feinberg ARI 31 October 2015 15:30

cargo throughout the cell. They are also responsible for chromosome separation during anaphase
and, in combination with larger microtubule assemblies, form flagella to enable single-cell motility
(25, 26). Myosins are a class of motor proteins that move along actin filaments. They can serve
various functions, from single motors for cargo transport similar to dynein or kinesin, to many
motors that assemble into contractile cytoskeletal filaments, including stress fibers and sarcomeres
(25, 27). There are also other motor proteins that couple to actin, such as filament end tracking
motors like formins and vasodilator-stimulated phosphoprotein that can be used for motility by
using polymerization to generate force (28). In fact, listeria bacteria can hijack this process to create
an actin “rocket tail,” which can push them through the cytoplasm (29). These molecular motors
operate within the cell and individually generate forces in the range of 1 to 10 pN over a spatial
range of a few nanometers to thousands of micrometers (such as for transport down neuronal
axons) (26, 30, 31). All of these molecular motors are relevant for biological soft robotics designed
to function at the nano- or microscale, where small ensembles of these motors can generate
Annu. Rev. Biomed. Eng. 2015.17:243-265. Downloaded from www.annualreviews.org

sufficient force for the intended application. The ability to generate greater force and move over
longer distances requires simultaneous action of motors in parallel and in series, respectively.
Access provided by 38.25.16.177 on 10/29/22. For personal use only.

2.2. Hierarchical Organization and Control of Molecular Motors


in Cells and Tissues
Molecular motors are flexible actuator systems because they are scalable. This is achieved through
the assembly of actin and myosin into contractile cytoskeleton filaments, such as stress fibers
and myofibrils. Even though they have distinctly different contractile properties, stress fibers and
myofibrils contain many similar structural elements, including bundles of actin filaments that are
coupled to myosin isoforms that also assemble into bipolar filament structures (32, 33). Other pro-
teins include tropomyosin, which regulates the crossbridge (binding) interaction between myosin
motors and actin filaments, and α-actinin, an actin-binding protein that helps spatially organize
actin filaments into the sarcomeres in striated muscle and sarcomere-like structures and dense
bodies in smooth muscle (34). For biological soft robotics, the sarcomere and sarcomere-like
structures are of particular interest because these are highly ordered arrays of actin and myosin
filaments with accessory regulatory proteins that dramatically scale the force and deformation that
can be achieved. For example, the peak stress generated by isolated myofibrils from cardiomyocytes
is >100 mN/mm2 (35), which for a typical myofibril 2 μm in diameter is ∼300 nN of force and
a five-order-of-magnitude increase over a single myosin motor. Stress fibers and myofibrils also
have different contractile properties. Stress fibers are found in nonmuscle cells, myofibroblasts,
and smooth muscle cells and generate slow tonic contractions (36–38). By contrast, myofibrils are
found in cardiac and skeletal striated muscle cells and generate rapid, cyclical contractions (32,
35, 39). Currently, there are no engineering methods available to synthetically organize any type
of molecular motor into larger arrays that can achieve the level of force generation achieved by
these contractile filaments. Thus, to use myofibrils and/or stress fibers as an actuator it is necessary
to use the cells that build them. For biological soft robotics, the temporal and spatial actuation
requirements dictate which specific type of muscle cell is needed.
Muscle cells can be categorized into three primary types with properties that differ at both
the cellular and tissue scales: smooth muscle cells, cardiomyocytes, and skeletal myotubes. It
should be noted that these are broad classifications, and within each category is a range of
more specific cell types with differences in structure and contractile function. However, for
all of these muscle cell types, a process termed excitation-contraction coupling translates an
excitatory signal into active contraction and force generation (40, 41). This can occur either via
electromechanical coupling, where an electrical potential depolarizes the membrane to cause

www.annualreviews.org • Biological Soft Robotics 247


BE17CH10-Feinberg ARI 31 October 2015 15:30

contraction, or by pharmacomechanical coupling, where a molecule interacting with a receptor


depolarizes the membrane to cause contraction. Membrane depolarization causes an influx
of Ca2+ from the sarcoplasmic reticulum and extracellular space that results in tropomyosin
changing conformation and enabling myosin to bind actin and crossbridge cycling, i.e., motor
function, to occur. When Ca2+ concentration is returned to baseline levels, which vary by cell
type, crossbridges are disrupted again by tropomyosin, and relaxation occurs. Although this is
an extremely brief overview of muscle contraction (see References 40 and 41 for more detail),
it highlights how excitatory signals can be used to control excitation-contraction coupling in
muscle cells to actuate large numbers of molecular motors in synchrony.
Also important are the different ways these types of muscle cells couple together to form tissues,
and thus larger biological actuators. Smooth muscle cells contain stress fibers that enable these
cells to maintain tonic contractions for long periods of time, which requires these cells to work
together. In vivo a primary example is the medial layer of blood vessels where smooth muscle cell
Annu. Rev. Biomed. Eng. 2015.17:243-265. Downloaded from www.annualreviews.org

contraction changes vessel diameter. Smooth muscle cells are fully differentiated as single cells
and couple together mechanically via adherens junctions and electrically via gap junctions (42).
Access provided by 38.25.16.177 on 10/29/22. For personal use only.

This coupling means that smooth muscle tissue is often a syncytium where electrical signals can
propagate between cells and regulate contraction. Regulation of smooth muscle excitation typically
occurs pharmacologically, such as via stimulation of adrenergic receptors from the nervous system
or via nitrous oxide secretion by endothelial cells in the vasculature. Cardiomyocytes are similar to
smooth muscle cells in some ways, as these fully differentiated cells couple together via adherens
and gap junctions to form a syncytium (32). However, excitation of cardiac muscle typically occurs
electrically, from the sinoatrial node that serves as the pacemaker for the heart or another part of the
cardiac conduction system (43). Furthermore, cardiac muscle contracts constantly at frequencies
that vary from ∼50 to 200 Hz in humans, making this type of muscle ideal in cyclical applications,
but a poor choice when periods of relaxation are required. Finally, skeletal myotubes are large
multinucleated cells that form via the differentiation and fusion of skeletal muscle myoblasts (44).
In vivo myotubes are organized into motor units that are innervated by motor neurons and operate
as a coordinated unit under neural stimulation (5). However, although chemically excited through
the neuromuscular junction in vivo (45), myotubes are easily stimulated electrically in vitro; such
stimulation serves as the preferred method when skeletal muscle is used as an actuator in soft robots.

3. BIOLOGICAL SOFT ROBOTS AT THE SUBCELLULAR


TO CELLULAR SCALE
Biological soft robots from the subcellular to cellular scale use a range of molecular motors to
generate force. The smallest of these devices are only a few molecules in size, and the largest
approach the size of cells at the micrometer length scale. A number of recent reviews have covered
this nano- and microscale area of biological soft robots in great detail, focusing on the engineering
applications of molecular motors (31), the engineering of synthetic molecular machines (46),
bacterial-based microswimmers (47), and the multiscale assembly of molecular motors (19). This
section briefly discusses key examples of these small, soft robotic devices to highlight the main
concepts; readers are referred to the aforementioned reviews for additional examples.
Actuators and devices powered by a single molecular motor or small ensembles of molecular
motors have enabled the engineering of a number of devices at the nanometer and micrometer
length scales. Use of these molecular motors requires that both the motor protein and the
cytoskeletal filament along which it moves to be used in the actuator. They are also generally used
within a microfluidic device or other microfabricated system to control the protein components
and soluble environment to regulate contraction. Of the microtubule motor proteins, kinesin is the

248 Feinberg
BE17CH10-Feinberg ARI 31 October 2015 15:30

one that has been used most widely in devices, in part because this motor has evolved for intracel-
lular transport and movement (48). In a number of devices, the kinesin motors are immobilized on
a surface and then used to move microtubules. For example, kinesin-driven transport can be used
for the directed movement of microtubules along a surface, and in turn, these microtubules can
be biotinylated or otherwise functionalized to carry a payload (49). Theoretical models have also
been developed that suggest a kinesin-coated microsphere could be used to drive fluid flow in a mi-
crofluidic device (50). Myosin motors exist in a variety of isoforms, including nonmuscle myosin II
in stress fibers; muscle myosin II in sarcomeres; and myosin V, which transports cargo along actin
filaments (27, 32, 39, 48, 51). However, myosin II is not possessive, making it challenging to use
in single-motor or small-ensemble applications (19). Myosin V, however, is processive and can
be used in combination with actin filaments similar to the way kinesin and microtubules are used
(52, 53). Actin end-tracking motors can also potentially be used to provide motility to nano- and
microscale objects, similar to the Listeria bacteria (29). In addition to these examples of cytoskeletal
Annu. Rev. Biomed. Eng. 2015.17:243-265. Downloaded from www.annualreviews.org

motor protein-based actuator systems, true molecular-scale robots have also been developed.
Lund et al. (54) have shown that, by adapting the 8–17 DNA enzyme in combination with DNA
Access provided by 38.25.16.177 on 10/29/22. For personal use only.

origami, it is possible create a device that can autonomously carry out sequences of actions such
as start, follow, turn, and stop. Thus, researchers have shown that different types of molecular
motors can be used as actuators; however, application in larger-scale systems that require more
force must use a living cell where coordinated control of large numbers of motors is possible.
Bacteria are highly adaptable single-cell organisms, and these prokaryotic cells can be harnessed
to perform a wide range of complex functions as semiautonomous or remotely controlled biological
soft robots (19, 47, 55). For example, Sokolov et al. (56) demonstrated that the aerobic bacteria
Bacillus subtilis could be used collectively to drive the rotation of submillimeter gears, with the
rotation velocity controlled by the available oxygen. Bacteria with magnetotactic properties can
also be remotely controlled using externally applied magnetic fields. For example, Gonzalez et al.
(57) used microfabricated metallic islands to create local magnetic field gradients to steer bacterial
motion, and Carlsen et al. (58) attached multiple bacteria to a paramagnetic microbead and used
external magnetic fields to steer the swimming direction. The bacteria can also be used as combined
sensors and actuators, as Kim et al. (59) demonstrated that ensembles of Serratia marcescens coupled
to a microbead can follow a chemical gradient through collective behavior, showing the potential
to deliver a payload. Similarly, Steager et al. (60) used the photosensitivity of S. marcescens to
UV light to phototactically control the movement of microfabricated structures over a surface
covered with these cells. These studies highlight the advantage of using a cellular system, as
intrinsic environmental sensing modalities expand the potential capabilities of these microrobotic
systems. Furthermore, bacteria also have useful properties that include the ability to be grown
in large numbers from standard cell lines, the ability to survive extremes in temperature beyond
what mammalian cells can tolerate, and the ability to be genetically modified to augment gene
expression and functionality. The challenge is that, although bacteria can be used in ensembles
to cooperatively scale force and functionality, this is still limited primarily to the microscale.
Achieving similar functionality at the submillimeter scale and beyond requires moving from single-
cell to multicellular systems.

4. MUSCLE-POWERED ACTUATORS FROM THE


MICRO- TO MACROSCALE
Muscle cells are used in biological soft robotics from the micro- to macroscale because of the
intrinsic ability to organize actin and myosin into scalable contractile motor units in the form of
stress fibers and myofibrils. As discussed above, the three primary types of muscle cells are smooth,

www.annualreviews.org • Biological Soft Robotics 249


BE17CH10-Feinberg ARI 31 October 2015 15:30

cardiac, and skeletal, with unique contractile properties as well as differences in how these cells
assemble into large-muscle tissues. To date, smooth muscle has not been used in soft robotics
devices, likely owing to the fact that the slow tonic contraction is not useful for walking, swimming,
or grasping, which typically requires actuation at 0.5 Hz and greater. However, there have been a
number of examples of cardiac (61) and skeletal (20) muscles used in biological soft robots, which
achieve a wide range of functionalities. The challenges in creating muscle-based actuators and soft
robotic devices are similar to those facing the field of tissue engineering, specifically the formation
of dense, innervated, and vascularized muscle in three dimensions. For this reason, many current
examples of muscle-powered devices are based on two-dimensional (2D) engineered muscle sheets
or three-dimensional (3D) constructs that, at a minimum, solve the vascularization challenge by
limiting the thickness so as to be within the diffusion distance of oxygen and that achieve control
of actuation through electrical field stimulation.
Annu. Rev. Biomed. Eng. 2015.17:243-265. Downloaded from www.annualreviews.org

4.1. Devices Based on Two-Dimensional Cantilever and Film Bending


Access provided by 38.25.16.177 on 10/29/22. For personal use only.

Muscle tissue needs to be aligned to generate maximal force, and microfabrication techniques
have been developed to guide muscle cells to form anisotropic 2D tissues on a range of sur-
faces. When the substrate is thin enough, contraction of the 2D muscle tissue causes bending
(increased curvature) of the cell/substrate bilayer. This enables 3D deformation of a 2D tissue
and can serve as an actuator for more complex functionality. For example, muscle cells have been
grown on microfabricated silicon cantilevers, where the narrow width of the cantilever directs cell
alignment longitudinally to create biological microelectromechanical systems (bioMEMS). Rat
neonatal myoblasts and C2C12 myoblasts have been seeded onto these cantilevers and differen-
tiated into myotubes to form contractile microtissues (62–65). These types of muscle bioMEMS
can generate twitch stresses of 0.5 to 5 kPa and can potentially be used as biosensors by measuring
changes in contractility resulting from exposure to environmental toxins (65). Muscle-powered
bioMEMS constructed of silicon and gold have also been used to create small walking robots (66),
demonstrating their utility as an actuator system at submillimeter length scales.
Although silicon cantilevers are viable at the micrometer length scale, they are a suboptimal
material for larger-scale devices because of their high elastic modulus and brittleness. Instead,
softer materials, such as hydrogels and elastomers, are used as these materials match the elastic
modulus of soft tissues, which is in the range of 1–100 kPa (67). For example, Kitamori and
coworkers (68, 69) have used microfabrication techniques to engineer surfaces with micropillars
consisting of polydimethylsiloxane (PDMS) elastomer or polyacrylamide hydrogel (70). They have
demonstrated that adhered cardiomyocytes or smooth muscle cells can generate contractile force to
bend the micropillars, forming surface-bound cantilever arrays. Akiyama et al. (71) used a similar
micropillar array to demonstrate that insect-derived muscle can also be used as a bioactuator
with better long-term viability in vitro than mammalian cells. Using larger PDMS cantilever
arrays on surfaces, Park et al. (72) showed that they can be used to measure the contractility of
adhered cardiomyocytes. In addition, Kim et al. (73) showed that these types of cantilevers can be
integrated as actuators into larger devices to enable walking in millimeter-scale soft robots. More
recently, Bashir and coworkers (74) have used stereolithography to engineer cantilevers from UV
polymerizable polyethylene glycol diacrylate (PEGDA) hydrogels and cultured cardiomyocytes
on these to create biohybrid actuators. This approach has been used to biofabricate entire soft
robots that can walk using integrated cardiac (75) or skeletal (76) muscle.
Larger-scale actuators based on 2D tissues are possible by engineering biohybrid systems where
cell sheets are coupled to carrier films. For example, Feinberg et al. (77) engineered a layer of
anisotropic cardiac muscle on a ∼15-μm-thick layer of PDMS elastomer to form a muscular thin

250 Feinberg
BE17CH10-Feinberg ARI 31 October 2015 15:30

film (MTF) that could be used as a soft robotic actuator system. In addition to being an actuator,
these MTFs were used as sensors to determine contractile function of the integrated cardiac
muscle tissue (78, 79). The MTFs also demonstrated that a range of cell types can be used to take
advantage of the contractile properties of different muscle and nonmuscle cells, including smooth
muscle cells (79, 80), skeletal muscle myotubes (81), embryonic stem cell–derived cardiomyocytes
(82, 83), and endothelial cells (84). Recent work also shows that MTFs are not limited to PDMS
and can be engineered using hydrogels, such as alginate (85) and gelatin (86), to improve long-term
viability of the cells. Although MTFs can be fabricated up to multiple centimeters in length, scaling
beyond this is limited because the muscle tissue layer is only 10–20 μm thick. Larger-scale soft
robotic actuators and devices that require greater force, and thus greater muscle cross-sectional
area, necessitate the engineering of 3D muscle tissue.
Annu. Rev. Biomed. Eng. 2015.17:243-265. Downloaded from www.annualreviews.org

4.2. Three-Dimensional Muscle Bundle–Based Constructs


Larger-scale soft robotics devices at the millimeter and centimeter scale and beyond require
Access provided by 38.25.16.177 on 10/29/22. For personal use only.

3D muscle tissue to generate larger forces. To do this requires biofabrication of muscle using
tissue-engineering techniques, which have been developed using a range of cell types and scaffold
technologies. A number of recent reviews highlight the different approaches for engineering
cardiac (87–89) and skeletal (20, 90) muscle tissue in vitro. Extracellular matrix (ECM) protein
hydrogels are the most common scaffold materials, where muscle cells are embedded within, and
over time, they remodel and compact the hydrogel to form a dense muscle tissue. This has been
used to engineer skeletal muscle tissue from primary avian skeletal muscle cells in collagen type I
gel (91) and has more recently been extended to engineer more mature muscle from C2C12 cells
(92). Other ECM gels, such as fibrin and Matrigel R
, can be used, either alone or in combination,
as demonstrated for engineered cardiac muscle tissues (93, 94). These results establish that it is
possible to engineer linear muscle bundles similar to those found in vivo, with anchor points on
both ends for potential integration in a soft robotic device. Unfortunately, these constructs are still
restricted to a diameter of <1 mm because of oxygen diffusion limitations. Bursac and coworkers
(95–97) have addressed this challenge by engineering thick muscle sheets with integrated pores to
provide oxygen and nutrient transport. To do this, they fabricated specially designed molds with
rectangular posts to guide alignment in to which they cast collagen type I, fibrin, and/or Matrigel
embedded with cardiomyocytes or skeletal muscle myoblasts that differentiated into myotubes
(95, 96). The rectangular posts were tuned in size and shape to maximize alignment and mass
transport (97), and the muscle constructs generated substantially greater forces owing to the large
increase in muscle mass compared to 2D engineered muscle tissues. However, overall thickness
was still limited to <1 mm.
Three-dimensional muscle tissue can also be engineered without using a scaffold, relying in-
stead on the constituent cells to synthesize and assemble the ECM components. For example,
self-organizing muscle tissue, often referred to as a myooid, starts as a 2D system with muscle
cells cultured on the bottom of a culture dish. Over time, the 2D tissue layer peels off of the dish
but remains anchored to two supports, and then slowly compacts into a 3D muscle bundle (98).
This has now been demonstrated using skeletal (99, 100) and cardiac (101) muscle cells and can
even be used to create an interface between muscle tissue and a self-organized tendon-like tissue
(102), potentially beneficial for integration into a robotic system. The limitations of the myooid
construct are comparable to those of the ECM gel-based scaffolds, i.e., limited in diameter to
<1 mm owing to the lack of a vascular system, and have only been reported as linear muscle
bundles. Another approach to engineering muscle tissue is the stacking of 2D muscle cell sheets
into thicker 3D muscle constructs. Okano and coworkers (103) have shown that muscle tissue can

www.annualreviews.org • Biological Soft Robotics 251


BE17CH10-Feinberg ARI 31 October 2015 15:30

be cultured on petri dishes coated with thermally responsive poly(N-isopropylacrylamide), which


when cooled from 37◦ C to room temperature releases the adhered monolayer as an intact tissue
layer. Once the muscle sheets are released, they are stacked, and then the cells couple together to
form an integrated tissue, as has been shown for cardiac muscle (104) as well as skeletal muscle
with alternating layers of endothelial cells (105). Scaffold-free engineered muscle tissues have also
been assembled directly in three dimensions. For example, C2C12 myoblasts that have endocy-
tosed magnetic nanoparticles can be engineered into muscle rings by using a donut-shaped mold
in combination with an applied magnetic field to hold the cells densely together (106). In addi-
tion, cardiac muscle tissue patches have been engineered by growing muscle cells in suspension
on a rotating orbital shaker (107). For application in biological soft robotics, these scaffold-free
techniques can produce very dense muscle tissue that should have high force-to-weight ratios.
However, thicknesses are typically limited to ∼200 μm, and always <1 mm, owing to the diffusion
limit of oxygen and lack of a vasculature.
Annu. Rev. Biomed. Eng. 2015.17:243-265. Downloaded from www.annualreviews.org

5. BIOLOGICAL SOFT ROBOTS WITH COMPLEX FUNCTIONALITY


Access provided by 38.25.16.177 on 10/29/22. For personal use only.

POWERED BY MUSCLE
Complex functionality in a biological soft robot is defined here as moving beyond simple movement
of a molecular motor-based actuator or contraction of a muscle construct to a more significant
interaction with the environment. Specifically, this section discusses examples that can pump,
grasp, walk, and swim and the muscle cells and system designs that have enabled these advances.
These examples demonstrate the potential to engineer biological soft robots with functional muscle
tissue that can achieve truly unique behaviors. Table 1 lists a number of the devices that have
been reported to date, comparing the muscle type used as the actuator, the biological structure
being mimicked by the overall design, and the primary performance metric for the system. Most of
the examples presented use cardiac muscle as the actuator system, rather than skeletal or smooth
muscle. The reason for this is that the standard neonatal rat cardiomyocytes used in these studies are
plastic cells, meaning that they readily respond to engineered environments to form muscle tissue,
yet still form interconnected, highly differentiated muscle tissue. However, one of the downsides
is that these cardiomyocytes are primary harvested and thus are not practical for scaling up, where
a cell line would be superior. Thus, these different examples of biological soft robots should be
viewed in terms of potential applications and design strategies, with the realization that the specific
cells used will eventually be transitioned to cell lines, and even different muscle types, as the field
progresses.

5.1. Muscle-Powered Pumps


Some of the simplest examples of biological soft robots powered by muscle tissue are pumps that
drive fluid flow as a result of cyclical muscle contraction. These typically operate as a component
in a microfluidic system in combination with valves that maintain unidirectional flow (Table 1).
For example, Tanaka et al. (104) have used 2D cardiac cell sheets, engineered using a thermally
sensitive poly(N-isopropylacrylamide)-coated petri dish, to serve as the actuator system to pump
fluid through a microfluidic system. In one case, a diaphragm pump was developed with a pusher
plate driven by contraction of an integrated cardiac muscle sheet, similar in basic overall design to
traditional microfluidic pumps (108). A later example used a cardiac muscle sheet wrapped around
an elastomeric sphere, creating a pump that mimics the structure and pumping behavior of a heart
ventricle (109). Park et al. (110) have also fabricated a microfluidic pump, using a thin membrane
covered by a layer of cardiomyocytes as a diaphragm to drive flow in a microfluidic channel.
Although these examples demonstrate feasibility, such devices are on the millimeter length scale yet

252 Feinberg
BE17CH10-Feinberg ARI 31 October 2015 15:30

Table 1 Types of biological soft robots powered by muscle tissue and key performance metrics
Robot Biological Synthetic
type Mimics component component Size Performance Reference
Pump Heart 2D cardiac muscle PDMS 5 mm 0.047 µL min−1 Tanaka et al.
(109)
Diaphragm 2D cardiac muscle PDMS ∼1 mm 0.24 µL min−1 Tanaka et al.
(108)
Diaphragm 2D cardiac muscle PDMS ∼1 mm 0.226 nL min−1 Park et al.
(110)
Gripper Claw 2D cardiac muscle PDMS (film) ∼5 mm ∼4-mm tip displacement Feinberg et al.
(77)
Claw Skeletal myotube Dog Hair 100 µm 0.6-µm tip displacement Hoshino &
(ion milled) Morishima
Annu. Rev. Biomed. Eng. 2015.17:243-265. Downloaded from www.annualreviews.org

(111)
Claw 3D skeletal muscle PDMS (frame) 6 mm ∼10-µm tip displacement Kabumoto
Access provided by 38.25.16.177 on 10/29/22. For personal use only.

et al. (112)
Claw Insect dorsal vessel PDMS (frame) 3 mm ∼15-µm tip displacement Akiyama et al.
tissue (113)
Walker Inchworm Cardiac muscle Au, Si (film) 138 µm 38 µm s−1 Xi et al. (66)
(stick slip) bundle ∼0.25 body lengths s−1
Inchworm 2D cardiac muscle PDMS (film) ∼3 mm 130 µm s−1 Feinberg et al.
(stick slip) ∼0.045 body lengths s−1 (77)
Inchworm 3D skeletal muscle PEGDA 6 mm 156 µm s−1 Cvetkovic
(stick slip) (3D printed) ∼0.025 body lengths s−1 et al. (76)
Inchworm 2D cardiac muscle PEGDA 7 mm 236 µm s−1 Chan et al.
(stick slip) (3D printed) ∼0.035 body lengths s−1 (75)
Crab like 2D cardiac muscle PDMS (film) ∼2 mm 100 µm s−1 Kim et al. (73)
∼0.05 body lengths s−1
Inchworm Insect dorsal vessel PDMS (frame) ∼10 mm 3.5 µm s−1 Akiyama et al.
(stick slip) tissue ∼0.00035 body lengths s−1 (114)
Inchworm Insect dorsal vessel PDMS (frame) ∼5 mm 257 µm s−1 Akiyama et al.
(stick slip) tissue 532 µm s−1 (stimulated) (115)
0.05 to 0.1 body lengths s−1
Swimmer Eel 2D cardiac muscle PDMS (film) ∼6 mm 400 µm s−1 Feinberg et al.
∼0.07 body lengths s−1 (77)
Jellyfish 2D cardiac muscle PDMS (film) ∼8 mm 1.5 mm s−1 Nawroth et al.
∼0.2 body lengths s−1 (14)
Sperm Cardiomyocyte PDMS (single ∼2 mm 5–10 µm s−1 Williams et al.
cluster filament) 0.0025 to 0.005 body (116)
lengths s−1
Sperm Cardiomyocyte PDMS (dual ∼2 mm 81 µm s−1 Williams et al.
cluster filament) ∼0.04 body lengths s−1 (116)

Abbreviations: 2D, two-dimensional; 3D, three-dimensional; PDMS, polydimethylsiloxane; PEGDA, polyethylene glycol diacrylate.

www.annualreviews.org • Biological Soft Robotics 253


BE17CH10-Feinberg ARI 31 October 2015 15:30

generate flow rates that are in the range of 100 nL min−1 , or much less. Performance improvements
are needed for these pumps to be useful beyond proof-of-concept, such as aligning the muscle
tissues to increase uniaxial force generation.

5.2. Muscle-Powered Grippers


The purpose of a gripper is to enable soft robotic devices to manipulate objects directly. A number
of soft robotic designs have been implemented using synthetic actuators (16), but very few examples
have been demonstrated using muscle-based actuators. A number of proof-of-concept soft robotic
devices with basic functionality have been created using engineered muscle tissue. In general, these
devices are fabricated from hydrogels or elastomers that provide a flexible structural framework
and then use the integrated muscle tissue to contract and cause deformation of part of the system
(Table 1). For example, at the micrometer scale, a compliant gripper was fabricated by using a
Annu. Rev. Biomed. Eng. 2015.17:243-265. Downloaded from www.annualreviews.org

focused ion beam to mill a dog hair into a pair of tweezers and attaching skeletal muscle myotubes
to actuate the device (111). At the millimeter scale, grippers have been fabricated using PDMS to
Access provided by 38.25.16.177 on 10/29/22. For personal use only.

form a compliant frame. Kabumoto et al. (112) showed that a PDMS gripper construct, with an
integrated strand of engineered skeletal muscle, could open when field stimulated. Feinberg et al.
(77) demonstrated that grippers can also be engineered from cardiac MTFs and established that the
stimulation rate could be used to modulate the open or closed state of the gripper. These examples
all used mammalian cells. To increase long-term viability, Akiyama et al. (113) engineered a gripper
actuated by insect dorsal vessel muscle tissue. Not only did this gripper operate for 5 days at room
temperature, but also by sealing the muscle tissue and media, the protruding gripper was able to
function in air. As a group, for these grippers to be useful more broadly, they need to achieve
larger tip displacements and larger forces so they can be used for actual grasping of objects. Also,
they need to be integrated into a larger soft robotic device that can position the grippers in 3D
space so that they can function as part of an appendage.

5.3. Robots That Walk or Crawl


The ability to engineer a macroscale biological soft robot that can move on its own has been a
long-term objective, and recent advances have achieved proof-of-concept devices that demonstrate
feasibility (Table 1). For example, Xi et al. (66) microfabricated walking robots from gold and
silicon and used cardiac muscle strands as the actuator. These were not entirely soft robots because
the silicon and gold frame was rigid; however, it demonstrated that engineered muscle could be
used to power walking at the microscale. At the millimeter scale, PDMS has been widely used to
create microfabricated frames that can be integrated with muscle. For example, Kim et al. (73)
developed a PDMS-based walker engineered from microfabricated cantilevers and cardiac muscle
that could walk along the bottom of a petri dish. Feinberg et al. (77) developed a similar device
from cardiac MTFs, termed a myopod, which also walked directionally. Bashir and coworkers (74)
expanded on these examples by using stereolithography to rapidly fabricate photocross-linkable
polyethylene glycol into different hydrogel frame designs, integrated with muscle tissue. They
have used this technology to engineer “bio-bots” that walk and are powered by both cardiac
(75) and skeletal muscle tissue (76). Walking soft robots have also been engineered using insect
muscle tissue to improve the environmental tolerance of the biological components. Akiyama et al.
(114) demonstrated the attachment of insect dorsal ventral muscle tissue to a PDMS frame that
spontaneously walked and could be stimulated to walk faster through the addition of a neuroactive
chemical to the media (115). It should be noted that all of these examples of walking biological soft
robots require certain conditions to function properly. These conditions include (a) maintenance

254 Feinberg
BE17CH10-Feinberg ARI 31 October 2015 15:30

at all times in cell culture media to provide an energy source and hydration, (b) an externally
applied electric field to stimulate contraction to walk at a constant velocity, and (c) a surface on
which a stick-slip walking motion can occur. It is also important to note that most of these walkers
are 1 to 10 mm in size, yet they walk at a rate of <1 mm s−1 . Thus improvements are needed
to enhance performance, including increased muscle mass to increase velocity as well as robot
designs that enable these walkers to operate outside cell culture media and in ambient conditions.

5.4. Swimming Robots


Biological soft robots that can swim have seen some of the most advanced functionality yet demon-
strated. Similar to the muscle-powered walkers, these swimmers operate within cell culture media
for nutrient supply and are electrically field stimulated to optimize contraction and swimming
speed (Table 1). The simplest swimmers have been engineered from cardiac MTFs shaped as
Annu. Rev. Biomed. Eng. 2015.17:243-265. Downloaded from www.annualreviews.org

isosceles triangles with the muscle aligned base to tip, which produces an anguilliform swimming
motion (77). Importantly, control of the stimulation rate was used to optimize deformation of
the construct and achieve a near constant swimming velocity, similar to real fish. Cardiac MTFs
Access provided by 38.25.16.177 on 10/29/22. For personal use only.

have also been used to engineer more complex swimming soft robots. Nawroth et al. (14) used
advanced imaging of a juvenile jellyfish to measure the body plan and architecture of the muscle
tissue and then used cardiac MTFs to engineer a biomimetic soft robot termed a medusoid. They
showed that properly recapitulating both the body plan and the tissue architecture were critical
to achieving a swimming behavior that closely matched that of the live animal. This is an im-
portant example that soft robots can mimic real organisms, even when using synthetic materials
and different cell types. Williams et al. (116) have demonstrated that this biomimetic approach
can also be used to recapitulate the swimming of sperm using flagella to produce thrust. Here
the body plan of the sperm was mimicked but scaled up from a length of ∼55 μm to 2 mm.
The flagella in the sperm are powered by dynein motors, but in the engineered swimmer, a small
cluster of cardiomyocytes at the base of the PDMS flagella powered the swimmer. Furthermore,
though not found in nature, adding two of the tails to the same device increased swimming ve-
locity through coordinated actuation. All of these PDMS-based swimmers have used engineered
cardiomyocytes to power them, but the limited size of engineered tissues has restricted the overall
size of these devices to the millimeter length scale. Larger muscle-powered swimmers have been
achieved by taking explanted semitendinosus muscles from frogs and attaching them to either side
of the flexible tail of a centimeter-scale robot (117). Alternating electrical stimulation of the two
muscles resulted in flapping of the tail and controlled swimming rates. What this demonstrates
is that a larger muscle results in larger robots that can function and generate larger forces. How-
ever, explanted muscle at this scale has a limited lifetime owing to the lack of a blood supply,
and as mentioned above, current tissue-engineering techniques cannot fabricate viable muscle at
this scale. The larger muscle-powered swimmer also demonstrates that selective stimulation of
different muscle bundles can enhance functionality, which could conceptually enable more ad-
vanced biomimetic robots using even 2D muscle actuators (Figure 2a). Thus, there is potential to
improve capabilities, but fabrication challenges remain before these technologies can be scaled up.

6. CURRENT AND LONG-TERM CHALLENGES


AND POTENTIAL SOLUTIONS

6.1. Sourcing Molecular Motors and Cells for Building Devices


Building biological soft robotic devices requires a source of the molecular motors and/or cells
and tissues to power these systems. The challenges in doing this lie in the structural complexity

www.annualreviews.org • Biological Soft Robotics 255


BE17CH10-Feinberg ARI 31 October 2015 15:30

a Controller for individual


b Top view Side view
limb actuation Relaxed
Patterned
smooth
muscle Low
“innervation”

Patterned
myotubes
High
“innervation”

Electrode
“innervation”

Joints
Computer- • Tonic contraction
(smooth muscle)
Annu. Rev. Biomed. Eng. 2015.17:243-265. Downloaded from www.annualreviews.org

controlled
actuations • Tetanus
(skeletal muscle)
Access provided by 38.25.16.177 on 10/29/22. For personal use only.

Figure 2
(a) Schematic of a soft robot with four muscular limbs (MTFs) that are independently controlled via stimulation from a microcontroller
to enable complex grasping, walking, and swimming behaviors. (b) Schematic of a combined skeletal and smooth muscle MTF with
microelectrodes and computer-controlled actuation for complex deformation. The selective stimulation of the patterned muscle
bundles permits changes in magnitude and position of the force generated. This enables changes in overall curvature as well as
formation of variable curvature along the length of the MTF, e.g., to create a temporary joint. Abbreviation: MTF, muscular thin film.

of even the simplest molecular motor and the limited number of methods that currently exist
to synthetically build these macromolecular assemblies (46). In the cell, molecular motors are
assembled by the ribosomes and then transported to specific locations throughout the cell for direct
action or assembly into larger-scale structures. The ribosome itself is a highly complex machine
specialized to translate mRNA into proteins, and to date, this activity has not been replicated
in vitro (118). This means we need to use cells to synthesize, assemble, and package molecular
motors. For research purposes, many studies have directly isolated cytoskeletal filament proteins,
such as tubulin from cow brains, and molecular motors, such as kinesin, can be recombinantly
expressed in bacteria (49). However, the costs are quite high, and the overall supply is limited,
especially if the goal is to scale up beyond a proof of concept in the lab.
The challenge of sourcing molecular motors increases when larger-scale assemblies are re-
quired. As described above, cardiac muscle and skeletal muscle are based on sarcomeres organized
into myofibrils that contain actin, myosin, titin, α-actinin, and a range of other cytoskeletal pro-
teins that provide structural and mechanical function. Assembling these proteins into myofibrils
outside of a cell is beyond the capability of current technology. Thus, cells need to assemble these
and the other contractile machinery inside themselves, including calcium handling and cellular
respiration to generate ATP. Mammalian cell culture at a large scale is possible (119) and is rou-
tinely done in the pharmaceutical industry for high-value therapeutics (120). However, growing
large numbers of cells for biological soft robotics is unproven and thus is costly because of the
media, serum supplementation, and sterility requirements. Furthermore, most muscle cells are not
amenable to long-term culture and need to be harvested from animals, including insect muscle tis-
sue, which has improved viability in vitro (121, 122). That adds an additional cost as well as ethical
and moral issues associated with scale-up of primary harvested tissues for use in robotics. Finally,
nearly all muscle cell types can be differentiated from embryonic and induced pluripotent stem
cells (123, 124), which are the primary sources of cells for many medical applications of engineered
muscle tissues (125). However, the cost of large-scale production using these cells is even higher

256 Feinberg
BE17CH10-Feinberg ARI 31 October 2015 15:30

and problematic owing to (a) the increased media and supplement costs of culturing pluripotent
stem cells, (b) the increased culture time needed for the specific differentiation protocols, (c) the
difficulty in achieving homogenous differentiation of a specific muscle cell type of interest, and
(d ) the contractile immaturity of stem cell–derived muscle cells and engineered tissues (82).

6.2. Limitations of Biological Soft Robots Based on Muscle Cell Type


The muscle cells that constitute smooth, skeletal, and cardiac tissue types have specific contractile
behaviors that place constraints on the kinds of actuators that can be fabricated and their achievable
performance. For example, cardiomyocytes have the advantage of being able to rapidly, electrome-
chanically couple and form a syncytium that can be large, on the centimeter length scale (14, 77).
The downside is that cardiomyocytes have evolved to contract continuously and cannot effectively
enter a resting state. The minimum contraction frequency is approximately the resting heart rate
Annu. Rev. Biomed. Eng. 2015.17:243-265. Downloaded from www.annualreviews.org

of the animal from which the cells are derived, or about 1 Hz for humans. There is also a maximum
contraction frequency in the range of 3 to 4 Hz for humans. Although this heart rate can vary by
Access provided by 38.25.16.177 on 10/29/22. For personal use only.

the developmental stage and species of the cardiomyocytes, there is still a relatively narrow tem-
poral range. An additional issue is operational temperature, as both the minimum and maximum
contraction frequency are decreased at ambient temperature (∼22◦ C) compared to body tem-
perature (37◦ C) (78, 83). This means that cardiomyocytes are appropriate for applications where
actuation rates and operational temperatures are within physiologic ranges but are a less than
ideal choice otherwise. Other mammalian muscle cells have similar temperature requirements but
offer different contractile properties. Skeletal muscle has evolved to operate over a much wider
temporal range from rest up to 100 Hz (tetanus). However, it is more difficult to form cellular
connections, which require strategies to synchronously actuate large numbers of myotubes either
through field stimulation (81) or by mimicking motor neuron innervation using microelectrodes
(126). Furthermore, engineered skeletal muscle tissue needs to be seeded as muscle stem cells or
myoblasts that then differentiate and fuse into myotubes. This extra differentiation step (compared
to cardiac and smooth muscles) often results in a subpopulation of undifferentiated cells in the final
engineered tissue that reduces overall muscle volume. Finally, smooth muscle cells are adapted
to contract slowly and hold a given contractile state for long periods of time relative to striated
muscle (127). The biological soft robotics devices reviewed in Table 1 all need to operate at ∼1 Hz
to function, which perhaps is why there are no examples of these types of devices using smooth
muscle. However, looking to the future, it is likely that researchers will start incorporating smooth
muscle to maintain changes in body shape for longer periods of time. For example, integrating
skeletal and smooth muscle in the same soft robot can enable tonic contraction in some locations
(such as for a temporary joint) and rapid contractions in other locations for motility (Figure 2b).

6.3. Long-Term Maintenance of Biological Components In Vitro


Biological molecules have a finite lifetime before they degrade to the point that they can no
longer perform their function. The time frame for this depends on a number of factors, including
mechanical, thermal, or light-induced damage as well as chemical degradation from enzymes or
reactive oxygen species. In the cell, molecular motors are replaced on a regular interval as part
of cellular homeostasis. Currently, it is not possible to implement this type of molecular repair
mechanism in a cell-free in vitro environment because of the complex intracellular signaling
networks that regulate this process. Furthermore, at the cell level, bacterial robotic systems have
the challenge of dealing with the life cycle of the bacteria as they normally divide in a relatively
short time frame. To date, most experiments using bacteria have been short lived, but if they

www.annualreviews.org • Biological Soft Robotics 257


BE17CH10-Feinberg ARI 31 October 2015 15:30

are ever to be used over longer time periods, this needs to be resolved. For mammalian cells,
the problem of cell division is not an issue for cardiomyocytes or skeletal myotubes, which are
terminally differentiated, but the issue of viability remains. These cells have stringent requirements
for temperature and sterility as well as a range of growth factors and other signaling molecules and
cell types (paracrine signaling) to maintain viability. Biological soft robots engineered from cardiac
muscle are viable only for a few weeks to a month (77), although engineering softer substrates
may prolong this somewhat (86). Fortunately, there are a wide range of nonmammalian animals
that have demonstrated the ability to exist in harsher environments. For example, Baryshyan et al.
(122) have developed methods to harvest insect muscles that were contractile but metabolically less
demanding than C2C12 cells. They also showed that these insect muscle cells could be engineered
into muscle bundles that could self-repair, survive for months without nutrient replenishment,
and produce contractile stresses of up to 2 kPa at ambient conditions (121). In addition, the
explanted insect muscle can survive extremes in temperature and pH for multiple days. The
Annu. Rev. Biomed. Eng. 2015.17:243-265. Downloaded from www.annualreviews.org

relatively long-term viability of insect muscle at ambient conditions is demonstrated by the gripper,
developed by Akiyama et al. (113), that functions for 5 days. Moreover, the insect muscle can be
Access provided by 38.25.16.177 on 10/29/22. For personal use only.

controlled through electrical stimulation to regulate contraction frequency, comparable to the


control schemes for mammalian muscle (128). What is needed moving forward is an insect muscle
cell line that can be expanded into a differentiated muscle on demand, eliminating the current
need to primary harvest the tissue.

6.4. Energy Storage, Supply Systems, and Handling of Metabolic Waste


Although standard robotic systems are primarily powered by electricity, biological soft robotics
systems are powered by chemical energy. As noted, there are advantages to chemical energy
storage, including high energy density; however, there are also specific challenges that need to be
addressed to maximize performance. First, molecular motors are all powered by ATP or GTP,
but these molecules can be used as is only in cell-free nano- or microscale devices. For larger
devices containing cells, ATP cannot be used directly as a supplement in the media because it can
act as an extracellular purinergic signaling molecule (129). Instead, the best energy sources for
cells are molecules that are normally broken down from food and used by cells during respiration,
including sugar, proteins, and fatty acids. For example, glucose is a common energy substrate for
eukaryotic cells used in soft robotics because it can be taken up from the media, and through
glycolysis, the citric acid cycle and oxidative phosphorylation can convert 30 ADP molecules back
into ATP. Even though many cell types can operate under anaerobic conditions (which generate
significantly less ATP), aerobic respiration is required for the high-energy demands of muscle
cells; thus, oxygen supply is a critical component to the design of biological soft robotics and the
systems in which they operate.
The need for oxygen as well as mass transport of nutrients and metabolic waste means that as
biological soft robotic devices become larger they need to be vascularized. Striated cardiac muscle
in vivo is highly vascularized with a capillary density of 2,000 to 3,000 per mm2 (130), highlighting
the metabolic demands that must be met. Current muscle constructs are not vascularized and, as
mentioned, are either 2D sheets or thicker constructs <1 mm in thickness. Unfortunately, creating
a functional vascular network within an engineered muscle tissue is a challenge that faces the entire
field of tissue engineering and still requires significant development before satisfactory results can
be achieved (131). It is possible to vascularize engineered muscle constructs by implanting them
in vivo in animals and using the host response to drive angiogenesis (132, 133). However, financial
and ethical concerns prevent the use of animals as living incubators for this type of nonmedical
application. The field of engineered vascular networks is advancing, with recent achievements in

258 Feinberg
BE17CH10-Feinberg ARI 31 October 2015 15:30

microfabrication and 3D printing demonstrating that it is possible to build microvessel networks


in vitro (134, 135). Thus, once these approaches mature, it should be possible to apply these
techniques to a 3D-engineered muscle construct with integrated microvasculature.

6.5. Motor and Muscle Control Systems for Selective


and Coordinated Actuation
A primary advantage of biological actuators, such as skeletal muscle, is the ability to modulate the
rate and force of contraction for a specific application. At the level of a muscle cell, contraction
is an all-or-nothing event with the rate of contraction determined by stimulation frequency.
For skeletal muscle, modulation of force is achieved through the contraction rate and spatial
recruitment of muscle fibers (100). Using field stimulation, controlling the stimulation rate of the
whole muscle tissue construct is straightforward, but controlling the selective recruitment of indi-
Annu. Rev. Biomed. Eng. 2015.17:243-265. Downloaded from www.annualreviews.org

vidual myofibers to modulate force is more complicated. Two potential solutions to this are the use
of microelectrodes (126, 136) or optogenetics (137) to activate specific myotubes (an example of the
Access provided by 38.25.16.177 on 10/29/22. For personal use only.

former is illustrated in Figure 2b), but these technologies require significant development before
they can be employed in more complex soft robotic devices. For example, channelrhodopsin-2, a
light-sensitive cation channel, transfected into muscle cells can cause depolarization and trigger
Ca2+ influx into the cytosol to initiate contraction. This was shown in C2C12 myotubes differen-
tiated in a collagen gel between PDMS pillars to produce contractile skeletal muscle controlled
by light pulses (137). Another option is to tissue engineer motor neuron networks into the muscle
tissue constructs, but this requires successfully recreating the neuromuscular junction, the inter-
face between the nerve and the muscle fiber. The potential to form a neuromuscular junction has
been shown between cocultured neurons and muscle cells in vitro (138, 139). But controlling this
process is difficult in two dimensions and increasingly challenging for 3D tissues. Furthermore,
even if innervation is achieved, the challenge then becomes stimulating the motor neurons instead.
Looking forward, researchers need to develop improved methods for controlling actuation of
motor units within a biological soft robot through patterned microelectrodes (Figure 2b)
or other approaches, as more complex functionality requires modulation of force and more
precise spatial control than that afforded by current field stimulation protocols.

7. CONCLUSIONS AND FUTURE DIRECTIONS


This review has presented an overview of molecular motors and the ways they are engineered
into biological soft robotic systems from the nanometer to centimeter length scales. Although
impressive proof-of-concept experiments have been demonstrated from nanoscale machines, to
bacteria actuated devices, to muscle-powered robots that can walk and swim, much work remains
as the field looks to improve performance and function. However, the potential for this technol-
ogy is exciting. For example, the tissue-engineered jellyfish (medusoid) by Nawroth et al. (14)
demonstrates that biomimetic design and advanced tissue-engineering techniques can replicate
the swimming motion of multicellular marine animals. The medusoid could swim, but it did not
perform more complex behavior such as feeding, where individual lappets are actuated. This high-
lights a major limitation of even the most complex biological soft robots to date: They perform a
single function and are not adaptable in their environment. Investigators also need denser muscle
that is vascularized or has an integrated fluidics network for oxygen and nutrient mass transport.
Looking to future directions, the most exciting aspect of biological soft robots is complexity.
Current muscle-powered robots have been engineered to perform a single function (Table 1).
However, the promise of robotics lies in complex systems that can perform useful tasks, and this

www.annualreviews.org • Biological Soft Robotics 259


BE17CH10-Feinberg ARI 31 October 2015 15:30

requires the development of multimodal robots with sensing. For example, possible multimodal
robots to be developed in the future include (a) a biological soft robot that can both walk and
swim, providing greater mobility in complex environments; (b) individual actuation of each arm of
Nawroth et al.’s medusoid (14), so that in addition to swimming it could walk or crawl similar to
an octopus (140); (c) integration of light, temperature, and/or chemical sensors so that the robot
can respond to environment cues; and (d ) the two-tailed PDMS swimmer of Williams et al. (116)
with the addition of a directional light sensor that could control which of the tails beat to steer
the robots toward or away from the light, enabling phototaxis. These types of advances are on
the horizon, but details on how they are to be implemented are unknown. Researchers need to
determine which structural and functional components of biological soft robots are fabricated from
living and synthetic components and how they are to be integrated. In terms of realistic translation,
these devices, or the technologies developed to create them, may have the most immediate impact
on bioprosthetics, where the drive to restore function to a lost limb necessitates a biomimetic,
Annu. Rev. Biomed. Eng. 2015.17:243-265. Downloaded from www.annualreviews.org

low-power system.
Access provided by 38.25.16.177 on 10/29/22. For personal use only.

DISCLOSURE STATEMENT
A.W.F. is a coinventor on US Patent 8,492,150 related to the muscular thin film technology.

LITERATURE CITED
1. Bustamante C, Chemla YR, Forde NR, Izhaky D. 2004. Mechanical processes in biochemistry. Annu.
Rev. Biochem. 73:705–48
2. Gelles J, Landick R. 1998. RNA polymerase as a molecular motor. Cell 93:13–16
3. Berg HC. 2003. The rotary motor of bacterial flagella. Annu. Rev. Biochem. 72:19–54
4. de Tombe PP. 2003. Cardiac myofilaments: mechanics and regulation. J. Biomech. 36:721–30
5. Kiehn O. 2006. Locomotor circuits in the mammalian spinal cord. Annu. Rev. Neurosci. 29:279–306
6. Brady AJ, Tan ST, Ricchiuti NV. 1979. Contractile force measured in unskinned isolated adult rat heart
fibres. Nature 282:728–29
7. Nishimura S, Yasuda S, Katoh M, Yamada KP, Yamashita H, et al. 2004. Single cell mechanics of rat
cardiomyocytes under isometric, unloaded, and physiologically loaded conditions. Am. J. Physiol. Heart
Circ. Physiol. 287:H196–202
8. Madden JDW, Vandesteeg NA, Anquetil PA, Madden PGA, Takshi A, et al. 2004. Artificial muscle
technology: physical principles and naval prospects. IEEE J. Ocean. Eng. 29:706–28
9. Van Ham R, Sugar TG, Vanderborght B, Hollander KW, Lefeber D. 2009. Compliant actuator designs
review of actuators with passive adjustable compliance/controllable stiffness for robotic applications.
IEEE Robot. Autom. Mag. 16:81–94
10. Pette D, Vrbová G. 1985. Invited review: neural control of phenotypic expression in mammalian muscle
fibers. Muscle Nerve 8:676–89
11. Kim S, Laschi C, Trimmer B. 2013. Soft robotics: a bioinspired evolution in robotics. Trends Biotechnol.
31:287–94
12. Trivedi DRC, Kier WK, Walker ID. 2008. Soft robotics: biological inspiration, state of the art, and
future research. Appl. Bionics Biomech. 5:99–117
13. Behkam B, Sitti M. 2007. Bacterial flagella-based propulsion and on/off motion control of microscale
objects. Appl. Phys. Lett. 90:023902–3
14. Nawroth JC, Lee H, Feinberg AW, Ripplinger CM, McCain ML, et al. 2012. A tissue-engineered
jellyfish with biomimetic propulsion. Nat. Biotechnol. 30:792–97
15. Tolley MT, Shephard RF, Mosadegh B, Galloway KC, Wehner M, et al. 2014. A resilient, untethered
soft robot. Soft Robot. 1:213–23
16. Ilievski F, Mazzeo AD, Shepherd RE, Chen X, Whitesides GM. 2011. Soft robotics for chemists. Angew.
Chem. Int. Ed. 50:1890–95

260 Feinberg
BE17CH10-Feinberg ARI 31 October 2015 15:30

17. Nawroth JC, Parker KK. 2013. Design standards for engineered tissues. Biotechnol. Adv. 31:632–37
18. Pfeifer R, Lungarella M, Iida F. 2007. Self-organization, embodiment, and biologically inspired robotics.
Science 318:1088–93
19. Chan V, Asada HH, Bashir R. 2014. Utilization and control of bioactuators across multiple length scales.
Lab Chip 14:653–70
20. Duffy RM, Feinberg AW. 2014. Engineered skeletal muscle tissue for soft robotics: fabrication strategies,
current applications, and future challenges. Wiley Interdiscip. Rev. Nanomed. Nanobiotechnol. 6:178–95
21. Mallik R, Gross SP. 2004. Molecular motors: strategies to get along. Curr. Biol. 14:R971–82
22. Kolomeisky AB, Fisher ME. 2007. Molecular motors: a theorist’s perspective. Annu. Rev. Phys. Chem.
58:675–95
23. Hirokawa N. 1998. Kinesin and dynein superfamily proteins and the mechanism of organelle transport.
Science 279:519–26
24. Sengupta S, Spiering MM, Dey KK, Duan WT, Patra D, et al. 2014. DNA polymerase as a molecular
motor and pump. ACS Nano 8:2410–18
Annu. Rev. Biomed. Eng. 2015.17:243-265. Downloaded from www.annualreviews.org

25. Vale RD. 2003. The molecular motor toolbox for intracellular transport. Cell 112:467–80
26. Goldstein LSB, Yang ZH. 2000. Microtubule-based transport systems in neurons: the roles of kinesins
and dyneins. Annu. Rev. Neurosci. 23:39–71
Access provided by 38.25.16.177 on 10/29/22. For personal use only.

27. Sweeney HL, Houdusse A. 2010. Structural and functional insights into the myosin motor mechanism.
Annu. Rev. Biophys. 39:539–57
28. Dickinson RB, Caro L, Purich DL. 2004. Force generation by cytoskeletal filament end-tracking pro-
teins. Biophys. J. 87:2838–54
29. Zeile WL, Zhang FL, Dickinson RB, Purich DL. 2005. Listeria’s right-handed helical rocket-tail tra-
jectories: mechanistic implications for force generation in actin-based motility. Cell Motil. Cytoskelet.
60:121–28
30. Fisher ME, Kolomeisky AB. 1999. The force exerted by a molecular motor. PNAS 96:6597–602
31. Hess H. 2011. Engineering applications of biomolecular motors. Annu. Rev. Biomed. Eng. 13:429–50
32. Bers DM. 2001. Excitation-Contraction Coupling and Cardiac Contractile Force. Boston: Kluwer Acad.
427 pp.
33. Kreis TE, Birchmeier W. 1980. Stress fiber sarcomeres of fibroblasts are contractile. Cell 22:555–61
34. Bond M, Somlyo AV. 1982. Dense bodies and actin polarity in vertebrate smooth muscle. J. Cell Biol.
95:403–13
35. Piroddi N, Belus A, Scellini B, Tesi C, Giunti G, et al. 2007. Tension generation and relaxation in single
myofibrils from human atrial and ventricular myocardium. Pflugers Arch. Eur. J. Physiol. 454:63–73
36. Katoh K, Kano Y, Masuda M, Onishi H, Fujiwara K. 1998. Isolation and contraction of the stress fiber.
Mol. Biol. Cell 9:1919–38
37. Hinz B. 2010. The myofibroblast: paradigm for a mechanically active cell. J. Biomech. 43:146–55
38. Tomasek JJ, Gabbiani G, Hinz B, Chaponnier C, Brown RA. 2002. Myofibroblasts and mechano-
regulation of connective tissue remodelling. Nat. Rev. Mol. Cell Biol. 3:349–63
39. Dabiri GA, Turnacioglu KK, Sanger JM, Sanger JW. 1997. Myofibrillogenesis visualized in living em-
bryonic cardiomyocytes. PNAS 94:9493–98
40. Somlyo AP, Somlyo AV. 1994. Signal-transduction and regulation in smooth muscle. Nature 372:231–36
41. Bers DM. 2002. Cardiac excitation-contraction coupling. Nature 415:198–205
42. Friend DS, Gilula NB. 1972. Variations in tight and gap junctions in mammalian tissues. J. Cell Biol.
53:758–76
43. Pennisi DJ, Rentschler S, Gourdie RG, Fishman GI, Mikawa T. 2002. Induction and patterning of the
cardiac conduction system. Int. J. Dev. Biol. 46:765–75
44. Close RI. 1972. Dynamic properties of mammalian skeletal muscles. Physiol. Rev. 52:129–97
45. Hirsch NP. 2007. Neuromuscular junction in health and disease. Br. J. Anaesth. 99:132–38
46. Browne WR, Feringa BL. 2006. Making molecular machines work. Nat. Nano 1:25–35
47. Carlsen RW, Sitti M. 2014. Bio-hybrid cell-based actuators for microsystems. Small 10:3831–51
48. Howard J. 2001. Mechanics of Motor Proteins and the Cytoskeleton. Sunderland, MA: Sinauer. 384 pp.
49. Lin C-T, Kao M-T, Kurabayashi K, Meyhofer E. 2008. Self-contained, biomolecular motor-driven
protein sorting and concentrating in an ultrasensitive microfluidic chip. Nano Lett. 8:1041–46

www.annualreviews.org • Biological Soft Robotics 261


BE17CH10-Feinberg ARI 31 October 2015 15:30

50. Bull J, Hunt A, Meyhöfer E. 2005. A theoretical model of a molecular-motor-powered pump. Biomed.
Microdevices 7:21–33
51. Vicente-Manzanares M, Ma X, Adelstein RS, Horwitz AR. 2009. Non-muscle myosin II takes centre
stage in cell adhesion and migration. Nat. Rev. Mol. Cell Biol. 10:778–90
52. Wakayama J, Shohara M, Yagi C, Ono H, Miyake N, et al. 2002. Zigzag motions of the myosin-coated
beads actively sliding along actin filaments suspended between immobilized beads. Biochim. Biophys. Acta
1573:93–99
53. Interliggi KA, Zeile WL, Ciftan-Hens SA, McGuire GE, Purich DL, Dickinson RB. 2007. Guidance of
actin filament elongation on filament-binding tracks. Langmuir 23:11911–16
54. Lund K, Manzo AJ, Dabby N, Michelotti N, Johnson-Buck A, et al. 2010. Molecular robots guided by
prescriptive landscapes. Nature 465:206–10
55. Kamm RD, Bashir R. 2014. Creating living cellular machines. Ann. Biomed. Eng. 42:445–59
56. Sokolov A, Apodaca MM, Grzybowski BA, Aranson IS. 2010. Swimming bacteria power microscopic
gears. PNAS 107:969–74
Annu. Rev. Biomed. Eng. 2015.17:243-265. Downloaded from www.annualreviews.org

57. González LM, Ruder WC, Leduc PR, Messner WC. 2014. Controlling magnetotactic bacteria through
an integrated nanofabricated metallic island and optical microscope approach. Sci. Rep. 4:4104
Access provided by 38.25.16.177 on 10/29/22. For personal use only.

58. Carlsen RW, Edwards MR, Zhuang J, Pacoret C, Sitti M. 2014. Magnetic steering control of multi-
cellular bio-hybrid microswimmers. Lab Chip 14:3850–59
59. Kim D, Liu A, Diller E, Sitti M. 2012. Chemotactic steering of bacteria propelled microbeads. Biomed.
Microdevices 14:1009–17
60. Steager E, Kim C-B, Patel J, Bith S, Naik C, et al. 2007. Control of microfabricated structures powered
by flagellated bacteria using phototaxis. Appl. Phys. Lett. 90:263901
61. Pilarek M, Neubauer P, Marx U. 2011. Biological cardio-micro-pumps for microbioreactors and ana-
lytical micro-systems. Sens. Actuators B Chem. 156:517–26
62. Das M, Gregory CA, Molnar P, Riedel LM, Wilson K, Hickman JJ. 2006. A defined system to al-
low skeletal muscle differentiation and subsequent integration with silicon microstructures. Biomaterials
27:4374–80
63. Wilson K, Molnar P, Hickman J. 2007. Integration of functional myotubes with a Bio-MEMS device
for non-invasive interrogation. Lab Chip 7:920–22
64. Das M, Wilson K, Molnar P, Hickman JJ. 2007. Differentiation of skeletal muscle and integration of
myotubes with silicon microstructures using serum-free medium and a synthetic silane substrate. Nat.
Protoc. 2:1795–801
65. Wilson K, Das M, Wahl KJ, Colton RJ, Hickman J. 2010. Measurement of contractile stress generated
by cultured rat muscle on silicon cantilevers for toxin detection and muscle performance enhancement.
PLOS ONE 5:e11042
66. Xi JZ, Schmidt JJ, Montemagno CD. 2005. Self-assembled microdevices driven by muscle. Nat. Mater.
4:180–84
67. Palchesko RN, Zhang L, Sun Y, Feinberg AW. 2012. Development of polydimethylsiloxane substrates
with tunable elastic modulus to study cell mechanobiology in muscle and nerve. PLOS ONE 7:1–13
68. Tanaka Y, Morishima K, Shimizu T, Kikuchi A, Yamato M, et al. 2006. Demonstration of a PDMS-based
bio-microactuator using cultured cardiomyocytes to drive polymer micropillars. Lab Chip 6:230–35
69. Tanaka Y, Sato K, Shimizu T, Yamato M, Okano T, et al. 2008. Demonstration of a bio-microactuator
powered by vascular smooth muscle cells coupled to polymer micropillars. Lab Chip 8:58–61
70. Morishima K, Tanaka Y, Ebara M, Shimizu T, Kikuchi A, et al. 2006. Demonstration of a bio-
microactuator powered by cultured cardiomyocytes coupled to hydrogel micropillars. Sens. Actuators
B Chem. 119:345–50
71. Akiyama Y, Iwabuchi K, Furukawa Y, Morishima K. 2009. Long-term and room temperature operable
bioactuator powered by insect dorsal vessel tissue. Lab Chip 9:140–4
72. Park J, Ryu J, Choi SK, Seo E, Cha JM, et al. 2005. Real-time measurement of the contractile forces of
self-organized cardiomyocytes on hybrid biopolymer microcantilevers. Anal. Chem. 77:6571–80
73. Kim J, Park J, Yang S, Baek J, Kim B, et al. 2007. Establishment of a fabrication method for a long-term
actuated hybrid cell robot. Lab Chip 7:1504–8

262 Feinberg
BE17CH10-Feinberg ARI 31 October 2015 15:30

74. Chan V, Jeong JH, Bajaj P, Collens M, Saif T, et al. 2012. Multi-material bio-fabrication of hydrogel
cantilevers and actuators with stereolithography. Lab Chip 12:88–98
75. Chan V, Park K, Collens MB, Kong H, Saif TA, Bashir R. 2012. Development of miniaturized walking
biological machines. Sci. Rep. 2:857
76. Cvetkovic C, Raman R, Chan V, Williams BJ, Tolish M, et al. 2014. Three-dimensionally printed
biological machines powered by skeletal muscle. PNAS 111:10125–30
77. Feinberg AW, Feigel A, Shevkoplyas SS, Sheehy S, Whitesides GM, Parker KK. 2007. Muscular thin
films for building actuators and powering devices. Science 317:1366–70
78. Feinberg AW, Alford PW, Jin H, Ripplinger CM, Werdich AA, et al. 2012. Controlling the contractile
strength of engineered cardiac muscle by hierarchal tissue architecture. Biomaterials 33:5732–41
79. Alford PW, Feinberg AW, Sheehy SP, Parker KK. 2010. Biohybrid thin films for measuring contractility
in engineered cardiovascular muscle. Biomaterials 31:3613–21
80. Alford PW, Nesmith AP, Seywerd JN, Grosberg A, Parker KK. 2011. Vascular smooth muscle contrac-
tility depends on cell shape. Integr. Biol. 3:1063–70
Annu. Rev. Biomed. Eng. 2015.17:243-265. Downloaded from www.annualreviews.org

81. Sun Y, Duffy R, Lee A, Feinberg AW. 2013. Optimizing the structure and contractility of engineered
skeletal muscle thin films. Acta Biomater. 9:7885–94
82. Feinberg AW, Ripplinger CM, van der Meer P, Sheehy SP, Domian I, et al. 2013. Functional differences
Access provided by 38.25.16.177 on 10/29/22. For personal use only.

in engineered myocardium from embryonic stem cell–derived versus neonatal cardiomyocytes. Stem Cell
Rep. 1:387–96
83. Domian IJ, Chiravuri M, van der Meer P, Feinberg AW, Shi X, et al. 2009. Generation of functional
ventricular heart muscle from mouse ventricular progenitor cells. Science 326:426–29
84. Balachandran K, Alford PW, Wylie-Sears J, Goss JA, Grosberg A, et al. 2011. Cyclic strain induces
dual-mode endothelial-mesenchymal transformation of the cardiac valve. PNAS 108:19943–48
85. Agarwal A, Farouz Y, Nesmith AP, Deravi LF, McCain ML, Parker KK. 2013. Micropatterning alginate
substrates for in vitro cardiovascular muscle on a chip. Adv. Funct. Mater. 23:3738–46
86. McCain ML, Agarwal A, Nesmith HW, Nesmith AP, Parker KK. 2014. Micromolded gelatin hydrogels
for extended culture of engineered cardiac tissues. Biomaterials 35:5462–71
87. Novakovic GV, Eschenhagen T, Mummery C. 2014. Myocardial tissue engineering: in vitro models.
Cold Spring Harb. Perspect. Med. 4:a014076
88. Jallerat Q, Szymanski JM, Feinberg AW. 2014. Nano- and microstructured ECM and biomimetic scaf-
folds for cardiac tissue engineering. In Bio-Inspired Materials for Biomedical Engineering, ed. AB Brennan,
CM Kirschner, pp. 195–226. Hoboken, NJ: Wiley
89. Feinberg AW. 2011. Engineered tissue grafts: opportunities and challenges in regenerative medicine.
Wiley Interdiscip. Rev. Syst. Biol. Med. 4:207–20
90. Koning M, Harmsen MC, van Luyn MJA, Werker PMN. 2009. Current opportunities and challenges
in skeletal muscle tissue engineering. J. Tissue Eng. Regen. Med. 3:407–15
91. Vandenburgh HH, Karlisch P, Farr L. 1988. Maintenance of highly contractile tissue-cultured avian
skeletal myotubes in collagen gel. In Vitro Cell. Dev. Biol. 24:166–74
92. Rhim C, Lowell DA, Reedy MC, Slentz DH, Zhang SJ, et al. 2007. Morphology and ultrastructure of
differentiating three-dimensional mammalian skeletal muscle in a collagen gel. Muscle Nerve 36:71–80
93. Turnbull IC, Karakikes I, Serrao GW, Backeris P, Lee JJ, et al. 2014. Advancing functional engineered
cardiac tissues toward a preclinical model of human myocardium. FASEB J. 28:644–54
94. Black LD, Meyers JD, Weinbaum JS, Shvelidze YA, Tranquillo RT. 2009. Cell-induced alignment
augments twitch force in fibrin gel–based engineered myocardium via gap junction modification. Tissue
Eng. Part A 15:3099–108
95. Bian W, Bursac N. 2009. Engineered skeletal muscle tissue networks with controllable architecture.
Biomaterials 30:1401–12
96. Bian W, Liau B, Badie N, Bursac N. 2009. Mesoscopic hydrogel molding to control the 3D geometry
of bioartificial muscle tissues. Nat. Protoc. 4:1522–34
97. Bian W, Juhas M, Pfeiler TW, Bursac N. 2012. Local tissue geometry determines contractile force
generation of engineered muscle networks. Tissue Eng. Part A 18:957–67
98. Dennis RG, Kosnik PE 2nd. 2000. Excitability and isometric contractile properties of mammalian skeletal
muscle constructs engineered in vitro. In Vitro Cell. Dev. Biol. Anim. 36:327–35

www.annualreviews.org • Biological Soft Robotics 263


BE17CH10-Feinberg ARI 31 October 2015 15:30

99. Dennis RG, Kosnik PE 2nd, Gilbert ME, Faulkner JA. 2001. Excitability and contractility of skeletal
muscle engineered from primary cultures and cell lines. Am. J. Physiol. Cell Physiol. 280:C288–95
100. Dennis RG, Dow DE. 2007. Excitability of skeletal muscle during development, denervation, and tissue
culture. Tissue Eng. 13:2395–404
101. Baar K, Birla R, Boluyt MO, Borschel GH, Arruda EM, Dennis RG. 2004. Self-organization of rat
cardiac cells into contractile 3-D cardiac tissue. FASEB J. 18:275–77
102. Larkin LM, Calve S, Kostrominova TY, Arruda EM. 2006. Structure and functional evaluation of tendon-
skeletal muscle constructs engineered in vitro. Tissue Eng. 12:3149–58
103. Akiyama Y, Kikuchi A, Yamato M, Okano T. 2004. Ultrathin poly(N-isopropylacrylamide) grafted layer
on polystyrene surfaces for cell adhesion/detachment control. Langmuir 20:5506–11
104. Shimizu T, Sekine H, Isoi Y, Yamato M, Kikuchi A, Okano T. 2006. Long-term survival and growth
of pulsatile myocardial tissue grafts engineered by the layering of cardiomyocyte sheets. Tissue Eng.
12:499–507
105. Haraguchi Y, Shimizu T, Sasagawa T, Sekine H, Sakaguchi K, et al. 2012. Fabrication of functional
Annu. Rev. Biomed. Eng. 2015.17:243-265. Downloaded from www.annualreviews.org

three-dimensional tissues by stacking cell sheets in vitro. Nat. Protoc. 7:850–58


106. Yamamoto Y, Ito A, Fujita H, Nagamori E, Kawabe Y, Kamihira M. 2011. Functional evaluation of arti-
ficial skeletal muscle tissue constructs fabricated by a magnetic force-based tissue engineering technique.
Access provided by 38.25.16.177 on 10/29/22. For personal use only.

Tissue Eng. Part A 17:107–14


107. Stevens KR, Kreutziger KL, Dupras SK, Korte FS, Regnier M, et al. 2009. Physiological function and
transplantation of scaffold-free and vascularized human cardiac muscle tissue. PNAS 106:16568–73
108. Tanaka Y, Morishima K, Shimizu T, Kikuchi A, Yamato M, et al. 2006. An actuated pump on-chip
powered by cultured cardiomyocytes. Lab Chip 6:362–68
109. Tanaka Y, Sato K, Shimizu T, Yamato M, Okano T, Kitamori T. 2007. A micro-spherical heart pump
powered by cultured cardiomyocytes. Lab Chip 7:207–12
110. Park J, Kim IC, Baek J, Cha M, Kim J, et al. 2007. Micro pumping with cardiomyocyte-polymer hybrid.
Lab Chip 7:1367–70
111. Hoshino T, Morishima K. 2010. Muscle-powered cantilever for microtweezers with an artificial micro
skeleton and rat primary myotubes. J. Biomech. Sci. Eng. 5:245–51
112. Kabumoto K, Hoshino T, Morishima K. 2010. Bio-robotics using interaction between neuron and
muscle for development of living prosthesis. Proc. Int. Conf. Biomed. Robot. Biomechatronics, 3rd , Univ.
Tokyo, Sept. 26–29, pp. 419–24. Piscataway, NJ: IEEE
113. Akiyama Y, Sakuma T, Funakoshi K, Hoshino T, Iwabuchi K, Morishima K. 2013. Atmospheric-operable
bioactuator powered by insect muscle packaged with medium. Lab Chip 13:4870–80
114. Akiyama Y, Hoshino T, Iwabuchi K, Morishima K. 2012. Room temperature operable autonomously
moving bio-microrobot powered by insect dorsal vessel tissue. PLOS ONE 7:e38274
115. Akiyama Y, Odaira K, Sakiyama K, Hoshino T, Iwabuchi K, Morishima K. 2012. Rapidly-moving insect
muscle-powered microrobot and its chemical acceleration. Biomed. Microdevices 14:979–86
116. Williams BJ, Anand SV, Rajagopalan J, Saif MTA. 2014. A self-propelled biohybrid swimmer at low
Reynolds number. Nat. Commun. 5:3081
117. Herr H, Dennis R. 2004. A swimming robot actuated by living muscle tissue. J. Neuroeng. Rehabil. 1:6
118. Shimizu Y, Kuruma Y, Ying BW, Umekage S, Ueda T. 2006. Cell-free translation systems for protein
engineering. FEBS J. 273:4133–40
119. Varley J, Birch J. 1999. Reactor design for large scale suspension animal cell culture. Cytotechnology
29:177–205
120. Grillberger L, Kreil TR, Nasr S, Reiter M. 2009. Emerging trends in plasma-free manufacturing of
recombinant protein therapeutics expressed in mammalian cells. Biotechnol. J. 4:186–201
121. Baryshyan AL, Domigan LJ, Hunt B, Trimmer BA, Kaplan DL. 2014. Self-assembled insect muscle
bioactuators with long term function under a range of environmental conditions. RSC Adv. 4:39962–68
122. Baryshyan AL, Woods W, Trimmer BA, Kaplan DL. 2012. Isolation and maintenance-free culture of
contractile myotubes from Manduca sexta embryos. PLOS ONE 7:e31598
123. Dinsmore J, Ratliff J, Deacon T, Pakzaban P, Jacoby D, et al. 1996. Embryonic stem cells differentiated
in vitro as a novel source of cells for transplantation. Cell Transplant. 5:131–43

264 Feinberg
BE17CH10-Feinberg ARI 31 October 2015 15:30

124. Lee T-H, Song S-H, Kim KL, Yi J-Y, Shin G-H, et al. 2010. Functional recapitulation of smooth muscle
cells via induced pluripotent stem cells from human aortic smooth muscle cells. Circ. Res. 106:120–28
125. Passier R, van Laake LW, Mummery CL. 2008. Stem-cell–based therapy and lessons from the heart.
Nature 453:322–29
126. Ahadian S, Ostrovidov S, Hosseini V, Kaji H, Ramalingam M, et al. 2013. Electrical stimulation as a
biomimicry tool for regulating muscle cell behavior. Organogenesis 9:87–92
127. Ogut O, Brozovich FV. 2003. Regulation of force in vascular smooth muscle. J. Mol. Cell. Cardiol.
35:347–55
128. Akiyama Y, Iwabuchi K, Furukawa Y, Morishima K. 2010. Electrical stimulation of cultured lepidopteran
dorsal vessel tissue: an experiment for development of bioactuators. In Vitro Cell. Dev. Biol. Anim. 46:411–
15
129. Bours MJL, Swennen ELR, Di Virgilio F, Cronstein BN, Dagnelie PC. 2006. Adenosine 5 -triphosphate
and adenosine as endogenous signaling molecules in immunity and inflammation. Pharmacol. Ther.
112:358–404
Annu. Rev. Biomed. Eng. 2015.17:243-265. Downloaded from www.annualreviews.org

130. Rakusan K, Flanagan MF, Geva T, Southern J, Van Praagh R. 1992. Morphometry of human coronary
capillaries during normal growth and the effect of age in left ventricular pressure-overload hypertrophy.
Circulation 86:38–46
Access provided by 38.25.16.177 on 10/29/22. For personal use only.

131. Lovett M, Lee K, Edwards A, Kaplan DL. 2009. Vascularization strategies for tissue engineering. Tissue
Eng. Part B Rev. 15:353–70
132. Sicari BM, Agrawal V, Siu BF, Medberry CJ, Dearth CL, et al. 2012. A murine model of volumetric
muscle loss and a regenerative medicine approach for tissue replacement. Tissue Eng. Part A 18:1941–48
133. Machingal MA, Corona BT, Walters TJ, Kesireddy V, Koval CN, et al. 2011. A tissue-engineered muscle
repair construct for functional restoration of an irrecoverable muscle injury in a murine model. Tissue
Eng. Part A 17:2291–303
134. Miller JS, Stevens KR, Yang MT, Baker BM, Nguyen D-HT, et al. 2012. Rapid casting of patterned
vascular networks for perfusable engineered three-dimensional tissues. Nat. Mater. 11:768–74
135. Zheng Y, Chen J, Craven M, Choi NW, Totorica S, et al. 2012. In vitro microvessels for the study of
angiogenesis and thrombosis. PNAS 109:9342–47
136. Nagamine K, Kawashima T, Sekine S, Ido Y, Kanzaki M, Nishizawa M. 2011. Spatiotemporally con-
trolled contraction of micropatterned skeletal muscle cells on a hydrogel sheet. Lab Chip 11:513–17
137. Sakar MS, Neal D, Boudou T, Borochin MA, Li Y, et al. 2012. Formation and optogenetic control of
engineered 3D skeletal muscle bioactuators. Lab Chip 12:4976–85
138. Langhammer CG, Kutzing MK, Luo V, Zahn JD, Firestein BL. 2011. Skeletal myotube integration with
planar microelectrode arrays in vitro for spatially selective recording and stimulation: a comparison of
neuronal and myotube extracellular action potentials. Biotechnol. Prog. 27:891–95
139. Langhammer CG, Kutzing MK, Luo V, Zahn JD, Firestein BL. 2013. A topographically modified
substrate-embedded MEA for directed myotube formation at electrode contact sites. Ann. Biomed. Eng.
41:408–20
140. Huffard CL, Boneka F, Full RJ. 2005. Underwater bipedal locomotion by octopuses in disguise. Science
307:1927

www.annualreviews.org • Biological Soft Robotics 265


BE17-FrontMatter ARI 2 November 2015 10:54

Annual Review of
Biomedical
Engineering
Contents Volume 17, 2015

Large-Volume Microfluidic Cell Sorting for Biomedical Applications


Annu. Rev. Biomed. Eng. 2015.17:243-265. Downloaded from www.annualreviews.org

Majid Ebrahimi Warkiani, Lidan Wu, Andy Kah Ping Tay, and Jongyoon Han p p p p p p p p p 1
High-Throughput Assessment of Cellular Mechanical Properties
Access provided by 38.25.16.177 on 10/29/22. For personal use only.

Eric M. Darling and Dino Di Carlo p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p35


Viral Vectors for Gene Therapy: Translational and Clinical Outlook
Melissa A. Kotterman, Thomas W. Chalberg, and David V. Schaffer p p p p p p p p p p p p p p p p p p p p p63
Digital Microfluidic Cell Culture
Alphonsus H.C. Ng, Bingyu Betty Li, M. Dean Chamberlain,
and Aaron R. Wheeler p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p91
The Mechanobiology of Aging
Jude M. Phillip, Ivie Aifuwa, Jeremy Walston, and Denis Wirtz p p p p p p p p p p p p p p p p p p p p p p p 113
Modeling Signaling Networks to Advance New Cancer Therapies
Julio Saez-Rodriguez, Aidan MacNamara, and Simon Cook p p p p p p p p p p p p p p p p p p p p p p p p p p p p 143
Biosensors for Cell Analysis
Qing Zhou, Kyungjin Son, Ying Liu, and Alexander Revzin p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 165
Advances in Antibody Design
Kathryn E. Tiller and Peter M. Tessier p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 191
Synergizing Engineering and Biology to Treat and Model Skeletal
Muscle Injury and Disease
Nenad Bursac, Mark Juhas, and Thomas A. Rando p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 217
Biological Soft Robotics
Adam W. Feinberg p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 243
Microfluidic Sample Preparation for Medical Diagnostics
Francis Cui, Minsoung Rhee, Anup Singh, and Anubhav Tripathi p p p p p p p p p p p p p p p p p p p p p p 267
Molecular-Scale Tools for Studying Mechanotransduction
Andrew S. LaCroix, Katheryn E. Rothenberg, and Brenton D. Hoffman p p p p p p p p p p p p p p p 287
Biomaterial Strategies for Immunomodulation
Nathan A. Hotaling, Li Tang, Darrell J. Irvine, and Julia E. Babensee p p p p p p p p p p p p p p p p 317

v
BE17-FrontMatter ARI 2 November 2015 10:54

Image-Based Predictive Modeling of Heart Mechanics


V.Y. Wang, P.M.F. Nielsen, and M.P. Nash p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 351
Positron Emission Tomography: Current Challenges and
Opportunities for Technological Advances in Clinical and
Preclinical Imaging Systems
Juan José Vaquero and Paul Kinahan p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 385
Coherent Raman Scattering Microscopy in Biology and Medicine
Chi Zhang, Delong Zhang, and Ji-Xin Cheng p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 415
Hamiltonian Systems and Optimal Control in Computational
Anatomy: 100 Years Since D’Arcy Thompson
Michael I. Miller, Alain Trouvé, and Laurent Younes p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 447
Annu. Rev. Biomed. Eng. 2015.17:243-265. Downloaded from www.annualreviews.org
Access provided by 38.25.16.177 on 10/29/22. For personal use only.

Indexes

Cumulative Index of Contributing Authors, Volumes 8–17 p p p p p p p p p p p p p p p p p p p p p p p p p p p p 511


Cumulative Index of Article Titles, Volumes 8–17 p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 515

Errata

An online log of corrections to Annual Review of Biomedical Engineering articles may be


found at http://www.annualreviews.org/errata/bioeng

vi Contents
Annu. Rev. Biomed. Eng. 2015.17:243-265. Downloaded from www.annualreviews.org
Access provided by 38.25.16.177 on 10/29/22. For personal use only.

You might also like