ll

Article
Living skin on a robot
Michio Kawai, Minghao Nie,
Haruka Oda, Yuya Morimoto,
Shoji Takeuchi

[email protected]

Highlights
The method for generating
seamless coverage of 3D objects
with skin equivalent

Three-joint robotic finger covered

with living skin equivalent

Wound repair of a robotic finger

covered with dermis equivalent

By replicating the appearances and functions of human beings, humanoids have

the potential to establish harmonic and natural human-robot interactions. Here,
we fabricated a three-joint robotic finger covered with living human skin. The robot
not only can achieve realistic tone and texture of humans, but it also can be healed
by grafting a hydrogel patch onto its wounds, resembling the process of skin graft
surgery of human beings.

Kawai et al., Matter 5, 1–19

July 6, 2022 ª 2022 Elsevier Inc.
https://doi.org/10.1016/j.matt.2022.05.019
 Please cite this article in press as: Kawai et al., Living skin on a robot, Matter (2022), https://doi.org/10.1016/j.matt.2022.05.019

ll

Article
Living skin on a robot
Michio Kawai,1 Minghao Nie,1 Haruka Oda,1 Yuya Morimoto,1 and Shoji Takeuchi1,2,3,4,*

SUMMARY Progress and potential

Humanoids are robots created with human forms or characteristics; The covering materials of
these robots also have the potential to seamlessly interact with hu- humanoid robots have gone
man beings. By replicating the appearances and functions (e.g., self- through transformations from stiff
healing) of human beings, humanoids have the potential to establish and heavy materials to soft and
more harmonic and natural human-robot interactions. Here, we pro- compliant materials to better
pose the use of skin equivalent, a living skin model consisting of cells mimic the appearance and
and extracellular matrix, as a human-like and self-healing coverage function of human beings. Here,
material for robots. We fabricated a three-joint robotic finger we report a biohybrid approach to
covered with skin equivalent by developing a method to cover generate robots covered with
three-dimensional objects with skin equivalent. Furthermore, tissue-engineered skin. The skin
inspired by the medical treatment of deeply burned skin using coverage not only results in a
grafted hydrogels, we demonstrated wound repair of a dermis human-like appearance but also
equivalent covering a robotic finger by culturing the wounded tissue enables self-healing functions. We
grafted with a collagen sheet. With the above results, this research demonstrate the seamless
shows the potential of using skin equivalent as human-like and self- coverage of a three-joint robotic
healing coverage material for robots. finger by culturing a single piece
of skin tissue surrounding the
robotic finger. Furthermore,
INTRODUCTION inspired by the medical treatment
Humanoids are robots that can perform a wide range of tasks to interact with humans of deeply burned skin using
in medical care settings,1 nursing care situations,2 and the service industry.3 Such grafted hydrogels, we
tasks for humanoids require a human-like appearance to improve the efficiency of demonstrate wound repair of a
information exchange with humans4 and to evoke likability.5 To mimic the appear- dermis equivalent covering a
ance of humans, the development of humanoid covering materials with realistic robotic finger by culturing the
tone and texture of human skin is imperative. Moreover, these covering materials wounded tissue grafted with a
are expected to have self-healing ability, as humanoids covered in these materials collagen sheet. Taken together,
are intended to operate in uncertain and dynamic task environments.6 Although these findings show the potential
silicone rubber, which has been traditionally used as the covering material for of a paradigm shift from
humanoids, has acquired a human-like appearance,7–9 it lacks many functionalities traditional robotics to the new
of living skin such as self-healing ability. scheme of biohybrid robotics that
leverage the advantages of both
Here, for the first time to our knowledge, we propose the use of skin equivalent as a living materials and artificial
covering material for robots. This living material not only gives robots a human-like materials.
appearance but also provides robots with self-healing ability (Figure 1A). Impor-
tantly, skin equivalents are in vitro tissues fabricated using living cells (fibroblasts,
keratinocytes, etc.) and extracellular matrix (ECM) hydrogels (collagen, etc.), which
can mimic unique skin characteristics, such as appearance and internal structure,
thus achieving a soft touch and self-healing capability. Skin equivalents have been
used as tools in biological research10–14 and as medical implants to treat severe
wounds and burns.15–17 Some studies have also shown the in vitro self-repair capa-
bility of skin equivalents after surface wounds have occurred.18–20 However, it has
been difficult to fabricate skin equivalents that can seamlessly cover the three-
dimensional (3D), curved, and uneven surfaces of humanoids. Pertinently, most
skin equivalents have been derived in the form of a two-dimensional flat sheet

Matter 5, 1–19, July 6, 2022 ª 2022 Elsevier Inc. 1

 Please cite this article in press as: Kawai et al., Living skin on a robot, Matter (2022), https://doi.org/10.1016/j.matt.2022.05.019

ll
Article

Figure 1. Conceptual illustration and fabrication process of the ‘‘living skin on a robot’’ proposed in this paper; see also Figures S1 and S10
(A) Conceptual illustration of a biohybrid robotic hand with its finger covered with human skin equivalent.
(B) Design of the three-joint robotic finger to be covered with skin equivalent.
(C) Fabrication process of the skin equivalent used to cover the robotic finger. Scale bar, 10 mm.

due to the inability to control the 3D shape of the dermis equivalent, which shrinks
and changes its shape drastically during culture, and the difficulty in uniformly seed-
ing keratinocytes onto dermis equivalent with 3D surfaces.

1Department of Mechano-Informatics, Graduate

In this study, we developed a method to construct a skin equivalent that uniformly
School of Information Science and Technology,
covers a three-joint robotic finger as the fundamental component for the hands of
The University of Tokyo, Tokyo 113-8656, Japan
humanoids. This method can be used to cover the 3D surface of a robotic finger 2Institute
of Industrial Science (IIS), The University
while controlling tissue shrinkage through anchor fixation. In addition, multidirec- of Tokyo, Tokyo 153-8505, Japan
tional seeding of keratinocytes enables us to uniformly form the epidermis layer. 3International Research Center for
In the experiments, we confirmed the effectiveness of our fabrication method of Neurointelligence (WPI-IRCN), The University of
Tokyo Institutes for Advanced Study, The
skin equivalent by histological analyses with staining and functional analyses with University of Tokyo, Tokyo 113-0033, Japan
electrical measurements and water retention tests. Moreover, to show the wound 4Lead contact
healing ability of the skin equivalent, we demonstrated that the wounded dermis *Correspondence:
equivalent covering the robotic finger can be repaired by grafting and culturing a [email protected]
collagen sheet onto the wound site. https://doi.org/10.1016/j.matt.2022.05.019

2 Matter 5, 1–19, July 6, 2022

 Please cite this article in press as: Kawai et al., Living skin on a robot, Matter (2022), https://doi.org/10.1016/j.matt.2022.05.019

ll
Article

RESULTS
Fabrication of the skin equivalent on the robotic finger
As shown in Figures 1B and 1C, the strategy for constructing the skin equivalent
covering the robotic finger contained two major steps: (1) the construction of a
dermis equivalent that covers the robotic finger and (2) the construction of the
epidermis on the surface of the dermis equivalent. First, to form the dermis equiva-
lent, collagen solution containing normal human dermal fibroblasts (NHDFs) was
poured into a cylindrical polydimethylsiloxane (PDMS) case and cultured to shrink
into a matured dermis equivalent. The anchoring structure at the stem of the finger
can hook and hold the skin equivalent at its edges to not only prevent the dermis
equivalent coverage on the surface of the fingertip device from retreating but also
to utilize the tension generated by shrinkage to cover the robotic finger tightly. Sec-
ond, the epidermal layer is constructed by seeding normal human epidermal kerati-
nocytes (NHEKs). Since the in vivo epidermis consists mainly of keratinocytes that
form tight cell-cell junctions, the epidermal layer is commonly formed by seeding
keratinocyte suspensions onto the surface of a planar-shaped dermis equivalent,
allowing the suspensions to settle and adhere to the dermis equivalent to achieve
a highly cellularized epidermis.21 To increase the chance of cell adhesion during
each seed-settle-adhere sequence, it was also necessary to prevent NHEKs from
completely slipping off the dermis surface using a PDMS case. Additionally, to
form an epidermal layer on a dermis equivalent covering a 3D object, it was neces-
sary to perform the seed-settle-adhere sequence multiple times, from multiple di-
rections, to cover the entire surface of the tissue. The skin equivalent can withstand
the deformation generated by the joint motions of the robotic finger. Detailed fabri-
cation procedures can be found in the Experimental procedures.

Experiments to evaluate the validity of this fabrication strategy were conducted us-
ing a simple object resembling a fingertip (referred to as a ‘‘fingertip device’’) (Fig-
ure 2A). As shown in Figure S2A, the fingertip device consists of a mold part and an
anchor part. The mold part defines the shape of the fingertip with a cylinder and a
spherical dome on its top, which is connected to a circular base plate. The relatively
simple shape of the mold part is useful for quantitatively evaluating the uniformity of
the constructed skin. The anchor part consists of an array of mushroom-shaped an-
chors circumferentially distributed around a donut-shaped baseplate. The skin
equivalent covering fingertip device can be fabricated using the same method as
that used for the robotic finger (Figure S2B). To evaluate the effectiveness of the an-
chors, we performed preliminary investigations of the shrinkage of the collagen gel
with various concentrations of NHDFs formed in Petri dishes. As a result, 45%–78%
shrinkage was found in the samples with NHDFs after a week of culture, while almost
no shrinkage was found in the samples without NHDFs (Figure S3). Moreover,
fingertip devices with and without anchors were then covered with the dermis equiv-
alent. After 7 days of culture, the coverage of dermis equivalent formed on the
fingertip devices without the anchors retreated, leaving a larger bare surface than
that left by those with anchors, indicating that it is necessary to hold the dermis
equivalent using anchors for gapless coverage of the device (Figures 2B and 2C).
Interestingly, complete retreat of the dermis equivalent coverage (the dermis equiv-
alent came off the mold part) was not found for the samples without anchors; this
outcome could have occurred due to the friction between the surfaces of the dermis
equivalent and the fingertip device.

To evaluate the effectiveness of the cap part on the uniform shaping of the dermis
equivalent, we compared the dermis equivalent fabricated with and without the

Matter 5, 1–19, July 6, 2022 3

 Please cite this article in press as: Kawai et al., Living skin on a robot, Matter (2022), https://doi.org/10.1016/j.matt.2022.05.019

ll
Article

Figure 2. Evaluation of the method to construct skin equivalent to cover 3D objects; see also Figures S2, S3, and S8
(A) Images of the fingertip device before and after being covered with skin equivalent.
(B and C) Evaluation of the effect of the anchor structure on the device. The graph shows the length of retreated tissue coverage for the samples
fabricated in the devices with and without anchor structures (n = 3). The results are shown as the mean G SD. *p < 0.05, unpaired Student’s t test.
(D and E) Comparison of the outline of the dermis equivalent fabricated in devices with and without the cap part (n = 3). Tissue thickness was
measured using images of frozen sections at three different points. The thicknesses of A and B were averaged for both sides. The results are shown as
the mean G SD.
(F and G) Conditions and results of the experiment on the effect of various culture periods after NHEK seeding (n = 3) are shown. The graphs show the
covering rate of the epidermis for each condition. The results are shown as the mean G SD. *p < 0.05, unpaired Student’s t test.
(H and I) Conditions and results of the experiment comparing the effect of different numbers of epidermal seeding directions (n = 3). The graph shows
the change in covering rate with the number of seeding directions. The results are shown as the mean G SD. *p < 0.05, unpaired Student’s t test. Scale
bars, (A) 10 mm (top), 5 mm (bottom); (B, C) 5 mm.

cap part (Figures 2D and 2E). The dermis equivalent fabricated without the cap part
was formed with nonuniform thickness, which is especially evident by comparing the
rounded/sharpened top corners of the tissues formed with/without the cap part.
These results confirm that it is necessary to adjust the initial thickness of the collagen
gel using the cap part so that the skin equivalent can be formed uniformly on the ro-
botic finger.

To evaluate the effectiveness of the multidirectional NHEK seeding method on the

formation of uniform epidermal coverage, we compared various samples with
different seeding directions and postseeding culture periods. Figures 2F and 2G

4 Matter 5, 1–19, July 6, 2022

 Please cite this article in press as: Kawai et al., Living skin on a robot, Matter (2022), https://doi.org/10.1016/j.matt.2022.05.019

ll
Article

shows the relationship between the length of postseeding cultivation periods and
the epidermal covering rate. The covering rate of the samples, in which NHEKs
were seeded from one direction onto the dermis equivalent and cultured for
3 days, was approximately 30%. The samples in which NHEKs were seeded from
one direction onto the dermis equivalent and cultured for 14 days showed a
coverage rate of approximately 40%. The higher covering rate of the sample
cultured for 14 days after seeding indicated that the NHEKs migrated and prolifer-
ated on the dermis equivalent within culture. On the other hand, it is inefficient to
build epidermis by seeding NHEKs from one direction because of the slow migration
of the epidermis. The covering rate of the samples, in which NHEKs were seeded
from two directions onto the dermis equivalent and cultured for 14 days, was
approximately 80%. Therefore, epidermal seeding from two directions was used
to fabricate the skin equivalent covering the robotic finger or fingertip device.
Figures 2H and 2I shows the relationship between the number of seeding directions
of NHEKs and the epidermal covering rate. Both samples were seeded with the same
total amount of NHEKs from one/two directions and cultured for 14 days. As a result,
samples in which NHEKs were seeded from two directions showed a coverage rate of
more than 80% on average. These results indicated that seeding NHEKs from two
directions was more effective than seeding NHEKs from one direction. Based on
these results, in our following experiments, the epidermal layers were constructed
by performing the NHEK seed-settle-adhere sequence from two directions and
culturing for 14 days.

Histological and functional analysis of the fabricated skin equivalent

To confirm the tissue morphology of skin equivalent, frozen sections (thickness:
8 mm) of the fabricated tissue were stained and analyzed. Figure 3A shows an image
of a frozen section with hematoxylin and eosin (H&E) staining. The image shows that
an epidermal layer seamlessly covered the surface of the dermis equivalent. In addi-
tion, both the dermis and epidermis layers demonstrated uniform thickness.
Figures 3B and 3C show immunostaining results of frozen sections using CK-15
and CK-10 markers, respectively. For the in vivo epidermis, CK-15 was used to stain
basal cells with higher stemness compared to that of other cells, and CK-10 was used
to indicate suprabasal cells with a higher degree of differentiation and expression of
barrier function compared to that of other cells. The expression of both CK-15 and
CK-10 in the fabricated skin equivalent indicated the reconstruction of a tissue bar-
rier that has the potential to mimic the turnover process of in vivo skin with preserved
stemness of CK-15-positive cells.

In the human body, the epidermis acts as a barrier that protects the body by keeping
exogenous substances from entering and preventing excessive water loss.22 Here,
we evaluated the barrier function of our in vitro fabricated skin equivalents with elec-
trical measurements and water retention tests. In electrical measurements, it is
known that the capacitance of the epidermis is much lower than that of the dermis;
thus, capacitance measurement is one of the methods used to assess the barrier
function of skin.23 To compare the difference in the electrical properties of the
dermis equivalent and those of the skin equivalent, we fabricated dermis equivalent
samples with only half of the surface covered with epidermis and measured the
capacitance and resistance of each half using a modified fingertip device with
embedded electrodes (Figures S4A–S4C). Figure 3D shows the comparison of
capacitance and resistance between the tissues with and without epidermis
coverage. The results show that the capacitances of the tissues with epidermis
were significantly lower than those of tissues without epidermis, and the resistances
of the tissues with epidermis were higher than those of tissues without epidermis.

Matter 5, 1–19, July 6, 2022 5

 Please cite this article in press as: Kawai et al., Living skin on a robot, Matter (2022), https://doi.org/10.1016/j.matt.2022.05.019

ll
Article

Figure 3. Histological and functional evaluation of the fabricated skin equivalent; see also Figure S4
(A) H&E staining of frozen sections of the tissue. The cell nuclei were stained violet, and the ECM and cytoplasm were stained pink.
(B) Immunostaining of the frozen tissue sections with CK15 antibody.
(C) Immunostaining of frozen tissue sections with CK10 antibody.
(D) Barrier function assessment by capacitance and resistance measurements (n = 3). The results are shown as the mean G SD. *p < 0.05, unpaired
Student’s t test.
(E) The result of the water retention test (n = 3). The results are shown as the mean G SD at each time point. *p < 0.05, unpaired Student’s t test. Scale
bars, (A, B, C) 2 mm, (insets) 200 mm.

These changes in electrical properties of tissues with and without the epidermal layer
were consistent with the results of previous studies,11,24 indicating the successful for-
mation of the functional epidermis.

For the water retention test, the mold part of the fingertip device was withdrawn
from the anchor part, leaving a pocket-shaped skin equivalent with an inner cavity.
After adding 4% paraformaldehyde solution in phosphate-buffered saline without
Mg2+ Ca2+ (PBS( )) to the inner cavity, leaked liquids were continuously collected
and weighed using an analytical balance for 1 h (Figures S4D and S4E). Figure 3E
shows the comparison of the amount of leaked PBS( ) from the inner cavity of the
skin equivalent and the dermis equivalent. The leakage from the dermis equivalent
sample was approximately 5.4 mg after 1 h, while the leakage from the skin equiva-
lent sample was approximately 0.13 mg after 1 h. The results show that our 3D-
shaped skin equivalents can serve as a permeability barrier to prevent water
leakages.

To assess the mechanical properties of the cultured skin material, we measured the
tensile strength of the dermis equivalent. Since the thickness of the dermis equiva-
lent is significantly thicker than that of the epidermis in skin equivalents and the
dermis equivalent is the main part responsible for the mechanical strength of the
skin equivalent, tensile tests were conducted using only dermis equivalent for con-
venience. As a result, the tensile strength of the dermis equivalent was 2.6, 4.9,

6 Matter 5, 1–19, July 6, 2022

 Please cite this article in press as: Kawai et al., Living skin on a robot, Matter (2022), https://doi.org/10.1016/j.matt.2022.05.019

ll
Article

Figure 4. Histological and functional evaluation of the fabricated skin equivalent; see also Videos S1, S2, S3, and S4
(A) The joint motion of the robotic finger covered with human skin equivalent. The photo on the left is the bent finger, and the photo on the right is the
straight finger.
(B) Photo of the robotic finger with and without the application of skin equivalent.

Matter 5, 1–19, July 6, 2022 7

 Please cite this article in press as: Kawai et al., Living skin on a robot, Matter (2022), https://doi.org/10.1016/j.matt.2022.05.019

ll
Article

Figure 4. Continued
(C) The upper graph shows the sectional width distribution of the robotic finger without skin equivalent and the robotic finger covered with skin
equivalent; the lower graph shows the thickness of the skin equivalent.
(D) Epidermal layer peeled off from the skin equivalent covering the robotic finger.
(E) Confirmation of the water repellency of fabricated skin equivalent.
(F) Manipulation of microbeads by the robotic finger covered with skin equivalent and dermis equivalent. Scale bars, (A, B, F) 10 mm; (D, E) 2 mm.

and 4.4 kPa for the samples cultured for 3, 7, and 14 days, respectively. The dermis
equivalent cultured for 14 days became significantly stronger than that cultured for
3 days (p = 0.10) (Figure S5A). Based on this result, the skin equivalent covering the
robotic finger was cultured for more than 14 days. It is also known that biomaterials
such as human skin have a strain-hardening feature, which serves as a mechanism to
protect themselves from excessive tension.25 Figure S5B shows the stress-strain
behavior of the skin equivalent cultured for 14 days. The J-shaped tensile curves
of the dermis equivalent cultured for 14 days exhibit a significant difference between
the tangent modulus of region 1 and region 3, indicating the strain-hardening
behavior of the skin equivalent (Figure S5C).

Characterization of skin equivalent on the robotic finger

Figure 4A and Video S1 show the joint motion of the fabricated robotic finger covered
with skin equivalent. The skin equivalent covered the robotic finger tightly and had
enough strength and elasticity to withstand the deformation caused by the joint mo-
tion of the robot. To confirm the uniformity in the thickness of the fabricated skin
equivalent, images of robotic fingers covered without or with the skin equivalent
were taken, and they are shown in Figure 4B. Based on the images, the sectional
widths of the robotic fingers at each horizontal position point, sliced along the vertical
direction, are plotted in Figure 4C. The sectional thickness of the skin equivalent, i.e.,
the difference in the sectional widths for the robotic finger with and without the skin
equivalent, was calculated to be 2.34 G 0.28 mm (mean G SD). In addition to the
decrease in sectional thickness due to the sharp slope of the finger’s outline at the
tip of the finger, it was confirmed that a uniform skin equivalent covering the robotic
finger tightly was constructed, reproducing the surface unevenness of the robot.

The skin equivalent covering the robotic finger had an epidermis with enough thickness
that can be peeled off and picked up with tweezers (Figure 4D and Video S2). Since it is
also known that skin equivalent with sufficient epidermis is water repellent and can form
water droplets on its surface,11 we tested our skin equivalent covering the robotic finger;
the outcome revealed sufficient water repellency to form water droplets on the surface
(Figure 4E). In addition, the water droplets could be wiped off from the skin equivalent, as
shown in Video S3. These properties confirmed the successful formation of the
epidermal layer as the outermost covering layer of the robotic finger.

The formation of a water-repellent epidermis has various advantages in the perfor-

mance of specific tasks, such as the handling of polystyrene foam beads. Polystyrene
foam beads are light, small, and easily electrostatically charged; therefore, they are
difficult for robots covered with wet materials (such as dermis equivalent) to handle
because the beads adhere to wet objects due to surface tension and electrostatic
forces. As shown in Figure 4F and Video S4, when a robotic finger covered with
dermis equivalent was used to flick the beads, the beads adhered on the dermal sur-
face and could not be repelled; on the other hand, when the skin equivalent with a
water-repellent epidermis was used, the beads were repelled without adherence to
the robotic finger. Therefore, the formation of this water-repellent epidermis has
expanded the range of possible operations and has made robots covered in this ma-
terial more suitable for practical applications.

8 Matter 5, 1–19, July 6, 2022

 Please cite this article in press as: Kawai et al., Living skin on a robot, Matter (2022), https://doi.org/10.1016/j.matt.2022.05.019

ll
Article

Figure 5. Repair of the dermis equivalent covering the robotic finger; see also Figures S6 and S7 and Video S5
(A) Images of the repair process. The wound punched on the dermis equivalent was repaired by grafting the collagen sheet onto it.
(B) The joint motion of the robotic finger after repair is shown.

Matter 5, 1–19, July 6, 2022 9

 Please cite this article in press as: Kawai et al., Living skin on a robot, Matter (2022), https://doi.org/10.1016/j.matt.2022.05.019

ll
Article

Figure 5. Continued
(C) Close-up image of the wounded area on the bent robotic finger.
(D) Photo of samples before and after grafting used for the cell migration test.
(E) Cell migration into the grafted collagen sheet after repair. The tissues were stained 3 and 7 days after repair.
(F) Various conditions were evaluated to test the strength recovery of the repaired tissue.
(G) The results of the strength recovery assessment of the repaired tissue (n = 3) are shown. The results are shown as the mean G SD. *p < 0.05, Tukey-
Kramer test, Scale bars, (A) 20 mm; (B) 10 mm; (C) 10 mm, 5 mm (inset); (D) 5 mm; (E) 1 mm, 100 mm (inset).

Repair of the dermis equivalent covering the robotic finger

Repair treatment of the dermis equivalent covering the robotic finger was performed
to show the repairability of the skin as a cover material for robots. Inspired by the
medical treatments of deeply burned skin using grafted hydrogels,26 repair treat-
ment was performed by applying an acellular collagen sheet on an intentionally
made wound at the center of the planar-shaped dermis equivalent (Figures S6A
and S6B) or on the dorsal surface of the robotic finger (Figure 5A, i).

Figure 5A shows the procedure that was used to repair the dermis equivalent
covering the robotic finger. The wound was made on the dorsal surface of the ro-
botic finger using a surgical knife (Figure 5A, i). The collagen sheet attached to
the dermis equivalent was able to seal the wound (Figure 5A, ii). After 7 days of cul-
ture, the boundary between the grafted acellular collagen sheet and the dermis
equivalent became unclear (Figure 5A, iii). The adhesion of the collagen sheet to
the dermis equivalent was strong enough to withstand the tensile strain produced
by repeated bending motions of the robotic finger (Figures 5B and 5C and Video S5).

We investigated cell migration into acellular collagen sheets of repaired planar-

shaped dermis equivalent. Figure 5D shows the wounded planar-shaped dermis
equivalent before and after the repair treatment. In the experiment, the planar-
shaped dermis equivalent was prepared in a frame-shaped device with anchors
and cultured for 3 days (Figure S6). A circular wound was punched at the center of
the dermis equivalent using biopsy (Figure 5D, i). The wound was then grafted using
an acellular collagen sheet (Figure 5D, ii), and cell migration into the acellular
collagen sheet was observed 3 and 7 days after the grafting of the collagen sheet
by staining the cells with phalloidin and Hoechst 33342 (Figure 5E). Images of the
sample (3 days after grafting) showed that the NHDFs had started to migrate into
the collagen sheet but had not reached the center of the collagen sheet (Figure 5E,
i). Images of the sample (7 days after grafting) revealed that the NHDFs had reached
the center of the sheet and that the overall cell density had increased (Figure 5E, ii).
With these results, it was confirmed that the NHDFs migrated into the collagen sheet
within culture, which resembled the phenomena observed at wounded sites in vivo,
such as cell migration into blood clots.27

To further confirm the repair treatment results, tensile tests were performed to check
the strength recovery of the repaired dermis equivalent. Figures S7A–S7C show the
setup that was used to measure the adhesion force between the grafted collagen
sheet and the dermis equivalent. The testing setup required two devices: a device
consisting of a frame with anchors (a.k.a. ‘‘dermis sheet holder’’) and a device for
hooking the collagen sheets (a.k.a. ‘‘collagen sheet jig’’). The collagen sheet, fabri-
cated to be attached to the collagen sheet jig, was grafted onto the dermis equivalent
fabricated in the dermis sheet holder. Following a designated incubation period after
grafting, the tensile test was performed in which the collagen sheet jig was pulled up
with the dermis sheet holder fixed (Figure S7D and Video S6). The collagen sheet jig
was attached to a rheometer probe connected to motors and force sensors, which
can generate the pulling motion and measure the tensile force simultaneously.

10 Matter 5, 1–19, July 6, 2022

 Please cite this article in press as: Kawai et al., Living skin on a robot, Matter (2022), https://doi.org/10.1016/j.matt.2022.05.019

ll
Article

Figure 5F illustrates the conditions for the three types of tissue samples that were
compared: acellular collagen without NHDFs (Sample A), dermis equivalent cultured
for 1 day (Sample B), and dermis equivalent cultured for 7 days after grafting of the
collagen sheet (Sample C). Graphs of the adhesion strength and the relationship of
tensile forces versus stretch width are shown in Figures 5G and S7E, respectively.
As a result, the presence of NHDFs in the dermis equivalent and the longer culture
time show higher adhesion strength. These results indicate that the repair process
is promoted by the activity of NHDFs.

DISCUSSION
The present work demonstrated the construction of a robotic finger covered with a
living material, i.e., skin equivalent. The anchoring structure located at the stem of
the finger allowed us to conformally cover the uneven surface of the robotic finger
by controlling tissue shrinkage. In addition, epidermal structures could be uniformly
formed by seeding cells from multiple directions. Furthermore, the wound was re-
paired by dermal fibroblast activity after the collagen sheet was applied to the
wound site. These results show the applicability of robots covered with living mate-
rials with biological functions and provide a new perspective on robotic materials.

The advantage of our method for covering 3D objects with skin equivalent is the uti-
lization of tissue shrinkage during culture, which enables conformal coverage of 3D
objects, especially those with curved and uneven surfaces. Alternatively, other
methods can also be considered to cover 3D objects such as tailoring the skin equiv-
alents fabricated in planar shapes; whole sheets of skin equivalents fabricated using
traditional methods can be measured, cut, and sutured/glued to enclose 3D objects.
However, this tailoring method has difficulty reproducing the subtle unevenness on
the surface of the object. It is difficult to cut, glue, or suture the endpoints of skin
equivalent without damaging the soft, fragile tissue. Furthermore, this approach re-
quires a substantial amount of time and labor to accomplish the task. Using our
method, the fabricated skin equivalent covering the robotic finger successfully repro-
duced the surface shape of the robotic finger with a thickness of 2.34 G 0.28 mm
(mean G SD). For these reasons, the method in this study is both theoretically and
practically suitable to use to cover 3D objects with skin equivalent. To further improve
fabrication efficiency, bioprinting methods, which have been used to fabricate skin
equivalents,28,29 could be modified to print dermis equivalents in a 3D shape and
to deposit keratinocytes onto the curved surfaces of dermis equivalents.

The shrinkage of the skin equivalent stemmed from the remodeling of the collagen
matrix by NHDFs. In theory, if NHDFs are uniformly distributed and have the same
level of activity, the localized shrinkage ratio should be the same across the whole
inner volume of the tissue. In practice, if this uniform localized shrinkage is countered
by the anchors set at the edges of a planar-shaped skin equivalent sample, the skin
equivalent will be stretched and become a thin sheet with uniform thickness (except
near the edges).11,12 However, judging from our results, the formation of stretched
tissues with a uniform thickness also requires tissues to have a uniform initial thick-
ness before shrinkage, which is achieved by the usage of the cap part to refine the
shape of the initially cross-linked collagen gel.

There is still room to improve the mechanical strength of the skin equivalent. The tensile
strength of our dermis equivalent was approximately 5.6 kPa (Figure S5), which was far
less than the tensile strength of in vivo skin tissue (10–30 MPa).30 We speculate that this
lack of tensile strength was mainly due to the low initial collagen concentration and the

Matter 5, 1–19, July 6, 2022 11

 Please cite this article in press as: Kawai et al., Living skin on a robot, Matter (2022), https://doi.org/10.1016/j.matt.2022.05.019

ll
Article

insufficient remodeling of the collagen matrix by the cells. The collagen density of in vivo
forearm skin was approximately 100–200 mg/mL.31 On the other hand, the initial
collagen density of the NHDF mixture solution used for the fabrication of dermis equiv-
alent was 3 mg/mL. Even considering the shrinkage of the skin equivalent (Figure S3),
the terminal density was approximately 10 mg/mL. It is expected that a higher collagen
concentration in the matured (and therefore stronger) tissue can be achieved with a
higher initial collagen concentration. For example, meniscal tissue with enhanced me-
chanical strength and biochemical properties was fabricated with a high initial collagen
concentration (10–20 mg/mL).32 The insufficient remodeling of the collagen matrix by
the cells could also affect the mechanical strength of skin equivalent. It has been re-
ported that in vivo collagenous tissue consists of collagen fibrils composed of a stag-
gered array of ultralong tropocollagen molecules that can maximize strength and pro-
vide large energy dissipation during deformation, thus creating tough and robust
tissue.33 Such nanostructures of in vivo collagen fibrils are believed to be mainly
produced by the cells, as indicated by the experimental results showing increased me-
chanical strength after in vivo implantation34 or demonstrating longer periods of in vitro
tissue maturation.35 Therefore, fine-tuning of fibroblast concentration and pheno-
types,36 as well as longer periods of maturation culture, could be adopted in our future
experiments to further enhance the mechanical properties of the skin equivalents for ro-
botic applications.

Durability is also important when considering the use of skin equivalent as a robotic
covering material. A previous study reported that cultured skin equivalents did not
exhibit damages after repeated stretch (10% deformation, 0.2 Hz) for 4 days.12

Other challenges remain in the longevity and functionalization of the skin. First, the
skin equivalent assessed in this study cannot survive in the air for a long period of
time. In our experiment, the robots were maintained in culture medium before being
taken out to perform tasks in an indoor room environment without intentionally
elevated or decreased humidity. In its current form, the robot should not be used
in a dry environment, since even the human skin cannot be exposed to a dry environ-
ment for a long period of time without proper water supply from the circulatory sys-
tems and localized wetting by the secretion of sweating glands. To avoid such drying
out, building perfusion channels within and beneath the dermis equivalent to mimic
blood vessels to supply water, as well as the integration of sweating glands in the
skin equivalent, are important directions for future research. Second, cultured skin
has the potential to be a multifunctional robot covering material because it can serve
as a matrix for various biological functions, such as sensing37 and temperature con-
trol,38 that such materials require. A future challenge is to reproduce the compo-
nents of the skin, including nerves and appendages (i.e., sweat glands, nails, etc.),
so that robots can acquire these skin-specific functions.

In this work, the robot finger was designed to be driven by pulling wires to avoid dete-
rioration of electronic actuators and wirings caused by the wet culture conditions. To
further improve the functionality and physical autonomy of the robots, the establish-
ment of a robotic architecture with coexisting perfusable artificial blood vessel network
and biocompatible waterproof electronics is an important future focus to not only main-
tain the viability of the living skin but also allow the integration of electrical processors,
motors, sensors, wires, batteries, communication modules, etc. Alternatively, with the
advancements on the biofabrication of functional muscle tissues39 and nerves,40 supe-
rior actuation and sensing features of the living organisms, such as quite actuation and
highly selective chemical detection, could be recruited to further improve the biohybrid
robots by the integration of more tissue-engineered elements.

12 Matter 5, 1–19, July 6, 2022

 Please cite this article in press as: Kawai et al., Living skin on a robot, Matter (2022), https://doi.org/10.1016/j.matt.2022.05.019

ll
Article

Taken together, this research is a major step toward biohybrid robots that have a
combination of living materials and artificial materials. Biohybrid robots would
exhibit superior sensory capabilities, highly efficient energy conversion, self-organi-
zation, and self-repair functions like living organisms, which have thus far been diffi-
cult to achieve for artificial materials alone. In this study, by covering a robotic finger
with skin equivalent, we demonstrated the possibility of not only reconstructing the
visual texture of skin but also replicating the unique capabilities of living organisms
using biomaterials that are able to perform in the air and exhibit self-repair. We
believe that these results will create a new paradigm for biohybrid robots.

EXPERIMENTAL PROCEDURES
Resource availability
Lead contact
Further information and requests for resources and reagents should be directed to
and will be fulfilled by the lead contact, Shoji Takeuchi ([email protected]
tokyo.ac.jp).

Material availability
This study did not generate new unique reagents.

Data and code availability

Any additional information required to reanalyze the data reported in this paper is
available from the lead contact upon request.

Cell culture
NHDFs (PromoCell, Heidelberg) were cultured in fibroblast growth medium (Dul-
becco’s modified Eagle’s medium [Sigma-Aldrich, St. Louis, MO] supplemented
with 10% fetal bovine serum [Biosera, Kansas City, MO], 1% penicillin-streptomycin
[Sigma-Aldrich, St. Louis, MO, USA], and 70 mg/mL L-ascorbic acid phosphate
magnesium salt n-hydrate [Wako Pure Chemical Industries, Osaka]). NHEKs
(PromoCell, Heidelberg) were cultured in keratinocyte growth medium (Keratino-
cyte Growth Medium 2 Kit [PromoCell, Heidelberg] supplemented with 1% peni-
cillin-streptomycin). For both NHDFs and NHEKs, media were replenished once
every 2 days, and the cells were subcultured at approximately 80% confluence.
NHDFs were used within 10 passages, and NHEKs were used within eight pas-
sages. Fibroblast growth medium was also used for the maturation of dermis
equivalent. Keratinocyte differentiation medium (1:1 mixture of fibroblast growth
medium and keratinocyte growth medium) was used when culturing NHEKs on
dermis equivalent samples.

Skin equivalent fabrication

In this paper, cross-linked and cultured collagen solution containing NHDFs is
referred to as ‘‘dermis equivalent,’’ and the dermis equivalent covered with migrated
and differentiated NHEKs is referred to as ‘‘skin equivalent.’’ The general process to
fabricate the skin equivalent began with casting a collagen hydrogel containing
dermal fibroblasts. Then, the hydrogel was cultured to mature into a dermis equiv-
alent, in which the fibroblasts adhered to and spread within the collagen matrix and
caused the gel to shrink. Next, the keratinocytes were seeded and cultured on top of
the dermis equivalent to complete the creation of the skin equivalent.21

In our experiment, a dermis equivalent was fabricated by crosslinking the collagen

solution containing NHDFs (a 9:1:5 mixed solution of type I collagen solution
[IAC-50] [Koken, Tokyo], 103 PBS( ) [Sigma-Aldrich, St. Louis, MO], and fibroblast

Matter 5, 1–19, July 6, 2022 13

 Please cite this article in press as: Kawai et al., Living skin on a robot, Matter (2022), https://doi.org/10.1016/j.matt.2022.05.019

ll
Article

growth medium containing 1.0 3 106 or 4.0 3 106 cells/mL NHDFs). Collagen solu-
tion crosslinking was achieved by placing the solution in an incubator (37 C, 5% CO2)
for more than 1 h. For the fabrication of skin equivalent covering the fingertip device
or the robotic finger, the NHDF concentration in the medium was 1.0 3 106 cells/mL.
For the experiments related to the repair of the dermis equivalent, the NHDF con-
centration in the medium was 4.0 3 106 cells/mL. The concentration of NHDFs for
the latter was increased to enhance the repair of the dermis equivalent.

Fabrication of the robotic finger covered with skin equivalent

As shown in Figure S1, the three-joint robotic finger consisted of five parts and had
an anchoring structure at the bottom part (i.e., part 5). All parts were printed by a 3D
printer (AGILISTA-3100, KEYENCE, Osaka) and conformally coated with parylene C
(4 mm thick) using a parylene evaporator (Parylene Deposition System 2010, Spe-
cialty Coating Systems, Indianapolis). This was done to prevent the release of cyto-
toxic substances from the 3D-printed devices. Each part was assembled through an
axis, and two nylon wires were threaded through the center for joint movement.
After assembly, the robotic finger was set into the cylindrical PDMS case for better
oxygen permeability. The cylindrical PDMS case was fabricated by solidifying the
PDMS in a 3D-printed mold. Then, the cap part was inserted into the top of the cy-
lindrical PDMS case to render a uniform space around the robotic finger. The surface
roughness of the 3D-printed parts was evaluated using a laser microscope as shown
in Figure S10.

As shown in Figure 1C, the procedures to cover the robotic finger with skin equiva-
lent consisted of two major steps: (1) the construction of a dermis equivalent that
covered the robotic finger and (2) the construction of the epidermis on the surface
of the dermis equivalent. First, to prepare the dermis equivalent, the collagen solu-
tion containing NHDFs, prepared on ice, was poured into the cylindrical PDMS case
through the hole of the cap part at room temperature (RT). Immediately after pour-
ing, the cylindrical PDMS case was placed upright in an incubator (37 C, 5% CO2) to
crosslink the collagen and subsequently culture the tissue for maturation (Figure 1C,
i). During maturation (R3 days), cells adhered to and remodeled the collagen ma-
trix,41 and the tissue shrank (Figure 1C, ii). In this paper, the shrunken tissue (with
R3 days of maturation culture) is referred to as ‘‘dermis equivalent.’’ Due to
shrinkage after 3 days of culture, the dermis equivalent covered the robotic finger
tightly. Second, to cover the dermis equivalent with the epidermis, 200 mL of
epidermal medium containing NHEKs at a concentration of 1.0 3 107 cells/mL
was gently poured onto the dorsal surface of the finger twice and cultured for
1 day (Figure 1C, iii). After that, the robot was rotated 180 in the PDMS case so
that the palmar surface of the finger faced upward, and the same amount of
NHEKs was seeded and cultured for an additional 2 weeks to complete the fabrica-
tion of the skin equivalent (Figure 1C, iv).

Fabrication of the fingertip device covered with skin equivalent

The fingertip device consists of a mold part and a circular anchor part (Figure S2A).
The mold part was inserted through the center hole of the circular anchor part. Both
parts were printed by a 3D printer and coated with parylene C (4 mm thick).

The process of tissue formation was the same as the formation of the tissue covering
the robotic finger, with the device replaced by the fingertip device, as illustrated in
Figure S2B. In detail, the collagen solution containing NHDFs was poured into a cy-
lindrical PDMS case containing a fingertip device and a cap part. Immediately after
pouring the collagen solution containing NHDFs, the cylindrical PDMS case was

14 Matter 5, 1–19, July 6, 2022

 Please cite this article in press as: Kawai et al., Living skin on a robot, Matter (2022), https://doi.org/10.1016/j.matt.2022.05.019

ll
Article

placed upright in an incubator. Then, NHEKs were seeded on the surface of the
dermis equivalent from two sides; the details of this procedure are as follows. First,
the fabricated dermis equivalent covering the fingertip device was set in the PDMS
case. Then, 100 mL of keratinocyte differentiation medium containing NHEKs at a
concentration of 1.0 3 107 cells/mL was added to the dermis equivalent. The next
day, the dermis equivalent (seeded with NHEKs from one side) was rotated 180
in the PDMS case and again seeded with the same amount of NHEKs. Dermis equiv-
alent (seeded with NHEKs from both sides) was then cultured in keratinocyte differ-
entiation medium for 2 weeks to allow the migration and differentiation of NHEKs; at
this point, the formation of skin equivalent covering the fingertip mold was
complete.

Histological analysis of the skin equivalent

The preparation of the frozen sections and the staining procedures for investigating
the morphology of the skin equivalent were as follows. First, the samples were fixed
by immersion in PBS( ) (Muto Pure Chemicals, Tokyo) for 12–24 h. The skin equiva-
lent covering fingertip mold was fixed while still attached to the fingertip device;
since the fixation resulted in mechanically stiffer tissues, detaching the skin equiva-
lent after the fixation was preferred to prevent disrupting the shape of the tissues
caused by handling. The fixed tissues were embedded in optimal cutting tempera-
ture compound (Sakura Finetek Japan, Japan), frozen using liquid nitrogen, and
sliced into 8-mm-thick frozen sections using a cryostat (CryoStar NX50 Cryostat,
Thermo Fisher Scientific, USA). For H&E staining, the frozen sections were stained
with Mayer’s hematoxylin and 0.5% eosin Y ethanol solution (Wako Pure Chemical
Industries, Osaka). For immunostaining, the frozen sections were washed with 13
PBS( ) (Cell Science & Technology Institute, Sendai, Japan), blocked with 1% BSA
(A2153-50G, Sigma-Aldrich) in PBS( ) (BSA solution), and incubated in cytokeratin
10 (CK10) antibody (rabbit, monoclonal) or cytokeratin 15 (CK15) antibody (mouse,
monoclonal) (Abcam, Cambridge) solution (1:200 diluted using BSA solution) at 4 C
for 1 day. Then, the sections were washed with BSA solution and incubated in Alexa
Fluor 488/568 goat anti-rabbit/mouse IgG secondary antibody (Invitrogen, Massa-
chusetts) solution (1:200 diluted using BSA solution) for 2 h at RT. Microscopic im-
ages of the stained sections were acquired with an inverted microscope (IX71,
Olympus, Tokyo).

Evaluation of the fabrication method of the 3D skin equivalent

We evaluated the effectiveness of the following three design points for the fabrica-
tion of uniformly shaped 3D skin equivalent: anchor structure, the cap part, and
multidirectional NHEK seeding.

To evaluate the effectiveness of the anchoring structure and the cap part on the for-
mation of uniformly shaped 3D dermis equivalent, a dermis equivalent was fabri-
cated to cover both the fingertip devices and the negative-control devices (i.e., de-
vices without anchoring structures or cap parts). All the samples were fabricated by
crosslinking collagen containing NHDFs and culturing for 7 days. For the anchoring
structures, the lengths of retreated tissue coverage on the fingertip devices and
negative-control devices (without anchoring structures) were measured using Im-
ageJ (NIH, Bethesda, MD). For the cap part, the thickness of the fabricated dermis
equivalent was measured at the side, corner, and top points (A, B, and C, respec-
tively, as shown in Figure 2E). Since side and corner points are not on the symmet-
rical axis, thickness values of the dermis equivalent on the left and right sides were
averaged. Thickness measurements were performed in ImageJ based on H&E stain-
ing images of the frozen sections of the tissues.

Matter 5, 1–19, July 6, 2022 15

 Please cite this article in press as: Kawai et al., Living skin on a robot, Matter (2022), https://doi.org/10.1016/j.matt.2022.05.019

ll
Article

In the experiments examining the seeding method of NHEKs, NHEKs were seeded
and cultured in three different conditions: 2.0 3 106 NHEKs seeded on one side and
cultured for 3 days (A), 2.0 3 106 NHEKs seeded on one side and cultured for 14 days
(B), and 1.0 3 106 NHEKs seeded on two sides and cultured for 14 days (C). In all
samples, NHEKs were seeded on dermis equivalent and cultured for 3 days after
collagen crosslinking. Representative H&E staining images of the frozen sections
are shown in Figure S9. For the calculation of the epidermal covering rate, images
of three frozen sections acquired at the tip, middle, and bottom of each finger sam-
ple were analyzed with averaged results to better represent the covering rate of the
whole finger. The coverage rate of each frozen section was determined by analyzing
the H&E-stained samples with ImageJ and measuring the percentage of epidermis
present in the outer edge of the sample.

Electrical measurements on the fabricated skin equivalent

For the measurements of the differences in electrical properties of samples due to
a properly formed epidermis, dermis equivalent samples that were half-covered
with epidermis were fabricated. The samples were prepared by seeding
2.0 3 106 NHEKs from only one side and culturing for 3 days after seeding. Rather
than culturing for 14 days, the shortened incubation time (3 days) was chosen to
suppress unwanted NHEK migration to partially cover the unseeded side. For
the electrical measurements, we fabricated a modified version of the mold part
of the fingertip device, which was made of PDMS, with two pairs of embedded
electrodes (referred to as the ‘‘PDMS fingertip mold’’) (Figure S4A). The PDMS
fingertip mold was fabricated by pouring liquid PDMS into a mold hole containing
electrodes beforehand. During measurement, the skin equivalent (together with
the anchor part) was detached from the mold part and was assembled with the
PDMS fingertip mold so that the electrodes in the PDMS fingertip mold could
be applied to the inner surface of the skin equivalent (Figure S4B). Next, with
another pair of electrodes placed on the outer surface of the skin equivalent,
the electrical characteristics of the tissue can be effectively measured (Figure S4C).
The capacitance and resistance between the electrodes were measured using an
LCR meter (IM3533-01, HIOKI E.E., Tokyo).

Water retention test

As shown in Figure S4D, the skin equivalent used for the water retention test was pre-
pared by seeding 1.0 3 106 NHEKs from two sides of the dermis equivalent and
culturing it for 14 days. The samples without epidermis were used as controls, which
comprised dermis equivalent cultured for 17 days after the crosslinking of collagen
solution containing NHDFs. After the prescribed incubation periods, both tissues
(with the fingertip device) were removed from the incubators and exposed to air
for 30 min. Then, the mold parts of the fingertip devices were removed, and 1 mL
of PBS( ) was added to the inner space of each dermis equivalent or skin equivalent.
Leaked liquids were collected by placing a Petri dish beneath the tissue, and the
liquid was weighed by an analytical balance (GX6100-00A00, A&D, Tokyo) once
every 10 min for 60 min (Figure S4E).

Measurement of dermis equivalent tensile strength

Tensile strength tests of the dermis equivalent were performed as follows. The
dermis equivalent samples were cultured using the devices illustrated in
Figures S8A–S8C. After the designated culture periods, the two anchor parts were
fitted to the shaft for disposable measuring systems of a rheometer (MCR rheometer
302, Anton Paar, Tokyo) and the 3D-printed base. A tensile test was performed by
pulling the probe upward at a speed of 500 mm/s, and the tensile force was recorded

16 Matter 5, 1–19, July 6, 2022

 Please cite this article in press as: Kawai et al., Living skin on a robot, Matter (2022), https://doi.org/10.1016/j.matt.2022.05.019

ll
Article

once every 0.25 s (Figure S8D). The stress was calculated by dividing the force by the
undeformed cross-sectional area of the skin tissue.

The repair treatment of the wounded dermis equivalent covering the robotic
finger
The robotic finger covered with dermis equivalent was fabricated by the process
adopted from the previous section. The pink color of the dermis equivalent was in-
herited from the fibroblast growth medium, which contains phenol red. The skin
equivalent was less pink since the keratinocyte differentiation medium contained
less phenol red. After fabrication of the dermis equivalent covering the robotic
finger, the tissue was wounded by cutting it with a scalpel. Then, an acellular
collagen hydrogel sheet was dipped in collagen solution and grafted onto the tissue
to cover the wounded area. The collagen sheets were found to adhere to the tissue
with the crosslinking of the dipped collagen solution and was shown to fuse with the
wounded tissue after 7 days of incubation.

To investigate cell migration from the dermis equivalent into the acellular collagen
sheet, NHDFs were stained with Alexa Fluor 488 phalloidin (phalloidin) (Thermo
Fisher Scientific, Tokyo) and Hoechst 33342 (Invitrogen, CA, Carlsbad). In detail,
two planar-shaped dermis equivalents (referred to as ‘‘dermis sheets’’) were fabri-
cated, as shown in Figure S6A, after 3 days of incubation after collagen crosslinking.
In addition, a hole (diameter: 6 mm) was punched into each tissue sample using a
biopsy punch (Figure S6B). Then, a collagen sheet was grafted onto each dermis
sheet to cover the holes. The samples were cultured for 3 or 7 days, fixed with 4%
PFA for 1 day at 4 C, and treated with 0.1% Triton X-100 (Thermo Fisher Scientific,
Tokyo) for 0.5 h at RT. Then, cell nuclei and cytoskeleton were stained with phalloidin
and Hoechst 33342. The stained tissues were imaged under an inverted microscope
(IX71, Olympus, Tokyo, Japan).

To show the restoration of the adhesion strength of the repaired tissues, tensile tests
were performed as follows. The dermis sheets were cultured using the devices illus-
trated in Figure S7A. After 3 days of gelation, the wound holes (diameter: 6 mm)
were punched at the center of the samples using a biopsy punch. Then, acellular
collagen sheets attached to the collagen sheet jigs (prepared according to the illus-
trated steps in Figure S7B) were grafted onto the dermis equivalent samples to cover
the wound holes. Before grafting, the collagen sheets were dipped in collagen solu-
tion to promote adherence to the dermis equivalent. To investigate the effect of the
activity of NHDFs on the changes in adhesion strength, three different types of sam-
ples were prepared. The first was the control sample consisting of collagen without
NHDFs (Sample A), which was fabricated by mixing the collagen solution and the
fibroblast growth medium containing no NHDFs. Sample A was grafted with acel-
lular collagen sheets (as described above) and cultured in fibroblast growth medium
for another 7 days. The second type of sample was the one with a short culture
period (Sample B), which was grafted with the acellular collagen sheet and then
cultured in fibroblast growth medium for 1 more day, for a total of 8 days. The third
type of sample was the one with a long culture period (Sample C), which was incu-
bated 7 days after the grafting of the acellular collagen sheet. After the designated
culture periods, the collagen sheet jig was fitted to the shaft for disposable
measuring systems (D-CP/PP7) of a rheometer, and the dermis sheet holder was
fixed to a 3D-printed base. A tensile test was performed to measure the adhesion
strength between the grafted collagen sheet and the dermis equivalent by pulling
the probe upward at a speed of 500 mm/s, and the tensile force was recorded
once every 0.25 s (Figure S7D).

Matter 5, 1–19, July 6, 2022 17

 Please cite this article in press as: Kawai et al., Living skin on a robot, Matter (2022), https://doi.org/10.1016/j.matt.2022.05.019

ll
Article

Statistical analysis
We evaluated statistically significant differences by independent Student’s t tests or
Tukey-Kramer tests. All statistical analyses were performed using the statistical soft-
ware R (http://www.R-project.org). All data are represented as the average G SD.

SUPPLEMENTAL INFORMATION
Supplemental information can be found online at https://doi.org/10.1016/j.matt.
2022.05.019.

ACKNOWLEDGMENTS
We thank Byeongwook Jo and Dina Myasnikova for the helpful discussions and
support.

Funding: JSPS Grants-in-Aid for Scientific (KAKENHI), Grant Number 16H06329.

JSPS Grants-in-Aid for Scientific (KAKENHI), Grant Number 21H00321. JSPS
Grants-in-Aid for Scientific (KAKENHI), Grant Number 21H05013. JSPS Grants-in-
Aid for Scientific (KAKENHI), Grant Number 19H04508. JSPS Grant-in-Aid for
Early-Career Scientists (KAKENHI), Grant Number 19K15415.

AUTHOR CONTRIBUTIONS
Conceptualization: M.K., M.N., Y.M., and S.T.; methodology: M.K., M.N., and H.O.;
investigation: M.K.; visualization: M.K., M.N., and H.O.; funding acquisition: M.N.,
H.O., Y.M., and S.T.; project administration: Y.M. and S.T.; supervision: Y.M. and
S.T.; writing – original draft: M.K. and N.M.; writing – review & editing: M.K.,
N.M., H.O., Y.M., and S.T.

DECLARATION OF INTERESTS
The authors declare no competing interests.

Received: December 10, 2021

Revised: March 19, 2022
Accepted: May 10, 2022
Published: June 9, 2022

REFERENCES
1. Salvador, M.J., Silver, S., and Mahoor, M.H. Computational Linguistics, pp. 553–561. Intelligence - WACI 2011: 2011 Workshop on
(2015). An emotion recognition comparative https://doi.org/10.3115/1075096.1075166. Affective Computational Intelligence,
study of autistic and typically-developing pp. 22–29. https://doi.org/10.1109/WACI.
children using the zeno robot. In 2015 IEEE 5. Mathur, M.B., and Reichling, D.B. (2016). 2011.5953147.
International Conference on Robotics and Navigating a social world with robot partners: a
Automation (ICRA), pp. 6128–6133. https://doi. quantitative cartography of the Uncanny 9. Rosenthal-von der Pütten, A.M., Krämer, N.C.,
org/10.1109/ICRA.2015.7140059. Valley. Cognition 146, 22–32. https://doi.org/ Maderwald, S., Brand, M., and Grabenhorst, F.
10.1016/j.cognition.2015.09.008. (2019). Neural mechanisms for accepting
2. Robinson, H., MacDonald, B., and Broadbent, and rejecting artificial social partners in the
6. Terryn, S., Brancart, J., Lefeber, D., Van Assche,
E. (2014). The role of healthcare robots for uncanny valley. J. Neurosci. 39, 6555–6570.
G., and Vanderborght, B. (2017). Self-healing
older people at home: a review. Int. J. Soc. https://doi.org/10.1523/JNEUROSCI.2956-18.
soft pneumatic robots. Sci. Robot. 2, eaan4268.
Robot. 6, 575–591. https://doi.org/10.1007/ 2019.
https://doi.org/10.1126/scirobotics.aan4268.
s12369-014-0242-2.
7. Kaneko, K., Kanehiro, F., Morisawa, M., Miura, 10. Bell, E., Ehrlich, H.P., Buttle, D.J., and
3. Tuomi, A., Tussyadiah, I.P., and Stienmetz, J. K., Nakaoka, S., and Kajita, S. (2009). Nakatsuji, T. (1981). Living tissue formed in vitro
(2021). Applications and implications of service Cybernetic human HRP-4C. In 9th IEEE-RAS and accepted as skin-equivalent tissue of full
robots in hospitality. Cornell Hosp. Q. 62, International Conference on Humanoid thickness. Science 211, 1052–1054. https://doi.
232–247. https://doi.org/10.1177/ Robots, HUMANOIDS09, pp. 7–14. https://doi. org/10.1126/science.7008197.
1938965520923961. org/10.1109/ICHR.2009.5379537.
11. Mori, N., Morimoto, Y., and Takeuchi, S. (2017).
4. Nakano, Y.I., Reinstein, G., Stocky, T., and 8. Becker-Asano, C., and Ishiguro, H. (2011). Skin integrated with perfusable vascular
Cassell, J. (2003). Towards a model of face-to- Evaluating facial displays of emotion for the channels on a chip. Biomaterials 116, 48–56.
face grounding. In Proceedings of the 41st android robot Geminoid F. In IEEE SSCI 2011 - https://doi.org/10.1016/j.biomaterials.2016.
annual meeting of the Association for Symposium Series on Computational 11.031.

18 Matter 5, 1–19, July 6, 2022

 Please cite this article in press as: Kawai et al., Living skin on a robot, Matter (2022), https://doi.org/10.1016/j.matt.2022.05.019

ll
Article

12. Mori, N., Morimoto, Y., and Takeuchi, S. (2018). development. Adv. Drug Deliv. Rev. 69–70, skin collagen and density. Br. J. Dermatol. 93,
Perfusable and stretchable 3D culture system 81–102. https://doi.org/10.1016/j.addr.2013. 639–643. https://doi.org/10.1111/j.1365-2133.
for skin-equivalent. Biofabrication 11, 011001. 12.006. 1975.tb05113.x.
https://doi.org/10.1088/1758-5090/aaed12.
22. Rawlings, A.V., and Harding, C.R. (2004). 32. Puetzer, J.L., and Bonassar, L.J. (2013). High
13. Lee, J., Rabbani, C.C., Gao, H., Steinhart, M.R., Moisturization and skin barrier function. density type I collagen gels for tissue
Woodruff, B.M., Pflum, Z.E., Kim, A., Heller, S., Dermatol. Ther. 17, 43–48. https://doi.org/10. engineering of whole menisci. Acta Biomater.
Liu, Y., Shipchandler, T.Z., and Koehler, K.R. 1111/j.1396-0296.2004.04s1005.x. 9, 7787–7795. https://doi.org/10.1016/j.actbio.
(2020). Hair-bearing human skin generated 2013.05.002.
entirely from pluripotent stem cells. Nature 23. Goretsky, M.J., Supp, A.P., Greenhalgh, D.G.,
582, 399–404. https://doi.org/10.1038/s41586- Warden, G.D., and Boyce, S.T. (1995). Surface 33. Buehler, M.J. (2006). Nature designs tough
020-2352-3. electrical capacitance as an index of epidermal collagen: explaining the nanostructure of
barrier properties of composite skin substitutes collagen fibrils. Proc. Natl. Acad. Sci. U S A 103,
14. Matsusaki, M., Fujimoto, K., Shirakata, Y., and skin autografts. Wound Repair Regen. 3, 12285–12290. https://doi.org/10.1073/pnas.
Hirakawa, S., Hashimoto, K., and Akashi, M. 419–425. https://doi.org/10.1046/j.1524-475X. 0603216103.
(2015). Development of full-thickness human 1995.30406.x.
skin equivalents with blood and lymph-like 34. Sander, E.A., Lynch, K.A., and Boyce, S.T.
capillary networks by cell coating technology. 24. Sriram, G., Alberti, M., Dancik, Y., Wu, B., Wu, (2014). Development of the mechanical
J. Biomed. Mater. Res. A 103, 3386–3396. R., Feng, Z., Ramasamy, S., Bigliardi, P.L., properties of engineered skin substitutes after
https://doi.org/10.1002/jbm.a.35473. Bigliardi-Qi, M., and Wang, Z. (2018). Full- grafting to full-thickness wounds. J. Biomech.
thickness human skin-on-chip with enhanced Eng. 136, 051008. https://doi.org/10.1115/1.
15. Eaglstein, W.H., and Falanga, V. (1997). Tissue epidermal morphogenesis and barrier 4026290.
engineering and the development of Apligraf, function. Mater. Today 21, 326–340. https://
a human skin equivalent. Cutis 19, 894–905. doi.org/10.1016/j.mattod.2017.11.002. 35. Lafrance, H., Guillot, M., Germain, L., and
https://doi.org/10.1016/S0149-2918(97)80043- Auger, F.A. (1995). A method for the evaluation
4. 25. Ma, Y., Feng, X., Rogers, J.A., Huang, Y., and of tensile properties of skin equivalents. Med.
Zhang, Y. (2017). Design and application of ‘‘J- Eng. Phys. 17, 537–543. https://doi.org/10.
16. Falanga, V., Margolis, D., Alvarez, O., Auletta, shaped’’ stress-strain behavior in stretchable 1016/1350-4533(95)00012-C.
M., Maggiacomo, F., Altman, M., Jensen, J., electronics: a review. Lab Chip 17, 1689–1704.
Sabolinski, M., and Hardin-Young, J. (1998). https://doi.org/10.1039/c7lc00289k. 36. Sriram, G., Bigliardi, P.L., and Bigliardi-Qi, M.
Rapid healing of venous ulcers and lack of (2015). Fibroblast heterogeneity and its
clinical rejection with an allogeneic cultured 26. Supp, D.M., and Boyce, S.T. (2005). Engineered
implications for engineering organotypic skin
human skin equivalent. Human Skin Equivalent skin substitutes: practices and potentials. Clin.
models in vitro. Eur. J. Cell Biol. 94, 483–512.
Investigators Group. Arch. Dermatol. 134, Dermatol. 23, 403–412. https://doi.org/10.
https://doi.org/10.1016/j.ejcb.2015.08.001.
293–300. https://doi.org/10.1001/archderm. 1016/j.clindermatol.2004.07.023.
134.3.293. 37. Osborn, L.E., Dragomir, A., Betthauser, J.L.,
27. Singer, A.J., and Clark, R.A. (1999). Cutaneous
wound healing. N. Engl. J. Med. 341, 738–746. Hunt, C.L., Nguyen, H.H., Kaliki, R.R., and
17. Brem, H., Balledux, J., Bloom, T., Kerstein,
https://doi.org/10.1056/ Thakor, N.V. (2018). Prosthesis with
M.D., and Hollier, L. (2000). Healing of diabetic
NEJM199909023411006. neuromorphic multilayered e-dermis perceives
foot ulcers and pressure ulcers with human skin
touch and pain. Sci. Robot. 3. https://doi.org/
equivalent: a new paradigm in wound healing.
28. Baltazar, T., Merola, J., Catarino, C., Xie, C.B., 10.1126/scirobotics.aat3818.
Arch. Surg. 135, 627–634. https://doi.org/10.
1001/archsurg.135.6.627. Kirkiles-Smith, N.C., Lee, V., Hotta, S., Dai, G.,
Xu, X., Ferreira, F.C., et al. (2020). Three 38. Mishra, A.K., Wallin, T.J., Pan, W., Xu, P., Wang,
18. Xie, Y., Rizzi, S.C., Dawson, R., Lynam, E., dimensional bioprinting of a vascularized and K., Giannelis, E.P., Mazzolai, B., and Shepherd,
Richards, S., Leavesley, D.I., and Upton, Z. perfusable skin graft using human R.F. (2020). Autonomic perspiration in 3D-
(2010). Development of a three-dimensional keratinocytes, fibroblasts, pericytes, and printed hydrogel actuators. Sci. Robot. 5, 1–10.
human skin equivalent wound model for endothelial cells. Tissue Eng. A 26, 227–238. https://doi.org/10.1126/scirobotics.aaz3918.
investigating novel wound healing therapies. https://doi.org/10.1089/ten.tea.2019.0201.
39. Morimoto, Y., Onoe, H., and Takeuchi, S.
Tissue Eng. Part C Methods 16, 1111–1123.
29. Hakimi, N., Cheng, R., Leng, L., Sotoudehfar, (2018). Biohybrid robot powered by an
https://doi.org/10.1089/ten.tec.2009.0725.
M., Ba, P.Q., Bakhtyar, N., Amini-Nik, S., antagonistic pair of skeletal muscle tissues. Sci.
19. Chioni, A.M., and Grose, R. (2008). Organotypic Jeschke, M.G., and Günther, A. (2018). Robot. 3, eaat4440. https://doi.org/10.1126/
modelling as a means of investigating Handheld skin printer: in situ formation of scirobotics.aat4440.
epithelial-stromal interactions during planar biomaterials and tissues. Lab Chip 18,
tumourigenesis. Fibrogenesis Tissue Repair 1, 1440–1451. https://doi.org/10.1039/ 40. Maksimovic, S., Nakatani, M., Baba, Y., Nelson,
8–9. https://doi.org/10.1186/1755-1536-1-8. c7lc01236e. A.M., Marshall, K.L., Wellnitz, S.A., Firozi, P.,
Woo, S.H., Ranade, S., Patapoutian, A., and
20. Küchler, S., Wolf, N.B., Heilmann, S., Weindl, 30. Nı́ Annaidh, A., Bruyère, K., Destrade, M., Lumpkin, E.A. (2014). Epidermal Merkel cells
G., Helfmann, J., Yahya, M.M., Stein, C., and Gilchrist, M.D., and Otténio, M. (2012). are mechanosensory cells that tune
Schäfer-Korting, M. (2010). 3D-Wound healing Characterization of the anisotropic mechanical mammalian touch receptors. Nature 509,
model: influence of morphine and solid lipid properties of excised human skin. J. Mech. 617–621. https://doi.org/10.1038/nature13250.
nanoparticles. J. Biotechnol. 148, 24–30. Behav. Biomed. Mater. 5, 139–148. https://doi.
https://doi.org/10.1016/j.jbiotec.2010.01.001. org/10.1016/j.jmbbm.2011.08.016. 41. Grinnell, F. (2003). Fibroblast biology in three-
dimensional collagen matrices. Trends Cell
21. Mathes, S.H., Ruffner, H., and Graf-Hausner, U. 31. Shuster, S., Black, M.M., and McVitie, E. (1975). Biol. 13, 264–269. https://doi.org/10.1016/
(2014). The use of skin models in drug The influence of age and sex on skin thickness, S0962-8924(03)00057-6.

Matter 5, 1–19, July 6, 2022 19