Adipose Tissue Engineering in Three-Dimensional Levitation
Tissue Culture System Based on Magnetic Nanoparticles
Alexes C. Daquinag, Ph.D.,1Glauco R. Souza, Ph.D.,2and Mikhail G. Kolonin, Ph.D.1
White adipose tissue (WAT) is becoming widely used in regenerative medicine/cell therapy applications, and its
physiological and pathological importance is increasingly appreciated. WAT is a complex organ composed of
differentiated adipocytes, stromal mesenchymal progenitors known as adipose stromal cells (ASC), as well as
endothelial vascular cells and infiltrating leukocytes. Two-dimensional (2D) culture that has been typically used
for studying adipose cells does not adequately recapitulate WAT complexity. Improved methods for recon-
struction of functional WAT ex vivo are instrumental for understanding of physiological interactions between the
composing cell populations. Here, we used a three-dimensional (3D) levitation tissue culture system based on
magnetic nanoparticle assembly to model WAT development and growth in organoids termed adipospheres.
We show that 3T3-L1 preadipocytes remain viable in spheroids for a long period of time, while in 2D culture,
they lose adherence and die after reaching confluence. Upon adipogenesis induction in 3T3-L1 adipospheres,
cells efficiently formed large lipid droplets typical of white adipocytes in vivo, while only smaller lipid droplet
formation is achievable in 2D. Adiposphere-based coculture of 3T3-L1 preadipocytes with murine endothelial
bEND.3 cells led to a vascular-like network assembly concomitantly with lipogenesis in perivascular cells.
Adipocyte-depleted stromal vascular fraction (SVF) of mouse WAT cultured in 3D underwent assembly into
organoids with vascular-like structures containing luminal endothelial and perivascular stromal cell layers.
Adipospheres made from primary WAT cells displayed robust proliferation and complex hierarchical organi-
zation reflected by a matricellular gradient incorporating ASC, endothelial cells, and leukocytes, while ASC
quickly outgrew other cell types in adherent culture. Upon adipogenesis induction, adipospheres derived from
the SVF displayed more efficient lipid droplet accumulation than 2D cultures. This indicates that 3D intercellular
signaling better recapitulates WAT organogenesis. Combined, our studies show that adipospheres are appro-
priate for WAT modeling ex vivo and provide a new platform for functional screens to identify molecules
bioactive toward individual adipose cell populations. This 3D methodology could be adopted for WAT trans-
plantation applications and aid approaches to WAT-based cell therapy.
integral for healing either after tissue reconstruction surger-
ies or post-tissue damage often caused by disease. Success of
tissue repair relies on stem cells and partially differentiated
progenitor cells present in the grafted tissue and/or recruited
from endogenous organs.1Bone marrow is a bona fide source
of progenitor cells activated in response to trauma.2How-
ever, because the quantity and ability of the bone marrow
progenitors to respond to mobilization stimuli appears to
decline with age, the contribution of cells recruited for injury
repair is likely to progressively decrease in parallel. Instead,
ecovery from various pathological conditions in-
volves tissue remodeling and repair. These processes are
some extramedullary organs, such as white adipose tissue
(WAT), have been shown to ectopically accumulate endo-
thelial and hematopoietic progenitors.3On the other hand,
the importance of stromal mesenchymal progenitors, com-
monly referred to as mesenchymal stromal cells (MSC), in
tissue repair has been increasingly appreciated.4,5MSC had
been originally characterized in the bone marrow as fibro-
blast colony-forming units. MSC are not only capable of dif-
ferentiating into adipocytes, osteoblasts, and chondrocytes,
which has resulted in the term mesenchymal stem cells,6,7
but also support vascularization as trophic pericytic cells and
suppress the immune response.8These combined features
have made bone marrow MSC as a cell type of choice for
numerous clinical trials that are currently in progress. In the
1Center for Stem Cell and Regenerative Medicine, The Brown Foundation Institute of Molecular Medicine for the Prevention of Human
Diseases, The University of Texas Health Science Center at Houston, Houston, Texas.
2Nano3D Biosciences, Houston, Texas.
TISSUE ENGINEERING: Part C
Volume 19, Number 5, 2012
ª Mary Ann Liebert, Inc.
meanwhile, organs such as WAT have been shown as a
considerably more abundant reservoir of mesenchymal
progenitors.9This has led to an explosion of interest in the
potential of WAT in regenerative medicine and cell therapy
applications.10The potential of using ex vivo engineered
WAT for angiogenic tissue grafting has become an emerging
WAT develops throughout the mammalian body in areas
of loose connective tissue, such as subcutaneous layers be-
tween muscle and dermis. In addition, visceral WAT depots
also form around the gut, heart, kidneys, and other internal
organs.13The main cellular components of WAT are adipo-
cytes, the large cells accumulating triglycerides in lipid
droplets.9The remaining cells composing the stromal vascu-
lar fraction (SVF) include perivascular adipose stromal cells
(ASC) serving as adipocyte progenitors, as well as vascular
endothelial cells and infiltrating leukocytes.14,15We and oth-
ers have shown that ASC display multipotency and prolifer-
ation capacity comparable to those of bone marrow MSC
while also serving as pericytes.16–19ASC promote endothelial
proliferation and blood vessel formation at least in part via
trophic effects of secreted growth factors, while also dis-
playing marked anti-inflammatory properties.8These fea-
tures of WAT have made grafts of adipose tissue fragments or
cells (lipotransfer) a promising approach to cosmetic and
functional tissue repair.20In parallel, approaches to de novo
tissue engineering based on ASC have been developed.21
With subcutaneous WAT being readily harvestable, hun-
dreds of regenerative therapy clinical trials are underway.22
Obesity is a result of WAT hypertrophy and hyperplasia,
with the latter relying on the expansion of ASC and pre-
adipocytes.9The increased accessibility of progenitor cells
has made WAT from obese individuals particularly attrac-
tive as a graft source. On the other hand, the emerging as-
sociation between WAT expansion in obesity and various
diseases has alerted for caution.8One of the key obesity
complications is the metabolic syndrome, a medical condi-
tion that is a risk and comorbidity factor for insulin resis-
tance, diabetes, dyslipidemia, and cardiovascular disease.23
Recently, obesity has surfaced as the condition associated
withrisk andprogression ofa number ofcancers.24,25This has
been attributed to the endocrine function of WAT that secrets
numerous growth and inflammatory factors and hormones
collectively termed adipokines.26Recent studies have shown
that WAT-derived cells can be mobilized27,28and recruited as
a component of tumor microenvironment,29,30suggesting
adipokine activity at the cancer site. Several leukocyte pop-
ulations, including macrophages and lymphocytes, progres-
sively concentrate in WAT as obesity progresses,31and their
contribution to the complex secretome of WAT may be key in
the obesity–cancer association. The migratory capacity and
plasticity of monocytes raise the possibility that they may also
traffic from WAT to tumors and contribute to cancer in pa-
tients. The apparent crosstalk between WAT and cancer
confirmed in a number of recent studies and tumorigenesis
associated with lipotransfer for cosmetic tissue reconstruc-
tion32sends a clear message that WAT needs to be better
investigated for its potential pathological functions.33
The prospective beneficial properties and pathogenesis
associated with WAT calls for streamlining efforts in sys-
tematic analysis of this tissue, which could be most efficient
in ex vivo screening formats. However, a challenge in adipose
tissue biology is the lack of adequate culture systems simu-
lating WAT physiology. While adherent cell cultures are
widely used to study the process of adipogenesis (differen-
tiation of ASC into adipocytes), two-dimensional (2D) cul-
tures do not recapitulate the native tissue complexity.
Indeed, when the SVF cells are plated in conventional culture
conditions, endothelial and hematopoietic cells are lost upon
cell passaging, while the mesenchymal compartment takes
over.15,17,19In addition, the morphology, immunophenotype,
and gene expression profile of cells instantly changes upon
plastic attachment of primary cells.8This has made it difficult
to assess the roles of individual WAT cell populations and to
identify functional molecules and markers of WAT cells that
could be used for therapeutic targeting.
Here, to establish a tissue culture model simulating the
complex intercellular interactions of WAT components, we
took advantage of a recently reported three-dimensional (3D)
cell culturing system based on magnetic levitation.34By fa-
cilitating aggregation of different populations of adipose
cells using a magnetic nanoparticle assembly and designed
magnetic fields, we achieved consistent generation of orga-
noids termed adipospheres that display key properties of
endogenous WAT. We demonstrate that this 3D culturing
system allows the retention of multicellular WAT complexity
and partial vascularization concomitant with efficient cell
proliferation and differentiation of preadipocytes.
Materials and Methods
3T3-L1 preadipocyte cells and bEND.3 endothelial cells
were purchased from the American Type Tissue Collection.
To obtain green fluorescent protein (GFP)-labeled bEND.3,
cells were transduced with lentivirus harboring GFP gener-
ated using the Lenti-X HTX system,19according to the
manufacturer’s protocol (Clontech). Cells were 2D cultured
in the Dulbecco’s modified Eagle medium (DMEM) con-
taining 10% (v/v) fetal bovine serum (FBS), 100U/mL of
penicillin, and 100mg/mL of streptomycin to obtain *80%
confluency. For coculture, counted cells were mixed at the
ratio of 95:5 (3T3-L1:bEND.3-GFP) before levitation.
Preparation of mouse primary SVF
The SVF from mixed subcutaneous and intraperitoneal
WAT was isolated by enzymatic digestion as described.19
Nucleated cells were counted and plated on uncoated culture
plate overnight in DMEM/10% FBS (v/v) to obtain *80%
cell confluency for levitation culture setup or to 100% con-
fluency for control 2D cultures.
Magnetic cell levitation
Three-dimensional levitation cell cultures were based on
previously established methodology34and were set up using
the 6- or 24-well Bio-Assembler? kit (Nano3D Biosciences?,
Inc.) consisting of nanoshuttle (NS) solution and a 6-well or
24-well plate magnetic drive. The NS is a nanoparticle as-
sembly of iron oxide (Fe2O3) and gold (Au) nanoparticles
cross-linked with poly-l-lysine to promote cellular uptake.
The 6- or 24-well plates used for 3D culture were flat bottom
ultralow attachment plates (Costar?, 3471 or 3473, respec-
tively). As shown in Figure 1A, indicated numbers of
2 DAQUINAG ET AL.
cultured or primary cells were mixed with 8mL of NS per
cm2plate area (or *10,000 cells per mL of NS) and placed in
a standard CO2cell culture incubator (37?C, 5% CO2[v/v] in
air) for 12h in standard adherent culture conditions. Cells
were then trypsinized to detach adherent cells from plate
and to obtain single-cell suspension. After trypsin inactiva-
tion with serum, cells were counted, centrifuged, and seeded
into a multiwell ultralow-attachment plate. The medium
volume in each well was 1.0mL and 0.3mL for the 6-well
and 24-well plate, respectively. A magnetic drive was im-
mediately placed above the culture to magnetically levitate
the cells and guide them to aggregate within hours of levi-
tation. The cells stay levitated just below the meniscus at the
center of the well, where they self-assemble into spheroids.
Levitated spheroids were incubated in a CO2incubator until
analysis at indicated time points. Cultures can be easily vi-
sualized while levitating with standard inverted microscopes
by just positioning the culture plates on the microscope stand
without removing the magnet drive and by allowing light to
be transmitted through the magnet opening (magnet drives
are made of ring magnets to allow light transmission). Fur-
thermore, because 3D cultures are magnetized, the same
magnet drives that are used to enable magnetic levitation can
be used to facilitate medium exchange throughout the 3D cell
culturing process. This is accomplished by removing the
magnet drive from the top and placing it at the bottom of the
6- or 24-well tissue culture plate. After medium exchange,
magnet drive can be placed back on the top of the culture to
All 2D or 3D cultures levitated for 1 day were induced for
adipogenenic differentiation with a conventional medium19
consisting of 0.5mM isobutylmethyxanthine, 1mM dexa-
methasone, 0.2mM indomethacin, and 1.7mM insulin (all
from Sigma-Aldrich) in DMEM/10% FBS (v/v) for 72h. After
this, the induction medium was replaced with DMEM/10%
FBS containing 1.7mM insulin, which was subsequently re-
placed every 2 days until analysis at theindicated time points.
Cell analysis by immunofluorescence
Cell cultures were fixed in 4% paraformaldehyde for
30min, washed with phosphate-buffered saline (PBS), and
then processed for whole-mount immunofluorescence anal-
ysis or embedded for frozen or paraffin sectioning. For
whole-mount immunostaining, cultures were permeabilized
with 0.3% Triton-X in PBS for 20min. Whole-mount samples
were magnetically held at the bottom of the slide by placing
a magnet (magnet drive from Bio-Assember) under the
microwell carrying the sample. For section analysis, citrate
buffer-based antigen retrieval (Thermo Scientific) was per-
formed before washing with 0.3% Triton-X/PBS. After
blocking in a serum-free Sea block buffer (Thermo Scientific)
for 30min, samples were exposed to primary antibodies
(12–16h at 4?C), and secondary antibodies (2h at room
temperature) in PBS/0.01% Triton-X. The following primary
antibodies were used: 1:100 rabbit anti-perillipin (Cell Sig-
naling), 1:200 goat anti-GFP (GenTex), 1:100 rabbit anti-CD31
and 1:50 goat anti-CD31 (Santa Cruz Biotechnology), 1:200
goat anti-decorin (R&D Systems), 1:100 rabbit anti-Ki67
(Neomarkers), 1:50 rabbit anti-platelet-derived growth factor
adipocyte culture. (A) Subsequent steps of the levitation 3D cell
culture setup where the steps are as follows: cells and magnetic
particles are placed into a culture plate well; a magnet driver is
placed above this well; cells levitate to the meniscus; and finally
,cells aggregate/self-assemble into a spheroid (see the Materials
and Methods section for details). (B) Representative wells with
spheroids of 3T3-L1 preadipocytes initiated by seeding 2.4·105
cells and cultured in a basal (left) or adipogenesis induction
(right) medium for 14 days. (C) Spheroids of 3T3-L1 pre-
adipocytes display buoyancy (right tube, solid arrow) in phos-
to lipid accumulation. The tube on the left shows a control
spheroid cultured without induction at the bottom of the tube
(no buoyancy, hollow arrow). (D) Phase-contrast micrographs
of 3T3-L1 spheroids initiated by seeding 2.4·105cells after 8
days of levitation in either a basal (left) or adipogenesis induc-
lipid droplets indicated (arrows). (E) Paraffin sections of spher-
oids in (D) subjected to immunofluorescence with perilipin an-
tibodies (red), indicating lipid droplet maturation in adipocytes
composing the spheroid upon culture in the adipogenesis in-
duction medium. (F) Frozen sections of spheroids in (D) stained
with Oil Red O (red arrows), indicating lipid accumulation
in adipocytes differentiating in spheroid upon culture in an
adipogenesis induction medium. Scale bar: 100mm. 3D, three-
dimensional. Color images available online at www.liebert
The setup and test of magnetic levitation system for
MAGNETIC LEVITATION 3D CULTURE OF ADIPOSE TISSUE3
receptor-b (PDGFRb; Epitomics), and 1:100 rat anti-CD45
(eBioscience). Secondary immunoglobulin G used was as
follows: donkey Alexa 488-conjugated (Invitrogen) and Cy3-
conjugated (Jackson Laboratories). Nuclei were stained with
Hoechst 33258 or TOPRO3 (Invitrogen). Fluorescence images
were acquired with Olympus IX70 and Magnafire software
(Olympus). Confocal images were acquired with TCS SP5
and LAS AF software (Leica).
Oil Red O staining on spheroids
Frozen sections of spheroids were fixed in 10% formalin
for 30min. Sections were rinsed with water and washed once
with 60% isopropanol for 30s, and then covered with filtered
Oil Red O solution (three parts of 0.3% Oil Red O [Sigma]/
isopropanol stock solution with two parts water) for 10min.
Sections were rinsed with water and mounted, and bright-
field images were acquired.
Magnetic levitation and differentiation of 3T3-L1 cells
To test whether magnetic nanoparticles and levitation can
be used for assembly of 3D adipose tissue cultures, we first
took advantage of the mouse 3T3-L1 cell line19convention-
ally used to study adipogenesis. We used polylysine-based
magnetic nanoparticle assembly to start 3D cultures from
60,000, 120,000, and 240,000 cells based on the technique
illustrated in Figure 1A. Spheroids successfully formed in
every case (Fig. 1B) and remained stable in culture in the
presence of magnetic field for the period of up to 45 days,
after which they were analyzed. As reported previously,
3T3-L1 cells detached and died upon reaching confluence
unless adipogenesis was induced. In contrast, as revealed by
analysis of spheroid sections, 3D-cultured 3T3-L1 cells re-
mained viable at the periphery, although cell death was
observed in the center of the sphere when more than 2.5·105
cells were aggregated and cultured extensively after onset or
levitation (data not shown). This observation suggests that
intercellular communication between cells grown in spher-
oids may differ from that achievable in conventional adher-
ent culture setting.
To determine whether magnetic levitation-based cultures
allow cells to undergo lipogenesis, we induced adipose dif-
ferentiation in 3T3-L1 spheroids. After 14 days of differen-
tiation, spheroids maintained in an adipogenesis induction
medium were buoyant in PBS, indicating lipid accumulation
(Fig. 1C). Spheroids in the induction medium were notice-
ably larger after 14 days after adipogenesis initiation, with
peripheral adipocytes visible by phase-contrast micrography
(Fig. 1D). We performed immunofluorescence analysis
on paraffin sections of spheroids, which detected robust
expression of perilipin-1, a marker of mature lipid droplets,
in an adipogenesis induction medium (Fig. 1E). Oil Red-O
staining of adiposphere frozen-sections confirmed lipid
accumulation upon adipogenesis induction, but not in un-
differentiated spheroids (Fig. 1F).
Vasculature simulation in magnetic levitation culture
Next, we tested whether magnetic levitation-based 3D
conditions enable coculture of distinct cell types. Interaction
of adipose stromal and endothelial cells is integral for the
formation of the basement membrane and mature vascula-
ture.35We therefore investigated the assembly of 3T3-L1
preadipocytes cocultured with murine endothelial bEND.3
cells that have been previously shown to form vascular
networks when cocultured with ASC in 2D conditions.36To
enable identification of endothelial structures among stromal
cells, we stably transduced bEND.3 cells with lentivirus
constitutively expressing GFP. Mixed 3T3-L1 and bEND.3-
GFP cells readily formed spheroids (Fig. 2A, left); an iden-
tical cell admixture plated at confluency in 2D formed a layer
also displaying cell heterogeneity (Fig. 2A, right). Fluorescent
microscopy on cultured cells revealed formation of circular
endothelial structures in spheroids (Fig. 2B, left), while in
adherent culture, only clusters of aggregated bEND.3-GFP
cells within the 3T3-L1 cell mass were observed (Fig. 2B,
right). As 3T3-L1 adipospheres, upon adipogenesis induc-
tion, spheroids formed by 3T3-L1 and bEND.3-GFP cells
accumulated lipids microscopically visible in differentiating
adipocytes (Fig. 2C). Whole-mount immunofluorescence
analysis of adipospheres revealed large perilipin-positive
lipid droplets clustering around GFP-positive networks
formed by bEND.3 cells (Fig. 2D). These data indicate that
magnetic levitation makes it possible to simultaneously
simulate vascularization and lipogenesis in cultured tissue
composed of endothelial cells and mesenchymal progenitors.
Having confirmed the functionality of the magnetic levi-
tation in application to cell lines, we have proceeded to test it
for simulation of WAT organogenesis from primary cells.
Cells derived from mouse WAT by enzymatic digestion were
used to isolate the SVF containing ASC, endothelium, and
infiltrating leukocytes as described previously.19,30Spheroids
readily formed upon levitation of disperse SVF preincubated
with NS and have been maintained for as long as 21 days
without signs of decomposition. Cells isolated from the
spheroids after 21 days displayed viability and phenotype
diversity representative of the original SVF (data not shown).
Paraffin sections of spheroids revealed striking hierarchical
organization with distinct capsule and internal large vessel-
like structures (Fig. 3). Immunofluorescence analysis with
anti-CD31 antibodies confirmed that the lumens of these
vascular cavities in spheroids were lined by endothelial cells
(Fig. 3A). Moreover, immunofluorescence with antibodies
against decorin, a protein abundantly secreted by stromal
cells in WAT,19demonstrated perivascular localization of
ASC (Fig. 3B). Secreted decorin deposited within the extra-
cellular matrix (ECM) was also abundant in other spheroid
compartments. Our data show that magnetic levitation
allows endothelial cells and mesenchymal cells of WAT to
assemble into structures recapitulating in vivo interactions
between these distinct cell types.
Simulation of WAT cellular hierarchy
in magnetic levitation culture
To identify the potential advantages offered by the mag-
netic levitation method, as opposed to conventional adherent
culture, 3D and confluent 2D cultures formed by identical
numbers of SVF cells from the same WAT preparation were
systematically compared. Cells were cultured in the same
medium for the same period of time and fixed using the
same protocol for 2D and 3D cultures to allow for adequate
4 DAQUINAG ET AL.
marker expression comparison. Whole-mount spheroid and
adherent cultures were subjected to confocal immunofluo-
rescence. Vascular network formation analyzed with two
distinct CD31 antibodies was markedly more robust in
spheroids compared to 2D cultures (Fig. 4A–C). Our quan-
tification data show that cultured WAT SVF display signifi-
cantly more CD31+ network branch points in spheroids
compared to 2D cultured cells (Supplementary Fig. S1A;
Supplementary Data are available online at www.liebertpub
Consistent with the data from sections (Fig. 3), the phe-
notypes of vascular-like structures were different depending
on their location within the spheroid (Fig. 4). Analysis of Ki-
67 expression revealed cell proliferation in both 3D and 2D
cultures, with areas of more intensive cell division observed
in peripheral areas of spheroids (Fig. 4A). In general, Ki-67
positive cells were more concentrated in areas of CD31-
positive cell networks. Tissue analysis with antibodies
against a pan-leukocyte marker CD45 revealed an abun-
dance of hematopoietic cells in spheroids, while in adherent
cultures, they were predominantly lost (Fig. 4B). Compara-
tively, higher CD31 expression by endothelium than by
leukocytes allowed us to distinguish the two cell types in
spheroids. Simulating their in vivo localization, leukocytes
were localized in association with CD31-positive endothe-
lium in the spheroids. To identify ASC, we used antibodies
against CD140b, also known as PDGFR-b, a marker of
mesenchymal cells. As expected, CD140b and CD31 signals
were mutually exclusive (Supplementary Fig. S1B). Forma-
tion of stromal/vascular network was observed in spheroids,
in particular in subperipheral layers (Fig. 4C).
Finally, we investigated whether levitated spheroids can
serve to model WAT development. Identical numbers of SVF
cells were used to establish spheroid cultures and confluent
2D cultures, after which adipogenesis was induced. Robust
lipid droplet formation was detected by perilipin immuno-
fluorescence in spheroid whole-mounts 4 days postinduc-
tion, while only traces of lipid accumulation were observed
in adherent culture at this point (Fig. 5A). The obvious dif-
ference in lipid droplet size remained at 8 days postinduction
(Fig. 5A). Importantly, in adipospheres, adipocytes formed
WAT cells. The SVF of mouse visceral WAT (6·104cells) was
used to start 3D levitation cultures. Paraffin sections of
spheroids harvested 8 days after adipogenesis induction were
subjected to immunofluorescence with antibodies against
CD31 (A) or decorin (DCN) (B) secreted by stromal cells.
Secondary antibody signal is in green. Note lumen formation
by endothelial (CD31+) cells (arrows) surrounded by DCN+
stroma (arrowheads). DNA stained blue. Scale bar: 50mm.
WAT, white adipose tissue; SVF, stromal vascular fraction.
Color images available online at www.liebertpub.com/tec
Vascularization in adipospheres formed by primary
coculture. 3T3-L1 preadipocytes (2.25·105cells) and GFP-ex-
pressing mouse bEND.3 endothelial cells (2.5·104cells) were
used to coseed levitation (3D) or adherent (2D) cultures. (A)
culture in an adipogenesis induction medium. (B) GFP fluo-
rescence (green) of vessel-like structures (indicated) composed
by bEND.3 cells in 3D culture contrasted with nonuniform
clustering of minimally organized bEND.3 cells in 2D culture.
(C) Phase-contrast micrographs of 3D cells after 14 days of
culture. Uninduced spheroids (left) and adipogenesis-induced
spheroids (right) with large adipocytes indicated. (D) Whole
cells) and GFP-expressing mouse bEND.3 endothelial cells
(2.5·104cells) 14 days postadipogenesis induction were sub-
jected to immunofluorescence with perilipin antibodies (red
arrows) and GFP antibodies (green arrows), indicating lipid
droplet maturation in adipocytes clustering around bEND.3
vessel-like structures (arrows). DNA stained blue. Scale bar:
50mm. GFP, green fluorescent protein; 2D, two-dimensional.
Color images available online at www.liebertpub.com/tec
Preadipocytes and endothelial cells cooperate in 3D
MAGNETIC LEVITATION 3D CULTURE OF ADIPOSE TISSUE5
continuous tissue mass, while only islets of differentiating
adipocytes were detectable in adherent 2D culture. Both 3D
and 2D cultures were compatible with retention of endo-
thelial cells and their assembly into networks during differ-
entiation; however, endothelial aggregates detected with
CD31 antibody were better organized in 3D. At day 14
postinduction, lipid droplets in adipocytes observed in adi-
posphere culture were comparable to those in WAT derived
from adult WAT of a C57BL/6 mouse (Fig. 5B). Combined,
our data show that organoids formed via magnetic levitation
of cells recapitulate key properties of endogenous WAT.
Although 2D cell culture has enabled the progress in un-
derstanding of the mechanisms driving adipogenesis, lipo-
genesis, and lipolysis in adipocyte progenitors, other cellular
components of WAT have remained relatively understudied.
An obstacle in basic research and drug discovery programs
focusing on WAT has been the lack of physiologically rele-
vant ex vivo tissue culture platforms. Here, to simulate the
complex 3D architecture of WAT, we aggregated different
adipose cell populations in levitation culture based on
magnetic field. By using 3T3-L1 preadipocytes as a model,
we show that while in 2D these cells do not survive con-
fluency unless adipogenesis is induced, they remain viable
indefinitely upon integration within the spheroid 3D struc-
ture. Both 3T3-L1- and primary WAT-derived cells grown in
adipospheres displayed more robust lipid droplet formation
and maturation upon adipogenesis induction than cells in
adherent monolayer. According to our comprehensive anal-
ysis, the levitated organoids maintain multicellular com-
plexity of endogenous WAT, unlike 2D cultures that are
prone to loss of leukocytes. Importantly, adipocyte differ-
entiation is compatible with proliferation and hierarchical
organization of cells within adipospheres. Vessel-like struc-
tures observed in SVF adipospheres were often larger than
venules and arterioles of WAT, and no clear lumens were
WAT cells. The SVF of mouse visceral WAT (3.5·105cells)
was used to start levitation (3D) or adherent (2D) cultures.
After 4 and 8 days (A) or 14 days (B) of adipogenesis in-
duction, whole mounts were subjected to confocal immu-
nofluorescence with perilipin antibodies (red arrows) and
CD31 antibodies (green arrows). Larger and more numerous
perilipin+ lipid droplets are observed in 3D (A). Compar-
ison of differentiated adipospheres to endogenous mouse
WAT whole mount (B) identifies similar sizes of adipocytes
and comparable appearance of CD31+ vessels. DNA stained
blue. Spheroid edge is on the left. Scale bar: 100mm. Color
images available online at www.liebertpub.com/tec
Adipogenesis in adipospheres made from primary
primary adipospheres. The SVF of mouse visceral WAT
(3.5·105cells) was used to start levitation (3D) or adherent (2D)
cultures. At day 4 postadipogenesis induction, 3D whole
mounts and 2D cultures were fixed in 4% paraformaldehyde
and processed for immunostaining with antibodies against in-
dicated markers and the corresponding secondary antibodies
(red/green). Note increased frequency of proliferating Ki67+
cells (arrows) (A) and hematopoietic (CD45+) cells (B) in
spheroid culture compared to 2D. In (B), arrows indicate leu-
dead cells trapping antibodies. The gradient of CD140b+ stro-
mal cells (C) enriched in outside layers (*) and robust vascular
network (CD31+) organization (A–C) are evident in 3D. DNA
stained blue. Spheroid edge is on the left. Scale bar: 100mm.
Color images available online at www.liebertpub.com/tec
Cell composition, organization, and proliferation in
6 DAQUINAG ET AL.
observed. Nevertheless, our data indicating that adipo-
spheres to an extent simulate vascularization take place in
endogenous WAT represent a significant advance for the
field of tissue engineering. Combined, our data indicate that
magnetically levitated cultures recapitulate key components
of WAT organogenesis.
While the advantages of levitated adipospheres over ad-
herent monolayers are apparent, more studies will be nec-
essary to determine the difference of this technique from
alternative approaches to 3D culture. In recent years, a
number of 3D cell culture techniques have been developed to
simulate complex tissue organization.37While some tissues
are capable of forming organoids spontaneously,38others
(including WAT) require scaffolds for the cells to integrate in
an organotypic manner. Most efforts have focused on bio-
polymer scaffolds with Matrigel or other ECM-based plat-
forms being commonly used for tissue microenvironment
simulation.39More rigid, sponge-like scaffolds made from
materials such as hydroxyapatite, or synthetic organic
polymers casted using foaming agents, have also been in-
troduced.40,41While artificial matrices are often used in tissue
modeling, the relevance of tissue composition achieved with
their help to in vivo settings has remained questionable.
Recently, matrix scaffolds based on decellularized native
tissues have gained popularity.42To generate substantial
tissue amounts, agitation-based bioreactors have been im-
plemented.43Such setups allow for expansion of certain cell
types in the context of perfusion simulating vasculature.
However, they are cumbersome and have other shortcom-
ings, such as inability to retain key cell types composing the
endogenous organ and difficulty of processing for subse-
quent in vivo application.
The initial attempts to simulate WAT organogenesis have
been based on the ceiling culture method.44Notable progress
in 3D tissue engineering has been recently achieved in
mimicking the in vivo WAT
However, so far reported studies have been based on scaf-
folds and custom-made bioreactors that do not allow for easy
translation.12,49,50Three-dimensional aqueous-derived silk
scaffolds have been previously used to coculture endothelial
cells and human ASC that could be induced to undergo
adipogenesis.51It has been shown that tissues engineered
ex vivo based on scaffolds can become vascularized upon
implantation into a host.40It was also reported that grafting
of human ASC grown in spheroids is similarly followed by
in vivo angiogenesis.52Interestingly, a recent study demon-
strated that scaffold-free spheroids preserve multilineage
potential of MSC.53While these studies represent important
advances, the capacity to recapitulate and integrate the
multiplicity of cell types composing WAT in vivo has not
been demonstrated to date, thus making our combined re-
sults an important step forward.
The approach to 3D culture by magnetic levitation has
been previously established based on magnetized phage
hydrogels seeded with tumor and neural stem cells.34Here,
we tested the application of the magnetic levitation setup
that does not depend on the phage hydrogel. In our phage-
free method, NS cellular uptake is necessary for cells to be
levitated and guided together to self-assemble into 3D
spheroids. However, as cells proliferate in the primer, most
of their progeny in the resulting spheroid end up nano-
particle-free, and the NS often localize in the ECM within the
3D structure. Therefore, because the NS stays within the
levitated spheroid, cultures do not have to be replenished
with NS to continue the levitation process as the culture as-
sembles and cells proliferate. As a part of this work, we
compared cell proliferation, differentiation, morphology, and
lipid distribution in 2D cultures with and without the pres-
ence of NS. Besides the initially observed dark coloration of
cells due to NS uptake, no changes in cell proliferation or
differentiation were detected in 2D cultures upon NS expo-
sure (Supplementary Fig. S2). While the NS are biocompatible
and in this study no adverse effects were detected for cells
treated with NS either in 2D or 3D, possible nanoparticle
effects in downstream applications are yet to be ruled out.
A unique feature of the magnetic levitation approach is the
expedited timeline of 3D spheroid formation driven by
magnetic field without the loss of cell populations upon
passive cell propagation in conditions favoring spheroid
formation. While simpler approaches to spheroid culture in
an agitated medium and/or low-adherence have been re-
ported for many cell types, including ASC,52,54,55the ratio-
nale behind using magnetic levitation for WAT-derived cells
was to enforce retention of cell types other than ASC com-
posing the organ, which tend to be dominated by ASC and
progressively lost in conventional culture settings. Further-
more, the magnetic levitation method assures the cell–cell
interaction between different cell types at the onset of levi-
tation by magnetically guiding cells together, in contrast to
relying on random cell interactions when using an agitated
medium and/or low-adherence methods.
This study showed that levitated cell spheroids are feasible
for long-term multicellular studies and recapitulate relative
cell positioning more closely than 2D culture. An important
advantage of this system is the dependence of cell adherence
on autocrine ECM molecules rather than on artificial sub-
strates serving as a foundation of other 3D culture designs.
We have not ruled out that some of the polylysine used for
assembly might remain external to cells and contributes to
cell interaction. However, compared to many other systems,
such a minimal tissue composition manipulation gives pri-
mary tissue cells an opportunity to better re-establish their
endogenous organ microenvironment. While normal WAT
does appear largely homogeneous on the millimeter scale,
like any other tissue, it faces interactions with other tissues.
This is relevant in both normal conditions and pathology: for
example, tumors tend to be surrounded by WAT, which may
be key in cancer progression. While investigating that this is
beyond the scope of this initial report, in future studies,
magnetic guidance will indeed likely to be useful for aggre-
gating tissue components with distinct cell contents in a de-
sired orientation to model intercellular interactions.
While this study focused on adipose tissue culture, our
results have important implications for research pursued
toward engineering of other tissues. They prove the principle
that forced aggregation of cells secreting an endogenous
matrix in 3D can be used as an approach to re-establish na-
tive tissue architecture. It remains to be determined whether
NS-based levitation can be applied to other organs composed
of different types of cells requiring different medium com-
positions for maintenance. In the future, careful comparison
of magnetic levitation with other 3D cell culture platforms
will help to define the nuances of each approach and identify
appropriate technologies for specific applications.
MAGNETIC LEVITATION 3D CULTURE OF ADIPOSE TISSUE7
Glauco R. Souza received support from the National Sci-
ence Foundation (NSF) SBIR award and the State of Texas
Emerging Technology Fund (ETF). Research in the Kolonin
laboratory is supported by the grants R01DK088131 and
1R21DK090752 from the NIH. Research in the Souza la-
boratory is supported by Phase I (0945954) and Phase II
(1127551) SBIR grant awards from NSF IIP Division of In-
dustrial Innovation and Partnerships and the State of Texas
Emerging Technology Fund (ETF).
Glauco R. Souza is employed by Nano3D Biosciences?,
1. Caplan, A.I., and Correa, D. The MSC: an injury drugstore.
Cell Stem Cell 9, 11, 2011.
2. Papayannopoulou, T., and Scadden, D.T. Stem-cell ecology
and stem cells in motion. Blood 111, 3923, 2008.
3. Han, J., et al. Adipose tissue is an extramedullary reservoir
for functional hematopoietic stem and progenitor cells.
Blood 115, 957, 2010.
4. Bianco, P., Robey, P.G., and Simmons, P.J. Mesenchymal
stem cells: revisiting history, concepts, and assays. Cell Stem
Cell 2, 313, 2008.
5. Kolonin, M.G., and Simmons, P.J. Combinatorial stem cell
mobilization. Nat Biotechnol 27, 252, 2009.
6. Pittenger, M.F., et al. Multilineage potential of adult human
mesenchymal stem cells. Science 284, 143, 1999.
7. Prockop, D.J. Marrow stromal cells as stem cells for non-
hematopoietic tissues. Science 276, 71, 1997.
8. Kolonin, M.G., Evans, K.W., Mani, S.A., and Gomer, R.H.
Alternative origins of stroma in normal organs and disease.
Stem Cell Res 8, 312, 2012.
9. Daquinag, A.C., Zhang, Y., and Kolonin, M.G. Vascular
targeting of adipose tissue as an anti-obesity approach.
Trends Pharmacol Sci 32, 300, 2011.
10. Choi, J.H., et al. Adipose tissue engineering for soft tissue
regeneration. Tissue Eng Part B Rev 16, 413, 2010.
11. Scherberich, A., Muller, A.M., Schafer, D.J., Banfi, A., and
Martin, I. Adipose tissue-derived progenitors for engineering
12. Guven, S., et al. Engineering of large osteogenic grafts with
rapid engraftment capacity using mesenchymal and endo-
thelial progenitors from human adipose tissue. Biomaterials
32, 5801, 2011.
13. Ailhaud, G., Grimaldi, P., and Negrel, R. Cellular and mo-
lecular aspects of adipose tissue development. Annu Rev
Nutr 12, 207, 1992.
14. Hausman, D.B., DiGirolamo, M., Bartness, T.J., Hausman,
G.J., and Martin, R.J. The biology of white adipocyte pro-
liferation. Obes Rev 2, 239, 2001.
15. Zuk, P.A., et al. Multilineage cells from human adipose tissue:
implications for cell-based therapies. Tissue Eng 7, 211, 2001.
16. Rodeheffer, M.S., Birsoy, K., and Friedman, J.M. Identifica-
tion of white adipocyte progenitor cells in vivo. Cell 135, 240,
17. Traktuev, D., et al. A population of multipotent CD34-
positive adipose stromal cells share pericyte and mesen-
chymal surface markers, reside in a periendothelial location,
and stabilize endothelial networks. Circ Res 102, 77, 2008.
18. Tang, W., et al. White fat progenitor cells reside in the adi-
pose vasculature. Science 322, 583, 2008.
19. Daquinag, A.C., Zhang, Y., Amaya-Manzanares, F., Sim-
mons, P.J., and Kolonin, M.G. An isoform of decorin is a
resistin receptor on the surface of adipose progenitor cells.
Cell Stem Cell 9, 74, 2011.
20. Park, H., et al. Three-dimensional hydrogel model using
adipose-derived stem cells for vocal fold augmentation.
Tissue Eng Part A 16, 535, 2010.
21. Gomillion, C.T., and Burg, K.J. Stem cells and adipose tissue
engineering. Biomaterials 27, 6052, 2006.
22. Gimble, J.M., Bunnell, B.A., Casteilla, L., Jung, J.S., and
Yoshimura, K. Phases I-III clinical trials using adult stem
cells. Stem Cells Int 2010, 604, 2011.
23. Lazar, M.A. How obesity causes diabetes: not a tall tale.
Science 307, 373, 2005.
24. Renehan, A.G., Tyson, M., Egger, M., Heller, R.F., and
Zwahlen, M. Body-mass index and incidence of cancer: a
systematic review and meta-analysis of prospective obser-
vational studies. Lancet 371, 569, 2008.
25. Flegal, K.M., Graubard, B.I., Williamson, D.F., and Gail,
M.H. Cause-specific excess deaths associated with under-
weight, overweight, and obesity. JAMA 298, 2028, 2007.
26. Park, J., Euhus, D.M., and Scherer, P.E. Paracrine and en-
docrine effects of adipose tissue on cancer development and
progression. Endocr Rev 32, 550, 2011.
27. Bellows, C.F., Zhang, Y., Simmons, P.J., Khalsa, A.S., and
Kolonin, M.G. Influence of BMI on level of circulating pro-
genitor cells. Obesity 19, 1722, 2011.
28. Bellows, C.F., Zhang, Y., Chen, J., Frazier, M.L., and Kolo-
nin, M.G. Circulation of progenitor cells in obese and lean
colorectal cancer patients. Cancer Epidemiol Biomarkers
Prev 20, 2461, 2011.
29. Klopp, A.H., et al. Omental adipose tissue-derived stromal
cells promote vascularization and growth of endometrial
tumors. Clin Cancer Res 18, 771, 2012.
30. Zhang, Y., et al. White adipose tissue cells are recruited by
experimental tumors and promote cancer progression in
mouse models. Cancer Res 69, 5259, 2009.
31. Weisberg, S.P., et al. Obesity is associated with macrophage
accumulation in adipose tissue. J Clin Invest 112, 1796, 2003.
32. Petit, J.Y., et al. Locoregional recurrence risk after lipofilling
in breast cancer patients. Ann Oncol 23, 582, 2012.
33. Zhang, Y., Bellows, C.F., and Kolonin, M.G. Adipose tissue-
derived progenitor cells and cancer. World J Stem Cells 2,
34. Souza, G.R., et al. Three-dimensional tissue culture based on
magnetic cell levitation. Nat Nanotechnol 5, 291, 2010.
35. Kolonin, M.G. Tissue-specific targeting based on markers
expressed outside endothelial cells. Adv Genet 67, 61, 2009.
36. Thangarajah, H., et al. IFATS collection: adipose stromal cells
adopt a proangiogenic phenotype under the influence of
hypoxia. Stem Cells 27, 266, 2009.
37. Haycock, J.W. 3D cell culture: a review of current ap-
proaches and techniques. Methods Mol Biol 695, 1, 2011.
38. Suga, H., et al. Self-formation of functional adenohypophysis
in three-dimensional culture. Nature 480, 57, 2011.
39. Dutta, R.C., and Dutta, A.K. Cell-interactive 3D-scaffold;
advances and applications. Biotechnol Adv 27, 334, 2009.
40. Wiggenhauser, P.S., et al. Engineering of vascularized adi-
pose constructs. Cell Tissue Res 327, 747, 2012.
41. Carletti, E., Motta, A., and Migliaresi, C. Scaffolds for tissue
engineering and 3D cell culture. Methods Mol Biol 695, 17,
8 DAQUINAG ET AL.
42. Song, J.J., and Ott, H.C. Organ engineering based on decel-
lularized matrix scaffolds. Trends Mol Med 17, 424, 2011.
43. Timmins, N.E., et al. Three-dimensional cell culture and tis-
sue engineering in a T-CUP (tissue culture under perfusion).
Tissue Eng 13, 2021, 2007.
44. Toda, S., et al. Adipose tissue-organotypic culture system as
a promising model for studying adipose tissue biology and
regeneration. Organogenesis 5, 50, 2009.
45. Hong, L., Peptan, I.A., Colpan, A., and Daw, J.L. Adipose
tissue engineering by human adipose-derived stromal cells.
Cells Tissues Organs 183, 133, 2006.
46. Flynn, L., and Woodhouse, K.A. Adipose tissue engineering
with cells in engineered matrices. Organogenesis 4, 228, 2008.
47. Hemmrich, K., et al. Autologous in vivo adipose tissue en-
gineering in hyaluronan-based gels—a pilot study. J Surg
Res 144, 82, 2008.
48. Shanti, R.M., et al. In vitro adipose tissue engineering using
an electrospun nanofibrous scaffold. Ann Plast Surg 61, 566,
49. Scherberich, A., Galli, R., Jaquiery, C., Farhadi, J., and
Martin, I. Three-dimensional perfusion culture of human
adipose tissue-derived endothelial and osteoblastic progen-
itors generates osteogenic constructs with intrinsic vascu-
larization capacity. Stem Cells 25, 1823, 2007.
50. Gerlach, J.C., et al. Adipogenesis of human adipose-derived
stem cells within three-dimensional hollow fiber-based bio-
reactors. Tissue Eng Part C Methods 18, 54, 2012.
51. Kang, J.H., Gimble, J.M., and Kaplan, D.L. In vitro 3D model
for human vascularized adipose tissue. Tissue Eng Part A
15, 2227, 2009.
52. Bhang, S.H., et al. Angiogenesis in ischemic tissue produced
by spheroid grafting of human adipose-derived stromal
cells. Biomaterials 32, 2734, 2011.
53. Baraniak, P.R., and McDevitt, T.C. Scaffold-free culture of
mesenchymal stem cell spheroids in suspension preserves
multilineage potential. Cell Tissue Res 347, 701, 2012.
54. Sanchez-Bustamante, C.D., Frey, U., Kelm, J.M., Hierle-
mann, A., and Fussenegger, M. Modulation of cardio-
morphogenetic protein-2 expression. Tissue Eng Part A 14,
55. Verseijden, F., et al. Adult human bone marrow- and adipose
tissue-derived stromal cells support the formation of pre-
vascular-like structures from endothelial cells in vitro. Tissue
Eng Part A 16, 101, 2010.
Address correspondence to:
Mikhail G. Kolonin, Ph.D.
Center for Stem Cell and Regenerative Medicine
The Brown Foundation Institute of Molecular Medicine for the
Prevention of Human Diseases
The University of Texas Health Science Center at Houston
Room 630-G, 1825 Pressler St.
Houston, TX 77030
Received: March 28, 2012
Accepted: September 18, 2012
Online Publication Date: November 2, 2012
MAGNETIC LEVITATION 3D CULTURE OF ADIPOSE TISSUE9