Induction of Pluripotent Stem Cells from Adult Human
Fibroblasts by Defined Factors
Takahashi, Kazutoshi; Tanabe, Koji; Ohnuki, Mari; Narita,
Megumi; Ichisaka, Tomoko; Tomoda, Kiichiro; Yamanaka,
CitationCell (2007), 131(5): 861-872
RightCopyright © 2007 Elsevier B.V. All rights reserved.
KURENAI : Kyoto University Research Information Repository
Induction of Pluripotent Stem Cells from Adult Human Fibroblasts by Defined
Kazutoshi Takahashi1, Koji Tanabe1, Mari Ohnuki1, Megumi Narita1,2, Tomoko Ichisaka1,2,
Kiichiro Tomoda3and Shinya Yamanaka1–4*
1Department of Stem Cell Biology, Institute for Frontier Medical Sciences, Kyoto
University, Kyoto 606-8507, Japan
2CREST, Japan Science and Technology Agency, Kawaguchi 332-0012, Japan
3Gladstone Institute of Cardiovascular Disease, San Francisco, CA 94158
4Institute for Integrated Cell-Material Sciences, Kyoto University, Kyoto 606-8507, Japan
* [D] Manuscript
Successful reprogramming of differentiated human somatic cells into a pluripotent state
would allow creation of patient- and disease-specific stem cells. We previously reported
generation of induced pluripotent stem (iPS) cells, capable of germline transmission, from
mouse somatic cells by transduction of four defined transcription factors. Here, we
demonstrate the generation of iPS cells from adult human dermal fibroblasts with the same
four factors: Oct3/4, Sox2, Klf4, and c-Myc. Human iPS cells were similar to human
embryonic stem (ES) cells in morphology, proliferation, surface antigens, gene expression,
epigenetic status of pluripotent cell-specific genes, and telomerase activity. Furthermore,
these cells could differentiate into cell types of the three germ layers in vitro and in
teratomas. These findings demonstrate that iPS cells can be generated from adult human
Embryonic stem (ES) cells, derived from the inner cell mass of mammalian blastocysts,
have the ability to grow indefinitely while maintaining pluripotency(Evans and Kaufman,
1981; Martin, 1981). These properties have led to expectations that human ES cells might
be useful to understand disease mechanisms, to screen effective and safe drugs, and to treat
patients of various diseases and injuries, such as juvenile diabetes and spinal cord
injury(Thomson et al., 1998). Use of human embryos, however, faces ethical controversies
that hinder the applications of human ES cells. In addition, it is difficult to generate patient-
or disease-specific ES cells, which are required for their effective application. One way to
circumvent these issues is to induce pluripotent status in somatic cells by direct
We showed that induced pluripotent stem (iPS) cells can be generated from mouse
embryonic fibroblasts (MEF) and adult mouse tail-tip fibroblasts by the retrovirus-mediated
transfection of four transcription factors, namely Oct3/4, Sox2, c-Myc, and Klf4(Takahashi
and Yamanaka, 2006). Mouse iPS cells are indistinguishable from ES cells in morphology,
proliferation, gene expression, and teratoma formation. Furthermore, when transplanted
into blastocysts, mouse iPS cells can give rise to adult chimeras, which are competent for
germline transmission(Maherali et al., 2007; Okita et al., 2007; Wernig et al., 2007). These
results are proof-of-principle that pluripotent stem cells can be generated from somatic cells
by the combination of a small number of factors.
In the current study, we sought to generate iPS cells from adult human somatic
cells by optimizing retroviral transduction in human fibroblasts and subsequent culture
conditions. These efforts have enabled us to generate iPS cells from adult human dermal
fibroblasts and other human somatic cells, which are comparable to human ES cells in their
differentiation potential in vitro and in teratomas.
Optimization of Retroviral Transduction for Generating Human iPS Cells
Induction of iPS cells from mouse fibroblasts requires retroviruses with high transduction
efficiencies(Takahashi and Yamanaka, 2006). We, therefore, optimized transduction
methods in adult human dermal fibroblasts (HDF). We first introduced green fluorescent
protein (GFP) into adult HDF with amphotropic retrovirus produced in PLAT-A packaging
cells. As a control, we introduced GFP to mouse embryonic fibroblasts (MEF) with
ecotropic retrovirus produced in PLAT-E packaging cells(Morita et al., 2000). In MEF,
more than 80% of cells expressed GFP (S-Figure 1). In contrast, less that 20% of HDF
expressed GFP with significantly lower intensity than in MEF. To improve the transduction
efficiency, we introduced the mouse receptor for retroviruses, Slc7a1(Verrey et al., 2004)
(also known as mCAT1), into HDF with lentivirus. We then introduced GFP to HDF-Slc7a1
with ecotropic retrovirus. This strategy yielded a transduction efficiency of 60%, with a
similar intensity to that in MEF.
Generation of iPS Cells from Adult HDF
The protocol for human iPS cell induction is summarized in Figure 1A. We introduced the
retroviruses containing human Oct3/4, Sox2, Klf4 and c-Myc into HDF-Slc7a1 (Figure 1B,
8 x 105 cells per 100 mm dish). The HDF derived from facial dermis of 36-year-old
Caucasian female. Six days after transduction, the cells were harvested by trypsinization
and plated onto mitomycin C-treated SNL feeder cells(McMahon and Bradley, 1990) at 5 x
104 or 5 x 105 cells per 100-mm dish. The next day, the medium (DMEM containing 10%
FBS) was replaced with a medium for primate ES cell culture supplemented with 4 ng/ml
basic fibroblast growth factor (bFGF).
Approximately two weeks later, some granulated colonies appeared that were not
similar to hES cells in morphology (Figure 1C). Around day 25, we observed distinct types
of colonies that were flat and resembled hES cell colonies (Figure 1D). From 5 x 104
fibroblasts, we observed ~10 hES cell–like colonies and ~100 granulated colonies (7/122,
8/84, 8/171, 5/73, 6/122 and 11/213 in six independent experiments, summarized in
Supplemental Table 1). At day 30, we picked hES cell–like colonies and mechanically
disaggregated them into small clumps without enzymatic digestion. When starting with 5 x
105 fibroblasts, the dish was nearly covered with more than 300 granulated colonies. We
occasionally observed some hES cell–like colonies in between the granulated cells, but it
was difficult to isolate hES cell–like colonies because of the high density of granulated
cells. The nature of the non- hES-like cells remains to be determined.
The hES-like cells expanded on SNL feeder cells with the primate ES cell medium
containing bFGF. They formed tightly packed and flat colonies (Figure 1E). Each cell
exhibited morphology similar to that of human ES cells, characterized by large nuclei and
scant cytoplasm (Figure 1F). As is the case with hES cells, we occasionally observed
spontaneous differentiation in the center of the colony (Fig. 1G).
These cells also showed similarity to hES cells in feeder dependency (S-Figure 2).
They did not attach to gelatin-coated tissue-culture plates. By contrast, they maintained an
undifferentiated state on Matrigel-coated plates in MEF-conditioned primate ES cell
medium, but not in non-conditioned medium.
Since these cells were similar to hES cells in morphology and other aspects noted
above, we will refer to the selected cells after transduction of HDF as human iPS cells, as
we describe the molecular and functional evidence for this claim. Human iPS cells clones
established in this study are summarized in S-Table 2.
Human iPS Cells Express hES Markers
In general, except for a few cells at the edge of the colonies, human iPS cells did not
express stage-specific embryonic antigen (SSEA)-1 (Figure 1H). In contrast, they expressed
hES cell–specific surface antigens(Adewumi et al., 2007), including SSEA-3, SSEA-4,
tumor-related antigen (TRA) -1-60, TRA-1-81 and TRA-2-49/6E (alkaline phosphatase),
and NANOG protein (Fig. 1I~N).
RT-PCR showed human iPS cells expressed many undifferentiated ES cell marker
genes(Adewumi et al., 2007), such as OCT3/4, SOX2, NANOG, growth and differentiation
factor 3 (GDF3), reduced expression 1 (REX1), fibroblast growth factor 4 (FGF4),
embryonic cell–specific gene 1 (ESG1), developmental pluripotency–associated 2 (DPPA2),
DPPA4 and telomerase reverse transcriptase (hTERT) at levels equivalent to or higher than
those in the hES cell line H9 and the human embryonic carcinoma cell line, NTERA-2
(Figure 2A). By western blotting, proteins levels of OCT3/4, SOX2, NANOG, SALL4,
E-CADHERIN, and hTERT were similar in human iPS cells and hES cells (Figure 2B).
Although the expression levels of Klf4 and c-Myc increased more than five fold in HDF
after the retroviral transduction (not shown), their expression levels in human iPS cells
were comparable to those in HDF (Figure 2A & B), indicating retroviral silencing. RT-PCR
using primers specific for retroviral transcripts confirmed efficient silencing of all the four
retroviruses (Figure 2C). DNA microarray analyses showed that the global gene expression
patters are similar, but not identical, between human iPS cells and hES cells (Figure 2D).
Among 32266 gene analyzed, 5107 genes showed more than 5-fold difference in
expression between HDF and human iPS cells, whereas 1267 genes between human iPS
cells and hES cells (S-Table 3 & 4).
Promoters of ES Cell–Specific Genes Are Active in Human iPS Cells
Bisulfite genomic sequencing analyses evaluating the methylation statuses of cytosine
guanine dinucleotides (CpG) in the promoter regions of pluripotent-associated genes, such
as OCT3/4, REX1 and NANOG, revealed that they were highly unmethylated, whereas the
CpG dinucleotides of the regions were highly methylated in parental HDFs (Figure 3A).
These findings indicate that these promoters are active in human iPS cells.
Luciferase reporter assays also showed that human OCT3/4 and REX1 promoters
had high levels of transcriptional activity in human iPS cells and EC cells (NTERA-2), but
not in HDF. The promoter activities of ubiquitously expressed genes, such as human RNA
polymerase II (PolII), showed similar activities in both human iPS cells and HDF (Figure
We also performed chromatin immunoprecipitation to analyze the histone
modification status in human iPS cells (Figure 3C). We found that histone H3 lysine 4 was
methylated whereas H3 lysine 27 was demethylated in the promoter regions of Oct3/4,
Sox2, and Nanog in human iPS cells. We also found that human iPS cells showed the
bivalent patterns of development-associated genes, such as Gata6, Msx2, Pax6, and Hand1.
These histone modification statuses are characteristic of hES cells (Pan et al., 2007).
High Telomerase Activity and Exponential Growth of Human iPS Cells
As predicted from the high expression levels of hTERT, human iPS cells showed high
telomerase activity (Figure 4A). They proliferated exponentially for as least 4 months
(Figure 4B). The calculated population doubling time of human iPS cells were 46.9 ± 12.4
(clone 201B2), 47.8 ± 6.6 (201B6) and 43.2 ± 11.5 (201B7) hours. These times are
equivalent to the reported doubling time of hES cells(Cowan et al., 2004).
Embryoid Body–Mediated Differentiation of Human iPS Cells
To determine the differentiation ability of human iPS cells in vitro, we used floating
cultivation to form embryoid bodies (EBs)(Itskovitz-Eldor et al., 2000). After 8 days in
suspension culture, iPS cells formed ball-shaped structures (Figure 5A). We transferred
these embryoid body–like structures to gelatin-coated plates and continued cultivation for
another 8 days. Attached cells showed various types of morphologies, such as those
resembling neuronal cells, cobblestone-like cells, and epithelial cells (Figure 5B–E).
Immunocytochemistry detected cells positive for βIII-tubulin (a marker of ectoderm), glial
fibrillary acidic protein (GFAP, ectoderm), α-smooth muscle actin (α-SMA, mesoderm),
desmin (mesoderm), α-fetoprotein (AFP, endoderm), and vimentin (mesoderm and parietal
endoderm) (Figure 5F–K). RT-PCR confirmed that these differentiated cells expressed
forkhead box A2 (FOXA2, a marker of endoderm), AFP (endoderm), cytokeratin 8 and 18
(endoderm), SRY-box containing gene 17 (SOX17, endoderm), BRACHYURY (mesoderm),
Msh homeobox 1 (MSX1, mesoderm), microtubule-associated protein 2 (MAP2, ectoderm)
and paired box 6 (PAX6, ectoderm) (Figure 5L). In contrast, expression of OCT3/4, SOX2
and NANOG was markedly decreased. These data demonstrated that iPS cells could
differentiate into three germ layers in vitro.
Directed Differentiation of Human iPS Cells into Neural Cells
We next examined whether lineage-directed differentiation of human iPS cells could be
induced by reported methods for hES cells. We seeded human iPS cells on PA6 feeder layer
and maintained them under differentiation conditions for two weeks(Kawasaki et al., 2000).
Cells spread drastically, and some neuronal structures were observed (Figure 6A).
Immunocytochemistry detected cells positive for tyrosine hydroxylase and βIII tubulin in
the culture (Figure 6B). PCR analysis revealed expression of dopaminergic neuron markers,
such as aromatic-L–amino acid decarboxylase (AADC), member 3 (DAT), choline
acetyltransferase (ChAT), and LIM homeobox transcription factor 1 beta (LMX1B), as well
as another neuron marker, MAP2 (Figure 6C). In contrast, GFAP expression was not
induced with this system. On the other hand, the expression of OCT3/4 and NANOG
decreased markedly whereas Sox decreased only slightly (Figure 6C). These data
demonstrated that iPS cells could differentiate into neuronal cells, including dopaminergic
neurons, by co-culture with PA6 cells.
Directed Differentiation of Human iPS Cells into Cardiac Cells
We next examined directed cardiac differentiation of human iPS cells with the recently
reported protocol which utilizes activin A and bone morphogenetic protein (BMP)
4(Laflamme et al., 2007). Twelve days after the induction of differentiation, clumps of cells
started beating (Figure 6D, Supplemental movie). RT-PCR showed that these cells
expressed cardiomyocyte markers, such as troponin T type 2 cardiac (TnTc), myocyte
enhancer factor 2C (MEF2C), myosin, light polypeptide 7, regulatory (MYL2A), myosin,
heavy polypeptide 7, cardiac muscle, beta (MYHCB), and NK2 transcription factor related,
locus 5 (NKX2.5) (Figure 6E). In contrast, the expression of Oct3/4, Sox2, and Nanog
markedly decreased. Thus, human iPS cells can differentiate into cardiac myocytes in vitro.
Teratoma Formation from Human iPS Cells
To test pluripotency in vivo, we transplanted human iPS cells (clone 201B7)
subcutaneously into dorsal flanks of immunodeficient (SCID) mice. Nine weeks after
injection, we observed tumor formation. Histological examination showed that the tumor
contained various tissues (Figure 7), including gut-like epithelial tissues (endoderm),
striated muscle (mesoderm), cartilage (mesoderm), neural tissues (ectoderm), and
keratin-containing epidermal tissues (ectoderm).
Human iPS Cells Are Derived from HDF, not Cross-Contamination
PCR of genomic DNA of human iPS cells showed that all clones have integration of all the
four retroviruses (S-Figure 3A). Southern blot analysis with a c-Myc cDNA probe revealed
that each clone had a unique pattern of retroviral integration sites (S-Figure 3B). In addition,
the patterns of 16 short tandem repeats were completely matched between human iPS
clones and parental HDF (S-Table 5). These patterns differed from any established hES cell
lines reported on National Institutes of Health website
(http://stemcells.nih.gov/research/nihresearch/scunit/genotyping.htm). In addition,
chromosomal G-band analyses showed that human iPS cells had a normal karyotype of
46XX (not shown). Thus, human iPS clones were derived from HDF and were not a result
of cross-contamination. Whether generation of human iPS cells depends on minor genetic
or epigenetic modification awaits further investigation.
Generation of iPS Cells from Other Human Somatic Cells
In addition to HDF, we used primary human fibroblast-like synoviocytes (HFLS) from
synovial tissue of 69-year-old Caucasian male and BJ cells, a cell line established from
neonate fibroblasts (S-Table 1 & 2). From HFLS (5 x 104 cells per 100-mm dish), we
obtained more than 600 hundred granulated colonies and 17 hES cell–like colonies
(S-Table 1). We picked six colonies, of which only two were expandable as iPS cells
(S-figure 4). Dishes seeded with 5 x 105 HFLS were covered with granulated cells, and no
hES cell-like colonies were distinguishable. In contrast, we obtained 7–8 and ~100 hES
cell-like colonies from 5 x 104 and 5 x 105 BJ cells, respectively, with only a few granulated
colonies (S-Table 1). We picked six hES cell–like colonies and generated iPS cells from
five colonies (S-Figure 4). Human iPS cells derived from HFLS and BJ expressed hES cell
marker genes at levels similar to or higher than those in hES cells (S-Figure 5). They
differentiated into all three germ layers through EBs (S-Figure 6). STR analyses confirmed
that iPS-HFLS cells and iPS-BJ cells were derived from HFLS and BJ fibroblasts,
respectively (S-Table 6 & 7).
In this study, we showed that iPS cells can be generated from adult HDF and other somatic
cells by retroviral transduction of the same four transcription factors with mouse iPS cells,
namely Oct3/4, Sox2, Klf4, and c-Myc. The established human iPS cells are similar to hES
cells in many aspects, including morphology, proliferation, feeder dependence, surface
markers, gene expression, promoter activities, telomerase activities, in vitro differentiation,
and teratoma formation. The four retroviruses are strongly silenced in human iPS cells,
indicating that these cells are efficiently reprogrammed and do not depend on continuous
expression of the transgenes for self-renewal.
hES cells are different from mouse counterparts in many respects(Rao, 2004). hES
cell colonies are flatter and do not override each other. hES cells depend on bFGF for
self-renewal(Amit et al., 2000), whereas mouse ES cells depend on the LIF/Stat3
pathway(Matsuda et al., 1999; Niwa et al., 1998). BMP induces differentiation in hES
cells(Xu et al., 2005) but is involved in self-renewal of mouse ES cells(Ying et al., 2003).
Despite these differences, our data show that the same four transcription factors induce iPS
cells in both human and mouse. The four factors, however, could not induce human iPS
cells when fibroblasts were kept under the culture condition for mouse ES cells after
retroviral transduction (data not shown). These data suggest that the fundamental
transcriptional network governing pluripotency is common in human and mice, but
extrinsic factors and signals maintaining pluripotency are unique for each species.
Deciphering of the mechanism by which the four factors induce pluripotency in
somatic cells remains elusive. The function of Oct3/4 and Sox2 as core transcription factors