In their letter to Nature, Hayashi et al. describe a new
method for the generation of ocular surface epitheliumlike
cells by incubating human induced pluripotent stem
cells (iPS) into four defined concentric zones mimicking the
developing eye, which the authors named as self-formed
ectodermal autonomous multi-zone (SEAM) (1). Different
SEAM zones contained cells with characteristics of the
ocular surface ectoderm, the lens, the neural retina and the
retinal pigment epithelium. Isolated corneal surface-like
cells successfully reinstated a corneal epithelium in a rabbit
limbal stem cell deficiency (LSCD) model. The paper offers
a new option for patients with LSCD.
Corneal epithelial integrity is essential for visual
function. The non-keratinized squamous corneal epithelium
is continuously regenerated by the limbal stem cells (LSCs)
that reside in the palisades of Vogt of the peripheral cornea.
Damage to this region can lead to irreversible LSCD,
resulting in impaired regeneration of corneal epithelial cells
and keratopathy (2,3).
In cases of unilateral LSCD, autologous LSC
transplantation from the healthy to the diseased eye can
be performed by limbal autografting, first described by
Kenyon and Tseng in 1989 (4). However, while success
rates between 45% and 100%, depending on the degree of
LSCD and other comorbidities (2,5), have been reported,
limitations remain: large limbal biopsies entail a risk of
LSCD for the donor eye, as well as minor complications,
e.g., discomfort, chronic inflammation, scarring and
infection (3). The ex vivo expansion of LSCs, established
by Pellegrini et al. in 1997 (6), showed the risk of donor
decompensation could be minimized and has provided a
road map for successful stem cell therapy use and approval
through regulation (5,6). However, treatment costs are high
with the need for clean-room facilities, trained staff as well
as GMP-qualified culture reagents (2,7). Finally simple
limbal epithelial transplantation (SLET), a one-step surgical
procedure combining the placement of healthy limbal
tissue fragments on hAM, which is directly anchored on the
recipient’s cornea, has shown promising results although
long-term studies are required (8).
In cases of bilateral LSCD, where autologous LSC
transplantation is impeded, other allogenic sources, such as
living related or cadaveric donors, are required. However,
the long-term success rates are not as good as autologous
tissue, and patients will require the use of long-term
immunosuppression (2).
Consequently, in recent years, researchers have been
focusing on alternative sources of corneal epitheliumlike
cells. Non-corneal autologous surface epithelia, e.g.,
conjunctival epithelial cells or cultivated oral mucosal
epithelium, have proven to be able to stabilize the corneal
surface in the short to mid-time range in clinical trials (9,10). However, the resulting ocular surface has been shown
to vary in its stratification and the number of cell layers,
which can lead to a mixed phenotype and suboptimal vision.
Further these epithelia do not express anti-angiogenic
factors, such as soluble FLT1, TIMP3, and TSP1 and the
majority of patients may develop corneal neovascularization
in the long-term (10). However, a distinct advantage is the
lack of use of oral immunosuppression.
In order to improve on this, alternative stem cell sources
have been investigated with the objective of generating
cells as closely resembling human corneal epithelial cells as
possible:
(I) Pluripotent embryonic SCs (ESCs) are selfrenewing
and represent a potentially infinite source
that can differentiate into virtually any cell type.
Human ESCs have been shown to exhibit a corneal
epithelial-like phenotype [expressing ΔNp63α (P63)
and cytokeratin (K) 3/12] when cultured in limbal
fibroblast-conditioned medium (11). However,
controversial ethical issues, differentiated cell
purity, identity and the risk of teratoma formation
have limited the implementation from experimental
results towards clinical use (12);
(II) Dental pulp stem cells (DPSCs), express markers
in common with LSCs, such as ABCG2, Integrin
b1, vimentin, connexin 43 and K3/12 (13).
Transplantation of a tissue-engineered cell sheet
has been shown to reconstruct rabbit corneas with
mild LSCD. However, in extensive LSCD, the
reconstructed epithelium consisted of unnatural
flattened cells (14). DPSCs also express proangiogenic
factors increasing the risk of corneal
neovascularization (15);
(III) Murine hair follicle bulge-derived stem cells have
been chemically induced to a corneal epithelial-like
phenotype expressing K12, and have shown 80%
repopulation efficiency of the corneal surface in a
mechanical mouse LSCD model (16);
(IV) Adult mesenchy malstem cells (MSCs) are
proliferative and multipotent stem cells that can
differentiate into cells of various lineages. They can
be harvested from allogenic sources e.g., umbilical
cord linings, but also from autologous sources,
such as bone marrow and adipose tissue, hence
avoiding ethical issues (17). Human bone marrow
and adipose-derived MSCs have been shown to
be able to be differentiated into corneal epithelial
lineage, improving corneal healing in rat alkali
burn models (17,18). However, reports have been
variable and feeder cells/conditioned medium
and several in vitro induction steps impede the
implementation of reliable protocols (3,7).
In vitro ocular organogenesis has been described
before. Eiraku et al. first reported in Nature in 2011
the autonomous formation of an optic cup in a threedimensional
culture (Matrigel, Corning Life Sciences,
USA) of mouse embryonic stem cell aggregates, that
generated stratified neural retinal tissue (19). Nakano et al.
similarly demonstrated the self-formation of optic cups and
stratified neural retina from human ESCs in 2012 (20). The
human ESC-derived optic cup was much larger than the
mouse ESC-derived one and the neural retina grew into
multilayered tissue containing both rods and cones, whereas
cone differentiation was rare in mouse ESC culture.
Reichman et al. demonstrated in 2014 the generation of
retinal pigmented epithelial cells and self-forming neural
retina-like structures containing retinal progenitor cells
from iPS in a floating culture technique (21).
Hayashi et al., however, are the first group to specifically
grow cells with ocular as well as non-ocular surface
ectoderm and lens characteristics within an in vitro
autonomously grown eye tissue (1). A strong appeal of
their approach is that the microenvironment of a SEAM
allows for a much more efficient and epigenetically stable
differentiation of cells. Further studies are required to show
the reproducibility of this technique.
The authors, like Reichman et al. before (21), relied on
iPS. These stem cells have the benefit of being derived from
somatic adult tissue, hence avoiding the controversial ethical
issues, and the need for aggressive post-transplantation
control for immune-mediated rejections, especially when
they can be obtained from autologous sources. However, the
widespread utilization of iPS is limited by their potential
risks of oncogenic transformation, the problematic
epigenetic memory and a low production efficacy.
Hayashi et al. previously reported the lentiviral generation
of iPS from human corneal limbal epithelial cells with a rate
as low as 0.0005% (22).
In this publication the authors report favorable efficacies
in their SEAM approach with 7.7% of incubated iPS
building colonies, of which 67.9% spontaneously separated
into four visible concentric zones, developing into all cell
types essential for the organogenesis of an entire eye (1).
Ocular surface epithelial cells have been derived from
iPS and used for treatment in animal models before.
Hayashi et al. were the first to reprogram iPS from human adult dermal fibroblasts and corneal limbal epithelial cells
and to induce them to corneal epithelial like character
in 2012. However, apart from the aforementioned low
reprogramming efficacy, cells were not tested in animal
models and DNA methylation analysis revealed epigenetic
differences between the iPS-generated and control human
corneal epithelium, even though no significant differences
in corneal epithelium-related genes such as K12, K3,
and PAX6 were detected (22). Epigenetic differences to
primary corneal limbal cells were confirmed by Sareen et al.;
limbal-derived iPS had fewer unique methylation changes
than fibroblast-derived iPS, suggesting retention of
epigenetic memory during reprogramming (23). CieślarPobuda
et al. found that differentiation of iPS to corneal
epithelium-like cells is a slow process (3 weeks) and that
pluripotency genes remained activated, implying a risk of
malignant transformation (24). Mikhailova et al. showed
that small molecules added to the medium (TGFβ inhibitor
SB-505124 , Wnt inhibitor IWP- 2 and b F G F ) could
significantly suppress pluripotency activity and improve
differentiation of iPS to corneal epithelium-like cells finding
P63 expression in 25–95% of induced cells, depending on
the medium composition and time point (25). However,
they also described variations in other corneal epithelial cell
markers questioning cell purity.
This issue also became apparent in this publication, as
cells for corneal epithelial therapy had to be meticulously
separated. After media changes and isolation of zone 3 cells
by pipetting, Hayashi et al. separated corneal epithelial
progenitor cells (SSEA-4+/ITGB4+) from the ocular surface
epithelial lineage by FACS (14.1%). Conjunctival epithelial
cells were obtained as SSEA-4− cells (16.6%) (1,26).
In addition cells expressed different ocular surface tissue
characteristics depending on their passage. While P2 cells
mainly mimicked conjunctival epithelium [Suppl. Table
in (1); PAX6+, K13+, p63+, K12−, HOX− and later also
MUC5AC+ and K7+ goblet cells appeared] and non-ocular
epithelium characteristics (PAX6−, K13+, p63+, K12−,
HOX+), cells from passage 3 expressed cell markers typical
of corneal epithelial (PAX6+, K13−, p63+, K12+, HOX−)
and limbal cells (PAX6+, K13−, p63+, K12+ low, HOX−).
Unfortunately further markers associated with the corneal
epithelium (K3, involucrin, connexin 43, ZO1, occludin,
CD98, 166 and 340) or limbus (ABCG2 and 5, K15 and 19,
vimentin, EGFR, integrin α9 and β1) were not specified.
In the organogenesis approach carried out by the
authors, however, ocular surface development proved
BMP4/TGFβ dependent, as inhibition of these factors disrupted SEAM formation (1). Induction of transplantable
cells was time-consuming (13–18 weeks) and cost intensive,
as cells slowly differentiated to mature PAX6+, P63+, K12+
corneal epithelium-like character.
The transplanted sheet reformed a stratified cornealike
epithelium in a surgically induced LSCD model. The
authors conclude that they are now in the position to
initiate first in human trials. However, further investigations
specifying epigenetic differences, e.g., DNA-methylation,
teratoma risk, proof of long-term phenotype stability and
an ideally simplified and xeno-free GMP differentiation
protocol might be necessary before treatment in patients
becomes reality.
