Identification of Nonepithelial Multipotent Cells in the Embryonic Olfactory Mucosa


  • Mercedes Tomé,

    1. Division of Clinical Neuroscience, Glasgow Biomedical Research Centre, University of Glasgow, Glasgow, Scotland
    Search for more papers by this author
  • Susan L. Lindsay,

    1. Division of Clinical Neuroscience, Glasgow Biomedical Research Centre, University of Glasgow, Glasgow, Scotland
    Search for more papers by this author
  • John S. Riddell,

    1. Division of Neuroscience and Biomedical Systems, Institute of Biomedical and Life Sciences, University of Glasgow, Glasgow, Scotland
    Search for more papers by this author
  • Susan C. Barnett

    Corresponding author
    1. Division of Clinical Neuroscience, Glasgow Biomedical Research Centre, University of Glasgow, Glasgow, Scotland
    • Division of Clinical Neuroscience, Biomedical Research Centre, Room B417, 120 University Place, University of Glasgow, Glasgow G12 8TA, Scotland
    Search for more papers by this author
    • Telephone: (0)141 330 8409/4353

  • Author contributions: M.T., J.S.R., and S.C.B.: contributed to manuscript writing; M.T. and S.C.B.: designed experiments; J.S.R. and S.C.B.: obtained financial support; M.T.: collected and assembled data; S.L.L.: carried out human tissue staining and fluorescence-activated cell sorting analysis; S.C.B.: approved manuscript.

  • First published online in STEM CELLS EXPRESS May 21, 2009.


Olfactory mucosal (OM) tissue, a potential source of stem cells, is currently being assessed in the clinic as a candidate tissue for transplant-mediated repair of spinal cord injury. We examined the ability of embryonic rat OM tissue to generate stem cells using culture conditions known to promote neural stem cell proliferation. Primary spheres formed that proliferated and exhibited two main morphologies: (a) CNS neurosphere-like (OM-I) and (b) small, tight spheroid-like (OM-II). The OM-I spheres expressed the neural stem cell marker nestin but also markers of peripheral glia, neurons, and connective tissue. Further studies demonstrated the presence of multipotential mesenchymal-like stem cells within OM-I spheres that differentiated into bone, adipose, and smooth muscle cells. In contrast, the OM-II spheres contained mainly cytokeratin-expressing cells. Immunolabeling of rat olfactory tissue with Stro-1, CD90, and CD105 showed the presence of multipotent mesenchymal cells in the lamina propria, whereas cytokeratin was expressed by the epithelial cells of the olfactory epithelium. In addition, a comparable pattern of immunoreactivity was detected in human tissue using Stro-1 and cytokeratin, suggesting the presence of similar cells in this tissue. The identification of a nonepithelial multipotent cell in the OM may explain the varied reports on olfactory stem cell differentiation capacity in vitro and in vivo and illustrates the cellular complexity of this tissue as a potential source of stem cells for transplantation and translation to the clinic. STEM CELLS 2009;27:2196–2208


The peripheral olfactory system is a useful source of cells for exploring cell transplant paradigms for central nervous system (CNS) repair due to its ability to renew olfactory receptor neurons (ORNs) throughout life [1–3]. It comprises the olfactory mucosa (OM), the peripheral component where external olfactory stimuli are detected, and the olfactory bulb, the central component where these signals are integrated and processed. The OM is made up of the olfactory epithelium (OE), where ORNs reside, and the lamina propria (LP), a layer of loose connective tissue that lies beneath the epithelium. The axons of ORNs, which are ensheathed by olfactory ensheathing cells (OECs), extend from the OE through the LP to the olfactory bulb.

It is thought that the continuous regenerative capacity of the olfactory system is due to stem cells present in the basal layer of the OE. The two stem cell candidates in the basal layer are the horizontal basal cells (HBCs) and globose basal cells (GBCs) [3, 4]. There have been conflicting views over the years regarding which of these is a stem cell [3–6]. Recent fate-mapping studies have suggested that the HBCs replenish the GBC progenitors, which in turn can reconstitute the cellular architecture of the OE [7, 8]. A great deal of effort is being made to expand these putative OE stem cells in vitro and elucidate their potential for transplant-mediated repair in a clinical setting. Several of these studies have examined the differentiation potential of these cultured human olfactory cells after transplantation in vivo and have reported that they form numerous non-epithelial-related tissue-specific cell types such as heart, liver, kidney, hair, and hematopoietic cells [9–12]. However, there is no direct evidence to suggest that HBCs or GBCs are the multipotent olfactory stem cells described in these reports. Therefore, it is important to determine to what extent the OE stem cell is really contributing to this multipotentiality.

In our effort to identify, isolate, and expand olfactory stem cells, we cultured the rat embryonic OM in conditions that generate CNS neural stem cells within neurospheres. Under these conditions, we identified two morphologically and antigenically distinct types of sphere-forming cells. One type, termed OM-I, contained multipotential cells with mesenchymal stem cell (MSC)-like characteristics and originated from the LP. The second type, termed OM-II, was cytokeratin-positive and likely to originate from the OE. Our results suggest that the OM contains a novel source of multipotent cells that are nonepithelial in origin. We hypothesize that these latter cells might contribute to the transplant-mediated repair of CNS injury when nasal mucosa is biopsied and transplanted into patients [13, 14].



Sprague-Dawley rats were used for the study. Embryonic day 15 (E15) rats were harvested by Caesarean section from pregnant rats, previously euthanized with carbon dioxide. Olfactory turbinates were obtained from postnatal day 7 (P7) and adult rats.

Primary OM Cell Culture

The brain and olfactory bulbs were removed from E15 embryos. Heads were cut longitudinally, and each turbinate was dissected with forceps under the microscope and placed in fresh cold Hanks' balanced saline solution (HBSS). The OM was microdissected to eliminate contamination of neighboring tissue and respiratory epithelium and dissociated with 1.33% collagenase (Sigma-Aldrich, St. Louis, for 20 minutes followed by 5 minutes in DNAse to reduce clumping (0.04 mg/ml bovine pancreas DNAse, 3.0 mg/ml bovine serum albumin-fraction A in L15; Sigma-Aldrich). Cells were mechanically dissociated by pipetting and plated in a 60-mm Petri dish with standard serum-free neural stem cell medium (NSM; see supporting information for details) [15]. Cells were plated at approximately 1-3 × 104 cells per milliliter to avoid cell aggregation (equivalent to six to seven embryos). The cultures were incubated at 37°C and 7% CO2, changing half of the medium every 2-3 days. After 2-3 days, the medium containing the cell suspension was transferred to a new Petri dish to remove the cells attached to the dish. For proliferation and differentiation studies, spheres cultured for up to 18 days were termed early spheres, whereas those maintained for longer periods were termed long-term spheres. Spheres were plated in noncoated flasks with Dulbecco's modified Eagle's medium and 10% fetal bovine serum (DMEM-FBS; Autogen Bioclear, Wiltshire, U.K., to generate monolayers.

Studies were carried out on purified sphere types. Single OM-I spheres were collected macroscopically or microscopically. OM-II spheres were collected after removing OM-I spheres by collagenase treatment. Collagenase only dissociated OM-I spheres, allowing the enrichment of OM-II spheres by filtration through a 40-μm diameter syringe (Millipore, Billerica, MA, to eliminate single OM-I cells, small spheres, and clumps. The OM-II spheres retained by the filter were filtered several times to ensure purity.


Immunocytochemistry was performed either on spheres or cell monolayers. Spheres were collected with Pasteur pipettes and stained in suspension in microplates (Jencons, Primary antibodies (supporting information Table 1) were used to detect neurons, glial cells, epithelial cells (HBCs and sustentacular cells), smooth muscle cells, fibroblasts, MSCs, and extracellular matrix proteins (fibronectin, laminin, and collagen I). Secondary antibodies conjugated to fluorochromes (Southern Biotechniques, Cambridge BioScience, Cambridge, U.K., were used at 1:100. HBSS containing 5% donor calf serum (Autogen Bioclear) with Hepes (Gibco, Grand Island, NY, was used to dilute the antibody stocks and wash between incubations.

Surface antigens were detected by incubating with the primary antibody followed by the appropriate secondary antibodies and then by fixing in methanol for 20 minutes at −20°C. For detecting intracellular antigens, cells were prefixed. For CD105 only, cells were prefixed with 4% paraformaldehyde (PF; Sigma-Aldrich) instead of methanol for 20 minutes, followed by 2% bovine serum albumin (Sigma-Aldrich) in phosphate-buffered saline for 30 minutes prior to antibody incubation. Primary and secondary antibodies were added for 60 minutes and 30 minutes, respectively, at room temperature (RT). Cells on coverslips were mounted in VectaShield (Vector Laboratories, Burlingame, CA, containing 4′6-diamidine-2-phenylindole-dihydrochloride (DAPI) to visualize nuclei. Spheres were passed onto a glass slide, mounted in VectaShield-DAPI, and covered with a coverslip. No pressure was applied to the coverslip to ensure preservation of spherical morphology for initial low magnification photomicrographs. For higher magnification images, the spheres were slightly distorted with pressure.

Treatment with 5-Bromo-2-Deoxyuridine

To identify proliferating cells within the OM spheres, 20 μM 5-bromo-2-deoxyuridine (BrdU) was added to 1-week and 6-week cultures for 18 hours. After this time, individual OM-I and OM-II spheres were collected and washed with fresh medium. BrdU uptake was visualized using indirect immunofluorescence on spheres in suspension. Spheres were fixed in methanol at −20°C for 20 minutes, washed, fixed in 0.2% PF for 1 minute, and then incubated in 0.07 M sodium hydroxide for 7-10 minutes at RT. Spheres were incubated with anti-BrdU (IgG1, 1:20; DAKO, Glostrup, Denmark, for 45 minutes and then with the fluorescein conjugate secondary antibody (1:100) for 30 minutes. Each sphere was placed onto a slide and mounted in VectaShield-DAPI.

Self-Renewal Assays

Twenty to 30 early primary OM-I spheres (8-15 days in vitro [DIV]), collected using a Pasteur pipette, were treated with collagenase. The resulting single-cell suspension was filtered and plated in NSM at a low density of 104 cells per square centimeter in 24-well microplates. Wells were checked every 2 days for secondary sphere formation. Secondary spheres were counted and then collected to assess tertiary sphere formation. This experiment was performed 10 times.

Monolayers from OM-I spheres were generated using DMEM-FBS. Early and long-term primary spheres and secondary spheres were studied. Single-cell suspensions from two to three spheres were obtained as described above. Cells were plated in a noncoated 25-cm3 flask (Nunc, Rochester, NY, with DMEM-FBS. Self-renewal was assessed by the ability to be passaged. Three experiments were performed for both early and long-term spheres.

Differentiation Assays

Osteogenesis and Adipogenesis.

The potential of the OM-I spheres to differentiate into adipogenic and osteogenic lineages was assayed on primary spheres and on monolayers. Primary spheres were plated directly into 24-well microplates for 2-3 days with DMEM-FBS before treatment (one to two spheres per well). Induction factors were added to six wells, and six wells were kept as controls. Early passaged monolayers (passaged once [p1]) were plated at 105 cells per square centimeter in noncoated 12.5-cm3 flasks for 3-4 days with DMEM-FBS until they reached confluency prior to induction. Three flasks were treated, and three flasks were controls. To test the persistence of the differentiation capacity at late passage, monolayers passaged 10 times (p10) were plated at 105 cells per square centimeter in noncoated 12.5-cm3 flasks incubated with DMEM-FBS. Induction factors were added to three flasks, and three flasks were kept as controls. All of the experiments were performed three times. Methods for adipogenic and osteogenic differentiation based on Caddick et al. [16] can be found in the supporting information.

Smooth Muscle Differentiation.

Differentiation of OM-I spheres into smooth muscle cells was induced by DMEM-FBS and assessed by α-smooth muscle actin (αSMA) and calponin immunoreactivity (IR) on p1 and p10 monolayers. Cells were plated on poly-L-Lysine-coated coverslips (13 μ/ml, Sigma) and stained after 24 hours. αSMA- and calponin-expressing cells were counted for each passage, and data were expressed as a percentage of total nuclei. A minimum of 300 cells were counted in three different fields from duplicate coverslips for each marker. Experiments were performed three times for each passage, and the data were presented as the mean ± SD.

Neuronal Differentiation.

To access the potential of OM-I spheres to differentiate into neurons, we cultured them using the protocol described by Yanjie et al. [17]. Briefly, cells were placed in Neurobasal A media (Invitrogen, Carlsbad, CA, containing 5% horse serum, 1% FBS, 0.5 μM retinoic acid (Sigma-Aldrich), 10 μg/ml brain-derived neurotrophic factor (BDNF; Peprotech, Rocky Hill, NJ,, and 10 ng/ml fibroblast growth factor 2 (FGF2; Peprotech). Cells were maintained for 6 days in these conditions, and lysates were generated for Western blotting using antibodies to the neuronal markers Tuj-1 (Chemicon, Temecula, CA,; Millipore; clone TU-20, anti-mouse, MAB1637) and neurofilament L (Cell Signaling Technology, Beverly, MA,; C28E10, anti-rabbit, catalog number 28375), glial fibrillary acidic protein (GFAP), and the housekeeping protein glyceraldehyde-3-phosphate dehydrogenase (GAPDH; Santa Cruz Biotechnology Inc., Santa Cruz, CA,; Insight Biotechnology Ltd., Wembley, U.K., Noninduced OM-I spheres and a neuroblastoma cell line (SH-SY5Y, ECACC catalog number 94030304) were used as a negative and positive controls, respectively.


OM turbinates from E15, P7, and adult animals were dissected. Human OM was collected during surgery where removal was part of normal surgical procedures for septoplasty and nasal polypectomy (approved by the Central Office for Research Ethics Committee). Tissue was collected from eight patients (four males and four females), ranging from 18-67 years old, of whom seven had septoplasty and one had nasal polypectomy. Samples were taken from as close to the posterior nasal cavity as possible. We confirmed the presence of olfactory tissue by immunocytochemistry with Tuj-1, which defines olfactory tissue [18].

Depending on the primary antibody, two different approaches were used to preserve tissue before immunohistochemistry (see supporting information). Tissue was placed in sucrose solution immediately prior to being frozen in liquid nitrogen and stored at −70°C until use. Longitudinal sections (15 μm thick) were cut on a Bright cryostat (Bright Instrument Co. Ltd., at −25°C and placed on Vectabond-coated slides (Vector Laboratories). Sections were left to air dry overnight at RT and stored at −70°C. For immunofluorescence procedures, see supporting information. For Stro-1 immunolabeling, the tissue had to be quickly immersed (snap frozen) in liquid nitrogen. Omission of primary antibody was used as a control.

Fluorescence-Activated Cell Sorting Analysis

To isolate and enrich for the potential mesenchymal-like stem/progenitor cells from the human OM, tissue was enzymatically dissociated from four biopsies, as described previously [19], and placed in culture in NSM for 1-2 weeks to propagate any potential stem cell populations. The resultant monolayer of cells was trypsinized and labeled with Stro-1 (1:50) for 90 minutes at 4°C followed by IgM-fluorescein isothiocyanate for 45 minutes, and 10,000 cells were analyzed using a fluorescence-activated cell sorting (FACS) Analyzer and FlowJo software (Becton, Dickinson and Company, Franklin Lakes, NJ, Controls were included in which the antibody was omitted.

Western Blotting

Protein extracts (12 μg per lane) were separated by electrophoresis in reducing conditions and transferred onto polyvinylidene difluoride membrane (Invitrogen). For immunoblotting, membranes were first incubated in blocking solution (10 mM Tris-HCl, pH 7.6, and 150 mM NaCl containing 5% milk powder and 0.5% Tween 20) for 1 hour followed by primary antibodies (Tuj-1, neurofilament L, GFAP, and GAPDH) incubated overnight at 4°C and the secondary mouse (Tuj-1) or rabbit IgG-horseradish peroxidase (Southern Biotechniques) antibodies incubated for 1 hour at RT. Protein-antibody binding was detected using chemiluminescence (ECL Plus; GE Healthcare UK Ltd., Chalfont St. Giles, U.K.,

Sphere Size Measurement and Image Analysis

Sphere diameter was measured from phase-contrast photomicrographs taken on OM cultures after 7, 12, and 20 div. Data were statistically analyzed from three separate experiments. Photomicrographs were captured with a Zeiss Axiovert 25 microscope (Carl Zeiss, Jena, Germany, equipped with a Fuji digital camera (Fujifilm, Tokyo, Immunofluorescence images were taken using either a Zeiss Axioskop 50B epi-fluorescence microscope (Carl Zeiss) equipped with a CCD camera (Princeton Instruments, Trenton, NJ, and Metamorph 6.1 software (Molecular Devices Ltd., Berkshire, U.K., or Olympus FV1000 inverted laser scanning confocal microscope equipped with the Olympus FV10-ASW software (Olympus, Tokyo, Images were imported to and processed with Adobe Photoshop 9.0 (Adobe, San Jose, CA,


Characterization of OM Cultures in NSM Growth Conditions

During the first 3 days after plating in NSM, embryonic OM cells of differing morphologies attached to the culture dish. A large proportion of these were likely to be differentiated cells, lacking proliferative capacity in these culture conditions. These cells were removed to avoid any possible environmental signals for attachment and differentiation. After 4-5 div, a population of spheres appeared in suspension. We identified two distinct morphologies under phase microscopy, which were termed OM-I and OM-II spheres. OM-I spheres were generally spherical, formed by round cells, and resembled CNS neurospheres (Fig. 1A). In contrast, the OM-II spheres were more heterogeneous in morphology (spherical, ovoid, and irregular) with a well-defined, phase bright perimeter (Fig. 1B, arrow). The two types could be distinguished from each other by the fact that individual cells were more obvious in the OM-I spheres. By day 7, both types of spheres took up BrdU, suggesting they were mitotically active (Fig. 1C, 1D). During the next 7 days, differences between both sphere types were more evident. The OM-I spheres increased in size until they became macroscopically detectable with an average sphere diameter of 345 ± 71 μm as measured after 12 div (Fig. 1E). Many of the OM-I spheres proliferated for longer periods and, after 20 div, had increased in diameter (699 ± 129 μm). After 20-25 div, long-term spheres had a less apparent increase in size even though BrdU was still incorporated (data not shown). As the OM-I spheres increased in size, they started to lose their spherical morphology, form aggregates between themselves, and attach to the culture dish (Fig. 1F).

Figure 1.

Two morphologically different olfactory spheres. OM-I (A) and OM-II (B) spheres; white arrow denotes phase bright perimeter. Proliferating cells in OM-I (C) and OM-II (D) spheres. (E): OM-I spheres increase in diameter with time in culture. (F): Two large OM-I spheres aggregated. (G): OM-II sphere (dotted outline) integrated with an OM-I sphere. Scale bars: (A, B, E-G) 10 μm (phase micrographs); (C, D) 50 μm (immunofluorescence). Blue: nuclei stained with 4′6-diamidine-2-phenylindole-dihydrochloride. Abbreviations: BrdU, 5-bromo-2-deoxyuridine; div, days in vitro.

OM-II spheres showed less cellular changes with time in culture. Their size increased during the first week but did not visibly change thereafter. This could be due to a slow growth rate or a balance between proliferation and death. OM-II spheres also became more adhesive and formed aggregates with the OM-I spheres (Fig. 1G). In some cases, the proliferating OM-I sphere appeared to integrate with the OM-II sphere (Fig. 1G); however, there was no morphological evidence that one type of sphere could progress into the other. Furthermore, this was later supported by immunofluorescence data in which antigenic markers were mutually exclusive. OM-II spheres never attached to the dish surface unless via its attachment to an OM-I sphere. The nonattached OM-II spheres could remain in suspension for 3-4 weeks, after which they began to disaggregate. Both types of OM spheres were characterized individually.

Self-Renewal of OM Spheres


Early primary spheres (8-15 div) were used because they were more proliferative and their increase in size was easier to detect. Older spheres (>20 div) tended to attach to the dish, formed aggregates more readily, and were difficult to enzymatically dissociate.

Embryonic OM cultures yielded 16-20 OM-I spheres per animal. Primary sphere cells were replated at a low density to avoid cell aggregation in NSM. Secondary spheres formed at a different time scale and pattern of formation compared with primary spheres. In the first week, cells with a fibroblast-like morphology were attached to the dish, forming colony-like structures. Spherical cells started to appear at the top of these colonies after 10-15 div. These cells grew in number, forming spheres but maintaining a semi-attachment to the dish. Spheres appeared in suspension after 20-25 div. A proportion of these spheres grew in suspension, but most of them increased in size only while they were attached to the dish. On average, the size of the secondary spheres was smaller than the size of the primary spheres, with an average diameter of 218 ± 14 μm. Secondary spheres in suspension could reattach to the dish with time in culture.

Secondary spheres were produced in all of the experiments, but the number of secondary spheres was always lower than expected. Assuming that each primary sphere contains at least one stem/progenitor cell, we expected that the number of secondary spheres formed would be equivalent to the number of primary spheres generated. We estimated a loss of 15%–40% of stem cells after the first passage in NSM conditions.

DMEM-FBS was used to passage primary OM-I spheres, which has generally been used for MSC culture. DMEM-FBS enhanced the attachment of the OM-I spheres to the dish surface. After 1 day, the cells spread out from the spheres, proliferated, and changed in morphology from round to largely flat. In DMEM-FBS, the OM-I-derived cells kept proliferating for at least 30 passages.


A similar approach was used to assess the stem cell potential of OM-II spheres. However, dissociation in collagenase did not disrupt their three-dimensional structure. Coating of collagen I, laminin, and fibronectin to dishes failed to promote sphere adhesion. The spheres were also cultured in DMEM-FBS, but this did not improve attachment to the dish, and the spheres disappeared after several weeks. Due to these distinct biological features, we were unable to study the stem cell characteristics of OM-II spheres using standard methodology.

Phenotype of the OM Spheres

OM-I Spheres.

Early primary OM-I spheres (12-18 div) expressed the neural stem cell marker nestin and markers that characterize neuronal and glial cells and their precursors. Nestin was widely expressed, with the strongest IR associated with cells with elongated morphology (Fig. 2A). p75NTR was expressed at lower levels than nestin and distributed heterogeneously throughout the spheres (Fig. 2B). Polysialic acid-neuron cell adhesion molecule (PSA-NCAM) was strongly expressed in groups of cells (Fig. 2C). A small subpopulation of cells expressed both PSA-NCAM and p75NTR, which may represent a subtype of OECs (Fig. 2C) [20]. Occasionally, after a long period in culture, cells from the spheres attached to the dish. Many of these cells expressed GFAP (Fig. 2D) and p75NTR, with even less expressing Tuj-1 (Fig. 2E), suggesting spontaneous differentiation. OM-I spheres also expressed non-neural markers. Intercellular adhesion molecule-1 (ICAM-1) was widely expressed in these spheres (Fig. 2F), and a large number of cells expressed αSMA, which out numbered those expressing p75NTR or PSA-NCAM (Fig. 2G). A small number of cells coexpressed αSMA and ICAM-1 (Fig. 2H).

Figure 2.

OM-I sphere phenotype. OM-I spheres express high levels of nestin (A), low p75NTR(B), and PSA-NCAM (C). They also express GFAP (D), Tuj-1 (E), high levels of ICAM-1 (F), and αSMA (G). (H): Coexpression of ICAM-1 and αSMA. Expression of mesenchymal stem cell markers: CD90 and Stro-1 (I) and CD105 (J). High expression of laminin (K) and fibronectin (L). Scale bars: (A–G, I–L) 50 μm; (H) 30 μm; (C, inset) 50 μm; (I, inset) 10 μm. (A–G, I—L): Fluorescence images. (H; I, inset): Confocal images. Blue: nuclei stained with 4′6-diamidine-2-phenylindole-dihydrochloride. *, Example of coexpression in cells. Abbreviations: GFAP, glial fibrillary acidic protein; ICAM-1, intercellular adhesion molecule-1; PSA-NCAM, polysialic acid-neuron cell adhesion molecule; SMA, smooth muscle actin.

The identification of cells with neural and non-neural phenotypes and the limited self-renewal in NSM suggest the presence of a non-neural-restricted stem/progenitor cell. OM-I spheres immunolabeled for three markers commonly used to identify MSCs—Stro-1, CD90, and CD105. The relative staining proportion for these markers was Stro-1 > CD90 > CD105 (Fig. 2I, 2J). Coexpression of Stro-1 and CD90 in a subpopulation of cells supports the presence of MSC-like cells in OM-I spheres. The tightly adhesive characteristics of the OM-I spheres suggested the expression of extracellular matrix molecules, and in accordance, we detected IR for laminin (Fig. 2K), fibronectin (Fig. 2L), and collagen I (data not shown). Secondary spheres had the same heterogeneous antigenic phenotype as primary spheres. Due to limitations of detecting IR in a three-dimensional sphere, quantification of several of these markers was carried out on p1 monolayers derived from the primary OM-I spheres (supporting information Table 1).1

Table 1. Cell type markers, dilution, and source
inline image

OM-II Spheres.

These spheres were immunolabeled with the same panel of antibodies as for OM-I spheres with differing results, which were clearly seen in aggregates of the two types of spheres. OM-II spheres labeled intensely for cytokeratins (Fig. 3A–3C, 3E, 3F). Tuj-1- and nestin-positive cells were occasionally seen within OM-II spheres (Fig. 3B, 3C). OM-II spheres did not express ICAM-1, p75NTR, αSMA (Fig. 3A), or the MSC markers (Fig. 3D, 3E). The IR for these markers was only associated with the OM-I spheres (Fig. 3). Central staining for CD90 in OM-II spheres (Fig. 3D) illustrates a possible connection point by which the two types of spheres interact and adhere to each other. Fibronectin IR was not detected in the OM-II spheres (Fig. 3F).

Figure 3.

OM-II sphere phenotype. OM-II spheres are immunoreactive for cytokeratins (A–C, E–F) but not for αSMA (A), nestin (C), CD90 (D), Stro-1 (E), and fibronectin (F). Occasional Tuj-1-positive cells within OM-II spheres are seen (arrow). Scale bar: 50 μm. Fluorescence images. Blue: nuclei stained with 4′6-diamidine-2-phenylindole-dihydrochloride. (A, inset) Detail of cytokeratin-expressing cells. *, OM-I sphere localization. Abbreviations: Cytk, cytokeratin; SMA, smooth muscle actin.

Differentiation of OM-I Spheres into Mesenchymal Lineages

OM-I spheres were induced to differentiate into osteocytes, adipocytes, and smooth muscle cells. We tested the mesenchymal differentiation potential of primary OM-I spheres only grown in NSM and the persistence of this multipotentiality in p1 and p10 monolayers subsequently grown in DMEM-FBS.


Under osteogenic induction, early OM-I spheres and p1 monolayers formed calcium deposits as visualized by alizarin red S staining (Fig. 4A–4D). Cultures maintained in osteogenic conditions proliferated faster than controls. Cell clusters above the monolayer or surrounding the periphery of the sphere appeared in induced cultures in which calcium staining was more intense (Fig. 4D, asterisk and arrows). Strong calcium depositions were observed at the top of these clusters (Fig. 4D). In contrast, even though the control cultures appeared healthy, they did not form cell clusters or produce calcium deposition (Fig. 4C). The initial time in culture in NSM conditions did not affect the osteogenic differentiation potential because the induced long-term OM-I spheres and their p1 monolayers displayed similar results as early ones. However, control (Fig. 4A) and induced (Fig. 4B) p10 monolayers, derived from either early or long-term spheres, did not form calcium deposits or cell clusters, although similar levels of confluency and viability were observed.

Figure 4.

Osteogenic and adipogenic differentiation of OM-I cells. Osteogenic (A–D) and adipogenic (E–H) induction of monolayers and spheres assessed by ARS and ORO staining, respectively. Only induced OM-I spheres and p1 monolayers produced (D) ARS-positive clusters (arrows) and calcium depositions (asterisks) and (H) ORO-positive lipid droplets (arrows). Scale bar: 10 μm. Phase images. (H, inset): Lipid droplets before ORO staining. *, ORO background staining. Abbreviations: ARS; alizarin red S staining; ORO, oil red O staining; p1, one passage; p10, 10 passages; Sph, sphere.


In induced spheres and p1 monolayers, large cells with bright, large droplets were seen under phase-contrast microscopy (Fig. 4H, inset). In sphere cultures, these cells appeared attached to the sphere's periphery. Induced cultures contained cells that stained intensely red with oil red O (ORO) typical of lipid droplets (Fig. 4H). The p1 monolayers and the sphere periphery of control cultures did not stain with ORO (Fig. 4G). Stressed cells with small pink droplets were found in both control and induced cultures, together with a background staining due to the cell density (Fig. 4G, 4H). Long-term OM-I spheres and monolayers gave similar results. However, control (Fig. 4E) and induced (Fig. 4F) p10 monolayers, although confluent and viable, did not contain ORO staining. Cells with the features shown in the inset in Figure 4H were never seen in p10 cultures.

Smooth Muscle Cell Differentiation.

The number of passages affected the antigenic phenotype of the monolayers derived from OM-I spheres (Fig. 5). The phenotype of p1 monolayers (Fig. 5A–5C) was consistent with the phenotype of OM-I spheres and was also quantified (Fig. 5G; supporting information Table 1). However, after 10 passages, the expression of some markers changed (Fig. 5D–5G). Coexpression of αSMA and calponin increased to almost 100% (Fig. 5D). Nestin expression was also higher after 10 passages, whereas p75NTR was no longer detected, which occurred after three to four passages (data not shown). However, the number of Stro-1-, CD90-, and ICAM-1-positive cells did not change noticeably (Fig. 5E, 5F). Despite the lasting expression of the MSC markers, the p10 monolayers expressed a phenotype characteristic of smooth muscle cells by coexpressing αSMA, calponin, and ICAM-1. This differentiated state correlated with their lack of osteogenic and adipogenic differentiation previously described in this section.

Figure 5.

Comparison of one passage (p1) and 10 passage (p10) OM-I monolayer phenotypes. (A–C): Nestin was expressed by a few glial (p75NTR) and smooth muscle cells (αSMA) in p1 monolayers. There was high expression of Stro-1 but low expression of calponin. (D–F): p10 cells highly expressed calponin, αSMA, ICAM-1, and Stro-1. (G): Quantification of cell markers at p1 and p10 (mean ± SD). (H): Western blot analysis showing no increase in levels of Tuj-1 after neuronal induction. Lane 1 = positive control (SH-SY5Y), lane 2 = experimental, and lane 3 = noninduced control. Scale bar: 50 μm. Fluorescence images except one confocal image (C). Blue: nuclei stained with 4′6-diamidine-2-phenylindole-dihydrochloride. Arrows in E indicate cell colabeling. *, Calponin and Stro-1 cell colabeling. Abbreviations: ICAM-1, intercellular adhesion molecule-1; SMA, smooth muscle actin.

Neuronal Differentiation.

The ability of the spheres to be induced into neurons under inducing culture conditions was assessed by Western blotting using antibodies to the neuronal markers Tuj-1 and neurofilament L and the glia marker GFAP. In Figure 5H, a lack of upregulation of Tuj-1 in the experimentally induced spheres (lane 2), when compared with the positive control (lane 1) or noninduced control (lane 3), can be seen. Similar results were obtained using antibodies to neurofilament L and to GFAP (data not shown), suggesting that under these induction factors, differentiation into other neuronal phenotypes was not possible.

In Vivo Localization of Antigens That Define the OM-I and OM-II Sphere-Forming Cells

The OM was immunolabeled with the same panel of antibodies used to characterize the OM spheres (Fig. 6). The pseudostratified distribution of the ORNs was identified by Tuj-1 IR together with their axons, which extend into the LP to form the nerve bundles (Fig. 6A, 6E). Cytokeratin, a marker of OM-II spheres, was strongly expressed in the OE but not in the LP. Cytokeratin IR was present in the basal (Fig. 6A, 6C) and apical layers (Fig. 6C). This apical IR is less obvious in the presence of intense Tuj-1 staining (Fig. 6A). The markers identified in OM-I spheres were mainly located in the LP. Nestin, an intense OM-I sphere marker, was present in the LP but also in the epithelium, mainly in its basal and medial layers (Fig. 6B). The MSC marker Stro-1 was intensely expressed by cells of the LP but not by the basal cytokeratin-positive cells or other cells of the OE (Fig. 6C). CD90 IR appeared in the LP, as punctuate, and perinuclear regions in some cells (Fig. 6D).

Figure 6.

Rat olfactory mucosa immunohistochemistry. (A, E): Tuj-1 expression in olfactory receptor neurons (ORNs). Arrows indicate ORN axons and nerve bundles. (A, C): Cytokeratin expression in OE basal and apical layers. (B): Nestin immunoreactivity (IR) in LP and OE. (C) Stro-1 expression only in LP. *, Apical expression of cytokeratin. (B, F): αSMA is not detected at E15 but is detected at P7 coexpressed with calponin. (D): CD90 and p75NTR IR in LP. Arrows indicate perinuclear localization of CD90 IR. *, Olfactory nerve surrounded by p75NTR-IR. (E): Fibronectin IR in LP. *, OE basal cells. (G): Cytokeratin expression in the adult rat OE and Stro-1 expression in the LP (arrow). Scale bars: (A) 50 μm; (B–D) 30 μm; (E–G) 20 μm. (A): Fluorescence image. (B–G) Confocal images. Blue: nuclei stained with 4′6-diamidine-2-phenylindole-dihydrochloride. Dotted line: demarcation of OE and LP regions. Abbreviations: Calp, calponin; Cytk, cytokeratin; E15, embryonic day 15; LP, lamina propria; OE, olfactory epithelium; P7, postantal day 7; αSMA, α-smooth muscle actin.

In the embryonic tissue, αSMA and calponin IR were never detected in either the OE or LP (Fig. 6B, calponin not shown, [21]). However, in postnatal tissue sections, αSMA and calponin were intensely coexpressed by cells of the LP (Fig. 6F). The peripheral glia (Schwann cells/OECs) was detected in the embryonic OM by expression of p75NTR IR associated with the LP as expected (Fig. 6D). Fibronectin IR was associated with the LP (Fig. 6E) and not detected in ORNs or basal cells of the OE (Fig. 6E). In all tissue sections, nonspecific immunofluorescent labeling was checked by omitting the primary antibody.

To assess whether Stro-1-positive cells were present in adult tissue, both rat and human OM sections were immunolabeled with cytokeratin and Stro-1 (Figs. 6 and 7, respectively). Fewer immunoreactive cells were found in the adult rat tissue when compared with embryonic tissue. The human tissue contained cytokeratin-positive cells in the OE and Stro-1-positive cells in the LP, as observed for the rat OM. Specific Stro-1 IR was localized to spindle-shaped cells in the LP. However, some staining was associated with small round cells, which were shown to be nonspecific when the primary antibody was omitted (Fig. 7B). The presence of olfactory tissue in the human samples was confirmed by Tuj-1 staining of the ORNs (Fig. 7C). Identification of potential mesenchymal-like stem/progenitor cells in the human OM was further supported by FACS analysis of the human OM cells, which yielded 6.7% ± 2.9% Stro-1-positive cells.

Figure 7.

Human olfactory mucosa (OM) immunohistochemistry and fluorescence-activated cell sorting (FACS) analysis. (A): Cytokeratin in olfactory epithelium (OE) basal and supporting cells and Stro-1 expression in the lamina propria (LP). Arrows illustrate specific Stro-1 immunoreactivity. *, Nonspecific staining. (B): Control tissue in which the primary antibody was omitted. *, Nonspecific staining. (C): Tuj-1 expression in olfactory receptor neurons. Blue: nuclei stained with 4′6-diamidine-2-phenylindole-dihydrochloride. Scale bars: (A, B) 20 μm; (C) 30 μm. (A–C): Fluorescence images. Dotted line: demarcation of OE and LP regions. (D): FACS profile of control (no antibody) and experimental (Stro-1-positive) cells obtained from OM tissue cultured for 1-2 weeks in neural stem cell media. Abbreviations: Cytok, cytokeratin; Exp, experimental.


In this study, we have shown that rat embryonic OM tissue grown in defined medium generated two types of spheres, which we termed a priori as OM-I and OM-II spheres, and that these appear to originate from the LP and the OE, respectively. We propose, based on this study, to rename these spheres in the future as LP spheres (OM-I) and OE spheres (OM-II), which clarifies their anatomical origin.

OM-I Spheres Contain Nonepithelial Multipotent Cells

We demonstrated that within the heterogeneous antigenic phenotype of LP spheres, there are proliferating cells that possess multipotential characteristics. These sphere-forming cells resembled mesenchymal-like stem/progenitor cells because they expressed mesenchymal-specific markers (Stro-1, CD90, and CD105) and, in the appropriate culture conditions, differentiated into three different mesenchymal lineages. Because these mesenchymal markers were also detected in the LP using immunohistochemistry, we suggest that there are multipotent mesenchymal-like stem/progenitor cells in the LP of the embryonic OM. It is these cells that, in vitro, contribute to the OM-I spheres described in the present study. These multipotent cells are likely to persist in the adult and in different species because we found Stro-1-expressing cells in the LP of the adult rat and human OM. Further work is ongoing to isolate and characterize these cells using a larger panel of markers and to assess their differentiation potential.

The existence of mesenchymal-like stem/progenitor cells in epithelial-containing tissues has been previously described for the lung and oral mucosa, where these cells have been associated with the supportive connective tissue and vasculature [22–24]. Studies on mesenchymal-like stem cells after long-term passage in serum demonstrated that their multipotentiality can be affected. For example, it was shown that self-renewal and multipotentiality of bone marrow MSCs were affected by long-term culture conditions [25, 26] and also that cell contact with differentiated cells can influence stem cell fate into specific cell lineages [27, 28]. This is in accordance with our data, which showed that the cells from the OM-I spheres became terminally differentiated into smooth muscle cells at high passage and lost their ability to differentiate into osteocytes and adipocytes. Surprisingly, this loss of multipotentiality of OM-I sphere-forming cells did not seem to correlate with a loss of Stro-1 and CD90 expression in vitro.

Of interest was the finding that the OM-I sphere antigenic phenotype was not restricted to mesenchymal cell types. Cells within the spheres also expressed nestin, neuronal markers (PSA-NCAM and Tuj-1), and peripheral nervous system glial (p75NTR and GFAP) markers. MSCs from differing origin have been reported to express markers of neuronal cell types under specific culture conditions (reviewed in [29]). However, when OM-I spheres were treated with well-known neuronal induction factors (BDNF, FGF2, and retinoic acid) that are used to induce neuronal differentiation of bone marrow MSCs [17, 30], they did not express neural markers such as Tuj-1 or GFAP. This may be either because the culture conditions did not meet the requirements for differentiating olfactory mesenchymal-like stem/progenitor cells into neurons or because these cells lack this potential. In addition, induction of neuronal markers on its own is not sufficient evidence that functional neurons have been produced. Furthermore, a recent study has suggested that the formation of neurons by non-neural stem cells could be an artifact of growth in culture rather than a differentiation process [31, 32]. Alternatively, expression of the various neural markers might suggest the presence of a wider multipotent progenitor cell within the OM-I sphere that can differentiate into mesenchymal-like stem/progenitor cells, as well as neural lineages. This feature is found in progenitors originating from the neural crest (NC), which migrate to diverse regions during embryogenesis and differentiate into neuroectodermal derivatives. In the head and neck, the NC also yields ectoderm-derived mesenchymal cells, which differentiate into smooth muscle cells, dermis, odontoblast, and bone [33, 34]. Furthermore, the frontonasal mesenchyme (future LP) adjacent to the olfactory placodal epithelium is primarily derived from cranial NC cells [35].

Embryonic and postnatal OM of mice has been shown to form CNS-like spheres in vitro with limited capacity for self-renewal [36–38]. We have shown that embryonic rat OM can form spheres in vitro with a similar capacity for self-renewal in serum-free defined medium. These reports suggest the presence of progenitor cells instead of stem cells in the OM cultures. Tissue-specific progenitor cells can proliferate and differentiate into multiple cell types but have a limited self-renewal capacity in contrast with the pluripotent embryonic stem cells. Moreover, the culture conditions typically used for CNS-neural stem cells might not be the appropriate growth conditions to maintain the self-renewal ability of the putative OM stem/progenitor cell for longer periods. Therefore, it may be more appropriate to suggest that the OM spheres contain cells that are more similar to progenitor cells.

Putative OE Stem Cell Versus Olfactory Nonepithelial Multipotent Progenitor Cells

Our in vivo immunohistochemical data demonstrated that the markers that label OM-I spheres also label cells situated in the LP. This suggests that the LP, and not the OE, is the origin of the OM-I multipotent cells. Furthermore, we show for the first time the existence of Stro-1-expressing cells in the LP of the OM, supporting the presence of mesenchymal-like stem/progenitor cells in this tissue. Interestingly, the absence of αSMA in the embryonic LP and its subsequent expression in later developmental ages in vivo are findings that are similar to our findings for calponin [21], which suggests an age-related differentiation process that seems to be induced in culture perhaps by removal of some local environmental tissue signals.

Immunohistochemistry of OE sections demonstrated that the OM-II cytokeratin-expressing spheres probably originated from the OE. In the OE, cytokeratins are expressed by two types of epithelial cells—the sustentacular cells, located in the apical layer, and the HBCs, situated in the basal layer of the OE [39, 40]. Recent reports using fate mapping analysis in vivo have shown that HBCs are progenitor cells with multipotential capacity yielding ORNs and non-neuronal cell types such as the sustentacular cells and cells of the Bowman's gland [7, 8]. These studies were carried out on postnatal animals, but recently, it has been suggested that the embryonic OE is largely devoid of HBCs and sustentacular cells and instead comprises a population of proliferating cell nuclear antigen-positive and nestin-positive apical and basal progenitors [37]. It is likely that these progenitors are the cells that form the OM-II spheres because they express cytokeratins and nestin and are located in the OE at the basal and apical surfaces. Therefore, the methodology used in our study appears to enrich for these OE progenitor cells in the OM-II spheres. However, the standard stem cell culture methodology failed to maintain these cells in vitro.

Our findings of a multipotential cell in the LP may have relevance to descriptions in the literature of an OE-derived multipotential stem cell. In general, it was not always clearly stated in the description of the preparations of olfactory stem cells whether the LP was present or absent [10, 36]. Contamination of OE cultures with cells from the LP (e.g., OM-I multipotent cells) could occur in these studies due to the difficulty in completely removing the LP during dissection. This may explain reports of multipotential human olfactory sphere-forming cells that differentiated into cell types of endoderm and mesoderm tissues both in vitro and when transplanted into the chick gastrula during in vivo development [10]. Although the authors claimed these were OE stem cells, they recognized that cross-contamination of the cultures could have occurred because no differences were seen in the biology of cells isolated from either the OE or LP. Furthermore, others have indicated that HBC stem cells can be purified by FACS sorting using antibodies to ICAM-1 [41]. Our results have shown that ICAM-1 labels other cell types in OM cultures such as αSMA-positive smooth muscle cells, suggesting that it may not be isolating just one cell type.


In this study, we have shown that there are two types of sphere-forming cells in the OM that could potentially contribute to the stem cell-like properties of the olfactory system. The multipotential capacity that has been attributed to the OE stem cell in studies in vitro and in vivo could be attributed instead to the multipotent nonepithelial progenitors we have identified in the LP. The idea that a useful cell type would exist in the LP of the olfactory system for the promotion of regeneration follows from this tissue's function of supporting neurogenesis throughout life. The regenerating ORN axons need to progress through the LP to reach their target in the olfactory bulb. It is tempting to speculate that the cells in the LP, together with the supporting OECs, play a role in promoting axonal outgrowth throughout life. It is this property that might makes cells within the LP a good choice for transplant-mediated repair for CNS lesions. Because in recent clinic trials the OM has been harvested from autologous human patients for the transplant-mediated repair of spinal cord injury, we suggest that these mesenchymal-like cells might be isolated and could contribute in part to the subsequent repair process.


We thank Thomas Gilbey, Margaret O'Prey, and Kurt Anderson for confocal assistance, Colin Nixon for immunohistological assistance, and Louise Clark and Saghir Sheikh for providing the human mucosal samples. This work was supported by a Medical Research Council strategic stem cell grant (to S.C.B. and J.S.R.), the Chief Scientist Office (S.C.B.), and the Multiple Sclerosis Society for Great Britain (S.C.B.). M.T. is currently affiliated with the Neuroscience and Molecular Pharmacology Group, Faculty of Biomedical and Life Sciences, University of Glasgow, Glasgow, Scotland.


The authors indicate no potential conflicts of interest.