Transcription and microRNA Profiling of Cultured Human Tympanic Membrane Epidermal Keratinocytes

Abstract

The human tympanic membrane (TM) has a thin outer epidermal layer which plays an important role in TM homeostasis and ear health. The specialised cells of the TM epidermis have a different physiology compared to normal skin epidermal keratinocytes, displaying a dynamic and constitutive migration that maintains a clear TM surface and assists in regeneration. Here, we characterise and compare molecular phenotypes in keratinocyte cultures from TM and normal skin. TM keratinocytes were isolated by enzymatic digestion and cultured in vitro. We compared global mRNA and microRNA expression of the cultured cells with that of human epidermal keratinocyte cultures. Genes with either relatively higher or lower expression were analysed further using the biostatistical tools g:Profiler and Ingenuity Pathway Analysis. Approximately 500 genes were found differentially expressed. Gene ontology enrichment and Ingenuity analyses identified cellular migration and closely related biological processes to be the most significant functions of the genes highly expressed in the TM keratinocytes. The genes of low expression showed a marked difference in homeobox (HOX) genes of clusters A and C, giving the TM keratinocytes a strikingly low HOX gene expression profile. An in vitro scratch wound assay showed a more individualised cell movement in cells from the tympanic membrane than normal epidermal keratinocytes. We identified 10 microRNAs with differential expression, several of which can also be linked to regulation of cell migration and expression of HOX genes. Our data provides clues to understanding the specific physiological properties of TM keratinocytes, including candidate genes for constitutive migration, and may thus help focus further research.

Electronic supplementary material

The online version of this article (10.1007/s10162-018-0660-1) contains supplementary material, which is available to authorized users.

Introduction

The tympanic membrane (TM) is a thin trilaminar tissue sheet separating the middle ear from the outer environment. It is covered on the lateral side by tympanic membrane keratinocytes (TMKs), which are specialised epidermal keratinocytes. The skin of the TM and medial part of the external auditory canal (EAC) lack all common skin appendages like hair follicles and sweat glands. This skin is particularly fragile, and the keratin layer easily desquamates as observed during aural toilet. The keratinocytes are recognised by their ability to constantly migrate centrifugally from a regenerative centre at the umbo region towards the periphery. The cells continue laterally along the EAC (Litton 1963; Alberti 1964). This action prevents accumulation of the shed keratin layer in the ear canal and removes debris, foreign bodies and wax. During wound healing of a TM perforation, migration is also of critical importance. Here, a keratin spur first protrudes from the edge to bridge the perforation. The spur then serves as a scaffold for the migrating keratinocytes. Thus, the proliferation and migration of the TM keratinocytes precedes ingrowth of fibroblasts during healing (Santa Maria et al. 2010; Unge and Hultcrantz 2011). For the study of TMK migration patterns in vivo, the most commonly used method is ink dots placed on healthy and diseased TMs (Alberti 1964; Santhi et al. 2015). An indirect measurement of the direction and velocity of the keratinocyte migration can be obtained by observing the ink displacement over time. Very little, however, is known about the cellular and molecular mechanisms responsible for this constitutive migration. Boxall et al. described that in vitro cultured TMKs and cholesteatoma keratinocytes were capable of en masse or collective migration (Boxall et al. 1988). This shows that the migration of TMKs is primarily due to cell-mediated processes irrespective of the interaction with neighbouring cells or extracellular matrix (ECM) structure and content.

The interest for regenerative medicine in otology evolved after reports on the effects of stem cell treatment on acute and chronic perforations (von Unge et al. 2003; Rahman et al. 2008). It was recently further stimulated by promising results from use of bioactive molecules such as basic fibroblast growth factor (FGF2) in combination with atelocollagen (Hakuba et al. 2010) and gelatine sponge/fibrin glue (Kanemaru et al. 2011). These studies have shown high rates of success in treating small- to medium-sized chronic perforations without the need for conventional surgery. The use of cultured cells for treatment of patients with chronic otitis was reported in 1997(Somers et al. 1997), but since then, there have been no reports on clinical use. However, an increasing number of papers have described culturing of TMKs for the purpose of potential transplantation (Levin et al. 2009; Mei Teh et al. 2013). The rationale for cell therapy must be dependent on the presence of TM-specific stem cells, and consequently that a stem cell deficiency may prevent regeneration. Our group has previously identified a distinct regeneration centre at the umbo where most of the proliferative activity takes place during regeneration after an acute perforation (Knutsson et al. 2011). Cells of the TM have been shown to generate neurospheres in culture that could be differentiated into neuronal lineages (Choi and Park 2014), which further supports the idea of stem cell presence.

To explore epidermal phenotype and functions specific to the TM, we carried out an unbiased experiment to compare the transcriptional messenger RNA (mRNA) and microRNA (miRNA) profile of cultured TMKs with those of normal human epidermal keratinocytes (NHEKs). This will provide a basis for understanding the homeostasis and healing process of the TM. The constitutive migration of the TMK is also thought to contribute to the development of cholesteatoma. This migration can either lead keratin the wrong way in a deep TM retraction pocket or it might be disrupted causing keratin accumulation. A characterisation of the TMKs might therefore also offer further insight into the pathophysiology of cholesteatoma development.

Materials and Methods

Ethics Statement

The study was performed according to the Declaration of Helsinki and was approved by the South-Eastern Norway Regional Committee for Medical Research Ethics (2010/1345). The use of cells derived from human TMs was approved by the Uppsala University Ethics Committee (Dnr. Ups. 99398). Material otherwise discarded and destroyed may be used for research purposes without written consent if data has been anonymised according to Swedish law.

Biological Material

Intact healthy human TMs were harvested from adult patients (n = 4) undergoing a translabyrinthine approach for surgical removal of vestibular schwannomas at Uppsala Academic Hospital, Sweden. NHEKs were purchased from ScienCell Research Laboratories, Carlsbad, CA, USA (HEK-a, Human Epidermal Keratinocytes-adult, cat. nr. 2110), originating from healthy adults undergoing reduction mammoplasty.

Cell Culture Media and Reagents

Keratinocyte medium (KM, cat. no. 2101) (ScienCell Research Laboratories, Carlsbad, CA, USA). Nunclon Δ-surface, gamma-irradiated culture flasks and multidishes (Thermo Fisher Scientific, Waltham, MA, USA). EDTA/trypsin (cat. no. T4049), HBSS (cat. no. 6648), DPBS (cat. no. D8537) (Sigma-Aldrich, St. Louis, MO, USA). QIAzol Lysis Reagent (Cat.no. 79306) and miRNeasy RNA purification kit (Cat.no. 217004) (Qiagen, Venlo, Netherlands).

Histology of the Human Tympanic Membrane

One TM was fixed in 4 % paraformaldehyde immediately after surgical removal. It was then dehydrated in increasing ethanol concentrations and embedded in paraffin. Five-mircometer sections were cut, the sections deparaffinised and rehydrated before haematoxylin and eosin staining. They were photographed on a Zeiss Observer.Z1 inverted microscope (Fig. 1).

Fig. 1.

Normal histology of the human TM. The section shows a part of the pars tensa. The right aspect of the TM was located close to the umbo region, while the pars tensa gradually gets thinner towards the left, more peripherally. The black arrows show the epidermal layer with detached keratin without adnexa covering the lateral side of the TM. The white arrow shows the largest among several blood vessels in the sub epidermal connective tissue on the lateral side, close to the umbo. The asterisk is placed in the middle layer of the TM, the fibrous lamina propria, where multiple fibroblasts are interspersed in the collagen fibres. The black arrow heads point to the medially covering mucosa, and the white arrowhead identifies a vessel on the medial side. Image courtesy of Dr. J. Knutsson, Västerås, Sweden. H&E, Scale bar = 100 μm

Establishing Primary Cell Culture of Tympanic Membrane Keratinocytes

The excised TMs were rinsed three times in DPBS. Traces of blood were removed, and the handle of malleus, the medial layer of the TM and traces of ear canal skin were mechanically trimmed away. The TM was then cut into small pieces (1 × 1 mm) and transferred to a 50-ml sterile tube containing 0.25 % trypsin EDTA solution and incubated for 30–40 min. The tube was turned every 5 min to allow cell detachment. Next, the supernatant containing dissolved cells was transferred to a tube containing keratinocyte medium (KM). Fresh trypsin EDTA solution was added to the tube containing the TM biopsies, and the cell harvesting procedure was repeated once more. The cell suspension was centrifuged at 100×g for 5 min, and the TM cells were re-suspended in KM and seeded at 10,000 cells/cm² in T-25 culture flasks. The cells were cultured in a humidified incubator at 37 °C with 5 % CO₂ until approximately 80 % confluent, and then cryopreserved as primary cultures (P0) in KM with the addition of 10 % DMSO.

After cryopreservation, the TMK samples (n = 3) and NHEK samples (n = 3) were seeded at a concentration of 10,000 cells/cm² (in total, 35,000 cells) in 12-well multidish plates. KM was again used for both cell types and the media was changed every 2 days. Upon reaching 90–100 % confluence, the cells were washed three times in HBSS followed by addition of 700 μl of QIAzol lysis buffer. Cell lysates were stored in − 80 °C until isolation of total RNA.

Cell Morphology by Phase Contrast Microscopy

For comparison of cell culture morphology, phase contrast micrographs were captured using a Leica DMIL LED microscope (Wetzlar, Germany) with a Canon EOS 5D Mark II camera (Tokyo, Japan) (Fig. 2).

Fig. 2.

Morphology of tympanic membrane keratinocytes (TMK) and normal human epidermal keratinocytes (NHEK) in culture. Phase contrast micrographs of a low cell density and confluent cell area of cultivated cells show typical epithelial morphology for both cell types. a, b Tympanic membrane keratinocytes. c, d Normal human epidermal keratinocytes. Original micrographs were converted to grey-scale and brightness was adjusted for presentation. Scale bar = 50 μm

Total RNA Isolation

RNA purification was performed using the miRNeasy purification kit according to the manufacturer’s protocol (QIAzol lysis buffer). The total RNA was purified from the aqueous phase using RNeasy Mini spin columns (Qiagen). RNA content was verified and levels quantified using NanoDrop ND1000 (Wilmington, DE, USA).

Gene Expression and microRNA Profiling

The Affymetrix gene expression analysis protocol was used for whole-genome expression analysis (Affymetrix, Santa Clara, CA, USA). One hundred fifty nanograms of total RNA was subjected to GeneChip HT One-Cycle complementary DNA (cDNA) Synthesis Kit and GeneChip HT IVT Labelling Kit, following the manufacturer’s protocol. Labelled and fragmented single-stranded cDNAs were hybridised to the GeneChip Human Gene 2.0 ST Array. The arrays were washed and stained using the FS-450 Fluidics Station (Affymetrix). For miRNA expression profiling, 300 ng of total RNA was used for biotin labelling of miRNA by the Genisphere FlashTag HSR Kit, following the manufacturer’s protocol (Genisphere, Hatfield, PA, USA). Labelled miRNAs were hybridised to the GeneChip miRNA 3.0 Array (Affymetrix). Signal intensities for both analyses were detected by Hewlett Packard Gene Array Scanner 3000 7G (Hewlett Packard, Palo Alto, CA, USA). Individual microarray chip data have been deposited in the Gene Expression Omnibus database (GSE76674).

Real-Time qRT-PCR

Results from the microarrays were validated by real-time qRT-PCR. Two hundred nanograms of total RNA from each sample was reverse transcribed using qScript cDNA SuperMix (Quantabio, Beverly, MA, USA) in a 20-μl volume. The assays were performed in duplicates using the Viia7 (Applied Biosystems, Thermo Fisher Scientific, Waltham, MA, USA). The following TaqMan gene expression assays (Applied Biosystems) were used: FOXC2 (Hs00270951_s1), PLAT (Hs00263492_m1), VCAN (Hs00171642_m1), KRT19 (Hs00761767_s1), SEMA3A (Hs00173810_m1), THBS2 (Hs01568063_m1), CXCL14 (Hs01557413_m1), HOXA1 (Hs00939046_m1), HOXA7 (Hs00600844_m1), HOXA9 (Hs00266821_m1), BMP4 (Hs03676628_s1), TGFB1 (Hs00998133_m1) and HEPHL1 (Hs01376180_m1) and the housekeeping genes ARIH1 (Hs00194934_m1) and UBA52 (Hs03004332_g1). For miRNA validation, the following TaqMan MicroRNA assays were used: MIR196a2 (241070_mat), MIR424 (000604), MIR503 (001048), MIR30a (000417) and MIR34a (000426) and the housekeeping miRNAs MIR205 (000509) and MIR24 (002441). The housekeeping genes for both mRNA and miRNA were selected on the basis of the least variance (CV %) between TMK and NHEK cells from the microarray experiment. In addition, the expression values for the housekeeping genes should be at least five times the mean value of all probes on the microarray chip. Forty cycles of PCR reactions were run using 1 μl of primers described above, 9 μl of cDNA diluted 1:10 and 10 μl TaqMan universal PCR mix (Applied Biosystems) in a 20-μl volume. Cycling conditions were 5 min at 25 °C, 30 min at 42 °C and 5 min at 85 °C.

Cell Motility Scratch Assay

Cells were seeded in 12-well culture plates at 15,000 cells/cm². When the cells reached 80 % confluence, media was replaced by KM basic media without adding keratinocyte growth supplement (KGS) that contains growth factors and pituitary extract. After 24 h, the media was aspirated and the cells were treated in the dark at 37 °C with the unsupplemented KM with the addition of 10 μg/ml Mitomycin C to block cell proliferation. After 1 h, the media was removed and the cells washed three times with unsupplemented media, 5 min each. The cells were scratched using a 200-μl micropipette tip, and pictures were taken at 0, 6, 12 and 18 h using a Leica DMIL LED microscope (Wetzlar, Germany) with a Canon EOS 5D Mark II camera (Tokyo, Japan). The area closure was calculated as follows:

% area closure = [(A _t = 0h − A _t = Δh)/A _t = 0h] × 100 %.

Data Analysis and Statistics

The scanned microarrays were processed using GeneChip Command Console® Software (AGCC) (Affymetrix), and the CEL files were imported into Partek Genomics Suite software (Partek, Inc. MO, USA). Robust multiarray analysis (RMA) was applied for normalisation. GeneChip IDs with an average expression level less than 4 for both cell types according to the Affymetrix algorithm were filtered out before further analysis. Based on the microarrays, genes were identified for differential expression between the TMKs and NHEKs using one-way ANOVA. In addition to the criterion of an unadjusted p value of < 0.05, genes that were expressed more or less than 1.5-fold relative to NHEK control cells were considered differentially expressed. Further bioinformatics analysis of the selected genes was conducted using two different, continuously updated, enrichment tools: g:Profiler (http://biit.cs.ut.ee/gprofiler/welcome.cgi) and Ingenuity Pathway Analysis (IPA) (Ingenuity Systems, Redwood City, CA, USA) (http://www.ingenuity.com/science/knowledge-base).

g:Profiler uses information from the Gene Ontology Consortium to identify biological processes and interpretation of gene lists generated in high throughput analyses like gene expression. g:Profiler uses a manufactured statistical method called g:SCS as default for multiple testing correction of p values for gene enrichment analyses. This method is more conservative compared to Benjamini-Hochberg correction and also takes into account the fact that the gene ontology annotations are not independent (Reimand et al. 2007).

Ingenuity is a licence-based analysis platform. Raw p values of the IPA analyses are calculated using the right-tailed Fisher’s exact test and corrections for multiple testing by Benjamini-Hochberg.

To calculate the p value for the real-time qRT-PCR analyses and the closure rates of the scratch assay, we used two-sided, unpaired t tests. Pearson’s correlation was used to compare the measured fold changes between microarray and qRT-PCR analyses.

Results

Tympanic Membrane Keratinocyte Isolation and Culturing Using a Standard Trypsin-Based Procedure

Figure 1 illustrates the thin epidermal keratinocyte layer (~ 10–30 μm) of the normal human TM. After surgical removal of the tympanic membrane, the TMKs were successfully cultured from harvested tissue using a trypsinization method. The cell cultures did not appear to differ morphologically between individual TMK donors (data not shown). The TMKs could be cultured at least until passage 8 with minor morphological changes compared to low passage cells (higher proportion of large, spread cells; data not shown). The TMKs and NHEKs both display typical epithelial morphology in the confluent cell cultures. However, at low concentrations, the TMKs appeared more stretched. Representative micrographs are presented in Fig. 2. Vimentin mRNA levels, commonly used as a marker of mesenchymal cells, are identical.

Differentially Expressed mRNA in the Tympanic Membrane Keratinocytes

Genome-wide gene expression microarray analysis identified 500 genes that were differentially expressed in TMK relative to NHEK. Of these, 251 demonstrated a high expression and 249 a low expression. The 30 genes of highest and lowest expression in TMK are listed in Table 1. A complete list of the genes including Affymetrix microarray chip IDs can be found in the supplementary material (Table S1). Selected genes for real-time qRT-PCR validation are presented in Table 2. The genes were chosen partly for showing markedly different expression between the cell types and for a potential biological importance. There was generally good correlation between the microarray and qRT-PCR in fold change calculations (Fig. 3). All genes selected for real-time qRT-PCR validation showed the same direction of expression as found in the microarrays, and differences were statistically significant. When the genes were undetected in either cell type, it was not possible to calculate a difference in expression. For forkhead box C2 (FOXC2), the threshold of qRT-PCR detection in NHEK was above 38 cycles which indicates that FOXC2 is unlikely to be present. Similarly, HOXA1 is unlikely to be present in TMK. The calculations of ΔΔCt (log2-fold change) will therefore be biased.

Table 1.

Most differentially expressed genes of tympanic membrane keratinocytes relative to normal human epidermal keratinocytes^a

Gene symbol	Entrez gene name	Affymetrix ID	Fold change	p valueb	TMK expressionc	NHEK expressionc
FOXC2	Forkhead box C2	16,821,660	18.05	0.0000	541	30
PLAT	Plasminogen activator, tissue	17,076,726	14.85	0.0006	1598	108
TMPRSS11D	Transmembrane protease, serine 11D	16,976,438	14.22	0.0080	269	19
KRT19	Keratin 19	16,844,775	13.71	0.0003	565	41
VCAN	Versican	16,986,913	11.38	0.0127	532	47
SEMA3A	Semaphorin 3A	17,059,323	11.27	0.0011	520	46
FAM27E3	Family with sequence similarity 27, member E3	17,118,415	10.56	0.0001	185	18
CEACAM6	Carcinoembryonic antigen-related cell adhesion molecule 6	16,862,563	9.59	0.0059	203	21
STC1	Stanniocalcin 1	17,075,553	8.59	0.0031	86	10
GDA	Guanine deaminase	17,085,829	8.20	0.0013	224	27
SLC6A14	Solute carrier family 6, member 14	17,106,398	7.35	0.0193	294	40
TMEM156	Transmembrane protein 156	16,975,234	7.07	0.0274	179	25
GALNT5	Galnac-T5	16,886,717	7.04	0.0025	790	112
EPHB2	Ephrin receptor B2	16,660,596	6.74	0.0051	219	33
EPB41L3	Erythrocyte membrane protein band 4.1-like 3	16,853,631	6.40	0.0036	222	35
IL1RL1	Interleukin 1 receptor-like 1	16,883,690	5.54	0.0004	88	16
TMEM204	Transmembrane protein 204	16,814,738	5.23	0.0106	344	66
KRT24	Keratin 24	16,844,419	4.79	0.0395	65	13
GPRC5A	G protein-coupled receptor, family C, group 5, member A	17,118,868	4.76	0.0024	507	107
GNG11	Guanine nucleotide binding protein, gamma 11	17,048,450	4.63	0.0009	317	68
MIG7	Migration Inducing gene-7	16,689,734	4.53	0.0067	282	62
PTGES	Prostaglandin E synthase	17,099,076	4.48	0.0232	585	131
NID2	Nidogen 2	16,792,954	4.48	0.0157	144	32
CENPV	Centromere protein V	16,841,768	4.26	0.0026	240	56
FYB	FYN binding protein	16,995,601	4.21	0.0218	49	12
UPK1B	Uroplakin 1B	16,944,325	4.14	0.0454	1859	449
IGFBP4	Insulin-like growth factor binding protein 4	16,834,091	3.99	0.0219	531	133
ENG	Endoglin	17,098,594	3.96	0.0047	386	97
SIX2	SIX homeobox 2	16,897,159	3.78	0.0001	297	79
MSLN	Mesothelin	16,814,366	3.76	0.0000	228	61
CNTN1	Contactin 1	16,750,154	− 4.45	0.0084	182	809
TP53I11	Tumour protein p53 inducible protein 11	16,737,543	− 4.50	0.0016	37	165
ITM2A	Integral membrane protein 2A	17,112,364	− 4.51	0.0006	26	116
DOCK11	Dedicator of cytokinesis 11	17,106,438	− 4.55	0.0083	25	112
HOXC9	Homeobox C9	16,751,952	− 4.59	0.0015	23	105
TMEM200A	Transmembrane protein 200A	17,012,546	− 4.64	0.0384	31	144
CDSN	Corneodesmosin	17,031,720	− 4.66	0.0005	41	192
COL8A1	Collagen, type VIII, alpha 1	16,943,241	− 4.73	0.0353	43	205
PLD5	Phospholipase D family, member 5	16,701,119	− 4.74	0.0073	54	256
HOXA2	Homeobox A2	17,056,105	− 5.31	0.0000	12	62
CLMP	CXADR-like membrane protein	16,745,563	− 5.33	0.0001	118	631
S1PR1	Sphingosine-1-phosphate receptor 1	16,667,760	− 5.43	0.0048	13	72
TLL1	Tolloid-like 1	16,972,198	− 5.54	0.0158	23	128
LOC400043	Uncharacterized LOC400043	16,751,973	− 5.82	0.0001	19	113
MOXD1	Monooxygenase, DBH-like 1	17,023,658	− 5.91	0.0343	12	68
HOXA4	Homeobox A4	17,056,125	− 6.01	0.0003	9	55
TBX3	T-box 3	16,770,685	− 6.27	0.0021	35	217
THBS2	Thrombospondin 2	17,025,844	− 6.30	0.0017	470	2961
SLC6A15	Solute carrier family 6 (neutral amino acid transporter), member 15	16,768,105	− 6.33	0.0093	57	360
BICC1	Bicaudal C homologue 1 (Drosophila)	16,705,089	− 6.35	0.0341	21	131
CTSK	Cathepsin K	16,692,846	− 6.36	0.0099	80	507
KRT34	Keratin 34	16,844,663	− 6.40	0.0036	21	135
SNORD84	Small nucleolar RNA, C/D box 84	17,036,580	− 6.64	0.0000	6	37
HOXA5	Homeobox A5	17,056,137	− 6.72	0.0001	21	142
HOXC10	Homeobox C10	16,751,917	− 6.82	0.0004	18	123
HOXA1	Homeobox A1	17,056,098	− 6.87	0.0009	40	272
HOXA7	Homeobox A7	17,056,152	− 8.00	0.0001	23	181
HEPHL1	Hephaestin-like 1	16,730,157	− 8.82	0.0032	58	510
CXCL14	Chemokine (C-X-C motif) ligand 14	17,000,168	− 9.45	0.0356	36	336
HOXA9	Homeobox A9	17,119,668	− 18.89	0.0000	16	309

^aThe fold change, p value and normalised expression values are derived from the Human Gene 2.0 ST arrays (gene expression microarrays)

^bT test (unpaired, two-tailed) limited to four decimal places

^cMean, normalised expression values using robust microarray analysis (RMA)

Table 2.

Real-time qRT-PCR expression levels of selected mRNAs and microRNAs

Gene name	ΔΔ Ct (log2-FC)	p valuea	Ct TMK mean	Ct NHEK mean
FOXC2	9.35	0.0000	29.09	38.30
PLAT	6.12	0.0005	25.31	31.29
KRT19	6.4	0.0002	23.16	29.42
VCAN	4.39	0.0128	26.67	30.92
SEMA3A	4.88	0.0008	27.67	32.41
TGFB1	1.26	0.0073	28.15	29.28
BMP4	0.49	0.6167	31.24	31.59
UBA52	0.05	0.5412	23.22	23.13
ARIH1b	− 0.05	0.7243	27.24	27.05
THBS2	− 3.61	0.0062	30.31	26.56
HOXA1	− 4.44	0.0000	38.85	34.28
HOXA7	–	–	Undetected	33.23
HEPHL1	− 2.75	0.0073	33.05	30.16
CXCL14	− 4.93	0.0105	33.68	28.61
HOXA9	–	–	Undetected	35.52
MIR30A	− 0.03	0.8675	27.06	27.14
MIR34A	0.52	0.2062	26.59	27.22
MIR205b	− 0.07	0.9577	21.89	21.93
MIR24b	0.07	0.7563	29.80	29.98
MIR503	− 3.26	0.0069	33.21	30.06
MIR424	− 2.78	0.0077	29.81	27.14
MIR196A2	− 12.15	0.0001	39.30	27.26

^aStudents t test (unpaired, two-tailed) limited to four decimal places

^bHousekeeping genes, ubiquitin A-52 residue ribosomal protein fusion product 1 (UBA52) and ariadne RBR E3 ubiquitin protein ligase 1 (ARIH1) were selected based on the least possible expression variation between cell types from the microarray analysis and ≥ 5× average expression levels. MicroRNA205 and 24 were selected on the basis of the least variation from microarray analysis. ΔΔCt = ΔCt_(NHEK) − ΔCt_(TMK) and ΔCt = Ct_(GOI) − Ct_{(AVG. HOUSEKEEPING GENES)}. Ct = cycle threshold, GOI = gene of interest

Fig. 3.

Correlation in fold change between microarray and qRT-PCR. Pearson’s correlations between the fold change measured by microarray and qRT-PCR for the genes selected for validation; see Table 2. Data are shown as log₂. Correlation coefficient R² = 0.90. p value < 0.0001

Ontology Enrichment Analyses of Highly Expressed Genes in Tympanic Membrane Keratinocytes Show Over-Representation of Genes Involved in Migration

Genes of high expression and genes of low expression in TMK were analysed separately. The lists of genes were uploaded using the g:Profiler tool for gene group functional profiling and in the IPA platform. The settings and results of the g:Profiler analysis are shown in Table 3. A complete list of enriched GO terms in hierarchical order is provided in the supplementary material (Table S2 for genes of high expression in the TMK, Table S3 for genes of low expression). The settings and results from the IPA are shown in Table 4.

Table 3.

g:Profiler analysis of over-represented GO terms and CORUM protein complexes of differentially expressed genes^a

Term type	GO term name	g:SCS corrected p value
	Genes of high expression in TMK
BP	Cell migration	1.98E-04
BP	Chondrocyte proliferation	0.0031
BP	Anatomical structure development	0.0060
BP	Multicellular organismal development	0.0139
BP	Cell surface receptor signalling pathway	0.0420
BP	Multicellular organismal process	0.0493
CC	Integral component of membrane	1.32E-06
CC	Integral component of plasma membrane	7.45E-05
CC	Extracellular space	2.08E-04
CO	MT1-MMP-claudin-1 complex	0.0060
	Genes of low expression in TMK
BP	Regionalization	4.91E-11
BP	Neuromuscular process	0.0145
CC	Integral component of membrane	3.14E-05
CC	Plasma membrane	0.0047
MF	Sequence-specific DNA binding	8.79E-04
MF	Collagen binding	0.0072

BP biological process, CC cellular component, MF molecular function, CO CORUM protein complex

^aAnalysis settings—associations retrieved 2015-11-19. organism: Homo Sapiens; query: Affymetrix Human Gene 2.0 ST array IDs of genes expressed in TMK higher and lower than fold change 1.5 relative to NHEK and p value < 0.05; options: significant only; ordered query; for the genes of high expression no electronic GO annotations are applied to strengthen the evidence (the evidence code: Inferred from Electronic Annotation, IEA, is excluded), Hierarchical sorting, Hierarchical filtering (best per parent—moderate filtering); all Affymetrix array IDs as statistical background; numerical IDs treated as AFFY_HUGENE_2_0_ST_V1; significance threshold: g:SCS threshold; Statistical domain size: Only annotated genes

Table 4.

Ingenuity Pathway Analysis downstream effects analysis of differentially expressed genes^a

^aAnalysis settings—ingenuity downstream analysis, associations retrieved 2015–11-13; reference set: Affymetrix Human Gene 2.0 ST Array; data source: all; confidence: all; species: mammal; tissues and cell lines: all; mutation: all; Affymetrix gene chip IDs chosen as Ingenuity IDs. In total, 448 individual genes were mapped by Ingenuity from the Affymetrix data after omission of duplicates

^bFunction annotation—a specific biological function or disease that is significant in the uploaded list of genes. We have presented the 15 most significant findings both for genes of high and low expression

^cAssigned genes—the number of genes associated with each functional annotation out of the 220 mapped individual genes of high expression and 228 mapped individual genes of low expression, respectively

^dZ-score—determines whether the function or disease annotation is either increased (positive values, red coloured) or decreased (negative values, blue coloured) in the dataset. A blank field indicates that for some functions or diseases, a z-score could not be calculated by Ingenuity. Genes may have been shown to influence on a function or process or disease, but it is not clear whether the function is increased or decreased. A value of zero indicates that the function is neither increased nor decreased in the dataset. We have added z-scores with colour codes for all diseases and function annotations listed for the genes of both high and low expression in TMK. The predicted biological functions of the genes of high expression in TMK, for example “migration of cells”, were increased as indicated by positive z-values/red colour. For the genes of low expression in TMK (high expression in NHEK), the same biological functions were expected to decrease as indicated by negative z-values/blue colour

Analyses in both platforms show a marked reduction of individually identified genes compared to the original microarray chip probe IDs considered (299 probe IDs of high expression and 328 probe IDs of low expression). Firstly, the number of genes is reduced due to the omission of duplicates, ambiguous IDs, mRNA from non-specified or potential genes, immature miRNA and other small non-coding RNA. Secondly, the analysis tools have different algorithms for recognition of genes. Across platforms, the ontological analyses show that many of the genes highly expressed in TMKs are important for cell migration. Among the 30 genes of highest expression that are listed in order of fold change in Table 1, the following contribute to this finding: FOXC2, PLAT, KRT19, VCAN, SEMA3A, CEACAM6, STC1, EPHB2, IL1RL1, PTGES, FYB, IGFBP4, and ENG. In addition to migration and locomotion, Ingenuity also identifies proliferation and cancer-related diseases and functions to be over-represented among the highly expressed genes. In particular, genes related to invasiveness and metastasis dominate Ingenuity’s cancer-related annotations. The GO term “Integral component of membrane” is enriched in genes of both high and low expression in TMK (Table 3). This means that many of the genes that separate TMK and NHEK in terms of relative expression code for proteins that are embedded in a cellular membrane.

Data from the CORUM database of protein complexes, supplied by g:Profiler (Table 3), show that the “MT1-matrix metalloproteinase -claudin-1 (MMP14-CLDN1)” complex (CORUM:2688) is significant (g:SCS corrected p value 0.0057). In the Ingenuity canonical pathways analysis, matrix metalloproteinase (MMP) regulation is also identified as “inhibition of matrix metalloproteinases” (Fisher’s exact p value 7.71E−4). The following four genes highly expressed in TMK in our data are related to this pathway: matrix metalloproteinase 14 (MMP14), TIMP metalloproteinase inhibitor 2 (TIMP2), tissue factor pathway inhibitor 2 (TFPI2) and low-density lipoprotein receptor-related protein 1 (LRP1).

Due to the finding of migration as the most important biological difference between the cultured TMK and NHEK, we followed up the results performing an in vitro scratch assay to see if there were any differences in the closure rates or patterns (Fig. 4). There was a slight trend towards faster closure rates for TMK but not significant. However, at 12 h and particularly 18 h, the gaps of the TMK samples were fully “populated” with individual cells and small cell groups with an archipelago-like appearance. The calculation of the precise closure rates at 18 h was therefore not possible (Fig. 4i).

Fig. 4.

Cell motility scratch assay. a–d Tympanic membrane keratinocytes scratch assay at 0, 6, 12 and 18 h respectively. e–h Normal human epidermal keratinocytes at 0, 6, 12 and 18 h respectively. i The percentage of scratch closure over time at 6 and 12 h. Data show the mean + s.e.m. No statistically significant differences were detected in the closure rate between the cells. (n = 9)

Striking Pattern of Homeobox A and C Cluster Genes Dominates the Genes of Low Expression in Tympanic Membrane Keratinocytes

Of particular interest among the genes of low expression is the large cluster of HOX genes of which several appear on the top 30 list. HOX genes are conserved transcription factors that regulate body development. Many genes from the HOXA cluster (with the exception of HOXA11 and 13), and HOXC cluster genes (except HOXC12), were expressed at significantly lower levels in the TMK cells. See the hierarchical cluster presentation of all expressed HOX genes (Fig. 5). HOXA9 accounts for the highest expression level among all HOX genes in the NHEK cells, and for the largest fold change (− 18.89). Of the HOXB cluster genes HOXB4, HOXB5 and HOXB6 showed low expression, and finally, none of the HOXD cluster genes were differentially expressed. No HOX genes were significantly more highly expressed in the TMKs compared to NHEKs. The low expression of the HOX genes strongly influences the ontological analyses of both g:Profiler and Ingenuity’s IPA presented here. The related functions “regionalisation” and “patterning of rostrocaudal axis” were found to be highly significant.

Fig. 5.

Hierarchical cluster heat map of HOX gene expression. The heat map shows the expression of all identified HOX genes in log2 transformed RMA normalised expression values. HOX genes that show significantly lower expression in TMK compared to NHEK (defined by uncorrected t test p ≤ 0.05 and fold change ≥ 1.5) are marked with asterisks (*)

Differentially Expressed microRNAs

In the miRNA array analysis, 10 miRNAs were found differentially expressed. These are listed in Table 5. The real-time qRT-PCR validation confirmed several miRNAs of low expression in TMK relative to NHEK, but the miRNAs that showed higher expression in TMK than NHEK in the microarrays (miRNA30a and miRNA 34a) were not confirmed and therefore excluded from further analysis (Table 2). For RNAs with low absolute expression levels, microarrays are less accurate than qRT-PCR. The combination of low absolute expression values, borderline significance values and a relatively low fold change may explain why some of the miRNAs of the array analysis were not confirmed by qRT-PCR. Due to the low number of differentially expressed miRNAs, the results of the miRNAs as a group may have less impact on the biological interpretation. Nonetheless, there was a trend for the following biological functions to be enriched in IPA analysis of the miRNAs: vasculogenesis (B-H corrected p value 0.0028), proliferation of cells (p value 0.0212), proliferation of tumour cell lines (p value 0.0189) and migration of cells (p value 0.0331). These results point in the same direction as the results from the gene expression/mRNA analysis.

Table 5.

Differentially expressed microRNA of tympanic membrane keratinocytes relative to normal human epidermal keratinocytes^a

Transcript ID	Probeset ID (Affymetrix)	Fold change	p valueb	TMK expressionc	NHEK expressionc
hsa-mir-3185	hsa-miR-3185_st	2.99	0.0401	125	42
hsa-mir-30a	hsa-miR-30a_st	2.36	0.0136	77	32
hsa-mir-34a	hsa-miR-34a_st	2.34	0.0388	148	63
hsa-mir-22	hsa-miR-22_st	1.71	0.0113	1223	713
hsa-mir-27b	hsa-miR-27b_st	1.60	0.0467	356	222
hsa-mir-92a-1//hsa-mir-92a-2	hsa-miR-92a_st	− 1.50	0.0111	1641	2461
hsa-mir-708	hsa-miR-708_st	− 1.85	0.0072	316	584
hsa-mir-503	hsa-miR-503_st	− 5.49	0.0154	56	307
hsa-mir-424	hsa-miR-424-star_st	− 10.30	0.0199	27	279
hsa-mir-196a-1//hsa-mir-196a-2	hsa-miR-196a_st	− 17.34	0.0024	2	38

^aThe fold change, p value and normalised expression values are derived from the GeneChip miRNA 3.0 Arrays (microRNA microarrays)

^bT test (unpaired, two-tailed) limited to four decimal places

^cMean, RMA normalised expression values

Of the individual miRNAs, miRNA196a and the clustered miRNAs 424/503 show a low expression in TMK. The miRNA196 family is located within the HOX clusters (Mansfield et al. 2004). MiRNA196a-1 is located adjacent to HOXB cluster at chromosome 17, and microRNA196a-2 is located within the HOXC cluster at chromosome 12. Their mature, active sequences are identical and are referred to as miRNA196a (Braig et al. 2010).

Discussion

This study presents the first direct comparison of the genome-wide gene expression and miRNA profiles in cultured keratinocytes of the TM and normal human skin epidermis. By analysing cultured cells, we were able to focus on the differences mainly in the keratinocytes, although some cell types other than keratinocytes will inevitably be present. When analysing intact tissue, the presence of many different cell types is known to create difficulties in analysis of transcriptional profiling (Mimoso et al. 2013). On the other hand, when culturing keratinocytes, some differences occur that are not present in the intact tissue. For example, genes that code for epidermal differentiation are normally not present in monolayer cultured cells. Isolated cells will not be allowed to interact with the niches cellular and non-cellular elements, which inevitably will change the behaviour of the keratinocytes. We do not know to what degree these aspects differ between the TMK and NHEK, and some caution should therefore be taken when interpreting the results in the context of in vivo homeostasis. Cultured cells are also activated and under selection for proliferation. Different keratinocytes in such conditions may display differential migratory phenotypes due to their intrinsic differences in growth and proliferation. The genes responsible for proliferation and migration also have considerable functional overlap. When proliferation is blocked by Mitomycin C, we still observe difference in the migration between the cell types.

The different bioinformatics tools used in this study show uniform results regarding cell migration, bearing in mind the complicated data-mining environment for such analyses (Huang et al. 2009). We know from observation of the skin on the tympanic membrane and in the ear canal that the outer keratin layer migrates constantly in the lateral direction in the ear canal. Normal epidermal keratinocytes only show this ability in activated state during wound healing and regeneration. The genes and miRNA that were highly expressed in the TMK compared to ordinary epidermal keratinocytes showed a very strong tendency towards migration as a biological function. This is thus “as expected”, but we believe we have come closer to determining the network of genes involved in the constitutive migration.

Cells were cultured until confluence to allow the formation of adhesion between the cells. In both g:Profiler and Ingenuity analyses, the annotation categories “cell migration” and “migration of cells” respectively, includes genes that are expected to both increase and decrease migration. Gene ontology terms such as “negative regulation of cell migration and motility” score significantly. Many of the genes found relating to migration can both promote and reduce the migration depending on the cell type and subcellular location. This finding reflects the complexity of cellular migration in which several pathways and regulating genes are involved, including epidermal and fibroblast growth factors (EGF/FGF), transforming growth factor-beta 1 (TGFB1), interleukin-pathways and tumour necrosis factor (TNF) (Barrientos et al. 2008). In addition, MMPs are important for the interaction between the cell and ECM during cell migration (Baker et al. 2002; Pastar et al. 2014). The many participants of MMP regulation in the highly expressed genes of TMKs indicate their role in the TMKs’ ability to migrate. MMP14 is inhibited by TIMP2 and TFPI2 and activated by claudin-1 (CLDN1, highly expressed in TMK, Table S1) through direct binding.

The scratch assay showed that there was a similar closure rate between the cells. There was, however, a clear qualitative difference in the cells’ migration behaviour with regard to the higher proportion of individual cells and small clusters of TMK cells that filled the scratched gap over the timespan of the assay.

Genes Related to Migration

Several of the most highly expressed genes in the present study are important regulators of cellular movement. Forkhead box C2 (FOXC2) specifically promotes mesenchymal differentiation during epithelial-mesenchymal transition (EMT), a process involving migration in wound healing, developmental processes and invasive cancer. When transfected with FOXC2, epithelial cells express mesenchymal-associated markers such as vimentin and fibronectin, without suppressing e-cadherin (CDH1) (Mani et al. 2007). The cells in the referenced study also became highly migratory and invasive. FOXC2 also is found to induce a marked increase in MMP2, which is activated by MMP14, as discussed above. Conversely, silencing of FOXC2 in oesophageal cancer cells significantly reduces invasiveness and migration (Nishida et al. 2010). The role of FOXC2 in EMT may also explain why skin wound healing of FOXC2-deficient mice is significantly delayed (Xue et al. 2008). Cultured TMKs have previously been shown to co-express epithelial and mesenchymal markers, and this was interpreted as an indication of EMT (Redmond et al. 2010). The authors found that the TMK cells already displayed the morphological characteristics of mesenchymal cells at primary culture, as opposed to the HaCaT control cells. The TMKs expressed increased levels of vimentin. In the microarray data of the present study, we found no difference in fibronectin, vimentin or CDH1 mRNA levels despite the high levels of FOXC2. The similar morphology between the compared cells (Fig. 2) does not indicate that an alteration of the TMKs towards a mesenchymal phenotype has occurred. Subtle differences in growth conditions and media, e.g. growth supplements and calcium levels, make it difficult to compare the phenotype of the cells in vitro across different studies. Also, the effects of culturing the cells in vitro adds bias to the interpretation of relevance in vivo. In sum, however, the current findings point towards EMT as a relevant process that can help us understand the migration of the TMKs.

Tissue-Type Plasminogen Activator

Tissue-Type Plasminogen Activator (PLAT) in particular, and also urokinase plasminogen activator receptor (PLAUR), are highly expressed in the TMK. Plasminogen (PLG) is necessary for the healing of the TM. Li et al. showed that mice lacking PLG had completely arrested TM wound healing after perforation (Li et al. 2006). Thick layers of keratin and clots of fibrin accumulated in the EAC (Eriksson et al. 2006) indicating, in addition, a role for PLG in the constitutive migration of cells since it appeared that the self-cleaning ability attributed to migration was affected. After systemic supplementation of PLG in mice, or when PLG was injected locally, the TMs regained their regenerative potential and showed almost normal healing (Shen et al. 2014). One study comparing the TM and cholesteatoma matrix did not detect PLG, PLAT or PLAU, or their inhibitors, plasminogen activator inhibitor 1 (PAI1) and 2 (PAI2) in the intact TMs (Schonermark et al. 1999). In cholesteatomas however, PLG and PLAU were strongly expressed in the basal epithelial layer. The serine peptidases plasmin and MMPs, as well as PAI1, closely interact in a wound microenvironment during re-epithelialisation (Simone et al. 2014), and this axis appears to be particularly important for TM homeostasis and regeneration.

Transmembrane protease, serine 11D

Transmembrane protease, serine 11D (TMPRSS11D) is also found highly expressed in the TMKs in our data and has been shown to be a potent fibrinolysis enzyme (Yoshinaga et al. 1998) and a modulator of PLAUR function (Beaufort et al. 2007).

Versican

Versican (VCAN) is a proteoglycan of the ECM. It is present in the basal layer of the epidermis (Bode-Lesniewska et al. 1996; Barrientos et al. 2008) and expressed in primary keratinocyte cultures (Zimmermann et al. 1994). VCAN influences migration of different cell types, and this is associated with an anti-adhesive function (Yang et al. 1999; Perissinotto et al. 2000). Arterial smooth muscle cells in vitro form VCAN and hyaluronic acid (HA)-rich pericellular matrix during detachment to allow the cells to migrate (Evanko et al. 2001). VCAN was secreted from the migrating cells. Levels of both VCAN and HA are also elevated around migrating epithelial cells during wound healing (Oksala et al. 1995; Wang et al. 2000). The very high expression of VCAN in the TMKs suggests that the cells themselves are capable of inducing changes in the ECM environment to facilitate their migration.

Semaphorin 3A

Semaphorin 3A (SEMA3A) is known for its role in vasculogenesis and axonal guidance during development (Raper 2000; Oh and Gu 2013). Reduced levels of SEMA3A expression have also been linked to the upregulation of epidermal nerve fibre density and thereby to pruritus in atopic dermatitis (Tominaga et al. 2008), psoriasis vulgaris and prurigo nodularis (Takada et al. 2013). Free nerve endings are very sparse in the outer epithelial layer of the TM, in contrast to the mucosal surface (Lim 1968; Yeh and Kruger 1984; Colin and Kruger 1986). The neuro-repulsive effect of SEMA3A might thus be part of the regulation of sensitivity in the TM.

Keratin 19

Keratin 19 (KRT19) is considered a potential stem cell marker in the epidermis (Michel et al. 1996). Michel found KRT19 positive cells in basal layers of glabrous skin of the palm and sole, in contrast to hairy skin where KRT19 positive cells were present almost exclusively in the bulge region of the hair follicle. The keratin expression in the epidermis of the cartilaginous parts of the EAC is typical of the epidermis in general (Lee et al. 1991). In the medial EAC and TM, KRT19 is highly expressed in basal epidermal cells (Broekaert and Boedts 1993; Vennix et al. 1996; Knutsson et al. 2011).

Specific HOX Gene Profile of the Tympanic Membrane Epidermal Phenotype

In 1993, Chuong introduced the idea that specific combinations or patterns of HOX expression, designated the “HOX code of skin appendages”, influences the site-specific epidermal differentiation in terms of epidermal thickness, hair types, etc. (Chuong 1993). This theory has later been strengthened by the findings of many authors (Stelnicki et al. 1998; Rinn et al. 2008; Johansson and Headon 2014). HOXC13 mutant mice which completely lack all pelage hair types (Godwin and Capecchi 1998) are one example of HOX gene influence on skin. Recently, mutations in the HOXC13 gene have been identified in human pure hair and nail ectodermal dysplasia, a genetic disease of alopecia and nail dystrophy (Lin et al. 2012).

The redundant nature and functional overlap between the HOX genes has made functional analysis difficult (Di-Poï et al. 2010). Di-Poï et al. investigated the morphological defects of mice lacking the entire HOXA cluster in the mesoderm lineages. With an additional deletion of genes within the HOXD cluster, the mice showed a fragile, oedematous skin without stratification and no differentiated appendices. The resemblance to the epidermal layer of the TM is interesting given the very low or absent expression of the majority of HOXA genes and additional HOXC and HOXB genes.

An increasing amount of evidence links HOXA cluster genes to the regulation of invasiveness and migration in multiple cancers. HOXA9 is inversely connected to migration in lung cancer cells (Hwang et al. 2014). When transfected into lung cancer cells, HOXA9 markedly decreased cellular migration, whereas knockdown enhanced cell migration. High expression of HOXA4 is found in invasive forms of ovarian cancer, but upon silencing of HOXA4, the migration rate of the cells increased in the presence of EGF (Ota et al. 2009). This finding led the authors to consider HOXA4 as a tumour suppressor in ovarian cancer cells, rather than a pro-neoplastic factor. Knockdown experiments of HOXA5 also promote invasiveness in vitro, and high levels of HOXA5 expression in tumours drastically increased the survival of patients with non-small cell lung carcinoma (NSCLC) making HOXA5 a strong prognostic marker (Wang et al. 2015).

MicroRNA Expression

The biological effects and the expression of miRNA196a in cancer cells are closely correlated to those of its directly targeted HOX genes: HOXB7 (Braig et al. 2010), HOXC8 (Li et al. 2010; Mueller and Bosserhoff 2011), HOXA5 (Liu et al. 2012), HOXA7 and HOXB8 (Yekta et al. 2004). These HOX genes and additional genes of the HOXA cluster are also among the top predicted targets judged on base complementarity according to the in silico tools offered by microRNA.org (http://www.microrna.org/microrna/getTargets.do?matureName=hsa-miR-196a&organism=9606) and TargetScanHuman (http://www.targetscan.org/cgi-bin/targetscan/vert_61/targetscan.cgi?mirg=hsa-miR-196a). In breast cancer cells, elevated miRNA196a leads to reduced levels of HOXC8 and thereby increased migration (Li et al. 2010). On the other hand, Braig et al. found that reduced levels of miRNA196a ultimately led to an activation of BMP4 and thereby induced migration of melanoma cells (Braig et al. 2010). This shows that in different experimental settings and different cells, miRNA196a can either increase or decrease the migration of cells. It also tells us that there are multiple regulatory systems interwoven that influence this important biological function. Articles have described common regulatory mechanisms imposed on the entire HOX clusters, including the embedded miRNAs (Mansfield et al. 2004). Furthermore, independent regulation is found in head and neck squamous carcinoma cell lines and normal oral keratinocytes. The mRNA expression of known and putative microRNA196a targets, including HOXA7, HOXB8, HOXC8 and HOXD8, remained unchanged after overexpression of miRNA196a (Severino et al. 2013). Our data suggests that both HOX genes and microRNA196a are regulated by common factors since transcription of both is heavily restricted.

The miRNAs424/503 are parts of the same cluster and are often co-regulated. At elevated levels, the miRNA424/503 cluster has been shown to suppress tumour development by acting anti-proliferative and anti-invasive via thyroid hormone receptors (Ruiz-Llorente et al. 2014). Suppressed levels of miRNA503 are associated with metastatic cancer disease. Opposite, a markedly reduced migration is found in numerous cancer cells overexpressing miRNA503 in vitro including NSCLC (Li et al. 2014), hepatocellular carcinoma (HCC) (Li et al. 2015) and gastric cancer (Peng et al. 2014). There are also indications of a miRNA503 driven suppression of the EMT programme in these cells as indicated by a decrease of mesenchymal markers when miRNA503 was overexpressed. Li et al. found that miRNA503 expression is restored in NSCLC by the demethylation agent 5-aza-20-deoxycytidine, suggesting important epigenetic regulation of its transcription by methylation (Li et al. 2014). Moreover, inhibition of miRNA424 is found to accelerate cell proliferation, migration and invasion of HCC (Yu et al. 2014) and osteosarcoma cells (Long et al. 2013).

In summary, suppressed transcription of miRNA196a and miRNA424/503 makes them candidates for regulation of TMK migration.

Conclusion

Several genes identified in this study, including FOXC2, several HOX genes and miRNA196a and miRNA424/503, are intimately associated with migration in activated cells. Further exploration of these specific genes and their regulation will help define the unique features of the TM epithelium and focus the quest for effective and safe non-surgical treatment alternatives for TM defects.

Reviewing the importance of HOX genes in skin homeostasis and site specificity, it seems likely that the differences in cell phenotype and gross morphology between the TM and ordinary skin is maintained partly through the regulation of HOX gene expression.

Funding information

The study was funded by Akershus University Hospital, The South-Eastern Norway Regional Health Authority and the University of Oslo.

Compliance with Ethical Standards

The study was performed according to the Declaration of Helsinki and was approved by the South-Eastern Norway Regional Committee for Medical Research Ethics (2010/1345).

Conflict of Interest

The authors declare that they have no conflict of interest.