Auditory canal skin and cholesteatoma share comparable histological features and harbour epidermal stem cells and fibroblasts in their connective tissue
The two different tissues differed macroscopically mainly in size and texture. While the small sample of healthy skin derived from the tympanomeatal flap displays a compact appearance, the cholesteatoma tissue is characterized by its frayed surface exhibiting abundance of keratin debris.
Histological imaging by H&E technique revealed the presence of comparable morphological features in the two different tissue samples. The dermis of healthy skin and the perimatrix of the cholesteatoma tissue is composed of different cell types of connective tissue. The dermis is followed by an epidermis in auditory canal skin and the matrix in cholesteatoma tissue. This histological feature differs between the two samples in various details. The epidermis exhibits a thickness between 50 µm and 100 µm, is structured with rete pegs reaching into the dermis, is clearly delimited from the underlying dermis, and shows a complex layered inner structure, beginning with the dense basal layer and ending with the stratum corneum. The cholesteatoma matrix shows a greater thickness of 100 µm to 200 µm, is not lined with process-like structures, is diffusely demarcated from the underlying perimatrix, and also shows only a diffuse inner structure. Another obvious difference between healthy epidermis and hyperproliferative cholesteatoma matrix is the thickness of the outmost layer of cornfield keratinocytes and keratin debris, which is with 20 µm to 100 µm 2-10x thicker in cholesteatoma tissue than the 10 µm layer in the auditory canal skin.
Using two different cell culture techniques, we were where able to derive two different subpopulations of cells from each of these two tissues. One type was fibroblasts growing as adherent cells in medium supplemented with serum and the other one was stem cells of epidermal origin growing as spheres under serum-free conditions [46]. When imaged by phase contrast microscopy, both kind of fibroblasts, from cholesteatoma tissue and auditory canal skin, showed a typical stromal physiognomy marked by spindle shaped morphology and signs of focal adhesion. The visualization of the spheres derived from both kinds of tissue also displayed a similar appearance. Particularly, a tight packaging of cells in the inside of the sphere, marked by high light absorption and scattering, and a surface structured by tightly adhered cells.
The cells differ in expression of some inflammatory mediators, growth factors and the expression von TLR4
Targets responsible for inflammation and wound healing were investigated by RT-qPCR in all cell types under standard culture conditions (Fig. 2). The chosen targets were different cytokines, chemokines and growth factors known to be overexpressed in cholesteatoma tissue and/or related to cholesteatoma development and its pathological exaggerated inflammatory niche. The relative expression levels for one specific cell type were distributed between the different donors over 3 orders of magnitude (OM) in case of TNF-α or IL-8 or just slightly different e.g. in the case if IL-1α. Despite this huge biological variance, a significantly higher expression in ME-CSCs and ME-CFs derived from cholesteatoma tissue could be observed for the growth factors KGF and IGF-2. A similar trend could be assumed for the growth factor HGF and the cytokine IL-1α even though it did not reach the level of statistical significance (p = 0.07 in both cases). More pronounced was the tissue-independent trend to higher expression of the inflammatory mediators IL-1β and IL-8 in stem cells (around 250 fold and 700 fold, respectively) and the growth factor IGF-2 in fibroblasts (between 6 and 60 fold). Since the TLR4 is known to play a fundamental role in the inflammatory environment of the cholesteatoma, we furthermore investigated its expression level in the cells of three different donors upon cultivation in FB-medium, as this medium was used in the following LPS stimulation experiments presented below.
For cholesteatoma tissue, the ME-CSCs and ME-CFs showed a similar distribution of TLR4 expression between 1% and 0.1%relative to GAPDH. In regards to the tissue of origin the obtained data showed an opposed effect for fibroblasts and stem cells. Interestingly, the TLR4 expression tended to be lower in fibroblasts and higher in stem cells derived from cholesteatoma compared to the same cell type from auditory canal skin. In numbers, the reduced expression in ME-CFs compared to ACFs range from rather insignificant 87% down to highly significant 15%, while all ME-CSCs showed a significant upregulation between 55 and 2 fold compared to ACSCs.
To investigate the reaction towards a bacterial infection in the four different cell types, we stimulated the cells with LPS and measured the transcription of different inflammatory mediators and growth factors. The stimulation was executed with 100 ng/ml, whereas the control contained only 0.37 ng/ml. We consider this as an LPS-free control since even at 1 ng/ml no significant stimulation could be detected (Fig. S1)
We grouped these transcripts into three classes, according to their characteristic reaction in the cell types upon stimulation. The first group comprised targets from the interleukin family (Fig. 4A). A tissue specific difference with and without LPS stimulation was detected in stem cells while in fibroblasts no such effect could be observed. The cell type specific difference between fibroblasts and stem cells derived from cholesteatoma tissue was rather marginal, showing relations between 200% and 25% which reached statistical significance infrequently. An exception was IL-8 exhibiting a downregulation to about 3% in ME-CFs. Since ACSFs also expressed low levels of IL-8, the high IL-8 expression was specific for ME-CSCs. Another interesting trend could be observed: the expression of IL-6 was heavily elevated for ME-CSCs.. A pattern of fibroblast-specific upregulation was measured for IL-1α .Since ME-CSCs did not show comparable expression levels as ME-CFs upon stimulation with LPS, IL-1α seems to be a rather fibroblast-specific target.
The second group of targets includes different inflammatory mediators, which reacted with a higher sensitivity upon LPS stimulation in all cell types derived from cholesteatoma tissue (Fig. 4B). The expression levels of different markers in ACSCs in relation to ME-CSCs lays at 2.5% (TNF-α), 3.5% (CXCL-5) and 30% (GM-CSF). This tissue specific difference is also distinctive for ACSFs, for which the expression levels were detected at around 2.2% (TNF-α, GM-CSF) and 10% (CXCL-5) of those values measured for ME-CFs. In this group, also the expression with and without LPS stimulation was much higher in fibroblasts independent of the tissue of origin. In average, the expression levels in stem cells reached 2% to 20% (TNF-a), 4% to 5% (GM-CSF) and 5% to 14% (CXCL-5) of the levels detected in fibroblasts, making all these targets specific for fibroblasts. The last group comprises all growth factors investigated in this study (Fig. 4C). The growth factors are characterised by a huge upregulation in expression in ME-CF and also in ACF, even though to a much lesser extent. In detail, the expression was elevated for ME-CFs and ACFs compared to their corresponding stem cells 160 fold and 30 fold (KGF), 530 fold and 110 fold (EGF), 13 fold and 11 fold (EREG), 340 fold and 4 fold (HGF), and 860 fold and 75 fold (IGF-2), respectively. In this group, only a random tissue specific response to the LPS stimuli could be detected. This response was rather weak for EREG in stem cells (3.5 fold) and more pronounced in fibroblasts for IGF-2 (13 fold), EGF (23 fold), and especially HGF (450 fold). Interestingly, HGF is the only target which seems to be specific in a tissue and cell type specific manner for ME-CFs.
Since we detected an abnormal expression of inflammatory mediators and growth factors for cells derived from cholesteatoma tissue upon stimulation with LPS, we decided to measure the effect of LPS on the metabolic activity and proliferative behaviour of ME-CSCs and ME-CFs.
To investigate the biological effect of the increased production of inflammatory mediators and growth factors on the two different cell types derived from cholesteatoma tissue, we measured the metabolic activity upon long-term exposure of ME-CSCs and ME-CFs to LPS. For ME-CSCs we could detect an increase in metabolic activity for one of the investigated three donors after 6 days (Fig. 5A). From an exponential curve fit, a doubling time for the metabolic activity of 68.5 ± 3.2 days for ME-CSCs and 91.4 ± 6.3 for ME-CFs days was derived. Repetition of this experiment resulted in a statistical significance of this effect. For ME-CFs, even after only two days of cultivation a significant change in metabolic activity was observed.
To investigate the mechanisms underlying the increased metabolic activity, we executed proliferation assays using cells of the same donors as investigated by the MTT assay (Fig. 5B). The examined ME-CSCs exhibited only a slight and insignificantly increased mitotic activity even after 6 days of stimulation with LPS. The exponential fit of the growth data resulted in a similar doubling time of 32.1 ± 1.8 hours without LPS and 30.2 ± 1.6 with stimulation by LPS. In contrast to that, the stimulation of ME-CFs with LPS lead to an significant increase in proliferation, with doubling times of 28.3 ± 0.9 hours and only 23.4 ± 1.4 hours without stimulation, detectable even 4 days after the addition of LPS into the medium.
To rescue the LPS-treated phenotype of ME-CSCs owning an enhanced proliferation, we repeated the proliferation assay with ME-CFs derived from three different donors with the application of the TLR4 antagonist LPS-RS, which was added into the LPS-supplemented medium (Fig. 5C). Again a significant increase in proliferation of ME-CFs was detected upon treatment with LPS. When executing the same experiment with ACFs derived from the same patients no such LPS-dependent stimulation of proliferation could be detected (Fig. S2).
By comparing the derived doubling times, we were able to show that LPS-RS is able to reduce the proliferation of ME-CFs cultured with LPS. To be more specific, the addition of LPS-RS into the medium returned the mitotic activity to the one observed at standard culture conditions. The downregulation of transcripts responsible for fibroblast proliferation upon blockage of the TLR4 receptor by LPS-RS (Fig. S3) even below the control level, illustrates one of the sources for this rescue effect.
To examine if the growth factors expressed by the ME-CFs exert a differentiating effect on the ME-CSCs we designed an indirect cell culture insert-based co-culture model of cholesteatoma progression and self-renewal. In this model, the fibroblasts and stem cells were co-cultivated in medium containing LPS. As controls served the same setup without LPS supplementation, ME-CSCs cultivated without the presence of ME-CFs (with or without LPS), and ME-CSCs cultivated under standard cell culture conditions. RT-qPCR analysis showed a remarkable and significant upregulation of cytokeratins in ME-CSCs after 14 days of co-culture stimulated with LPS (Fig. 6A). The expression of cytokeratin 14 was upregulated 15-fold compared to ME-CSCs co-cultured without LPS and 30-fold relative to culture conditions without LPS and co-cultivation. For cytokeratin 16, cytokeratin 18 and cytokeratin 19 the corresponding fold changes were 25-fold and 210-fold, 9-fold and 45-fold, and 12 fold and 150 fold, respectively. In the co-culture with LPS the relative expression level was highest for cytokeratin 16 and cytokeratin 18 with relative expression compared to the house keeping gene of 1.6‰ and 4‰, respectively, and lower for cytokeratin 14 and cytokeratin 19 with both showing a relative expression of only 0.3‰. The expression of Ki-67 was significantly reduced in all samples cultivated over 14 days from 3 to 10 fold (Fig. 6B). Intriguingly, the expression in the co-culture system comprising LPS was shown to be elevated compared to the other ME-CSCs samples by a factor of 3. In ME-CSCs co-cultured with LPS treated ME-CFs, this decrease of proliferation was less pronounced. The subsequent immunofluorescence for these cytokeratins revealed, that cytokeratins 16 and 19 are heavily upregulated in ME-CSCs co-cultivated with LPS-stimulated fibroblasts (Fig. 6C). Even though cytokeratin 19 was also irregularly expressed in ME-CSCs cultivated with non-stimulated ME-CFs, but this was a rare observation. Cytokeratin 18 was also homogenously induced in the control cells on protein level, but to a smaller extent.