Reconstructed human keloid models show heterogeneity within keloid scars

After removal of subcutaneous fat and any other soft tissue until the typical firm and rubbery keloid consistency was reached, keloids were further subdivided into peripheral (P-Kscar), central superficial (Cs-Kscar) and central deep (Cd-Kscar) regions (Fig. 1a–c). Dermal tissue until ± 0.5 cm depth was included for cell isolation, Cs-Kscar and Cd-Kscar samples were obtained from the upper and lower central half, respectively. If present, any extralesional normal skin directly adjacent to the keloid (sNskin) was also included. sNskin extended to approximately ± 0.5 cm beyond P-Kscar (see area between edge of keloid and dotted line in Fig. 1a, b). In contrast to sNskin which was always derived from keloid patients, normal skin (Nskin) was obtained from unaffected control subjects. Keratinocytes and fibroblasts were isolated and cultured essentially as described previously [32, 34]. Skin models were constructed in duplicate from keratinocytes (P2) and fibroblasts (P2-3) essentially as previously described [32] ²¹ (Fig. 1d). In brief, 4 × 10⁵ Fibroblasts were seeded onto 2.2 × 2.2 cm squares of MatriDerm® (dr. Suwelack Skin & Health Care, Billerbeck, Germany) with FSM-I and cultured submerged in FSM-II for 3 weeks in 0.4 µm pore size transwells (Costar Corning Inc., New York, NY, USA) in a 37 °C, 5% CO₂ atmosphere. Keratinocytes (P2) were then seeded on top of the fibroblast-populated MatriDerm® and cultured submerged in KC-I for 3–4 days, prior to culturing at an air–liquid interface in deep-well plates (BD Biosciences, Bedford, MA, USA) in KC-II for an additional 10 days. Upon addition of keratinocytes, SE were cultured in a 37 °C, 7.5% CO₂ atmosphere. Medium was changed twice a week. See supplementary table 1 for contents of culture media (KC-I, KC-II, FSM-I, and FSM-II) used.

Wound contraction is expressed as a reduction in surface area of the skin models at the end of the culture period. Surface area (mm²) was determined in photographs of skin models at the time of harvesting (5 weeks culture), using NIS-Elements AR 2.10 imaging software.

Paraffin tissue sections (5 µm) were stained with Haematoxylin & Eosin (HE) for histologic evaluation and determination of epidermal and dermal thickness. Epidermal thickness quantified by counting the number of keratinocyte cell layers at three random points in each skin model section (200× magnification). Dermal thickness measured using NIS-elements software to calculate length in µm at five random points per skin model section (100× magnification).

Immunohistochemical stains were performed on deparaffinized, formalin-fixed tissue sections to assess epidermal proliferation (Ki67: clone MIB-1, Dakocytomation, Glostrup, Denmark; 1:50), epidermal differentiation (K10: keratin 10, clone DE-K10, Progen, Heidelberg, Germany; 1:500 and involucrin: clone SY5, Novocastra, New Castle, United Kingdom; 1:1000), presence of fibroblasts (vimentin: clone V9, Dakocytomation) and myofibroblasts (α-SMA: clone 1A4, Dakocytomation). Supplementary antigen retrieval treatments were performed prior to incubation with the primary antibody using a 15 min. incubation step with pepsin (K10) and/or heat-induced antigen retrieval with 0.01M citrate buffer pH 6.0 (Ki67, K10, K17, vimentin). Immunohistochemical staining scoring (−) absence of staining; (+) normal staining pattern; (++) increased number of positively stained cells; (+++) strongly increased number of positively stained cells. Ki67 Proliferation index 100 basal cells were counted in three random locations in a tissue section (100× magnification), after which the number of positive cells along this length of the epidermis was counted. The proliferation index was defined as the percentage of Ki67 positive nuclei within these regions.

Previously, we have identified a panel of wound healing mediators that are secreted predominantly by the epidermis (IL-1α, TNF-α, CCL5, VEGF), the dermis (TIMP2, HGF), or those significantly increased in the full thickness skin equivalents (CCL2, CXCL1, CXCL8, IL-6, sST2) [25]. CCL27 is found in burn wound exudates and has been implicated in the increased secretion of many of the aforementioned proteins [31]. IL-18 was also included in this panel because it has previously been implicated in keloid formation [4] and is known to be expressed in reconstructed human skin models also [11].

Culture supernatants (1.5 ml KC-II without hydrocortisone) derived from the skin models were collected over a 24-h period at the end of the culture period (5 weeks) to measure the levels of IL-6 and CXCL8 (PeliKine Sanguin Reagents, Amsterdam, The Netherlands); CCL2, CCL5, CCL20, CCL27, CXCL1, HGF and VEGF (R&D System Inc., Minneapolis, MN, USA); and IL-18 (MBL International, Woburn, MA, USA) secreted by the SE, using enzyme-linked immunosorbent assays (ELISA).

For RNA isolation, the epidermis was removed from the dermis using a slide-warmer (40 °C), the dermis was then flash frozen and stored in liquid nitrogen until further processing. Samples were disrupted and homogenized in a TissueLyser, then flash frozen for storage at − 80 °C. RNA isolation was performed using QiaShredder™ kits and RNeasy® Mini kits with on-column DNAse digestion and stored at − 80 °C. The Nanodrop spectrophotometer (Nanodrop technologies, Wilmington, DE, USA) was used to measure total RNA concentration. cDNA was synthesized using the RT² First Strand Kit, while the RT² SYBR Green Fluor qPCR Mastermix was used to run the real-time PCR reactions for the following genes (Table 2): COL4A2, HAS1 and MMP3. These three genes were selected because they showed differential expression between Nskin and Kscar in previous work [16]. The geometric mean of two housekeeping genes (ACTB and HPRT1) was used to normalize expression. Unless stated otherwise, all RNA and qPCR reagents were obtained from Qiagen GmbH (Hilden, Germany).

A colorimetric (MTT based) assay was used to quantify cell proliferation and viability (Roche Applied Science, Penzberg, Germany) of the skin models, as described previously [34].

Experiments were performed in duplicate with n ≥ 4 different donors, except for some of the immunohistochemical staining (one of duplicate skin models was stained: keratin 10, vimentin, α-SMA) and qPCR experiments; see Tables 1 and 3 for an overview of donor matching and the number of biological replicates used per experiment. All results in graphs and tables were expressed as the mean ± standard error of the mean, PCR scatter plots showed the median. Normality testing (Shapiro–Wilk test) was performed on the residuals (errors); an ordinary one-way ANOVA with Tukey’s multiple comparisons tests was employed if the residuals passed the normality test (epidermal thickness; dermal thickness; contraction; MTT; ECM gene expression; secretion of VEGF, CCL5, CXCL8), otherwise the Kruskall-Wallis test with Dunn’s multiple comparisons tests (Ki67; α-SMA; secretion of CCL2, CCL20, IL-18, CXCL1, IL-6, HGF, CCL27) was applied. For analysis of the PCR data, gene expression (2^−∆Ct) was normalized with the geometric mean of two housekeeping genes (ACTB and HRPT1). Differences were considered significant if p < 0.05 (*), p < 0.01(**) or p < 0.001 (***). It should also be noted that due to our use of stringent statistical analysis for these small sample sizes, our results very likely show underestimation of true statistical significance. A correction for performing multiple comparisons was made using an ANOVA. Additionally, the use of non-parametric tests in the event of non-normal distribution of the residuals, although statistically correct, further reduced power. For this reason, the term ‘trend’ was used when a clear pattern in graph data was observed without significance being reached; the exact p value was listed in the graph if 0.05 < p < 0.08. GraphPad Prism 6 software (GraphPad Software Inc., San Diego, CA, USA) was used to construct all graphs and tables and perform statistical analysis.