Design and analysis of a squamous cell carcinoma in vitro model system

Biomaterials 34 (2013) 7401e7407

Contents lists available at SciVerse ScienceDirect

Biomaterials journal homepage: www.elsevier.com/locate/biomaterials

Design and analysis of a squamous cell carcinoma in vitro model systemq Eva Brauchle a, b, c,1, Hannah Johannsen a,1, Samantha Nolan a, Sibylle Thude a, Katja Schenke-Layland a, b, * a Fraunhofer Institute for Interfacial Engineering and Biotechnology (IGB), Department of Cell and Tissue Engineering, Nobelstr. 12, 70569 Stuttgart, Germany b University Women’s Hospital Tübingen and Interuniversity Centre for Medical Technology (IZST), Eberhard Karls University Tübingen, Silcherstr. 7/1, 72076 Tübingen, Germany c University of Stuttgart, Institute for Interfacial Engineering and Plasma Technology (IGVP), Nobelstr. 12, 70569 Stuttgart, Germany

a r t i c l e i n f o

a b s t r a c t

Article history: Received 5 May 2013 Accepted 11 June 2013 Available online 1 July 2013

Tissue-engineered skin equivalents based on primary isolated fibroblasts and keratinocytes have been shown to be useful tools for functional in vitro tests, including toxicological screenings and drug development. In this study, a commercially available squamous cell carcinoma (SCC) cell line SCC-25 was introduced into epidermal and full-thickness skin equivalents to generate human-based disease-in-a-dish model systems. Interestingly, when cultured either in the epidermis or dermis of full-thickness skin equivalents, SCC-25 cells formed hyper-keratinized tumor cell nests, a phenomenon that is frequently seen in the skin of patients afflicted with SCC. Raman spectroscopy was employed for the label-free cell phenotype characterization within the engineered skin equivalents and revealed the presence of differential protein patterns in keratinocytes and SCC-25 cells. To conclude, the here presented SSC diseasein-a-dish approaches offer the unique opportunity to model SSC in human skin in vitro, which will allow further insight into SSC disease progression, and the development of therapeutic strategies. Ó 2013 The Authors. Published by Elsevier Ltd. All rights reserved.

Keywords: Co-culture Epithelium In vitro test Organ culture Keratinocyte

1. Introduction Among basal cell carcinoma, squamous cell carcinoma (SCC) is the most common type of skin cancer. Every year, an estimated 400,000e600,000 new cases of cutaneous SCC are diagnosed worldwide, predominantly occurring in sun-exposed skin areas of elderly patients [1]. Although SCC is curable in most cases, treatment at early disease stages is advisable since SCC has the potential to spread and form metastases [1,2]. The development and expansion of SCC is considered to be an accumulative process of cellular malignant alterations within the epidermal skin layer [3]. SCC can arise from so-called precancerous lesions, the most frequent being

q This is an open-access article distributed under the terms of the Creative Commons Attribution-NonCommercial-ShareAlike License, which permits noncommercial use, distribution, and reproduction in any medium, provided the original author and source are credited. * Corresponding author. Fraunhofer IGB Stuttgart, Department of Cell and Tissue Engineering, Nobelstr. 12, 70569 Stuttgart, Germany. Tel.: þ49 711 970 4082; fax: þ49 711 970 4158. E-mail address: [email protected] (K. Schenke-Layland). 1 Both authors contributed equally.

actinic keratosis, which is induced by UV-radiation. Although actinic keratosis may stay inconspicuous for several years, it is estimated that 8e20% of untreated actinic keratosis turn into SCC [1]. Histopathologically, actinic keratosis is associated with atypical keratinocytes that are present within a thickened epidermal layer due to abnormal keratin expression [1]. An advanced, pre-invasive stage of cutaneous SCC is known as Bowen’s disease, which occurs as an asymptotic, but well-defined scaly plaque that affects all layers of the epidermis [1]. Morphologically, the spatial organization of the epidermis is destructed in Bowen’s disease; parakeratosis and sometimes hyperkeratosis can be observed [1]. Pathological keratinocytes proliferate, laterally expand and can appear as multinucleated, dyskeratotic cells [1]. An invasive, aggressive SCC is characterized by cancerous keratinocytes passing the basal lamina and infiltrating the dermis, in which case a high risk for reoccurrence and the formation of metastases was reported [3]. Various risk factors have been identified to be involved in tumor formation and progression, including exposure to UV-radiation, immunosuppression, individual skin type or previous therapies [4,5]. Less is known about the alterations that occur in the skin due to risk factor exposure, which can finally lead to a malignant highrisk SCC. In vitro toxicological screening on malignant cancer cells

0142-9612/$ e see front matter Ó 2013 The Authors. Published by Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.biomaterials.2013.06.016

7402

E. Brauchle et al. / Biomaterials 34 (2013) 7401e7407

have shown that a three-dimensional (3D) environment impacts the response of malignant cells significantly [6]. Furthermore, previous work from our group has demonstrated the critical importance of 3D environments for in vitro models of the intestine, trachea, and myocardium [7e9]. The establishment of a threedimensional in vitro test system for SCC would be of great benefit for the evaluation of therapies and a general better understanding of the underlying mechanisms of the disease. In vitro skin equivalents have been shown to be suitable tools to mimic functional and structural key characteristics of human native skin [10]. Accordingly, various tissue-engineered products are already commercially available for topical substance testing or toxicity screening [10]. This skin-in-a-dish approach has the potential to be modified to meet different application requirements. In vitro SCC skin models were previously generated by the incorporation of different SCC cell lines into the epidermis. However, these SCC in vitro models could only be weakly discriminated from standard skin equivalents using morphological and tumor marker expression analyses [11,12]. Here, we generated a simplified epidermal skin equivalent that omits the dermis as well as a full-thickness skin equivalent in order to mimic the different aspects of SCC in vivo. An SCC cell line (SCC-25), derived from tongue tissue of a 70-year-old patient afflicted with SCC [13], was introduced in either epidermal or full-thickness skin equivalents (Fig. 1). Although the epidermal disease-in-a-dish models enabled the study of 3D culture effects on SCC-25 cells, they did not allow for the monitoring of the impact of fibroblast signaling, which is crucial for the development of an invasive SCC state in vivo. In previous studies, spontaneous invasion of SCC cell lines could not be accomplished within an in vitro skin equivalent [11,12]. Here, SCC cells were directly diffused within the dermal collagen gel to establish an SSC model system that also mimics the invasive SCC morphology. The co-culture of both healthy human skin cells and malignant cells is important for the reproduction of in vivo cellecell interactions and the in vivo microenvironment that influences and directs cell behavior [14]. Since cellular markers were not sufficient to discriminate SCC in vitro models from standard skin equivalents [11], we employed Raman spectroscopy on full-thickness standard skin-in-a-dish and late-stage disease-in-adish models. Raman spectroscopy is an emerging technology in the field of biomedical research that possesses great potential for the non-invasive discrimination between healthy and pathological cells in vitro [15]. In skin cancer diagnosis, Raman spectroscopy was explored in vivo showing its suitability to differentiate between

cancerous skin tissue and healthy skin regions [16]. Raman spectroscopy generates fingerprints of the molecular constitution of a biological sample; monochromatic laser light is thereby used to excite vibration modes in molecules. Dependent on the excitation mode, the photons from the incident beam can experience a frequency-shift when scattered [17]. The frequency-shifted scattered light is detected as a spectrum, where the resulting peaks are specific for molecular bonds and their secondary structure. Moreover, Raman spectroscopy permits non-destructive measurements on living cells and native tissues without the need for staining or processing steps prior to the measurements [15]. In cancer research, achievements using Raman spectroscopy as a diagnostic tool were presented in various in vitro and in vivo studies [18,19]. In a previous study, Raman spectroscopy enabled an accuracy of more than 90% for the differentiation between non-tumorigenic and tumorigenic skin cell lines [20]. Krishna et al. investigated Raman spectroscopy on biopsy sections of oral squamous carcinoma and non-malignant tissues. Here, spectral differences revealed increased protein and DNA signals in the epidermal cell layer of malignant tissue [21]. Our previous studies indicated that Raman spectroscopy is a promising technology to analyze skin cells within their 3D microenvironment [22], as well as key extracellular matrix proteins such as collagen and elastin [15]. 2. Materials and methods 2.1. Human tissue samples All research was carried out in compliance with the rules for investigation of human subjects as defined in the Declaration of Helsinki. Written informed consent of the patients or parents of the patients was obtained according to the approval of the Landesärztekammer Baden-Württemberg (F-2012-078; for normal foreskin from elective surgeries) or the clinical ethics committee of the University Hospital Tübingen (2012B02; for squamous cell carcinoma biopsies).

2.2. Cell isolation and culture Primary human fibroblasts and keratinocytes were isolated from foreskin samples (age: 2 monthse9 years). The isolation of primary cells was performed as previously described [22]. SCC-25 cells were purchased from LGC Standards (Wesel, Germany (¼ATCC; CRL1628)). SCC-25 cells were cultured in DMEM/F-12 medium (Invitrogen, Karslruhe, Germany), supplemented with 1 mM sodium pyruvate (Life Technologies GmbH, Darmstadt, Germany), 1.1 mM hydrocortisone (Sigma Aldrich, Steinheim, Germany), 10% fetal calf serum (Life Technologies) and 1% gentamycin (Invitrogen). Media exchange was performed every 2e3 days and cells were passaged after reaching 70e80% confluence.

Fig. 1. Schematic of (A) epidermal and (B) full-thickness skin equivalents depicting the standard skin-in-a-dish model and the modified disease-in-a-dish models with SCC-25 cells.

E. Brauchle et al. / Biomaterials 34 (2013) 7401e7407

7403

2.3. Epidermal skin equivalents

2.8. Principal Component Analysis (PCA)

Skin equivalents were generated from primary isolated keratinocytes and SCC25 cells. A total amount of 3.15  105 cells was seeded on culture inserts (Millicell, 0.4 mm pores, Merck Chemicals GmbH, Schwalbach, Germany) in a standard 24-well plate either as a) SCC-25 cells, b) keratinocytes or c) in a co-culture of both cell types at a ratio of 100:1, 50:1, 10:1 and 1:1. Due to increasing the percentage of SCC-25 cells (100:1 to 1:1), early- and late-stage SCC disease-in-a-dish model systems were generated (Fig. 1). For all cultures, we used EpiLife medium (Life Technologies), supplemented with 1.5 mM CaCl2 (Merck, Darmstadt, Germany) and 1% gentamycin. After two days of culture, the seeded inserts were transferred into a 6-well plate, where they were lifted to the air-liquid interface and cultured in EpiLife medium containing 250 mM ascorbic acid (Sigma Aldrich) to allow stratification of the uppermost cell layer. Medium was exchanged every 2e3 days. Epidermal skin equivalents were processed for histological staining after a 14-day period of airlift-culture.

PCA is a multivariate method utilized to analyze the variances in a spectral dataset. PCA is valuable to identify significant shifts in the spectra between two sample groups [25]. Prior to the PCA computation, all Raman spectra were imported to the UnscramblerX 10.2 Software (CAMO, Oslo, Norway). PCA was performed on vector-normalized Raman spectra of primary keratinocytes and SCC-25 cells as well as the respective in vitro skin models. In total, seven principal components (PCs) were calculated using the NIPALS algorithm [26]. As a result of the PCA, every spectrum is described by seven score values. Different score values for one component indicate spectral shifts between sample groups.

3. Results 3.1. Integration of SCC-25 cells into epidermal skin equivalents

2.4. Full-thickness skin equivalents Full-thickness skin-in-a-dish models, consisting of epidermis and dermis, were generated and cultured according to a previously described modified protocol [22]. In this study, 1  105 cells, either keratinocytes or SCC-25 cells, were seeded onto the collagen type-I gel (Fraunhofer IGB, Stuttgart, Germany). Late-stage SCC disease-ina-dish model systems were created by seeding primary dermal fibroblasts and SCC25 cells at a ratio of 10:1 into the dermal collagen gel (Fig. 1). A total amount of 0.4  105 cells was loaded into 400 ml of collagen gel. 2.5. Histological and immunohistochemical staining Hematoxylin and eosin (HE) as well as immunohistochemical staining were performed on epidermal and full-thickness skin equivalents, as well as on native human SCC biopsies (male, 72 and 81 years). Immunohistochemical staining was employed as previously described using the Dako EnVision kit (Dako, Hamburg, Germany) [23], employing the primary antibodies: Ki67 (1:60, #M7240, Dako), cytokeratin 5/6 (1:100, #M7237, Dako), cytokeratin 14 (1:100, #sc-53253, Santa Cruz, Heidelberg, Germany) and E-cadherin (1:500, BD Biosciences, Heidelberg, Germany). Bright-field microscopy was performed with a Zeiss Axiovert 200 microscope using either a 20 or a 40 objective. 2.6. Set up of the Raman spectrometer A custom-built Raman spectrometer was used for all measurements [24]. The laser beam was set on 85 mW power and was guided through a 60 water immersion objective (NA 1.2; Olympus, Hamburg, Germany). For the spectral detection, a spectrograph (Kaiser Optical Systems Inc., Ann Arbor, MI, USA) as well as a nearinfrared-optimized and cooled CCD camera (Andor iDus, Belfast, Northern Ireland) were integrated at the output port of the system. Raman spectra were acquired in the wavenumber range of 0e2000 cm1. 2.7. Data acquisition and processing of Raman spectra Raman spectra were recorded using the Andor Solis software package (Andor iDus). Measurements were performed on single primary keratinocytes and SCC-25 cells in suspension after passaging as well as on the epidermis of full-thickness standard skin-in-a-dish and late-stage disease-in-a-dish models. The acquisition time per spectrum was 100 s. In order to generate a representative set of spectra, 30 Raman spectra were taken per sample. All measurements were performed from specimens on glass surfaces. After every five to ten measurements, a background signal was taken from the glass. The background signal was subtracted from the sample signal using the OPUSÒ software 4.2 (Bruker Optik GmbH, Ettlingen, Germany). Additionally, OPUSÒ software 4.2 was used to cut the Raman spectra into the spectral range of 400e1800 cm1 and to perform the baseline correction.

Epidermal skin equivalents were generated from keratinocytes, co-cultured keratinocytes and SCC-25 cells at varying ratios, as well as SCC-25 cells alone (Figs. 1 and 2; Suppl. Fig. 1). In standard epidermal skin-in-a-dish models, cell stratification and differentiation was observed, due to the exposure of keratinocytes to the airliquid interface (Fig. 2A). Early-stage disease-in-a-dish models, where keratinocytes and SCC-25 cells were seeded at a ratio of 10:1, formed a similar stratified epidermal layer (Fig. 2B). In these earlystage disease-in-a-dish models, a cancerous epidermal morphology was not observed. In contrast, late-stage disease-in-a-dish models containing only SCC-25 cells showed less stratification, with cells retaining their nucleus (Fig. 2C). Interestingly, none of the epidermal early-stage disease-in-a-dish models displayed prominent tumor characteristics such as tumor cell aggregation or the formation of keratin pearls. When varying the ratio of keratinocytes and SCC-25 cells, we found that increasing numbers of SCC-25 cells disturbed the formation of a regular structured epidermis (Suppl. Fig. 1). A lack of spatial organization was previously described for SCC in vivo [27]. Based on these results, we used SCC25 cells in the epidermis alone in order to establish a late-stage fullthickness disease-in-a-dish model. 3.2. Integration of SCC-25 cells into full-thickness skin equivalents Keratinocytes benefit from soluble factors released by fibroblasts when both cell types are co-cultured in standard fullthickness skin equivalents [28,29]. To mimic the in vivo tumor microenvironment, SCC-25 cells were integrated into full-thickness skin equivalents, consisting of epidermal and dermal layers. In early in vivo stages of the disease, SCC cells do not infiltrate the dermis, but expand within the epidermal skin layer [1]. In order to generate a disease-in-a-dish model system for this pre-invasive stage, SCC-25 cells were directly seeded onto the fibroblast-containing dermis (Fig. 3A and B). It has been reported that SCC cells in vivo produce and accumulate keratins that lead to the formation of encapsulated keratin pearls in the epidermis (Fig. 3C) [1]. Similar to the in vivo milieu, we identified the presence of keratin accumulations (Fig. 3A

Fig. 2. HE-stained sections of epidermal skin equivalents: (A) Standard epidermal skin-in-a-dish, (B) epidermal early-stage disease-in-a-dish model with keratinocytes and SCC-25 cells, co-cultured at a ratio of 10:1, and (C) epidermal late-stage disease-in-a-dish model. Scale bar equals 50 mm.

7404

E. Brauchle et al. / Biomaterials 34 (2013) 7401e7407

and B; black arrows). Moreover, the SCC-25 cells in the epidermis formed multiple cell layers, but lacked spatial organization and a stratified corneous layer. Additionally, this epidermal layer displayed small gaps, which are due to a lack of tight cellecell contact formation, indicated by discontinuous staining of E-cadherin (Fig. 3D). During the progression of the disease in vivo, SCC cells migrate into the dermis and form typical tumor cell nests that contain keratin accumulations (Fig. 3E). However, we did not observe spontaneous invasion of the SCC-25 cells into the dermal layer. In order to mimic this invasive disease stage in vitro, SCC-25 cells were seeded and co-cultured with fibroblasts in the dermal layer of full-thickness skin equivalents (Fig. 3F). In this late-stage disease-in-a-dish model, we identified the formation of round cell nests due to a high proliferative capacity of the SCC-25 cells, as demonstrated by Ki67 staining (Fig. 4A). Additionally, the tumor cell nests were positively stained for cytokeratin 5/6 (Fig. 4B)

and cytokeratin 14 (Fig. 4C), resembling keratin accumulations observed in vivo [1]. 3.3. Discrimination of full-thickness skin-in-a-dish and disease-ina-dish models by Raman spectroscopy Raman spectroscopy has previously been demonstrated to be a suitable technology for the label-free detection of molecular differences between tumorigenic and healthy primary cells in vitro [22,23]. Due to the lack of appropriate cellular marker proteins, Raman spectroscopy was explored to analyze the epidermal morphology of full-thickness skin-in-a-dish and late-stage diseasein-a-dish models. The mean Raman spectra of skin-in-a-dish and disease-in-a-dish models exhibited typical patterns of cellular molecules, including signals from phenylalanine at 1003 cm1, CeH vibrations at 1440 cm1 and amide I at 1660 cm1 [30]. The

Fig. 3. HE-(AeC, E, F) and immunohistochemically (D) stained sections of in vivo squamous cell carcinoma (SCC) and full-thickness skin equivalents. (A) Full-thickness late-stage disease-in-a-dish model with SCC-25 cells cultured in the epidermis. Arrows indicate keratin accumulations. (B) Magnification of (A). (C) Section of SCC from a 72-year-old male patient. The arrow points to a keratin pearl in the epidermis. (D) E-cadherin staining (brown) shows disrupted cellecell contacts. Hematoxylin was used as nuclear counterstain (blue). (E) Late-stage SCC from an 81-year-old male patient. Arrows highlight tumor cell nests and keratin accumulations in the dermis. (F) Section of the dermis of a disease-in-adish model composed of fibroblasts and SCC-25 cells that were seeded in a ratio of 10:1. Scale bars equal 100 mm.

E. Brauchle et al. / Biomaterials 34 (2013) 7401e7407

7405

Fig. 4. Immunohistochemical staining of full-thickness late-stage disease-in-a-dish models. (A) Ki67 staining indicates the presence of highly proliferative SCC-25 cells within the tumour nests, where also (B) cytokeratin 5/6 and (C) cytokeratin 14 are expressed (all in brown). Hematoxylin served as nuclear counterstain (blue). Scale bar equals 100 mm.

standard deviation for each spectral group, skin-in-a-dish and disease-in-a-dish models, was comparable. Notably, the spectra of disease-in-a-dish models showed higher intensities in proteinspecific Raman shifts (1300e1350 cm1, 1440 cm1, 1660 cm1) when compared to skin-in-a-dish models. Additionally, the baseline in Raman spectra from full-thickness skin-in-a-dish displayed some fluorescence background, which was not present in the Raman spectra of disease-in-a-dish models (Fig. 5A). In order to generate a model to classify unknown spectra, we employed PCA on the spectral data sets. Seven PCs were calculated to describe the spectral variances. The seven PCs explained a total variance of 97.4%. Based on the score values that were assigned for every single Raman spectrum, a separation of full-thickness skin-in-a-dish and disease-in-a-dish models was observed due to PC1, which accounted for 86% of the spectral variance. Raman spectra of skin-in-a-dish models exhibited negative PC1 score values. In contrast, the spectra of disease-in-a-dish models were assigned to positive score values for PC1 (Fig. 5B). Interestingly, when Raman spectroscopy was conducted on 2D-cultured primary keratinocytes and SCC-25 cells in suspension, the spectra showed no significant differences and could thus not be used for a classification model based on the PCA (Suppl. Fig. 2). Therefore, we conclude that the spectral patterns of both cell types are highly dependent on a physiological 3D microenvironment as it is provided in native tissues. Raman spectra of SCC-25 cells that were incorporated in full-thickness disease-ina-dish models reflected an altered morphology that was also detected histologically.

4. Discussion In this study, SCC-25 cells were integrated into epidermal and full-thickness skin equivalents. Early-stage epidermal disease-in-adish models were generated by co-culturing keratinocytes and SCC-25 cells. Hereby we observed that the differentiation and stratification of the epidermis is highly impacted by the number of incorporated SCC-25 cells. Accordingly, increasing percentages of SCC-25 cells resulted in a less organized epidermal structure with insufficient stratification of the uppermost cell layer. Although no formation of defined tumor cell nests was detected in the epidermal disease-in-a-dish models, the co-culture of keratinocytes with SCC-25 cells allowed interaction and signaling between healthy and malignant cells, as was previously described [14]. Insufficient stratification is a hallmark of precancerous lesions; therefore, the here described epidermal disease-in-a-dish models may enable the in-depth investigation of precancerous lesions such as actinic keratosis and Bowen’s disease. A major challenge for the detection of SCC is the lack of suitable methods to visualize malignant cells in the epidermis. Various marker proteins have been screened for the discrimination between keratinocytes and SCC-25 cells, but these cells share similar protein expression patterns [11]. To overcome this problem, we explored Raman spectroscopy as a non-invasive optical monitoring technology. Raman spectroscopy was previously shown to be a suitable tool for the label-free detection of molecular differences between tumorigenic cells and primary cells in vitro [23]. Here,

Fig. 5. Raman spectral analysis of full-thickness skin-in-a-dish and late-stage disease-in-a-dish models. (A) Mean Raman spectra of skin-in-a-dish and disease-in-a-dish models. Differences in signal intensities are highlighted. Grey lines depict standard deviations. Spectra are shifted vertically for clarity. (B) Principal Component Analysis (PCA) allows discrimination between full-thickness skin-in-a-dish and disease-in-a-dish models. PC 1 accounts for 87% of the total variance.

7406

E. Brauchle et al. / Biomaterials 34 (2013) 7401e7407

different molecular patterns of keratinocytes and SCC-25 cells were revealed by Raman spectra collected of full-thickness skin-in-a-dish and late-stage disease-in-a-dish models. Both models exhibited a specific spectral pattern, consisting of molecular signals from proteins, nucleic acids and lipids [30]. Significant differences in the Raman spectra between the two models were detected by employing PCA on the spectra. Protein-related signals at the Raman shifts 1300e1350 cm1, 1440 cm1 and 1660 cm1 displayed higher intensities in disease-in-a-dish models. Previously, increased protein signals in Raman spectra were related to higher proliferation of tumor cells when compared to healthy cells [31,32]. In our study, no differences between the two cell types were detectable in 2D cell cultures. Consequently, we hypothesize that the differential peak intensities, seen in full-thickness skin-in-a-dish and disease-in-adish models, were not only related to cell proliferation, but to differences in the expression or the folding of epidermal-specific keratins. In this context, we demonstrated that high amounts of cytokeratin 5/6 and 14 were expressed in the accumulations formed within the full-thickness disease-in-a-dish models. Interestingly, the identified Raman signals were previously assigned to keratin signals in skin [33], which strengthens our hypothesis. Additionally, the Raman band at 1660 cm1, which is also known as the amide I peak, was associated with malignant protein conformations in other tissues [34]. Spectral regions around 1420 cm1 and 1640 cm1 were outlined in an in vivo study where Raman spectroscopy was used to classify different skin cancer types including SCC [35]. It was also shown that Raman spectra of the epidermis differed in their signal intensities dependent on the functionality and the thickness at the measured skin region [36]. According to the findings known from Raman spectroscopy on healthy skin and SCC in vivo, we hypothesize that the molecular changes that cause the discussed spectral differences are specific for SCC. Molecular abnormalities detected in disease-in-a-dish models together with histological staining indicate that the here presented SCC in vitro models correlate with SCC in vivo. PCA, which was employed to compare these Raman spectra with those of skinin-a-dish models, showed significant spectral shifts. Potential classification of in vitro skin models based on the PCA results was indicated due to differential PC1 score values for standard skin-ina-dish and disease-in-a-dish models. In a previous report, similar PCA-based classifications were generated for the visualization of cells within porcine epidermis [37]. Notably, this is the first study reporting on Raman measurements for the comparison of keratinocytes and SCC-25 cells within tissue-engineered skin equivalents. In standard full-thickness skin equivalents, keratinocytes benefit from soluble factors released by fibroblasts that support generation of a stratified and regular epidermis [28,29]. In this study, full-thickness skin equivalents were generated with an epidermis built from SCC-25 cells and a dermis composed either of fibroblasts alone, or co-cultured fibroblasts and SCC-25 cells. In contrast to other studies [11,12] where a different SCC cell line was incorporated into 3D skin equivalents, we did not observe spontaneous invasion of SCC-25 cells into the dermal layer. However, the epidermis containing SCC-25 cells displayed morphological similarities to SCC in vivo, such as the presence of keratin pearls. Additionally, small gaps in-between the epidermal SCC-25 cells indicated weak cellecell contacts, which were confirmed by discontinuous immunohistochemical staining for E-cadherin. It was previously shown that the expression of proteins related to cellecell contacts, such as occludin, is down-regulated in squamous cell carcinoma [38]. In our study, SCC-25 cells in the dermis formed capsular tumor cell nests, containing cells at various stages of differentiation, resembling SCC in vivo. Our observations indicate that this behavior of SCC-25 cells is highly dependent on

soluble factors released by fibroblasts, which play a pivotal role for signaling that can lead to tumor progression in squamous head and neck cancers [39]. 5. Conclusions In this study, SCC-25 cells were successfully integrated into epidermal and full-thickness 3D skin equivalents to mimic an SSC phenotype in vitro. Raman spectroscopy of full-thickness skin-in-adish and disease-in-a-dish models displayed differential molecular patterns for both models. Moreover, our results implicate an important role of fibroblasts for SCC progression and tumor cell behavior. The presented epidermal and full-thickness disease-in-adish models containing SCC-25 cells in the epidermis may enable in-depth investigations on precancerous lesions in vitro. However, in order to study invasive SCC, infiltration into the dermis had to be simulated. Therefore, SCC-25 cells were seeded together with fibroblasts directly into the dermal layer. Previously reported skin equivalents generated with SCC cells exhibited only minor invasion, without the presence of tumor cell nests, as observed in our experiments. Altogether, the here presented SSC disease-in-a-dish approaches offer the unique opportunity to model SSC in vitro, which will allow further insight in SSC disease progression, and facilitate the development of therapeutic strategies. Acknowledgments We thank Shannon Lee Layland (Fraunhofer IGB Stuttgart) for his thoughtful comments on the manuscript. The authors are grateful for the financial support by the Fraunhofer-Gesellschaft Internal programs (Attract 692263 to KSL), the BMBF (0316059 to KSL) and the Ministry of Science, Research and the Arts of BadenWürttemberg (33-729.55-3/214 to KSL). Appendix A. Supplementary data Supplementary data related to this article can be found online at http://dx.doi.org/10.1016/j.biomaterials.2013.06.016. References [1] Weedon D. Tumors of the epidermis. In: Weedon D, editor. Weedon’s skin pathology. 3rd ed. Edinburgh: Churchill Livingstone; 2010. p. 667e708. [2] Sidoroff A, Thaler P. Taking treatment decisions in non-melanoma skin cancerethe place for topical photodynamic therapy (PDT). Photodiagnosis Photodyn Ther 2010;7:24e32. [3] Boukamp P. Non-melanoma skin cancer: what drives tumor development and progression? Carcinogenesis 2005;26:1657e67. [4] Glogau RG. The risk of progression to invasive disease. J Am Acad Dermatol 2000;42:23e4. [5] Schmook T, Stockfleth E. Current treatment patterns in non-melanoma skin cancer across Europe. J Dermatolog Treat 2003;3:3e10. [6] Godugu C, Patel AR, Desai U, Andey T, Sams A, Singh M. AlgiMatrix based 3D cell culture system as an in-vitro tumor model for anticancer studies. PLoS One. http://dx.doi.org/10.1371/journal.pone.0053708. Available from URL: http://www.ncbi.nlm.nih.gov/pubmed/23349734; 2013. [7] Hinderer S, Schesny M, Bayrak A, Ibold B, Hampel M, Walles T, et al. Engineering of fibrillar decorin matrices for a tissue-engineered trachea. Biomaterials 2012;33:5259e66. [8] Pusch J, Votteler M, Gohler S, Engl J, Hampel M, Walles H, et al. The physiological performance of a three-dimensional model that mimics the microenvironment of the small intestine. Biomaterials 2011;32:7469e78. [9] Schenke-Layland K, Nsair A, Van Handel B, Angelis E, Gluck JM, Votteler M, et al. Recapitulation of the embryonic cardiovascular progenitor cell niche. Biomaterials 2011;32:2748e56. [10] Groeber F, Holeiter M, Hampel M, Hinderer S, Schenke-Layland K. Skin tissue engineering d In vivo and in vitro applications. Adv Drug Deliv Rev 2011;63: 352e66. [11] Commandeur S, De Gruijl FR, Willemze R, Tensen CP, El Ghalbzouri A. An in vitro three-dimensional model of primary human cutaneous squamous cell carcinoma. Exp Dermatol 2009;18:849e56.

E. Brauchle et al. / Biomaterials 34 (2013) 7401e7407 [12] Obrigkeit DH, Jugert FK, Beermann T, Baron JM, Frank J, Merk HF, et al. Effects of photodynamic therapy evaluated in a novel three-dimensional squamous cell carcinoma organ construct of the skin. Photochem Photobiol 2009;85:272e8. [13] Rheinwald JG, Beckett MA. Tumorigenic keratinocyte lines requiring anchorage and fibroblast support cultures from human squamous cell carcinomas. Cancer Res 1981;41:1657e63. [14] Javaherian A, Vaccariello M, Fusenig NE, Garlick JA. Normal keratinocytes suppress early stages of neoplastic progression in stratified epithelium. Cancer Res 1998;58:2200e8. [15] Brauchle E, Schenke-Layland K. Raman spectroscopy in biomedicine - noninvasive in vitro analysis of cells and extracellular matrix components in tissues. Biotechnol J 2013;8:288e97. [16] Larraona-Puy M, Ghita A, Zoladek A, Perkins W, Varma S, Leach IH, et al. Development of Raman microspectroscopy for automated detection and imaging of basal cell carcinoma. J Biomed Opt 2009;14:054031e054031-10. [17] Raman CV, Krishnan KS. A new type of secondary radiation. Nature 1928;121: 501e2. [18] Harris A, Rennie A, Waqar-Uddin H, Wheatley S, Ghosh S, Martin-Hirsch D, et al. Raman spectroscopy in head and neck cancer. Head Neck Oncol 2010;2:26. [19] Tu Q, Chang C. Diagnostic applications of Raman spectroscopy. Nanomedicine 2012;8:545e58. [20] Wang H, Tsai TH, Zhao J, Lee AM, Lo BK, Yu M, et al. Differentiation of HaCaT cell and melanocyte from their malignant counterparts using micro-Raman spectroscopy guided by confocal imaging. Photodermatol Photoimmunol Photomed 2012;28:147e52. [21] Krishna CM, Sockalingum GD, Kurien J, Rao L, Venteo L, Pluot M, et al. MicroRaman spectroscopy for optical pathology of oral squamous cell carcinoma. Appl Spectrosc 2004;58:1128e35. [22] Pudlas M, Koch S, Bolwien C, Thude S, Jenne N, Hirth T, et al. Raman spectroscopy: a noninvasive analysis tool for the discrimination of human skin cells. Tissue Eng Part C Methods 2011;17:1027e40. [23] Pudlas M, Brauchle E, Klein TJ, Hutmacher DW, Schenke-Layland K. Noninvasive identification of proteoglycans and chondrocyte differentiation state by Raman microspectroscopy. J Biophotonics 2013;6:205e11. [24] Votteler M, Carvajal Berrio DA, Pudlas M, Walles H, Stock UA, SchenkeLayland K. Raman spectroscopy for the non-contact and non-destructive monitoring of collagen damage within tissues. J Biophotonics 2012;5:47e56. [25] Wold S, Esbensen K, Geladi P. Principal component analysis. Chemometrics Intell Lab Syst 1987;2:37e52.

7407

[26] Martens H, Naes T. Multivariate calibration. Chichester: John Wiley & Sons; 1989. [27] Janes SM, Watt FM. New roles for integrins in squamous-cell carcinoma. Nat Rev Cancer 2006;6:175e83. [28] El-Ghalbzouri A, Gibbs S, Lamme E, Van Blitterswijk CA, Ponec M. Effect of fibroblasts on epidermal regeneration. Br J Dermatol 2002;147:230e43. [29] Fuchs E. Epidermal differentiation: the bare essentials. J Cell Biol 1990;111: 2807e14. [30] Notingher I, Hench LL. Raman microspectroscopy: a noninvasive tool for studies of individual living cells in vitro. Expert Rev Med Devices 2006;3: 215e34. [31] Oshima Y, Shinzawa H, Takenaka T, Furihata C, Sato H. Discrimination analysis of human lung cancer cells associated with histological type and malignancy using Raman spectroscopy. J Biomed Opt 2010;15:017009e017009-8. [32] Taleb A, Diamond J, Mcgarvey JJ, Beattie JR, Toland C, Hamilton PW. Raman microscopy for the chemometric analysis of tumor cells. J Phys Chem B 2006;110:19625e31. [33] Barry BW, Edwards HGM, Williams AC. Fourier transform Raman and infrared vibrational study of human skin: assignment of spectral bands. J Raman Spectrosc 1992;23:641e5. [34] Gniadecka M, Wulf HC, Nielsen OF, Christensen DH, Hercogova J. Distinctive molecular abnormalities in benign and malignant skin lesions: studies by Raman spectroscopy. Photochem Photobiol 1997;66:418e23. [35] Lieber CA, Majumder SK, Ellis DL, Billheimer DD, Mahadevan-Jansen A. In vivo nonmelanoma skin cancer diagnosis using Raman microspectroscopy. Lasers Surg Med 2008;40:461e7. [36] Caspers PJ, Lucassen GW, Wolthuis R, Bruining HA, Puppels GJ. In vitro and in vivo Raman spectroscopy of human skin. Biospectroscopy 1998;4:31e9. [37] Zhang G, Moore DJ, Flach CR, Mendelsohn R. Vibrational microscopy and imaging of skin: from single cells to intact tissue. Anal Bioanal Chem 2007;387:1591e9. [38] Rachow S, Zorn-Kruppa M, Ohnemus U, Kirschner N, Vidal YSS, Von Den Driesch P, et al. Occludin is involved in adhesion, apoptosis, differentiation and ca(2þ)-homeostasis of human keratinocytes: implications for tumorigenesis. PLoS One. http://dx.doi.org/10.1371/journal.pone.0055116. Available from URL: http://www.ncbi.nlm.nih.gov/pmc/articles/PMC3563667/; 2013. [39] Leef G, Thomas SM. Molecular communication between tumor-associated fibroblasts and head and neck squamous cell carcinoma. Oral Oncol. http:// dx.doi.org/10.1016/j.oraloncology.2012.12.014. Available from URL: http:// www.ncbi.nlm.nih.gov/pubmed/23357526; 2013.