Presented here is a stepwise procedure for in vitro differentiation of human primary keratinocytes by contact inhibition followed by characterization at the molecular level by RNA-seq analysis.
Human primary keratinocytes are often used as in vitro models for studies on epidermal differentiation and related diseases. Methods have been reported for in vitro differentiation of keratinocytes cultured in two-dimensional (2D) submerged manners using various induction conditions. Described here is a procedure for 2D in vitro keratinocyte differentiation method by contact inhibition and subsequent molecular characterization by RNA-seq. In brief, keratinocytes are grown in defined keratinocyte medium supplemented with growth factors until they are fully confluent. Differentiation is induced by close contacts between the keratinocytes and further stimulated by excluding growth factors in the medium. Using RNA-seq analyses, it is shown that both 1) differentiated keratinocytes exhibit distinct molecular signatures during differentiation and 2) the dynamic gene expression pattern largely resembles cells during epidermal stratification. As for comparison to normal keratinocyte differentiation, keratinocytes carrying mutations of the transcription factor p63 exhibit altered morphology and molecular signatures, consistent with their differentiation defects. In conclusion, this protocol details the steps for 2D in vitro keratinocyte differentiation and its molecular characterization, with an emphasis on bioinformatic analysis of RNA-seq data. Because RNA extraction and RNA-seq procedures have been well-documented, it is not the focus of this protocol. The experimental procedure of in vitro keratinocyte differentiation and bioinformatic analysis pipeline can be used to study molecular events during epidermal differentiation in healthy and diseased keratinocytes.
Human primary keratinocytes derived from the human skin are often used as a cellular model to study the biology of the epidermis1,2,3,4. The stratification of the epidermis can be modeled by keratinocyte differentiation, either in a 2D submerged monolayer fashion or 3D air-lift organotypic model2,3,5,6,7. Although 3D models have become increasingly important to assess the epidermal structure and function, 2D differentiation models are still widely used, due to their convenience and the possibility to generate large numbers of cells for analyses.
Various conditions have been applied for inducing keratinocyte differentiation in 2D, including addition of serum, high concentration of calcium, lower temperature and inhibition of epidermal growth factor receptors2,3. Each of these methods has been validated by a number of keratinocyte differentiation marker genes and shown to be effective in assessing keratinocyte differentiation, including under pathological conditions. However, these induction conditions also show differences in their differentiation efficiency and kinetics when specific panels of marker genes are examined2,3.
One of these methods involves keratinocyte contact inhibition and depletion of growth factors in the culture medium8. It has been shown that keratinocytes can differentiate spontaneously when cells reach full density. Excluding growth factors in the culture medium can further enhance differentiation. The method combining contact inhibition and depleting growth factors has been shown to generate differentiated keratinocytes with gene expression patterns similar to the normal stratified epidermis when using several epidermal markers3, suggesting that this model is suitable for studying normal keratinocyte differentiation. Recently, two comprehensive gene expression analyses of keratinocyte differentiation using this model have been reported9,10. Researchers validated this model at the molecular level and showed that it can be used to study normal and diseased keratinocyte differentiation.
This protocol describes the procedure for the in vitro differentiation method and molecular analysis of differentiated cells using RNA-seq. It also illustrates characterization of the transcriptome of cells on differentiation day 0 (proliferation stage), day 2, day 4, and day 7 (early, middle, and late differentiation, respectively). It is shown that differentiated keratinocytes display gene expression patterns that largely resemble cells during epidermal stratification. To examine whether this method can be used for studying skin pathology, we applied the same experimental and analysis pipeline to investigate keratinocytes carrying mutations of the transcription factor p63 that are derived from patients with ectrodactyly, ectodermal displasia and cleft lip/palate (EEC) syndrome11,12. This protocol focuses on the in vitro differentiation of keratinocytes as well as subsequent bioinformatic analysis of RNA-seq. Other steps in the complete procedure such as RNA extraction, RNA-seq sample preparation and library construction, are well documented and can be easily followed, especially when using many commonly used commercial kits. Therefore, these steps are only briefly described in the protocol. The data show that this pipeline is suitable for studying molecular events during epidermal differentiation in healthy and diseased keratinocytes.
Skin biopsies were taken from the trunk of healthy volunteers or patients with p63 mutations, to set up the primary keratinocyte culture. All procedures regarding establishing human primary keratinocytes were approved by the ethical committee of the Radboud University Nijmegen Medical Centre (“Commissie Mensgebonden Onderzoek Arnhem-Nijmegen”). Informed consent was obtained.
1. Human primary keratinocyte differentiation by contact inhibition
2. RNA extraction
3. RNA quality check
4. RNA-seq library preparation
NOTE: The RNA-seq library preparation is often performed with a commercial kit or under commercial settings. The described protocol is adapted from a commercial kit, KAPA RNA HyperPrep Kit with RiboErase (Illumina), with a brief description of all required steps: rRNA depletion with oligo hybridization to human ribosomal RNAs, RNA fragmentation, first-strand synthesis, second-strand synthesis and A-tailing, and cleanup after each step15. Other library preparation kits can also be used for this purpose. It is recommended to perform this step using a commercially available kit, as the quality of generated cDNA library is often more consistent. 2) The following steps are described for 1x library preparation. If preparing several samples, make master mixes with 10% extra volume.
5. Data pre-processing
6. RNA-seq data analysis
Normal keratinocyte differentiation and RNA-seq analysis
In this experiment, keratinocyte lines derived from five individuals were used for differentiation and RNA-seq analyses. Figure 1 summarizes the experimental procedure of differentiation and RNA-seq analysis results. An overview of in vitro differentiation procedures of normal keratinocytes and cell morphology changes during differentiation are illustrated in Figure 1A. Principle component analysis (PCA) showed that keratinocytes undergoing differentiation had connected but distinct overall gene expression profiles (Figure 1B). Highly variable genes were clustered by kmeans to visualize gene expression dynamics and patterns during differentiation (Figure 1C).
Each cluster of genes was represented by the keratinocyte differentiation hallmark genes (e.g., KRT5 for proliferation, KRT1 and KRT10 for early and mid-differentiation, and IVL/LOR/FLG for late differentiation). Gene ontology (GO) annotation analysis of highly variable gene clusters (Figure 1C) showed a clear difference in gene functions of these gene clusters (e.g., keratinization in the mid-stages of differentiation; and epidermal cell differentiation, keratinocyte differentiation, and peptide cross-linking in the late stages; Figure 1D). Protein expression of several differentiation markers were measured by western blotting (Figure 1E).
P63 mutant keratinocyte differentiation and RNA-seq analysis:
In the second experiment, cell morphology and gene expression differences were compared between keratinocytes from healthy controls and three lines derived from patients carrying p63 mutations (mutants R204W, R279H, and R304W). Figure 2A shows an overview of differentiation procedures and cell morphology changes. Mutant keratinocytes remained flat on the surface of the dish and did not become crowded or overlap growth as control keratinocytes on day 7.
In PCA analysis, the control cell lines clearly followed a differentiation pattern, as compared to Figure 1B. However, the pattern of gene expression of mutant cells during differentiation stays largely similar to those of proliferating/undifferentiated cells. Among the three mutant lines, differentiating R279 samples moved along PC1 and PC2 to some extent, indicating that its differentiation was less impaired, compared to R204W and R304W.
In the clustering analysis (Figure 2C), genes downregulated in control cells (cluster 1) were partially downregulated in R204W and R279W, but their gene expression was not drastically changed in R304W. These genes likely play roles in cell proliferation, as shown by GO annotation (Figure 2D). In cluster 2 (Figure 2C), genes were first induced and subsequently downregulated in control cells. These genes are likely involved in keratinocyte differentiation, as epidermal differentiation and keratinization functions were highly enriched for this cluster genes (Figure 2D). These genes were not induced in R204W and R304W, whereas in R279H cells, these genes were induced but not downregulated as much as the control cells.
Genes in cluster 3 were only induced at the end of differentiation in control cells (Figure 2D). Consistent with this, these genes may have a role in the outer most layer of the epidermis, as they have been shown to be responsive to external stimuli and inflammation (Figure 2D). The expression pattern of these genes did not change much in all three mutant cell lines. The visible differences in gene expression patterns between control and mutant cells demonstrate that mutant cells cannot differentiate properly in these in vitro differentiation models.
Figure 1: Keratinocyte differentiation and analysis. (A) Overview of the keratinocyte differentiation protocol and cell morphology. Scale bar = 100 µm. (B) Principle component analysis of differentiating control keratinocytes. (C) Heatmap of the top 500 highly variable genes during keratinocyte differentiation. Genes are clustered in three clusters using kmean clustering. Representative differentiation marker genes for each gene cluster are indicated at the side. (D) GO term enrichment analysis of overrepresented functions for genes in the highly variable gene clusters, as compared to a background of all expressed genes (counts of >10). GOrilla was used for the enrichment test. (E) Western blot of keratinocyte differentiation markers during differentiation. Please click here to view a larger version of this figure.
Figure 2: Comparison of control and p63 mutant keratinocytes. (A) Overview of the keratinocyte differentiation protocol and cell morphology of control and p63 mutant keratinocytes. Scale = 100 µm. (B) Principle component analysis of differentiating control and patient (R204W, R279H & R304W) keratinocytes. (C) Heatmap of differential genes between control and mutant keratinocytes. Genes are clustered in three clusters using kmean clustering. Representative differentiation marker genes for each gene cluster are indicated. (D) GO term enrichment analysis of overrepresented functions for genes in the highly variable gene clusters, as compared to a background of all expressed genes (counts of >10). GOrilla was used for the enrichment test. Please click here to view a larger version of this figure.
KGM component | Stock | Medium | Volume |
KBM | 500 mL | ||
Pen/Strep | 100,000 units/mL | 100 units/mL | 5 mL |
BPE | ~13 mg/mL | 0.4% | 2 mL |
Ethanolamine | 0.1 M | 0.1 mM | 500 μL |
o-phosphoethanolamine | 0.1 M | 0.1 mM | 500 μL |
Hydrocortisone | 0.5 mg/mL | 0.5 µg/mL | 500 μL |
Insulin | 5 mg/mL | 5 µg/mL | 500 μL |
EGF | 10 μg/mL | 10 ng/mL | 500 μL |
Table 1. KGM-pro medium supplements.
KGM component | Stock | Medium | Volume |
KBM | 500 mL | ||
Pen/Strep | 100,000 units/mL | 100 units/mL | 5 mL |
Ethanolamine | 0.1 M | 0.1 mM | 500 μL |
o-phosphoethanolamine | 0.1 M | 0.1 mM | 500 μL |
Table 2. KGM-diff medium supplements.
Supplemental Coding Files: Folder_structure.txt; Generate_genome.txt; Map_fastq.txt; RNA_seq_kc_differentiation_patient.html; RNA_seq_kc_differentiation_wt.html; Sample_data_example.csv; TrimGalore.txt; and Wig2bw.txt. Please click here to download this file.
This work describes a method for inducing human keratinocyte differentiation and subsequent characterization using RNA-seq analyses. In the current literature, many studies on human keratinocyte differentiation use two other methods, with a high calcium concentration or with serum as methods to induce differentiation2,3,23. A previous report carefully compared these three different methods3 and showed that these methods can represent distinct biology of keratinocyte differentiation. In this same report, authors showed that differentiated cells induced by serum have high proliferative potential and express genes such as KRT16 and SKALP/ PI3 that are expressed in the psoriasis skin but not in normal epidermis, and that therefore serum-induced differentiation model can be used to study psoriasis. In contrast, the method of contact inhibition plus growth factor exclusion can induce KRT1 and KRT10, which are normally expressed in the epidermis and resemble normal keratinocyte differentiation. High calcium induction gives rise to a gene expression profile that is between serum induction and contact inhibition, as the least specific method. The analyses using RNA-seq analyses during differentiation confirmed that the method of contact inhibition plus growth factor exclusion results in differentiated keratinocytes with similar gene expression profiles as those expected for epidermal differentiation (e.g., KRT5 in proliferating cells, KRT1 and KRT10 in early differentiation induction, and LOR and FLG in late differentiation; Figure 1C).
These findings demonstrate that this differentiation technique is an easy-to-use and reliable method to study keratinocyte differentiation. Furthermore, this work on keratinocytes derived from p63 mutant keratinocytes also show that the method can be used to study affected keratinocyte differentiation under diseased conditions. In this in vitro differentiation approach, KGM purchased from Lonza is used, as this medium yields consistent results. In principle, other epidermal medium with similar composition such as keratinocyte serum-free medium (KSFM) from Thermo Fisher Scientific is also likely suitable, although this needs to be tested.
It should be noted that this method has some limitations. As it is based on contact inhibition, the confluency of cell density is required. In our experiments with p63 mutant keratinocytes, a higher initial seeding density has been necessary to ensure cells to reach confluency. In addition, when cells do not grow with the same speed, differentiation sometimes must be induced on different days. These considerations should be taken into account when setting up experiments. In experimental settings where cells need to be induced at the same time, other differentiation methods should be considered, including addition of serum, high concentrations of calcium, and inhibition of epidermal growth factor receptors2,3. Nevertheless, all in vitro differentiation methods have pros and cons, as they probably represent partial differentiation induction signals that are present during in vivo skin development2.
A comprehensive molecular analysis that compares these different methods will be highly informative and can instruct the choice of method most suitable for different studies on biological processes. In addition, validating changes measured at the transcriptomic level is highly advised, for example, via western blotting or proteomic analyses. Furthermore, in vitro data should be used with caution, and conclusions from these models should be validated in vivo, preferably in human skin development.
Basic principles of RNA extraction and RNA-seq library preparation have been well described previously24,25,26,27,28. In this protocol, the RNA extraction and RNA-seq library preparation procedures are based on workflows of commercially available kits. In principle, different methods for RNA extraction should not have a major influence on RNA-seq analyses if the RNA quality is good. In cases where RNA quality is poor (i.e., when RNA is extracted formalin-fixed paraffin-embedded tissues), RNA-seq can still be performed; however, the RNA fragmentation step should be adjusted. The RNA-seq library preparation covers from ribosomal RNA (rRNA) removal to the cDNA library construction for sequencing. The major procedures may be performed by using various kits, or using individual enzymes and homemade buffers29,30,31,32. It should be noted that, if RNA-seq analysis is performed using different basic principles (e.g., either ribosomal RNA depletion or polyA mRNA enrichment by hybridization to polyT oligos), the outcome of RNA-seq analyses may differ. Furthermore, the protocol utilizes ribosomal RNA removal by hybridization of DNA oligos to human, mouse, and rat rRNA. When working with different species, an alternative oligo set should be employed.
Sequencing of the generated library can be performed either on both ends of the fragments (paired end) or at one end of the fragment (single end). In general, paired end sequencing vastly improves mappability and provides more information regarding transcript variants. However, for a relatively simple differential gene analysis as described here, single end sequencing can also provide sufficient information.
For the important step of data analysis, a relatively simple method for quality control of the fastq files is described, followed by mapping reads to the genome. Even though the provided bash code does not have ideal scalability, it has the advantage of transparency. The choice of software, which steps it performs, and version of the genome used during data preprocessing are all nontrivial steps that should be well-documented, which is essential for the repeatability.
For more advanced users, an automated pipeline can be used to perform the steps of fastQ file quality check, adapter trimming and mapping, e.g. the ARMOR snakemake workflow33 or the ‘RNA-seq’ Snakemake workflow from the van Heeringen lab34. However, these completely automated pipelines are less transparent and more difficult to change. When using these tools, it is vital to understand the functions within these automated pipelines. Finally, the protocol includes RNA-seq data analysis with an emphasis on variable gene expression over time. Variable gene expression is preferred over pairwise differential testing when looking at processes with multiple timepoints, such as differentiation. In conclusion, this analysis pipeline introduces relevant tools for bioinformatic analysis of RNAseq data. It contains tools that can thoroughly assess keratinocyte differentiation, both under normal and diseased conditions.
The authors have nothing to disclose.
This research was supported by Netherlands Organisation for Scientific Research (NWO/ALW/MEERVOUD/836.12.010, H.Z.) (NWO/ALW/Open Competition/ALWOP 376, H.Z., J.G.A.S.); Radboud University fellowship (H.Z.); and Chinese Scholarship Council grant 201406330059 (J.Q.).
Bioanalyzer 2100 | Agilent | G2929BA | |
Bovine pituitary extract (BPE) | Lonza | Part of the bulletKit | |
CFX96 Real-Time system | Bio-Rad | qPCR machine | |
Dulbecco's Phosphate-Buffered Saline (DPBS) | Sigma-Aldrich | D8537 | |
Epidermal Growth Factor (EGF) | Lonza | Part of the bulletKit | |
Ethanolamine >= 98% | Sigma-Aldrich | E9508 | |
High Sensitivity DNA chips | Agilent | 5067-4626 | |
Hydrocortison | Lonza | Part of the bulletKit | |
Insulin | Lonza | Part of the bulletKit | |
iQ SYBR Green Kit | BioRad | 170-8886 | |
iScript cDNA synthesis | Bio rad | 1708890 | |
KAPA Library Quant Kit | Roche | 07960255001 | Low concentration measure kit |
KAPA RNA HyperPrep Kit with RiboErase | Roche | KK8540 | RNAseq kit |
KGM Gold Keratinocyte Growth Medium BulletKit | Lonza | 192060 | |
Nanodrop | deNovix | DS-11 FX (model) | Nanodrop and Qbit for DNA and RNA measurements |
NEXTflex DNA barcodes -24 | Illumnia | NOVA-514103 | 6 bp long primers |
Penicillin-Streptomycin | Gibco | 15140122 | |
RNA Pico Chip | Agilent | 5067-1513 |