Transcriptional profiling of human dendritic cell populations and models--unique profiles of in vitro dendritic cells and implications on functionality and applicability

Abstract

Background: Dendritic cells (DCs) comprise heterogeneous populations of cells, which act as central orchestrators of the immune response. Applicability of primary DCs is restricted due to their scarcity and therefore DC models are commonly employed in DC-based immunotherapy strategies and in vitro tests assessing DC function. However, the interrelationship between the individual in vitro DC models and their relative resemblance to specific primary DC populations remain elusive.

Objective: To describe and assess functionality and applicability of the available in vitro DC models by using a genome-wide transcriptional approach.

Methods: Transcriptional profiling was performed with four commonly used in vitro DC models (MUTZ-3-DCs, monocyte-derived DCs, CD34-derived DCs and Langerhans cells (LCs)) and nine primary DC populations (dermal DCs, LCs, blood and tonsillar CD123(+), CD1c(+) and CD141(+) DCs, and blood CD16(+) DCs).

Results: Principal Component Analysis showed that transcriptional profiles of each in vitro DC model most closely resembled CD1c(+) and CD141(+) tonsillar myeloid DCs (mDCs) among primary DC populations. Thus, additional differentiation factors may be required to generate model DCs that more closely resemble other primary DC populations. Also, no model DC stood out in terms of primary DC resemblance. Nevertheless, hierarchical clustering showed clusters of differentially expressed genes among individual DC models as well as primary DC populations. Furthermore, model DCs were shown to differentially express immunologically relevant transcripts and transcriptional signatures identified for each model DC included several immune-associated transcripts.

Conclusion: The unique transcriptional profiles of in vitro DC models suggest distinct functionality in immune applications. The presented results will aid in the selection of an appropriate DC model for in vitro assays and assist development of DC-based immunotherapy.

Figure 1. Resemblance of in vitro DC models with skin, tonsillar and peripheral blood DC populations.

Resemblance demonstrated by principal component analysis (PCA) of expressed transcripts (A. 51,191) and differentially expressed transcripts (B, 18,590) identified by ANOVA p<10⁻⁶ (corresponding to p<0,05 after Bonferroni correction). Replicate similarities visualized using k-Nearest Neighbors (k-NN) algorithm (k = 2 in this case) and relationships between cell types demonstrated by minimal spanning tree analysis (lines connecting the different populations). Axes (marked 1, 2 and 3) correspond to the three main components in the PCA analysis and numbers in brackets correspond to percentage of total variation contained within each component. C) Heatmap visualizing gene expression profiles of differentially expressed genes (18,590) upon hierarchical clustering with complete linkage and Euclidean measure. Colors represent high (red) and low (green) normalized intensity.

Figure 2. Heatmaps visualizing gene expression profiles of in vitro DC models and ex vivo DC populations.

Hierarchical clustering on differentially expressed transcripts (ANOVA p<10⁻⁶, corresponding to p<0,05 upon Bonferroni correction) among in vitro DCs (A. 892 transcripts) and ex vivo DCs (B. 9,055 transcripts), using complete linkage and Euclidean measure. Colors represent high (red) and low (green) normalized intensity, respectively.

Figure 3. Hierarchical clustering of in vitro DCs.

Clustering using complete linkage algorithm on differentially expressed transcripts (892), identified by ANOVA (p<10⁻⁶, corresponding to p<0,05 upon Bonferroni correction), demonstrates relationships among in vitro DC models.

Figure 4. Transcriptional signatures of individual in vitro DC subsets.

Signatures identified by expression level >200 and differential expression in one model as compared to all other in vitro DC models (based on fold difference >2 and statistical significance p<0,05; student's T-test). Comparisons performed on MAS5-normalized data and expression ratio calculated on average of replicates as described in Methods. Selected immununologically associated transcripts are highlighted in red. Transcripts lacking official gene symbols are identified with respective Affymetrix Probe Set ID.