Fig 1.
Experimental groups illustrated.
A single vial of P0 hMSCs were expanded at low density to greater than 25 million cells at P1. These cells were trypsinized and a portion of them replated at 500,000 cells / 75cm2 flask for 2D samples. Another portion of those cells was encapsulated in poly(ethylene glycol) diacrylate (MW 4000 Da). Two encapsulation methods were used, radical polymerization with ammonium persulfate (APS) and tetramethylethylenediamine (TEMED), and conjugate addition with a pentaerythritol tetrakis(3-mercaptopropionate) (PETMP) crosslinker. Several samples of each type were created and half from each type were irradiated with a Black Ray UV bench lamp, peak wavelength 365nm.
Fig 2.
Principle components analysis shows tightly segregated clustering based on culture condition, not by UV exposure.
(alternate viewing angle and axis values available in S2 Fig). PCA is a statistical analysis tool which reduces the dimensionality of data by determining the key variables resulting in differences seen between samples[25]. Each axis of this PCA map represents a linear combination of expression levels from many thousands of gene transcripts such that, combined, the maximum variation among all data points is achieved on only three axes. The result gives a visual representation of which samples behave similarly to each other by their physical closeness in three dimensions, while including information from many thousands of variables (gene expression levels). This is a bird’s eye view of the entire set of gene array data, for which absolute units and values can be considered arbitrary. The following in-depth analysis of pathway enrichment and specific gene expression provide insight into how each group differs from the others in their gene expression.
Fig 3.
UV effects in 2D groups are insignificant.
(a) This Venn diagram shows that whether combining raw data across months or using repetition power between months, no significant pathways or biological functions are found in the differentially expressed genes from the Kyoto Encyclopedia of Genes and Genomes (KEGG) or Gene Ontology (GO) except for a slight change in cell cycle. (b) The most extreme gene changes for each comparison have relatively weak fold changes, and a heat map (unsupervised hierarchical clustering) of the 148 gene transcripts (103 unique genes) differentially expressed between 2D2UV vs. 2D2 shows that none of the other UV comparisons in 3D have any differences in expression level for these genes. (red is upregulated, blue is downregulated, genes included in the heat map are those differentially expressed between denoted * groups) Also of note is the stark similarity between samples 2D1 and 2D2. The 3D groups look very similar to each other, while they are very different from the 2D groups. The most repeated changes of significance were in the genes DHRS3 and FOXQ1. Their individual scatter plots, (c) and (d), reinforce the similarities between the months as well as the direction of change. However, for so few genes and so small a change, the effects could easily be found by chance. (RMA = robust multi-array average).
Fig 4.
UV effects in 3D are insignificant, but environmental effects are dramatic.
(a) Venn diagram showing overlapping genes between cell cycle differences and supposed UV effects in 3D. UV exposure in the radically polymerized gels results in 352 unique differentially expressed genes, which are enriched for two KEGG pathways, p53 and cell cycle, each with 4.5% of the participating genes in those pathways. These genes overlap with 50% of the 2D control genes and 17 regular cell cycle genes described by others [37]. (b) Venn diagram showing similarity of gene lists from inclusion or exclusion of UV samples when comparing 3D polymerization. Purely comparing polymerization mechanisms (3DR vs. 3DC) yields 2423 unique genes. Making the equivalent comparison in UV (3DRUV vs. 3DC UV) yields 1652 genes. These two comparisons share 1221 genes, all included in the 1847 genes identified by inclusion of UV samples. (c) The heat map of 507 differentially expressed genes between 3DRUV vs. 3DR (* in (a) and (c)) (unsupervised hierarchical clustering). The 3DR samples were completely separate from the other 3D groups, while all the samples for the 3DC conditions were co-mingled regardless of UV condition. No differences in these genes were visible between any of the 2D conditions. (d) The heat map of 3030 differentially expressed genes between 3DR±UV vs. 3DC±UV (* in (b) and (d)) (unsupervised hierarchical clustering) reveals stark differences between polymerization conditions. 2D samples are not different for these genes. 3D samples within polymerization methods co-mingle without regard for UV exposure.
Table 1.
IPA summary for 3DR±UV vs. 3DC±UV, (>2fold, p<0.05).
Fig 5.
The top 20 canonical pathways identified from the 3DR±UV vs. 3DC±UV comparison ranked by p-value.
Red dotted line marks p-value threshold,-log(0.05) = 1.3. The ratio of genes from our gene list found in the pathway to the total number of genes in the pathway is given as ratio x 10, and marked as the orange squares and line, and fluctuate between 15% and 25%.