Expression arrays developed from the AFT024-subtracted library were interrogated with probes derived from stromal cell lines. A k-means clustering algorithm was used to profile the data and informative clusters were selected. The contents of the clusters revealed in the cDNA microarray (StroChip) analysis (clusters 2, 3, 7, 9, 10, 12, 14, 15) were compared to those selected in the custom filter array analysis (clusters 8, 9 ,11, 12, 17, 19). The overlapping gene products are displated as their relative signal to background ratios in red to green coloration for both the StroChip (left panel) and filter arrays (right panel).

Clusters generated from the StroChip experiment.

Clusters generated from the membrane array experiment..

Pearson Correlation Coefficients (R) for all elements, 3932, (in duplicate) on the membrane arrays.

Cell line	AFT024	2012	2018	2058
AFT024	-----	0.630	0.191	0.496
2012	0.630	-------	0.084	0.550
2018	0.191	0.084	-------	0.240
2058	0.496	0.550	0.240	-------

Pearson Correlation Coefficients (R) for all elements, 3601, (average of two hybridizations) on the Stromal Cell Chip.

Cell Line	2012	2018	BFC012	2058
2012	-------	0.893	0.601	0.561
2018	0.893	-------	0.640	0.544
BFC012	0.601	0.640	-------	0.407
2058	0.561	0.544	0.407	-------

Pearson Correlation Coefficients (R) for informative, non-redundant, averaged elements (201) selected from the cluster analysis of the membrane array hybridizations.

Cell line	AFT024	2012	2018	2058
AFT024	-----	0.403	0.016	0.443
2012	0.403	-------	0.084	0.697
2018	0.016	0.084	-------	0.044
2058	0.443	0.697	0.044	-------

Pearson Correlation Coefficients (R) for informative, non-redundant, averaged elements (143) selected from the cluster analysis of the Stromal Cell Chip hybridizations.

Cell Line	2012	2018	BFC012	2058
2012	-------	0.944	0.598	0.777
2018	0.944	-------	0.680	0.738
BFC012	0.598	0.680	-------	0.409
2058	0.777	0.738	0.409	-------

Pearson Correlation Coefficients (R) for the overlapping elements (33) in the final analysis of the membrane array data.

Cell line	AFT024	2012	2018	2058
AFT024	-----	0.317	0.048	0.420
2012	0.317	-------	0.324	0.442
2018	0.048	0.324	-------	0.010
2058	0.420	0.442	0.010	-------

Pearson Correlation Coefficients (R) for the overlapping elements (33) in the final analysis of the Stromal Cell Chip data.

Cell Line	2012	2018	BFC012	2058
2012	-------	0.946	0.415	0.694
2018	0.946	-------	0.543	0.592
BFC012	0.415	0.543	-------	0.107
2058	0.694	0.592	0.107	-------

Total and poly A+ RNA were prepared from stromal cell monolayers using commercial reagents (Life Technologies and Ambion). The quantity and quality of the mRNA was evaluated by Northern blotting (data not shown). The Stromal Cell cDNA Microarray (StroChip) was developed and printed by Incyte Genomics Inc. from non-redundant cDNAs in the AFT024-subtracted library. To determine the non-redundant clone set, individual sequences were compared by BLAST (basic local alignment sequence tool, (Altschul, et al. 1997)) against the mouse and human UniGene databases (http://www.ncbi.nlm.nih.gov/UniGene). Highly significant matches (e < 10^-15) were assembled into the same cluster. For entries with multiple representatives, the one containing the largest sequence was selected. 3600 PCR-amplified DNAs were arrayed on glass slides and correspond to an annotated entry in StroCDB. Controls on the cDNA arrays included housekeeping genes (alpha tubulin, 23 kD HBP, and ribosomal subunit S9), a complex target to measure probe complexity, as well as sensitivity controls and fluorescence intensity controls (doped spots), spotted in triplicate on each chip. Cy3 and Cy5 fluorescent labels were incorporated into cDNA transcribed from RNA templates, hybridized to microarrays, and quantitated. Probe labeling, microarray hybridizations and image quantitation were performed by Incyte Genomics Inc. The stromal cell lines, 2012, 2018, 2058, and BFC012 were each compared to AFT024. Each comparison was performed in two labeling orientations, allowing compensation for different dye incorporation efficiencies and an experimental replicate. The resulting data were processed as follows. Cy5 signal was normalized relative to the Cy3 signal by whole chip balancing to differences in the intensity of each fluorochrome. In order to assess dataset quality, each type of control was grouped and asked, by t-test, if the means of each signal (Cy3 and Cy5, after balancing) were significantly different. In no case was there a significant difference. The data were filtered as log₂ > 1 across all four comparisons and 381 cDNAs passed this filter. log₂ transformed, balanced signals were expressed as a ratio (AFT024 vs. the compared cell line) and averaged. These results were grouped into 18 clusters using a k-means clustering algorithm (adapted by JAH from the cclust R package written by E. Dimitriadou), using Euclidean distance as the similarity metric. Clusters were chosen with patterns of expression that correlated with a biological phenotype, i.e. higher expression in AFT024 compared to known non-supporting stromal cell lines.

Approximately 2000 clones from the subtracted library were arrayed on nylon membranes where each duplicate spot corresponds to an entry in StroCDB. The membranes were hybridized with ³³P-dCTP-labeled subtracted probes produced using the PCR Select technology (Clontech). Probes were generated from AFT024, 2012, 2018, and 2058 where each test line was subtracted with BFC012. We have previously shown that BFC012 cells do not support stem cells in culture (Moore, et al. 1997). Image data were collected by a Molecular Probes Phosphoimager and quantitated by Incyte Genomics Inc. using Arrayvision software. Raw signal/background intensities were log₂ transformed and values for a given probe were averaged. These values were also subjected to the same k-means clustering algorithm. Representative clusters that showed higher expression in AFT024 than 2018 were selected. The log-transformed, averaged data were used to generate red to green pseudo-colored images of the signal/background intensity.

Copyright © 2005 Princeton University