Figure 1.
“Matching” DNA fragment size distributions are necessary for optimal aCGH data.
(A,C,E,G) Agarose gel electrophoresis images and ImageJ gel intensity analysis plots of reference gDNA (Promega) after heat fragmentation. Mode fragment size is indicated in blue (bp) relative to DNA ladder. Heat times were adjusted to produce four mode fragment size combinations (225/225, 525/225, 525/140, 225/140). (B,D,F,H) Plot of results from chromosome 1 following self-hybridization of specific combinations of mode size. Differentially labeled aliquots (cy5/cy3) were coded as follows: green;log2ratio<-0.3, black;-0.3≤log2ratio≤0.3, red;log2ratio>0.3. Data quality was assessed by dLRsd on Agilent 180 K arrays.
Figure 2.
Determination of optimal size among matched DNA fragment size distributions.
(A,C,E,G) Agarose gel electrophoresis of reference gDNA (Promega) aliquots after various heat fragmentation times shown adjacent to ImageJ gel analysis of same lanes, molecular weight indicated in bp. Mode fragment size of each smear, as measured with ImageJ, indicated in blue. (B,D,F,H) Agilent 180 K array results of self-hybridizations using reference gDNA (left) and characterized by matching fragment size distributions (B;250/250, D;315/315, F;400/400, H;525/525). Log2 ratios for signal intensities of differentially labeled aliquots (cy5/cy3) are plotted for probes corresponding to chromosome 1 (green;log2ratio<−0.3, black;-0.3≤log2ratio≤0.3, red;log2ratio>0.3). Data quality was assessed by dLRsd. (I) Mean dLRsd of duplicate (n = 5) or triplicate (n = 2) size-matched self-hybridizations representing seven fragment size distributions plotted by mode fragment length (225, 250, 315, 400, 525, 625, and 680 bp). Error bars indicate SEM.
Figure 3.
DNA fragmentation and thermodegradation are unpredictably variable.
(A) Gel electrophoresis image of DNA extracted from 22 FFPE tissue specimens stored in paraffin from one to 13 years. (B) Mode fragment size of samples in (A) plotted by age of paraffin block, linear regression of data indicated by dashed line. (C) Gel electrophoresis image of DNA from six FFPE specimens intact prior to labeling (i), after ULS labeling only (0), or after ULS labeling plus 1 min heat fragmentation (1). (D) Mode fragment size of lanes marked 0 and 1 plotted for the six FFPE samples from the gel shown in (C). (E) Gel electrophoresis image of DNA from three frozen specimens with i, 0, and 1 indicating same conditions as in (C), and another samples after ULS labeling conditions plus 2 min heat fragmentation (2). (F) Plot of mode fragment size for lanes marked 0, 1, and 2 plotted for the three frozen samples in (E).
Figure 4.
A Fragmentation Simulation Method (FSM) enables accurate prediction and precise control of labeled DNA fragment sizes.
(A–F) Vertical axes indicate DNA bp. (A,D) Gel image of DNA from three FFPE specimens (A) or three frozen specimens (D) either intact, (i), after ULS labeling conditions only, (0), or ULS labeling conditions and 0.5, 1, 2, 4, 6, or eight minutes heat fragmentation (0.5, 1, 2, 4, 6, 8). (B,E) Utilizing mode fragment size of lanes in (A) or (D) as data points, FSM regression curves fit to data from each sample. Intersection with target size (dashed line) reveals FSM prediction for optimal time of heat fragmentation for each sample. (C,F) Agarose gel electrophoresis of samples in (A) or (D) after heat fragmentation for time predicted by FSM in (B) or (E) and ULS labeling conditions, shown adjacent to ImageJ gel analysis of same lanes. The mode fragment size of each smear, as measured with ImageJ, is indicated by arrows and solid horizontal lines.
Figure 5.
Application of FSM ULS method to FFPE samples creates equivalent results to those from fresh-frozen samples.
(A) Plot showing dLRsd for 122 FFPE tumor specimens processed according to either standard ULS or FSM ULS protocols and analyzed on Agilent 1 M arrays. (B) Data quality (dLRsd) from (A) plotted by FFPE block age and method. Dashed lines indicate linear regression. Statistics indicate magnitude and significance of correlation between block age and aCGH data quality. (C) Quality (dLRsd) of Agilent 1 M aCGH data of 78 fresh-frozen tissue specimens or frozen tumorsphere cell cultures processed according to either standard ULS or FSM ULS protocols. (D) FFPE and Frozen FSM ULS subsets from (A) and (C) compared to 206 fresh-frozen GBM specimens analyzed on Agilent 244 k arrays from the glioblastoma TCGA study. Statistical significance was assessed by t test and ANOVA, (****;p<.0001, ns;p>0.05), and error bars indicate mean and standard deviation. Additional QC metrics data for all samples are provided in Table S2.
Figure 6.
Size matching using FSM is a more critical determinant of array quality than other known variables.
(A–H) Probe log2 ratio (signal intensity test DNA/signal intensity reference DNA) data plotted for a single chromosome (chr.13 or chr.1) from eight Agilent 1 M arrays (green;log2ratio<-0.3, black;-0.3≤log2ratio≤0.3, red;log2ratio>0.3). (A–C) Chromosome 13 plotted log2 ratios are representative profiles of three Agilent 1 M arrays of a single FFPE GBM specimen (GBM1) processed with the FSM ULS protocol (A), standard ULS protocol (B), or FSM ULS protocol after altered proteinase K digestion during DNA extraction (C) (plotted log2 ratio data for all chromosomes provided in Figure S2). (D–H) Chromosome 1 plotted log2 ratios are representative profiles of five Agilent 1 M arrays of a single FFPE GBM specimen (GBM2) processed using the FSM ULS protocol, with reduced DNA input in (E) and (F) (see Figure S3 and Figure S4 for detailed copy number analysis). Increased hybridization time (G) improved quality to a modest degree. Use of FFPE brain tissue as reference DNA (H) did not significantly improve results (dLRsd of 0.21 vs. 0.20 for standard reference).
Figure 7.
Overview of proposed methods for FSM ULS processing of FFPE specimens.
Summary of methods and timeline for aCGH using the FSM method. Following DNA extraction, the workflow and protocol for preparation of fresh or frozen samples is identical to FFPE workflow shown.
Figure 8.
Predicted hierarchy of known variables contributing to aCGH data quality.