Figure 1.
5 blood samples were drawn from each of 8 patients and 7 controls were extracted by three methods: LEUKOLOCK, LL; PAXGENE, PAX(x2); TEMPUS, TEM(x2). Samples were extracted separately and not pooled. Additional aliquots of RNA extracted from PAX and TEM samples were depleted for globin RNA using GlobinClear. The resulting 75 samples were amplified by IVT and run on Affymetrix U133 Plus 2.0 whole human genome arrays and analysed with MAS5.0, GCRMA or with the R package PUMA.
Figure 2.
Representative Bioanalyzer electropherograms after initial RNA extractions.
High quality RNA (RIN >7.0) can be extracted from all three methods. PAXGENE however requires a DNase step as traces from initial extractions show genomic DNA contamination at high molecular weights and as a “shoulder” to the 28S peak (indicated by asterisks). Good quality RNA can be detected after globin depletion with both TEM and PAXGENE. However, PAXGENE samples showed consistently lower concentration levels, as indicated by the smaller 18S and 28S peaks (FU, fluorescent units). Traces are representative and from single samples from each extraction method.
Figure 3.
RNA quality and quantity, pre and post in vitro transcription.
Diagrams showing averages of all samples in the extent of yield from the 5 experimental conditions (with standard deviation) for (A) RNA extracted, and (C) aRNA produced post IVT, as well as (B) average RIN values of extracted RNA.
Figure 4.
Representative Bioanalyzer electropherograms after IVT and labelling, pre- and post- fragmentation.
After IVT, the aRNA profiles from each method showed distinctly different profiles. Generally, IVT resulted in a wide range of detectable products. However in the cases of TEM and PAX especially, distinct high intensity peaks were observed in all cases (black arrows). These disappeared in all cases in the globin mRNA depleted samples. After fragmentation, aRNA profiles were indistinguishable from each other (FU, fluorescent units). Traces are representative and from single samples from each extraction method.
Figure 5.
Initial quality control for the 75 microarrays.
Analysis of array data of all 75 samples run was carried out in Affymetrix Expression Console. PolyA (A) and Hyb (B) controls show that the preparation steps were consistent, and that in general all hybridisations were consistent, but that hybridisation of PAX samples differed from the rest. The % present call (C) values were generally consistent in LL, TEM GC and PAX GC samples, but were clearly lower with TEM and PAX gene samples. The scale factor between chips were generally low, except for PAX hybridisations (D) while the relative log expression (E) of all chips were generally similar with a few exceptions, and a slight general increase in PAX samples (FU, fluorescent units).
Table 1.
Selected QC metrics for LL, PAX GC, and TEM GC arrays.
Table 2.
Selected QC metrics for PAX and TEM arrays.
Figure 6.
Quality control: globin levels in arrays.
Data were GCRMA normalised and levels of 3 representative alpha globin probe sets (A, B and C) and one beta probe set (D) were averaged for each condition. Expression level box plots indicate that in both the cases of alpha and beta globin, TEM and PAX exhibit highest levels of globin mRNA while TEM GC and PAX GC were the lowest, LL levels were generally lower than PAX and TEM, but higher than PAX GC and TEM GC.
Figure 7.
Number of probe sets called present increases after globin depletion.
MAS5.0 normalised data indicated that the presence of globin affects the number of probes sets called present. Comparison of all probe sets called present in every array for each condition reveals higher numbers in the case of LL, TEM GC and PAX GC (“globin negative”) as compared to TEM and PAX (“globin positive”; A). Comparison between probe sets called present in LL, TEM and PAX reveals 9632 probe sets in common (B), and 15576 probe sets were found to be in common between LL, TEM GC and PAX GC (C). By comparing these 2 populations 6022 probe sets were found to be unmasked by globin depletion (D).
Figure 8.
PCA plotted using QLUCORE OMICS Explorer on GCRMA normalised data: each dot represents an array and takes into account the expression levels of every probe set and shows 2 main clusters (A). One tightly clustering ‘globin negative’ cluster includes the following conditions: LL, TEM GC and PAX GC, while a second more variable cluster “globin positive” includes arrays hybridised with PAX and TEM extracted RNA. Focusing on the globin negative group PUMA analysis identified differentially regulated probe sets (ALS patients vs controls) for the following (B; LL, 3047; TEM GC, 3619; PAX GC, 4511). A Venn diagram comparing these lists shows that TEM and PAX share more genes in common, than either with LL, and 142 genes in common between all 3 methods).
Figure 9.
Real-time validation PCR for genes in LL and TEM GC gene lists.
ΔCTs plotted for 6 genes chosen for validation using the LL method of RNA extraction from blood (A). Four out of 6 genes (CDK1, EGR1, SSPN and IL23) could be validated from the Affymetrix microarray studies (C). ΔCTs plotted for 6 genes chosen for validation using the TEMPUS/GC method of RNA extraction from blood (B). Four out of 6 genes (EP400, E2F2, EHBP1 and KAZ) could be validated from the Affymetrix microarray studies (D). Blue bars represent levels of gene expression in ALS patients (n = 6) and red bars the levels in normal control individuals (n = 6).
Table 3.
Functional annotation clustering carried out in DAVID for differentially regulated genes from a cohort of ALS patients and controls, using RNA extracted from blood with LL and TEM GC undergoing PUMA GEP analysis.
Table 4.
PCR primer sequences designed for genes validated in this study.
Table 5.
Characteristics of subjects taking part in the study.