This paper investigated the effects of sampling methods (surgical resection versus biopsy or cytology), tumor specimen heterogeneity (74-93% versus 37-73% and 17-35%), fixation delays (0 versus 24 and 48 h at room temperature), preservative (FFPE versus frozen), RNA degradation (intact versus degraded or heavily degraded), library preparation kit (9 kits), amplification kit (4 kits), RNA quantity (1-500 ng), detection platforms (two Illumina platforms), and site of laboratory analysis (3 sites) on RNA expression in tumor specimens using data from more than 800 paired specimens included in public datasets (Gene Expression Omnibus, GEO; The Cancer Genome Atlas, TCGA; and Sequence Read Archive, SRA).
Exploring the influence of pre-analytical variables on gene expression measurements and relative expression orderings in cancer research.
Tian T, Lai G, He M, Liu X, Cai H, Guo Y, Hong G, Li H, Song K, Luo Y
Purpose
Conclusion
While 1,000-10,388 RNAs were differentially expressed between case-matched biopsy and resection specimens, cytological and resection specimens, specimens with high (74-93%) versus low (14-35% or 14-73%) tumor content, specimens fixed immediately and those fixed after a 24 or 48 h delay at room temperature, FFPE and frozen specimens, intact and degraded RNA, specimens analyzed at different facilities, most library preparation kits, and between amplification kits, the relative expression ordering for gene pairs (REO) remained very high (>85%). Generally, the REO consistency scores were high among the input volumes evaluated, but REO consistency scores were lower (75-85%) when 1 or 2 ng RNA was used rather than 20 or 100 ng, and REO consistency was even lower when 250 pg was used compared to 200 ng. REO consistency was also lower when different analytical platforms were compared (Illumina Beadchip versus HiSeq platform), when total RNA libraries were compared to mRNA-specific libraries, and when specimens with different levels of degradation were used with different library kits. While REO consistency between the Nugen Ovation FFPE WTA System Kit and Affymetrix IVT Kits was high for diffuse large B-cell lymphoma specimens (81.81%), it was lower for colon cancer (77.68%). When two or more of the investigated variables changed, the average consistency scores remained >70%. Analysis of all data found that the average percentage of differentially expressed RNAs was greater than the average percentage of gene pairs with different REOs for paired samples (P<0.05) and the average percentage of reversal of gene pairs (P<0.05).
Studies (1)
Purpose
This study investigated the effects of sampling methods (surgical resection versus biopsy or cytology), tumor specimen heterogeneity (74-93% versus 37-73% and 17-35%), fixation delays (0 versus 24 and 48 h at room temperature), preservative (FFPE versus frozen), RNA degradation (intact versus degraded or heavily degraded), library preparation kit (9 kits), amplification kit (4 kits), RNA quantity (1-500 ng), detection platforms (two Illumina platforms), and site of laboratory analysis (3 sites) on RNA expression in tumor specimens using data from more than 800 paired specimens included in public datasets (Gene Expression Omnibus, GEO; The Cancer Genome Atlas, TCGA; and Sequence Read Archive, SRA). Collection method effects were investigated using data from 20 paired surgical resection and biopsy specimens of esophageal cancer and 116 paired surgical resection and cytology specimens from breast cancer. Heterogeneity was examined using data from 7 paired gastric cancer specimens, with 74-93% versus 37-73% tumor cells; 7 paired gastric cancer specimens, with 74-93% versus 14-35% tumor cells; and 3 paired colon tumor specimens (details not specified). The effects of delay to fixation were investigated using three paired bone marrow specimens that were fixed immediately, fixed after 24 h and fixed after 48 h at room temperature. Effect of formalin fixation and paraffin embedding were investigated using paired formalin-fixed, paraffin-embedded (FFPE) and frozen specimens of 8 lung cancers, 59 diffuse large B-cell lymphomas and four different sets of breast cancer specimens that consisted of 6, 9, 8, and 16 specimens, respectively. Effects of RNA degradation were investigated using data from reference RNAs, including 72 matched intact and degraded RNA samples and 48 matched intact and highly degraded (peak size ~200 bp) RNA samples. The effects of library preparation kit was investigated using human reference RNA to compare the mRNA TruSeq Kit versus the SsmRNA TruSeq Kit (8 samples) and the SsTotRNA TruSeq Kit (8 samples); the SsmRNA TruSeq Kit versus the SsTotRNA TruSeq Kit (8 samples); the SsTotRNA RiboZero Kit versus SsTotRNA RiboGone Kit (8 samples); the ssTotRNA TruSeq Kit versus the SsTotRNA RiboGone Kit (8 samples) and the SsTotRNA RiboZero Kit (8 samples); the TruSeq Kit versus the RNA Access Kit (24 samples); the TruSeq kit versus the RiboZero Kit (30 samples); the RiboZero Kit versus the RNA Access Kit (36 samples), the RiboZero Kit versus the TruSeq Kit (12 samples); and the RNA Access Kit versus the RiboZero Kit (12 samples). Poly-A enriched libraries and ribo-depleted libraries (16 samples) were compared using RNA from five breast cancer specimens and the TruSeq Kit versus the RiboZero Kit. The effects of amplification kit choice were investigated using RNA from 59 diffuse large B-cell lymphoma specimens and by comparing the Affymetrix IVT Kit and Nugen WT-Ovation System; RNA from 8 colon cancer specimens to compare the Nugen Ovation WTA System and the Affymetrix IVT Kit; and RNA from 8 breast cancer specimens to compare the Nugen Ovation WTA System and the Affymetrix Sensation Kit. The effects of RNA input amount were extrapolated by comparing the following RNA quantities: 130 pg, 250 pg, 750 pg, 1 ng, 2 ng, 5 ng, 10 ng, 20 ng, 100 ng, 200 ng, and 1 µg using 4-24 specimens or reference RNA; 200 ng and 250 pg, 1 ng, 5 ng, 10 ng, 50 ng of 4 ovarian cancer RNAs were compared; between 200 ng and 50 ng RNA from 8 ovarian cancer RNAs were compared; and 100 and 500 ng of RNA from 8 breast cancer specimens were compared. The results obtained using the Illumina HiSeq 2000 and the Illumina HumanHT-12 V4.0 expression Beadchip assays were compared using 55 breast cancer specimens. Additionally, expression profiles of 6 reference RNA specimens that were obtained at three different sites were compared. The consistency of the ordering based on relative expression of genes between pairs was assessed.
Summary of Findings
An average of 3,286 RNAs were differentially expressed (≥2-fold change) between case-matched biopsy and surgical resection specimens; however, the average consistency of REO was 86%, and this increased to 89.90% when only the top 90% of gene pairs were included. Among the 116 paired cytological and surgical resection samples of breast cancer, the average consistency of REO was also approximately 85% and increased slightly when only the top 90% of gene pairs were included. An average of 5,707 genes had ≥2-fold change in expression values when specimens with high (93%–74%) and low (73%–14% and 35%–14%) tumor content were compared. The consistency of the REO between case-matched specimens with high and low tumor content was 89.24% and increased to 92.46% when only the top 90% of gene pairs were included. Similarly, the consistency of the REO among two different laboratory sites that analyzed the same three colon specimens was approximately 85%. A total of 2,113 and 2,970 RNAs were differentially expressed between the six immediately fixed bone marrow specimens and the case-matched specimens fixed after a 24 and 48 h room temperature delay, respectively, but the consistency of the REOs was 88.94% and 85.63%, respectively; this increased to 92.27% and 88.84%, respectively, for the top 90% of genes. An average of 5,009 and 10,388 mRNAs were differentially expressed between paired FFPE and frozen lung and breast cancer specimens, respectively. The consistency of the REOs between matched FFPE and frozen specimens was 86.42% for lung tumors and 84.64% for breast cancer; REOs were also high (>75%) for FFPE and frozen diffuse large B-cell lymphoma. An average of 5,180 and 3,743 RNAs were differentially expressed between intact reference RNA and matched degraded and highly degraded reference RNA, respectively. The consistency of the REOs was higher for matched intact and degraded reference RNA than matched intact and highly degraded reference RNA (89.50% and 87.25%, P<0.05). Using reference RNA, the number of RNAs that were differentially expressed between high input (20 or 100 ng) and low input (1-20 ng) samples ranged from 3,620 to 5,730 with REO consistency scores of 77.57%. 80.42%, 83.52%, 91.95% and 93.22% for the 100 ng and the 1, 2, 5, 10, and 20 ng inputs, respectively and 78.65%, 84.72%, 94.15% and 94.54% for the 20 ng and 1, 2, 5 and 10 ng inputs, respectively. Using ovarian cancer RNA, average REO consistency scores were >82.5% between 200 ng and 5, 10 or 50 ng, >80% for 200 ng and 1 ng, but <75% for 250 pg and 200 ng. An average of 9,833 RNAs were differentially expressed between the Illumina Beadchip and HiSeq platforms with an average REO consistency score of 73.28%. Although an average of 5,983-6,062 RNAs were differentially expressed when samples were analyzed at different sites, the average REO consistency score was 80.59-80.64%. Among the six different pair assessments of library preparation methods, an average of 5,656 genes were differentially expressed, with average REO consistency scores of 82.21%; however, scores were lower (<80%) when libraries for total RNA were compared to mRNA-specific libraries and when specimens with different levels of degradation were used. When amplification was with the Nugen Ovation FFPE WTA System Kit rather than the Affymetrix IVT Kits, 315 RNAs were differentially expressed in diffuse large B-cell lymphoma specimens and 1,256 RNAs were differentially expressed in colon cancer specimens with stable REOs for 81.81% and 77.68% of gene pairs, respectively. When two or more variables were changed, average consistency scores remained >70%. When all data was considered, the average percentage of differentially expressed RNAs was greater than the average percentage of gene pairs with different REO for paired samples (P<0.05) and average percentage reversal of gene pairs (P<0.05).
Biospecimens
TissueBone MarrowTissueBreastTissueColorectalTissueEsophagusTissueLymph NodeTissueLungTissueStomachTissueOvaryPreservation Types
FrozenFormalinDiagnoses
NeoplasticCarcinomaNeoplasticLymphomaPlatforms
RNANext generation sequencingRNADNA microarrayPre-analytical Factors
Classification Pre-Analytical Factor Value(s) Biospecimen Acquisition Cold ischemia time 0 h24 h48 hBiospecimen Acquisition Method of tissue acquisition AspirateBiopsySurgical resectionBiospecimen Aliquots and Components Biospecimen heterogeneity A range of tumor cell contents examinedMultiple specimens analyzedBiospecimen Preservation Type of fixation/preservation Formalin (buffered)FrozenAnalyte Extraction and Purification Between site variability Multiple processing centers comparedNext generation sequencing Specific Nucleic acid amplification Affymetrix IVT kitNugen WT-Ovation SystemNext generation sequencing Specific Technology platform Different library kits (mRNA TruSeq kit, SsmRNA TruSeq kit, SsTotRNA TruSeq, SsTotRNA RiboZero kit, SsTotRNA RiboGone kit, RNA Access kit, Poly-A enriched libraries, Ribo-depleted libraries)Illumina HiSeq 2000 versus Illumina HumanHT-12 V4.0 expression BeadchipNext generation sequencing Specific Template modification Degraded RNAHighly degraded RNAIntact RNANext generation sequencing Specific Template/input amount 1 ng1 µg10 ng100 ng130 pg2 ng20 ng200 ng250 pg5 ng500 ng750 pg