NIH, National Cancer Institute, Division of Cancer Treatment and Diagnosis (DCTD) NIH - National Institutes of Health National Cancer Institute DCTD - Division of Cancer Treatment and Diagnosis

Assessment of stained direct cytology smears of breast cancer for whole transcriptome and targeted messenger RNA sequencing.

Author(s): Marczyk M, Fu C, Lau R, Du L, Trevarton AJ, Sinn BV, Gould RE, Pusztai L, Hatzis C, Symmans WF

Publication: Cancer Cytopathol, 2023, Vol. 131, Page 289-299

PubMed ID: 36650408 PubMed Review Paper? No

Purpose of Paper

This paper compared RNA integrity, whole transcriptome RNA sequencing (wtRNA-Seq), and targeted RNA sequencing results from RNAlater-preserved frozen breast tumor, formalin-fixed paraffin-embedded (FFPE) breast tumor, and cell smears obtained from the same breast tumor post-resection; breast tumor specimens were fixed in Carnoy's fixative or 95% ethanol and stained with Papanicolaou stain or air dried and then fixed and stained with DiffQuik. The correlation between RNA integrity and specimen storage duration was also evaluated for all specimen types.

Conclusion of Paper

As expected, RNA extracted from the FFPE specimen and the cytology smears (fixed in Carnoy’s or ethanol and stained with DiffQuick) had significantly lower DV200 and RIN values than were obtained from frozen specimens. The library yield from RNA extracted from DIffQuick-stained smears was not sufficient for sequencing. The transcript integrity number (TIN), a measure of uniformity, was only slightly albeit significantly higher for RNA extracted from FFPE specimens and Carnoy’s-fixed and ethanol-fixed cytology slides than RNA from frozen specimens. For all cytology smears, the storage duration (15-528 days) was modestly and negatively correlated with RIN and DV200 (r=-0.58 to -0.71), but no correlation was observed for frozen (12-250 days) or FFPE specimens (27-296 days). The percentage of reads mapping to an intron was higher for FFPE and cytology smears than frozen specimens. However, specimens clustered by patient source rather than specimen type/preservation method, and gene expression was very strongly to strongly correlated between frozen and FFPE specimens or cytology smears. The concordance correlation coefficient (CCC) across all genes was higher when the frozen and FFPE specimen were compared than between the frozen and Carnoy’s- or ethanol-fixed cytology smears. Cytosmears had lower stromal scores than frozen and FFPE specimens. The scores for all nine breast cancer signatures using wtRNA-Seq were generally concordant (moderately to strongly) when the frozen specimen was compared to the other specimen types, but the strength of the CCC was dependent on both the signature and specimen type. The highest CCC for all three specimen types when compared to the frozen was observed with the sensitivity to endocrine therapy index (SETER/PR, mean CCC=0.935) signature followed by the PIK3CA activation prediction signature (PI3Kges, mean CCC=0.904), but the ranking of the other signatures were dependent on specimen type.  The scores for the two signatures included in the targeted sequencing data (PI3Kges and SETER/PR) were strongly correlated between frozen specimens and the other specimen types evaluated. The variant allele frequency was strongly correlated between frozen specimens and the matched FFPE specimen, ethanol-fixed cytospin specimens, and Carnoy’s fixed cytospin specimen by wtRNA-Seq and targeted sequencing (6 variants). Concordance of allele frequencies between sequencing platforms was dependent on allele fraction with higher concordances observed for mutations with high allele frequencies than low allele frequencies.

Studies

  1. Study Purpose

    This study compared RNA integrity, wtRNA-Seq, and targeted RNA sequencing results from RNAlater-preserved frozen breast tumor, FFPE breast tumor, and cell smears that were all obtained from the same breast tumor after resection; match specimens were fixed in Carnoy's fixative or 95% ethanol and stained with Papanicolaou stain or air dried and then stained with DiffQuik. A potential correlation between RNA integrity and specimen storage duration was also investigated for all specimen types.  Breast tumor specimens were collected during resection from eleven patients with treatment naïve breast cancer. The cut surface of the resected specimen was scraped with a scalpel and the resulting cell-rich droplets were smeared on glass slides. The smears were then: 1) fixed in Carnoy's fixative and stained with Papanicolaou stain, 2) fixed in 95% ethanol and then stained with Papanicolaou, or 3) air dried and then fixed and stained with DiffQuik. The slides were then stored at room temperature until RNA extraction using a PicoPure RNA Isolation Kit. Multiple samples were obtained from the remaining tumor specimen, diced, mixed, and divided. One pile of tissue was placed in RNAlater and stored at 4°C for 6-72 h before freezing at -80°C and the other was fixed in 10% neutral buffered formalin overnight before paraffin embedding and block storage at 4°C.  RNA was extracted from the frozen tumor specimens using the Qiagen RNeasy Kit and from xylene deparaffinized unstained 5 µm thick sections using the Norgen FFPE RNA Purification Kit. RNA was quantified by NanoDrop and integrity was assessed using a bioanalyzer. Whole transcriptome sequencing libraries were prepared using the RNA HyperPrep Kit with RiboErase and sequenced using the Illumina HiSeq 4000. Targeted RNA-sequence libraries were prepared by incorporating unique molecular identifiers and sequenced using an Illumina Miseq platform.  WT-Seq reads were quality assessed using FastqQC v 0.11.5, adapters were removed, and aligned to the human genome using STAR v2.5.3a. Only reads in exonic regions were used to calculate gene expression. Targeted sequencing reads were aligned using blastn and gene expression was counted based on the number of unique molecular identifier per target. A single nucleotide variant (SNV) was counted if the sequencing depth was >10 reads, VAF was >5%, and the read balance was 0.8-1.2. The percentage of stromal cells was calculated from the WT-Seq data using the ESTIMATE algorithm and the Winslow Score. Expression was compared using R to nine different breast cancer signatures: EndoPredict, the 70 genes prognosis profile (GENE70), the 3‐gene expression prognostic index using subtypes (GENIUS), the genomic grade index (GGI), the PIK3CA activation signature (PIK3CAGS), the reduced PIK3CA activation prediction signature (PI3Kges), the recurrence score (RS), the sensitivity to endocrine therapy index (SETER/PR), and the tamoxifen resistance signature (TAMR).

    Summary of Findings:

    As expected, RNA extracted from the FFPE specimen and the cytology smears (Carnoy’s-fixed, ethanol-fixed, and DiffQuick) had significantly lower DV200 and RIN values than were obtained from frozen specimens (DV200: 74-80 versus 91, P≤ 8.43E-07; RIN: 1.8-2.4 versus 7.2, P≤ 7.67E-18). The library yield from RNA extracted from DIffQuick-fixed smears was not sufficient for sequencing. The transcript integrity number (TIN), a measure of uniformity, was slightly albeit significantly higher in RNA extracted from FFPE specimens, and Carnoy’s-fixed and ethanol-fixed cytology slides than in RNA from frozen specimens (P≤2.23E-05). For all cytology smears, the storage duration (15-528 days) was modestly and negatively correlated with RIN and DV200 (r=-0.58 to -0.71); storage was not correlated with RIN or DV200 in frozen (12-250 days) or FFPE specimens (27-296 days). The percentage of reads mapping to an intron was higher for FFPE and cytology smears than frozen specimens. However, after normalization, gene expression metrics were comparable across specimen types and specimens clustered by patient source rather than specimen type/preservation method. Further, gene expression was very strongly correlated between frozen and FFPE specimens (r=0.91) and strongly correlated between frozen specimens and cytology smears (r≥0.88). The concordance correlation coefficient (CCC) across all genes was higher between the frozen and FFPE specimen (0.627) than between the frozen and Carnoy’s- or ethanol-fixed smears (0.513 and 0.499, respectively), which the authors attributed to lower correlation coefficients and a lower bias coefficient between frozen and cytology specimens (R=0.740-0.747, and 0.818, respectively) than frozen and FFPE specimens (R=0.831 and 0.818, respectively). The CCC values were generally very strongly correlated between Carnoy’s- and ethanol-fixed smears (r=0.87), but there was a pool of genes with higher CCC between smears and FFPE specimens. In pathways analysis, 35 of the 185 pathways were enriched for genes that had a higher CCC between FFPE specimens and smears, which included immune‐related pathways, immune‐related disease pathways, and cancer and drug metabolism pathways. Cytosmears had lower stromal scores than frozen and FFPE specimens.  The scores for all nine breast cancer signatures using wt-SEQ were generally concordant (moderately to strongly) between the frozen specimen and the other specimen types, but the strength of the CCC was dependent on both the signature and specimen type. The highest CCC for all three specimen types with the frozen specimen was observed with the sensitivity to endocrine therapy index (SETER/PR, mean 0.935) signature followed by the PIK3CA activation prediction signature (PI3Kges mean 0.904); the ranking of the other signatures was dependent on specimen type.  The scores for the two signatures included in the targeted sequencing data (PI3Kges and SETER/PR) were strongly correlated between frozen specimens and matched FFPE specimens (0.90 and 0.93), ethanol-fixed cytospins (0.86 and 0.76), or Carnoy’s-fixed cytospins (0.88 and 0.80). The variant allele frequency in wtSeq data was strongly correlated between frozen specimens and matched FFPE specimens, ethanol-fixed cytospins, and Carnoy’s-fixed cytospins (CCC=0.98, all). Only six of the identified variants were found by targeted sequencing and frequencies were strongly correlated between frozen specimens and matched FFPE specimens, ethanol-fixed cytospins, and Carnoy’s-fixed cytospins (r=0.89, r=0.95, and r=0.91, respectively). Concordance of allele frequencies between platforms was dependent on allele fraction, with higher concordance observed for mutations with high allele frequencies than low allele frequencies.

    Biospecimens
    Preservative Types
    • Formalin
    • Ethanol
    • RNAlater
    • Other Preservative
    • Frozen
    Diagnoses:
    • Neoplastic - Carcinoma
    Platform:
    AnalyteTechnology Platform
    RNA Next generation sequencing
    RNA Automated electrophoresis/Bioanalyzer
    Pre-analytical Factors:
    ClassificationPre-analytical FactorValue(s)
    Biospecimen Preservation Type of fixation/preservation Formalin (buffered)
    Ethanol
    Frozen
    RNAlater
    Carnoy's solution
    Air-dried
    Storage Storage duration 15-528 days (Cytospins)
    12-250 days (frozen at -80°C)
    27-296 days (FFPE at 4°C)
    Next generation sequencing Specific Technology platform Targeted RNA sequencing,
    wtRNA sequencing
    Next generation sequencing Specific Type of tissue stain Papanicolaou stained cytospin
    DiffQuick Stained cytospin
    Unstained FFPE section
    Unstained frozen specimen
    Biospecimen Aliquots and Components Biospecimen components Tumor pieces
    Tumor scraping

You Recently Viewed  

News and Announcements

  • New Articles on the GTEx Project are Now FREELY Available!

  • Just Published!

  • April 24, 2024: Biobanking for Precision Medicine Seminar

  • More...