Skip to main content

Long-read RNA sequencing of archival tissues reveals novel genes and transcripts associated with clear cell renal cell carcinoma recurrence and immune evasion.

Lee J, Snell EA, Brown J, Booth CE, Banks RE, Turner DJ, Vasudev NS, Lagos D

Publication: Genome Res (2024) • Vol 34 • Page 1849-1864PMID: 39284687View on PubMed

Purpose

 This paper compared direct RNA sequencing (DRS) and PCR cDNA sequencing (PCS) results in archival frozen clear cell renal cell carcinoma (ccRCC) specimens from patients who relapsed and those who did not.  Importantly, the comparison of DRS and PCS included analysis of known genes, novel genes and alternate transcripts in specimens from patients who relapsed and those who did not and select results were confirmed by real-time PCR in a validation cohort.

Conclusion

PCS generated a higher median number of reads, longer median read length, higher median read accuracy and a longer longest aligned read than DRS.  When analysis was by PCS there were more mapped genes (median of 26,203 versus 18,047), a higher percentage of genes that mapped to protein coding genes (45.47% versus 32.48%), a lower percentage of genes that mapped to pseudogenes  (12.34% versus 16.50%) and a comparable percentage of genes that mapped to long noncoding RNAs (lncRNA) (29.48% versus 28.81% ) compared to analysis by DRS.  For the  Gene expression of the 13,423 genes identified using both PCS and DRS was strongly correlated (R2=0.8966, P<0.001). Out of the 26,834 novel transcript isoforms that were mapped by PCS, 14,544 were also found by DRS in at least one specimen; when detected using both PCS and DRS, novel transcript isoform levels were modestly correlated between PCS and DRS (R =0.6832, P≤0.0001). PCS confirmed the existence of 15 of 16 splice variants previously identified only by proteomic analysis, and 9 were confirmed by DRS.

RNA integrity numbers (RINs) were comparable between frozen archived ccRCC specimens from patients who relapsed and those who did not.  Principal component analysis (PCA) based on PCS or DRS data did not cluster specimens based on relapse, patient sex or the number of ccRCC prognostic marker mutations. However, 159 and 68 genes were significantly and differentially expressed (log fold change ≥2, P≤0.1) between recurrent and nonrecurrent tumors using PCS and DRS, respectively, and the direction of the differential expression was strongly correlated between PCS and DRS for the 57 identified using both methods (R2=0.8170, P<0.001). When PCS data was down-sampled to have an equivalent number of reads relative to DRS, 92 RNAs were found to be differentially expressed between patients that relapsed in both the down-sampled PCS data and the DRS data.  Gene Ontology analysis showed suppression of adaptive immune response-related pathways in frozen ccRCC specimens from patients who subsequently relapsed.  The inferred tumor purity and immune cell abundance based on PCS data were lower in specimens from patients who relapsed than those who did not, and a similar trend was observed using the DRS data.  Using multiple additional deconvolution algorithms (CIBERSORTx, EPIC), it was found that the fraction of CD8+ T cells was lower in specimens from patients with recurrence than those without. Confirming these findings, real-time PCR amplification of CD8B (CD8+ T cell marker), GZMK (T-cell activation marker) and TOX (T-cell exhaustion marker) were lower in specimens from patients in the validation cohort who went on to relapse. However, only one (LINC04216) of three genes found by PCS was the only gene that was differentially expressed in specimens that went on to relapse and was confirmed to be differentially expressed in the validation cohort using real-time PCR. Using the PCS and DRS data, 31 and 2 genes, respectively, were found to have differential transcript usage between specimens from patients who went subsequently relapsed and those who did not. Among the novel isoforms identified was a novel soluble PD-L1 isoform, which was supported by both PCS and DRS data, and whose existence was then confirmed in RCC4 cells as well as in normal kidney and lung sequencing data from the Genotype-Tissue Expression (GTEx) project. This PD-L1 isoform, and PD-L1 in general, were not differentially expressed between specimens from patients who relapsed and those who did not, but another PD-L1 isoform was differentially expressed. Analysis of novel genes in PCS and DRS data revealed 40 and 4 differentially expressed genes, respectively, in specimens from patients who relapsed and those who did not. Of the two transcripts evaluated in the validation cohort, only one was confirmed to be differentially expressed in patients who relapsed.

Studies (1)

Purpose

This study compared direct RNA sequencing (DRS) and PCR cDNA sequencing (PCS) results in archival frozen clear cell renal cell carcinoma (ccRCC) specimens from patients who subsequently relapsed and those who did not.  Importantly, the comparison of DRS and PCS included analysis of known genes, novel genes and alternate transcripts in specimens from patients who went on to relapse and those who did not, and select results were confirmed by real-time PCR in a validation cohort. Snap-frozen archival nephrectomy specimens were obtained from ccRCC patients, 16 of whom went on to relapse and 16 who did not. Twelve specimens (6 from those with recurrence and 6 from those without) were used for sequencing, and the remaining 20 were used for real-time PCR in the validation cohort. RNA was extracted from ccRCC specimens and cultured RCC4 cells using QIAzol. RNA integrity was qualified using a Bioanalyzer, and RNA was quantified using the Qubit RNA HS assay Kit. Direct RNA (DRS) and PCR-cDNA (PCS) sequencing libraries were prepared using the SQK-PCS111 and SQK-RNA002 Kits from Oxford Nanopure Technologies, respectively. Differential gene expression analysis was conducted using featureCounts using a ≥2 log2 fold change threshold and an adjusted P<0.1.  Transcript isoform quantification was performed using Salmon. Estimates of tumor purity and tumor-infiltrating immune cell population abundance that was based on gene signature were obtained using ESTIMATE (v1.0.13) and xCell (v1.1.0); tumor CIBERSORTx and EPIC were used for tumor-infiltrating immune cell type deconvolution. Differential transcript usage was analyzed using RNAseqDTU (version 3.14) and an adjusted P<0.1.  Novel genes were identified using StringTie2 annotation, characterized using SQANTI3 (v5.1.2) and verified by analysis of GTEx project RNA-seq data from normal tissues. Expression of select transcripts was confirmed in the 20 specimens in the validation cohort.

Summary of Findings

PCS generated a higher median number of reads, longer median read length, higher median read accuracy and a longer longest aligned read than DRS.  When analysis was by PCS there were more mapped genes (median of 26,203 versus 18,047), a higher percentage of genes that mapped to protein coding genes (45.47% versus 32.48%), a lower percentage of genes that mapped to pseudogenes  (12.34% versus 16.50%) and a comparable percentage of genes that mapped to long noncoding RNAs (lncRNA) (29.48% versus 28.81% ) compared to analysis by DRS.  For the  Gene expression of the 13,423 genes identified using both PCS and DRS was strongly correlated (R2=0.8966, P<0.001). Out of the 26,834 novel transcript isoforms that were mapped by PCS, 14,544 were also found by DRS in at least one specimen; when detected using both PCS and DRS, novel transcript isoform levels were modestly correlated between PCS and DRS (R =0.6832, P≤0.0001). PCS confirmed the existence of 15 of 16 splice variants previously identified only by proteomic analysis, and 9 were confirmed by DRS.

RINs were comparable between frozen archived ccRCC specimens from patients who relapsed and those who did not.  PCA based on PCS or DRS data did not cluster specimens based on relapse, patient sex or the number of ccRCC prognostic marker mutations. However, 159 and 68 genes were significantly and differentially expressed (log fold change ≥2, P≤0.1) between recurrent and nonrecurrent tumors using PCS and DRS, respectively, and the direction of the differential expression was strongly correlated between PCS and DRS for the 57 identified using both methods (R2=0.8170, P<0.001). When PCS data was down-sampled to have an equivalent number of reads relative to DRS, 92 RNAs were found to be differentially expressed between patients that relapsed in both the down-sampled PCS data and the DRS data.  Gene Ontology analysis showed suppression of adaptive immune response-related pathways in frozen ccRCC specimens from patients who subsequently relapsed.  The inferred tumor purity and immune cell abundance based on PCS data were lower in specimens from patients who relapsed than those who did not, and a similar trend was observed using the DRS data.  Using multiple additional deconvolution algorithms (CIBERSORTx, EPIC), it was found that the fraction of CD8+ T cells was lower in specimens from patients with recurrence than those without. Confirming these findings, real-time PCR amplification of CD8B (CD8+ T cell marker), GZMK (T-cell activation marker) and TOX (T-cell exhaustion marker) were lower in specimens from patients in the validation cohort who went on to relapse. However, only one (LINC04216) of three genes found by PCS was the only gene that was differentially expressed in specimens that went on to relapse and was confirmed to be differentially expressed in the validation cohort using real-time PCR. Using the PCS and DRS data, 31 and 2 genes, respectively, were found to have differential transcript usage between specimens from patients who went subsequently relapsed and those who did not. Among the novel isoforms identified was a novel soluble PD-L1 isoform, which was supported by both PCS and DRS data, and whose existence was then confirmed in RCC4 cells as well as in normal kidney and lung sequencing data from the Genotype-Tissue Expression (GTEx) project. This PD-L1 isoform, and PD-L1 in general, were not differentially expressed between specimens from patients who relapsed and those who did not, but another PD-L1 isoform was differentially expressed. Analysis of novel genes in PCS and DRS data revealed 40 and 4 differentially expressed genes, respectively, in specimens from patients who relapsed and those who did not. Of the two transcripts evaluated in the validation cohort, only one was confirmed to be differentially expressed in patients who relapsed.

  • Biospecimens

    TissueKidney
  • Preservation Types

    Frozen
  • Diagnoses

    NeoplasticCarcinoma
  • Platforms

    RNANext generation sequencing
    RNAFluorometry
    RNAReal-time qRT-PCR
    RNAAutomated electrophoresis/Bioanalyzer
  • Pre-analytical Factors

    ClassificationPre-Analytical FactorValue(s)
    PreaquisitionDiagnosis/ patient condition
    Future relapse ccRCC
    Non-relapse ccRCC
    Next generation sequencing SpecificTechnology platform
    Direct RNA sequencing (DRS)
    PCR cDNA sequencing (PCS)
    Real-time PCR