A quantitative reverse transcriptase-polymerase chain reaction assay to identify metastatic carcinoma tissue of origin

Abstract

Identifying the primary site in patients with metastatic carcinoma of unknown primary origin can enable more specific therapeutic regimens and may prolong survival. Twenty-three putative tissue-specific markers for lung, colon, pancreatic, breast, prostate, and ovarian carcinomas were nominated by querying a gene expression profile database and by performing a literature search. Ten of these marker candidates were then selected based on validation by reverse transcriptase-polymerase chain reaction (RT-PCR) on 205 formalin-fixed, paraffin-embedded metastatic carcinoma specimens originating from these six and from other cancer types. Next, we optimized the RNA isolation and quantitative RT-PCR methods for these 10 markers and applied the quantitative RT-PCR assay to a set of 260 metastatic tumors. We then built a gene-based algorithm that predicted the tissue of origin of metastatic carcinomas with an overall leave-one-out cross-validation accuracy of 78%. Lastly, our assay demonstrated an accuracy of 76% when tested on an independent set of 48 metastatic samples, 37 of which were either a known primary or initially presented as carcinoma of unknown primary but were subsequently resolved.

Figure 1

Experimental workflow: marker candidate selection (A) and assay optimization and prediction algorithm building and testing (B).

Figure 2

Expression of 10 selected tissue-specific gene marker candidates in FFPE metastatic carcinomas and prostate primary adenocarcinomas. For each plot, the y axis represents the normalized marker expression value.

Figure 3

Assay optimization. A and B: Electropherograms obtained from an Agilent Bioanalyzer. RNA was isolated from formalin-fixed, paraffin-embedded tissue using a 3-hour (A) or 16-hour (B) proteinase K digestion. Sample C22 (red) was a 1-year-old block, whereas sample C23 (blue) was a 5-year-old block. A size ladder is shown in green. C and D: Comparison of Ct values obtained from three different qRT-PCR methods: random hexamer priming in the reverse transcription followed by qPCR with the resulting cDNA (RH 2 step), gene-specific (reverse primer) priming in the reverse transcription followed by qPCR with the resulting cDNA (GSP 2 step), or gene-specific priming and qRT-PCR in a one-step reaction (GSP 1 step). RNA from 11 samples was divided into the three methods, and RNA levels for two genes were measured: β-actin (C), HUMSPB (D). The median Ct value obtained with each method is indicated by the solid line.

Figure 4

Heat map showing the relative expression levels of the 10 marker panel across 239 samples. Red indicates higher expression. For each sample, ΔCt was calculated by taking the mean Ct of each CUP marker and subtracting the mean Ct of an average of our housekeeping markers (ΔCt = Ct[CUP marker] − Ct[ave. HK marker]). The minimal ΔCt for each tissue of origin marker set (lung, breast, prostate, colon, ovarian, and pancreas) was determined for each sample. The tissue of origin with the overall minimal ΔCt was scored one, and all other tissues of origin were scored zero. Data were sorted according to pathological diagnosis. Partek Pro was populated with the modified feasibility data, and an heat map was generated.