Bladder cancer detection and surveillance: How urine cell-free DNA stacks up against cytology

David Wild

June 2019—A high-throughput sequencing panel was found to be more than 90 percent sensitive in detecting urinary tumor DNA in early-stage bladder cancer and in post-treatment surveillance. The approach, reported in April in Cancer Discovery, overcomes some of the challenges urinary cell-free DNA analysis poses, said one of its developers, and is far more sensitive than cytology and cystoscopy.

“We were able to demonstrate an approximately sixfold improvement in sensitivity over cytology, the current standard of care for noninvasive early detection of bladder cancer, as well as a significant improvement in sensitivity over the combination of cytology and cystoscopy in a surveillance context,” said Jonathan Dudley, MD, who conducted the research as a molecular genetic pathology fellow at Stanford University School of Medicine (Dudley JC, et al. Cancer Discov. 2019;9[4]:500–509).

In a presentation at the Association for Molecular Pathology annual meeting last year, Dr. Dudley, who is now a postdoctoral fellow in Bert Vogelstein’s lab at the Ludwig Center at Johns Hopkins Hospital, said he and his colleagues set out to develop a better instrument to detect and surveil bladder cancer because current methods for doing so “have pretty significant limitations.” The sensitivity of cytology is below 50 percent, he noted, and “cystoscopy is invasive and relatively expensive.”

Could sequencing cell-free DNA in the urine achieve high performance characteristics, he wondered, and potentially outperform current modalities for detection of bladder cancer?

To explore this question, he joined the lab of Max Diehn and Ash Alizadeh at Stanford, where researchers had previously developed a workflow called cancer personalized profiling by deep sequencing, or CAPP-Seq, to analyze circulating tumor DNA. They had used this approach to detect minimal residual disease in patients with localized lung cancer (Chaudhuri AA, et al. Cancer Discov. 2017;7[12]:1394–1403). “We completely reoptimized this workflow to look at cell-free DNA in the urine, from the wet lab workflow to the informatics,” Dr. Dudley said, crediting coauthor Joseph Schroers-Martin, MD, with extensive work on the bioinformatics pipeline. They termed the new approach urine tumor DNA CAPP-Seq, or uCAPP-Seq.

As per standard processing protocol in the cytology lab, the samples were centrifuged so cells could be collected for microscopy, with the supernatant discarded. “We saved those discarded supernatants for cfDNA extraction so we could compare our approach to cytology from the same biological sample,” Dr. Dudley explained.

One of the difficulties of working with urine, he said, is that the volume of fluid is much larger than the volume typically obtained from plasma. This makes it challenging to purify nucleic acids, as he and his colleagues noted in their published report. “We had to develop an extraction protocol that could pull the DNA out of that large volume,” Dr. Dudley said.

To do so, they adapted and modified a previously described approach using adsorption of cell-free nucleic acids on Q-Sepharose resin and ion exchange chromatography (Shekhtman EM, et al. Clin Chem. 2009;55[4]:723–729).

Another challenge of urinary cfDNA, he said, is that the distribution of urinary cfDNA fragment sizes is much broader than the distribution typically encountered in plasma. Dr. Dudley and his team used enzymatic fragmentation to convert the distribution “to a form more amenable for next-generation sequencing, without suffering massive yield loss.” Enzymatic fragmentation, Dr. Dudley and coauthors noted in their published report, yielded significantly higher DNA recovery than acoustic shearing.

The team then designed a hybrid capture panel focusing on genes commonly mutated in bladder cancer, using genetic data from 412 patients with muscle-invasive bladder cancers in The Cancer Genome Atlas data set, among other sources. They subsequently reoptimized their bioinformatics pipeline to handle the different distribution of fragment sizes resulting from enzymatic fragmentation.

“Within that TCGA data set, we expected to see a median of about seven mutations per patient within our panel space,” Dr. Dudley said. “Across the cohorts we looked at, we saw a median of about six mutations per patient.” About 98 percent of patients would be predicted to have at least one mutation covered in their panel.

They tested the concordance of mutations detected in tumor tissue and urine using 18 patients for whom paired urine and tumor tissue were available. Across cases, they found a median of 73.2 percent of the mutations first identified in urinary tumor DNA were also detected in paired tumor tissue. “And moving in the opposite direction, we saw a median of 67 percent of the mutations from tumor cases in the urine as well,” Dr. Dudley said.

The team discovered that mutations found in both tumor tissue and in urine had higher median allele fractions than mutations found in tumor tissue but not in urine samples (27.2 percent versus 9.2 percent for both versus tumor only, respectively). They also found that the highest number of mutations, 74 percent, were located in the TERT promoter region, specifically in two hotspots previously described in the literature.

The second most commonly mutated region was the PLEKHS1 promoter, with 46 percent of these regions mutated, Dr. Dudley said, noting that PLEKHS1 promoter mutations have been observed previously (Weinhold N, et al. Nat Genet. 2014;46[11]:1160–1165).

“To our knowledge,” he said, “this is the largest series of bladder cancers that have been profiled for this region [PLEKHS1],” and the findings provide confirmatory evidence that “this is one of the most common mutations in bladder cancer.”

Once they had developed their workflow and identified the most common mutations, Dr. Dudley further validated the diagnostic capability of the panel using two common approaches to diagnosing patients using cfDNA, which he termed tumor-naive and tumor-informed profiling.

“In the first case, you analyze cell-free DNA by sequencing and you classify the sample as positive if you detect a driver mutation,” he explained.