In transplantation, detecting CMV antiviral resistance

Dr. Binnicker and colleagues conducted a multiple-month validation of the test to ensure its accuracy, during which time they compared it with Sanger sequencing. They defined the limit of detection, confirming the lowest viral load that produced reproducible and accurate sequencing data.

“We were interested in assessing whether the method could identify whether mixed viral populations were present in the sample and be able to tell us whether there was a minor subpopulation that could harbor resistance in the presence of a majority wild-type background,” he said.

Dr. Binnicker’s team also studied analytical specificity and precision using prospective clinical samples. The team found complete agreement between the Sanger and NGS methods for predicting antiviral susceptibility in 11 clinical samples (Fig. 2). “We also saw good agreement for resistance calling for the drugs ganciclovir, cidofovir, and foscarnet, so about 97 percent overall agreement between the two platforms at the level of susceptibility and resistance calling.”

In one patient sample, the NGS method gave a call of ganciclovir resistance, whereas the Sanger method called it as foscarnet resistance. “We followed up on this with additional sequencing and weren’t able to confirm the results of the Sanger method calling out foscarnet resistance, so we felt that the ganciclovir call was most likely accurate in that setting,” Dr. Binnicker said.

A deeper dive into the comparison data revealed how well each method performed at specific mutation calling in the UL97 and UL54 genes. “The vast majority of mutations detected in our studies showed agreement between the two sequencing methods,” he said. Some mutations detected by NGS were not detected by Sanger, “and when we looked at the prevalence of those mutations, they were less than 20 percent in most cases, which is typically the threshold for Sanger to be able to identify those,” Dr. Binnicker said. “This likely calls out some increased sensitivity or discriminatory power of next-generation sequencing.”

Two samples showed mutations detected by Sanger but not by NGS, but “we weren’t able to confirm those with our subsequent sequencing studies,” he added.

Additional validation studies confirmed that the LOD is reproducible and highly accurate at 500 IU/mL.

For mixed viral population studies, Dr. Binnicker and his colleagues took clinical samples with known UL97 and UL54 mutations and mixed them in varying ratios. “We were able to detect and differentiate the expected mutations by our next-generation sequencing method,” he said. They also wanted to find out how low prevalence a minor subpopulation can be with a mutation in the background of a majority wild type and whether NGS can differentiate them.

The team mixed wild-type CMV with CMV known to harbor various UL97 or UL54 mutations at different ratios. “We identified that if the minor subpopulation is present at a level of 15 percent or above, we were able to reproducibly and accurately identify that mutation and report it out,” he said.

“One of the fascinating aspects we’ve discovered as we started to use next-generation sequencing is the ability of the technology to identify low levels or minor subpopulations of the virus that can have emerging resistance mutations,” Dr. Binnicker said. “With Sanger, you need to have at least 20 percent of a viral population present with resistance to be able to tease it apart from a wild-type strain that lacks resistance.” With NGS, “we can get down to prevalence levels of 10 to 15 percent and be able to discriminate viral populations from each other.” Minor subpopulations may emerge to become predominant populations if not caught early.

When it comes to sequence coverage and reads, “Sanger typically falls apart in terms of its quality of sequencing after about 600 to 700 base pairs,” he said. NGS can sequence thousands of base pairs, “even the full-length UL97 and UL54 genes,” with high-quality results.

NGS offers a lower LOD compared with Sanger—500 IU/mL for NGS, “and in some cases even lower”—compared with greater than 1,000 IU/mL for Sanger. Throughput is much higher for NGS compared with Sanger.

In comparing the two sequencing methods, the only advantage found in the Sanger method was financial. “Sanger is definitely less expensive than next-generation sequencing,” Dr. Binnicker said, though he predicts NGS costs will decrease and notes the benefits to transplant patients are likely worth the higher cost of NGS. “When we’re dealing with complex cases and need to determine early on whether resistance is emerging, next-generation sequencing is a good application in these high-risk individuals.”

Since implementing the CMV NGS method for routine testing in May 2019, Dr. Binnicker’s laboratory is seeing some level of resistance-associated mutations in 20 to 25 percent of all samples tested. “That speaks to the fact that a lot of times this test is being ordered when there’s evidence of treatment failure or suspicion for resistance, and we’re definitely seeing a fairly good rate of resistance being detected by the next-generation sequencing method,” he said.

By tracking the specific mutations identified using NGS, they have found that the L595S mutation “is the one we have seen with the highest frequency,” Dr. Binnicker said. “About 28 percent of all the samples that show resistance—and 4.7 percent of all samples tested—have that specific mutation.”

Mutations at the A594V/P/T location occur in about 26.5 percent of the resistant samples, or about 4.5 percent of all samples tested, he said. The M460I/V location occurs in about 20 percent of resistance samples, or about 3.2 percent of all samples tested. Less common mutations are found at the C603W (16.2 percent of samples showing resistance), C592G (5.9 percent), and H520Q (4.4 percent) locations. “We do observe samples that have multiple resistance-associated mutations in the same sample,” Dr. Binnicker said. “That occurs about 22 percent of the time among those who have resistance; we see at least two or more specific mutations.” The most they’ve seen is four resistance-associated mutations in the same sample.

Less common mutations occur in the UL54 gene, he said. The K513N mutation occurs in about 4.4 percent of the samples showing resistance, or less than one percent of all samples tested. “We see some samples that have UL54 mutations present in them but only about one percent of the total samples tested.”

Returning to the case study, the 67-year-old kidney transplant recipient who presented with fever and diarrhea and had initially responded to ganciclovir had a viral load of 280,000 IU/mL 11 or 12 weeks after transplantation.

Using the NGS method, “we detected two mutations in UL97 [L595S and L595F] that conferred ganciclovir resistance, so the patient was switched to foscarnet,” Dr. Binnicker said.

One month later, the patient’s viral load increased to 1.8 million IU/mL. The prevalence of the two mutations seen a month earlier had now increased dramatically. “But we saw a new mutation show up—the E756N mutation—right at the threshold [15 percent] for the test that confers resistance to foscarnet, which would explain the increasing viral load,” Dr. Binnicker said.

Dr. Binnicker’s team plans to add UL56 to its NGS assay to determine possible letermovir resistance and to study whether routine use of NGS leads to improved outcomes when compared with Sanger sequencing. His team will also explore whether baseline sequencing should be performed in all transplant cases before therapy is initiated, and whether NGS can be used to identify new resistance-associated mutations.

Amy Carpenter Aquino is CAP TODAY senior editor.