Karen Titus
October 2026—In the never-ending search for new cancer biomarkers, there is always at least one new hopeful waiting in the wings, readying for a potential elevation (so to speak) to star status.
Among the latest to generate buzz (at least in the form of oncologists asking about testing) are MTAP deletions. Could these be the latest to have a breakout role?
There are many good reasons to think so, says Eric Vail, MD, director of molecular pathology, Cedars-Sinai Medical Center, and associate professor and director of the molecular genetic pathology fellowship program.
Methylthioadenosine phosphorylase deficiency is seen across multiple cancers. Located on chromosome 9p21, it is frequently codeleted with CDKN2A, a tumor suppressor. Homozygous MTAP deletions occur in some nine percent of cancers overall; 9p21 deletion (including MTAP) is associated with resistance to immunotherapy in a number of tumor types. Deletion of MTAP is most often seen in patients with mesothelioma (33.1 percent), urothelial carcinoma (23.8 percent), central nervous system tumors (19 percent), pancreatic cancer (18.4 percent), thymomas (15.7 percent), cholangiocarcinoma (15.6 percent), and non-small cell lung cancer (14.3 percent), according to the Center for Cancer Genomics and Advanced Therapeutics database. It may also be quite high in other cancers, according to other studies, including malignant peripheral nerve sheath tumor (55 percent) and melanoma (25 percent). And some estimate it might be as high as 60 percent in glioblastoma. It’s also possible there is a higher risk of death in patients with NSCLC or glioblastoma with MTAP deletions.
“Imagine if we got a new drug for glioblastoma,” says Dr. Vail. “Tons of tumors have this 9p21 deletion. So it’s common in common tumors, but it’s also common in uncommon ones as well, relatively speaking.” If testing for MTAP deletions becomes commonplace, and if targeted therapies emerge, the rewards could be enormous.
Both “ifs” are big, but also attainable, say observers who’ve been involved in MTAP research and testing. At Memorial Sloan Kettering Cancer Center, “We first studied MTAP deletion many years ago, in mesothelioma,” says Marc Ladanyi, MD, chief of molecular diagnostics, Department of Pathology and Laboratory Medicine. That was in the FISH era, before NGS became widely available. Now, with NGS typically the first line of testing for MTAP deletion, it’s important to understand the pitfalls and limitations of assigning homozygous deletion, using either large panel or whole exome NGS assays, he says.
These are early days—the Lexington and Concord of what will likely be a much longer engagement with MTAP-related testing and treatments. Initial trials, of course, look at single-agent treatments, and the current MTAP targeted trials involve patients who’ve failed all previous lines of therapy. But, says Dr. Ladanyi, “Intuitively, I think it’s likely to be more effective in combination with other targeted therapies.” Trials are moving in that direction, he says, including those involving PRMT5 and MAT2A inhibitors as targets for MTAP-deleted tumors. He also sees plenty of interest in combinations of therapies, such as an MTAP targeted therapy with, say, an EGFR tyrosine kinase inhibitor.
As he surveys the knowledge emerging around MTAP deletions and related developments, Dr. Ladanyi says laboratories would be wise to muster in now, even if approvals aren’t yet imminent. “It’s important for us in molecular diagnostics to be anticipating the clinician questions.” At MSKCC, whenever he and his colleagues report a CDKN2A deletion, for example, they will add a note that clearly states whether or not both copies of MTAP are lost.
“I kind of semi-joke, Let’s save ourselves an email or phone call from the clinician,” he says. “Just put the information in the report very clearly.” It works. “We rarely get a follow-up question from the clinician.” Reports are automatically annotated using MSKCC’s OncoKB database, so clinicians receive information about the significance of the deletion, that it’s potentially targetable, et cetera.
In a similar vein, says Dr. Vail, “It is critical that pathologists are aware of the pipeline and the clinical practice that’s coming out in oncology. Not only do you have to talk the talk with the oncologists and the surgeons, but you also have to anticipate what is going to be needed in the future,” well before therapies are approved, and even before clinical trials are completed.
That means taking a close look at the nuances of MTAP testing.
Or, as Dr. Vail puts it: “We have an action plan for cases that have wishy-washy deletions.”
One limitation of current testing, Dr. Ladanyi says, is related to the varying degrees of contamination by non-neoplastic tissue when extracting DNA from tumor tissue. This can make it more difficult to be certain there is a homozygous deletion of p16/CDKN2A, and, secondarily, of MTAP, he says.
It’s important to realize that MTAP is essentially never homozygously deleted unless CDKN2A is also deleted, he says. When isolated MTAP deletion is detected, it’s most likely because the bioinformatics used by a given laboratory was unable to call the CDKN2A deletion (though it will be evident on further review).
“That’s an important limitation,” says Dr. Ladanyi. Since there’s always some degree of contamination by non-neoplastic tissue, it may be difficult to confidently call a homozygous deletion. There are, he notes, computational pipelines, like FACETS, that can correct for tumor content and then, based on that, produce an integer copy number for MTAP and CDKN2A.
At MSKCC, Dr. Ladanyi and colleagues use the FACETS software pipeline in addition to their standard copy number pipeline to help confirm homozygous deletion in cases involving excessive contamination by non-neoplastic cells. “And to be honest,” he adds, “we always review the actual copy number profile from the NGS assay. That’s an important practice point—don’t just rely on the output of the software.
“We often do what we call a rescue call,” he continues. “We’ll review the numbers, and if it’s just below the cutoff used in the bioinformatic pipeline, and it’s very clearly visible on the copy number profile, we will ‘rescue’ the call.”
“That’s what clinicians need to know,” he says. “Is there a functional MTAP gene, or not?” If not, the result counts as a positive, so to speak, and it indicates that the tumor will respond to a targeted therapy that is based on MTAP deletion.
At Cedars-Sinai, Dr. Vail and colleagues use comprehensive genomic profiling, with one assay for solid tumors and one for hematologic malignancies; both can detect homozygous loss, he says.
He adds: “If there’s some evidence of a deletion, we actually include in our NGS report a caveat when we have equivocal calls,” suggesting that the clinician use IHC to confirm the diagnosis if they’re considering enrolling the patient in a clinical trial.
In fact, “Right now we’re exclusively using [MTAP results] for clinical trials,” Dr. Vail says. “I haven’t done much clinical reach out on this, just because we are pretty early in clinical trial development.”
That has interesting implications for results reporting. The institution has three different layers of oncology practice, he explains. One involves what he calls “the super elite, academic, trial-running oncologists.” The second includes affiliated faculty, and the third are nonaffiliated oncologists who have admitting privileges. Each group has different needs.
An emerging biomarker such as MTAP is, at least for now, primarily of interest to the first group. Until it’s clear that a clinical trial shows benefits, he says, “Why bombard others with information that they might never need and potentially displace something that’s actually useful right now?”
He suspects that if “amazing” phase three data are presented at ASCO next year, and that physicians feel certain of a Food and Drug Administration approval, it will then make sense for the laboratory to start distributing information to the other oncology groups likely to use it.
As of late summer, Dr. Vail reports, the majority of MTAP deletion-related clinical trials listed in the clinicaltrials.gov database were phase one trials: two early phase one, 31 phase one. Nineteen were phase two, and two were phase three. In that last category, one involves pancreatic cancer and the other non-small cell lung cancer.
The pancreatic trial is set for primary completion in 2029. “If they have excellent results,” says Dr. Vail, “there will probably be some interim analysis in 2027,” which might be made available at ASCO or ESMO. The lung cancer trial began in January and is still recruiting, with primary completion set for 2031.
Any FDA approvals of new targeted treatments will raise additional questions about the most appropriate testing approach.
Most of the clinical trials are recruiting based on immunohistochemistry expression, not on NGS data, Dr. Vail says. “That doesn’t mean we just slavishly do exactly what the FDA or the trial says. We’re clinicians as well, and we’re allowed to use clinical medical judgment and our expertise in techniques—and in clinical care on the oncology side—to determine what’s right for our patients.”
“I’d really like to have a comprehensive look at the concordance between” the two, Dr. Vail adds. “I think that’s going to be an important question.”
Even with clinical trial reliance on IHC, he continues, a well-validated NGS assay “can absolutely work—with the understanding that you may miss some cases” involving samples with low tumor cellularity. Dr. Vail calls this issue “a major, major factor with copy number detection, especially copy number loss.” As tumor cellularity decreases, both the accuracy and sensitivity of bulk sequencing NGS assays drop dramatically. In traditional NGS assays, 20 to 40 percent cellularity is the cutoff, he says. “Ours is around the 30 to 40 percent range, because it’s amplicon based.” (Hybrid capture panels have a small advantage over amplicon-based panels, he says, though the tradeoff is more nucleic acid input.)
The result is a specificity problem, he continues. “Say you have a heterozygous deletion, and you input 10 percent tumor cellularity into the pipeline—but it really has, say, 40 percent tumor cellularity.” The pipeline will overcorrect that signal to noise, he says, and identify a deep deletion. “Without manual review of data” in such cases, he says, “you can get into trouble.”
He and colleagues at Cedars-Sinai did a study (they presented the data in an abstract at the AMP meeting last November) comparing results with and without manual review. “Just using the caller, there was significantly less accuracy than with the caller and manual review,” he says. In the latter case, sensitivity was in the 80 to 90 percent range, he says.
This was only a small (31 samples) cohort, he acknowledges. “But the positive predictive value went from 61 percent to 93 percent when manual review was attached on it.”
IHC has its own limitations, Dr. Vail says. In the study, four samples had heterogeneous staining. “We think these are actually subclonal populations.” This would be hard for bulk sequencing to differentiate. “If there’s no morphologic difference between the two tumor populations,” he says, it will be impossible to discern which will require microdissection, leading to more equivocal calls.
The benefits of IHC testing are genuine, but running IHC on every sample would be a tall order.
Depending on the clinical trial data, Dr. Vail says, “As people are operationalizing this, it’s really going to depend on what your local institution has. What are the capabilities?”
Using IHC or FISH as a screening approach would take “a lot of work,” agrees Dr. Ladanyi, “and a lot of resources.” But either might provide clarity in cases where the NGS is unclear or uncertain.
While there are advantages to doing single-gene IHC biomarker testing, Dr. Vail urges caution. “Every time you go back to the block, you’re losing something to refacing. It’s just one slide—until it’s 10 slides.” Clinicians don’t always appreciate the toll, especially when the tests are ordered asynchronously.
FISH for deletion on paraffin sections could be another option. Because the nuclei are sectioned, it won’t be clear if the two signals for MTAP are in the part of the nucleus that got sliced away, Dr. Ladanyi says. “It can be done, but it’s tricky. If you have the opportunity to study whole nuclei—as in hematological malignancies—then the problem doesn’t arise.”
Dr. Ladanyi suggests there might be a place for a more standardized workflow for cases where the MTAP deletion is difficult to call because of normal tissue contamination. “Do you rerun the NGS data through a different software package, like we do? Or does a borderline or ambiguous call trigger IHC or FISH to confirm?” he asks. “Those cases probably could use some consensus approach.”
In a similar vein, Dr. Vail says, “How do you design your systems to work out the gray” of equivocal cases? “This is something you need to take into account when you’re validating your assays. The interesting thing about MTAP is that this is on a ton of assays that are already out there, and nobody was thinking about this” until recently.
He talks about reviewing calls he’s made in the past, where he reported a loss without distinguishing between a homozygous or heterozygous state. “It didn’t matter clinically at the time,” he says.
“But now it does.”
At least for now, molecular testing for MTAP deletion alone does not make sense, Dr. Vail says. The NGS panels are, obviously, looking for oncogenic drivers such as KRAS, EGFR, ALK fusions, etc., that are targetable as well as prognostic/diagnostic. “Since we’re already doing this, if we can come to an agreement on how to best use NGS [for MTAP], there’s no reason to throw out the good cases that come out of it.”
In other words, “If I have a very accurate way of saying, ‘This is specifically an MTAP deletion, and these are very likely not to be MTAP deleted,’ then I only need to confirm those gray area cases,” Dr. Vail continues. “That saves another slide, it saves tissue, it saves time if I’m already doing the testing to begin with.”
As noted earlier, though, providers often presume that the same methods used in a clinical trial are also best for clinical practice. “That’s not necessarily wrong,” Dr. Vail concedes, especially if insurers insist on following what was used in the trials.
But, he adds, labs still need to discuss with, and explain to, their clinical colleagues the different types of testing, their benefits and pitfalls, and their role in choosing treatment. “Clinical colleagues don’t know lab. That’s not their job. Their job is to take the results you provide them and then act on it. Your job is to help them understand what the results mean.”
If the promises of MTAP testing bear fruit, there will also be questions about when to test and for what.
“It gets back to some of the same questions we had to deal with in terms of EGFR testing,” Dr. Ladanyi says. Uncertainty about when to test eventually gave way to the understanding that doing EGFR testing at diagnosis made the most sense.
Similarly, he says, it’s possible the clinicians will want MTAP status sooner rather than later. Most large-panel assays include CDKN2A, “but if you’re using a reference lab, you might want to make sure that MTAP is also included, since that’s not always the case.”
At MSKCC, “We’re strong believers in broad testing at diagnosis,” he says, “so the oncologist knows from the beginning what the options are going to be for the patient.” For example, even if they start with an EGFR TKI for a patient with EGFR-mutant lung cancer with p16 and MTAP deletion, he says, they know that other options might be available down the line if needed. “And I think eventually they’ll be thinking about combining the EGFR TKI with the MTAP-directed therapy.”
Recalling his early work in MTAP and mesothelioma, Dr. Ladanyi says it makes sense to look at the prevalence of p16 deletion across other cancers. Again, that’s where doing broad panel testing comes into play, as at MSKCC. “But if one were not running large-panel NGS routinely, then consideration of which cancers are likeliest to have MTAP deletion becomes more important.”
Like Othello and Iago, MTAP and CDKN2A tend to be spoken of in the same breath. Which is the true bad actor?
In a recently published paper (Ross JS, et al. J Thorac Oncol. 2026;21[3]:103520), Dr. Ladanyi and MSKCC colleagues looked at the clinical significance of MTAP deletions and concurrent oncogenic driver alterations (including EGFR) in NSCLC. The MTAP deletion was associated with a shorter time to osimertinib discontinuation. But, as he points out, “Those clinical correlates are really the correlates of CDKN2A deletion, and not MTAP deletion by itself. We never see MTAP deletion without CDKN2A deletion, and that was the case with this study as well.”
In the vast majority of cases where CDKN2A is lost, it’s lost by homozygous deletion, says Dr. Ladanyi, though occasionally one copy will be inactivated by a point mutation. That means the MTAP gene cannot be homozygously deleted.
Dr. Ladanyi chooses his words carefully. There is nothing pleasant about any cancer pathway, of course, so he settles on “appealing” to describe one feature of MTAP deletion: If the cell has lost both copies of MTAP, it cannot restore a copy. So while there may be other, as-yet-unknown ways for the cell to become resistant to PRMT5 and MAT2A inhibitors, restoring MTAP is not an option.
As noted, the action takes place on chromosome 9p21, with MTAP, CDKN2A, and CDKN2B all very close to one another, explains Dr. Vail. “The most common event that happens in these tumors that have MTAP loss is concurrent deletion of all three of those.” In his experience, this happens in 70 to 80 percent of the cases where MTAP is lost.
But there are also a large number of cases that have CDKN2A loss without MTAP loss. That raises not only a small sensitivity issue, Dr. Vail says, but also a specificity issue that comes from using CDKN2A as a surrogate marker.
CDKN2A is the true oncogenic factor, says Dr. Vail. It can be lost in three main ways. One is deep deletion, which is the most common. The second is promoter methylation. In this scenario, MTAP loss is probably not occurring, but the methylation of CDKN2A is.
CDKN2A can also be inactivated by a frameshift, or nonsense, or small mutation, Dr. Vail continues. “You might have a heterozygous deletion of 9p21, then a loss of function mutation in CDKN2A. Again, this is one of those cases where if you have, say, low cellularity, you might mistake that heterozygous mutation for a homozygous mutation,” resulting in a false-positive.
(As for CDKN2B, “It’s not clinically actionable now or in the near future,” Dr. Vail says. “The vast majority of the time, they’re lost together.”)
As further background, Dr. Vail explains that MTAP loss leads to accumulation of the MTAP substrate MTA. This in turn inhibits PRMT5—the buildup of MTA competes with S-adenosylmethionine, or SAM, which binds to PRMT5, crippling its function. “So if you knock out the rest of PRMT5, or you knock out MAT2A, which blocks SAM synthesis, then that ends up leading to that cell dying. But the other cells that have a functional MTAP don’t die.” This is similar to the synthetic lethality of PARP inhibitors. MTAP, he says, “will be the second sort of pathway with synthetic lethality. And there’s a ton more coming down the pipe.
“And it’s such an amazing route of tumor suppression,” he continues. “Because if you can truly find ways that target the cancer biology uniquely, you develop effective low-side-effect drugs.”
There is wonder in Dr. Vail’s voice as he talks about all of this. “I was actually surprised about the biological significance of MTAP,” he says. “It’s just fascinating. MTAP is almost an accidental participant in this whole thing. It’s hanging around with CDKN2A. It’s getting codeleted. It’s not really the driver or the problem here.
“And to be able to target something that’s not even participating in the tumor genesis is amazing,” he continues. “It’s one of those things that makes me think that as a species, we can solve entropy.” He adds: “I am ever-impressed, as well as being humbled by the complexity but also humanity’s ability to deconvolute that complexity of biology. It’s a really, really cool finding, and I hope it leads to clinical benefit.”
Karen Titus is CAP TODAY contributing editor and co-managing editor.