Marshall Makes the Case for Molecular Sorting in Gastrointestinal Disease

OncologyLive, Vol. 19/No. 13, Volume 19, Issue 13

Oncologists have been slow to introduce targeted and immune therapies into the treatment of gastrointestinal cancers, but that is changing, and poor patient survival statistics call for a continuation of this trend, according to John L. Marshall, MD.

John L. Marshall, MD

Oncologists have been slow to introduce targeted and immune therapies into the treatment of gastrointestinal (GI) cancers, but that is changing, and poor patient survival statistics call for a continuation of this trend, according to John L. Marshall, MD, chief of the Division of Hematology and Oncology and a professor of medicine and oncology at Georgetown Lombardi Comprehensive Cancer Center in Washington, DC.

In remarks at the 3rd Annual School of Gastrointestinal Oncology™ (SOGO®), held in New York, New York, in April, Marshall described a push to differentiate gastrointestinal cancer types based on tumor mutational load (TML) or tumor mutational burden (TMB), mismatch repair deficiency (dMMR) status, microsatellite instability (MSI) status, and protein expression. Although testing for these biomarkers has become prevalent for other tumor sites, cancers of the stomach and intestines still tend to be lumped together into a single category, adding to misconceptions on how to classify and treat.

Marshall explained that gastric cancers are distinct diseases based on their molecular subtypes and should be treated differently. Clinically, treating these tumors the same produces mixed results, he said.

It is clear that upper GI cancers can be sorted by their anatomic locations and then by pathology. Knowledge of molecular subtypes has added a third tier of sorting, Marshall said. “It makes more sense to use [molecular distinctions] than the right versus left colon. A long time ago, we treated all of these squamous and adeno carcinomas and all of the locations as 1 set of cancers, and clearly we’re dividing it up.” Molecular stratification into subgroups leads to more accurate classification of tumors, better treatment strategies, and, ultimately, better survival rates.

Although the overall estimated incidence of gastric cancers has declined by about 2.3% from 1995 to 2013 and survival rates are increasing, the numbers still “aren’t that great,” Marshall said, adding that for patients younger than 50 years, the incidence of gastric cancers increased by 1.3% annually during that same period.1 Assessing each patient for molecular biomarkers could help physicians make better decisions about therapy.

Each actionable biomarker represents a strong case for classifying gastric and upper GI tumors differently. HER2 expression has significant value in characterizing gastric tumors, Marshall noted. “We know that HER2 expression in this disease is much more common in the proximal cancers and almost unheard of in squamous cancers.”

Among all gastric cancers, HER2 expression is found in 17.9% of tumors. In gastroesophageal junction (GEJ) tumors, it is found at a considerably higher rate: 25% to 34% of tumors. In chromosomal instable gastric tumors, HER2 expression is 16% to 34%; in genomically stable tumors, 6% to 7%.

In a comparative molecular analysis of esophageal squamous cell carcinoma, esophageal adenocarcinoma, and gastric adenocarcinoma using next-generation sequencing (NGS), investigators demonstrated that each type of gastric tumor has different expression of various genes.2 “There are some common threads,” Marshall said, “but there is differential expression.”

GEJ adenocarcinoma had higher expression of CDH1 and RNF43 mutations than esophageal adenocarcinoma and squamous cell esophageal cancer. GEJ adenocarcinoma and esophageal adenocarcinoma tumors had relatively high levels of ARID1A and APC mutations. Squamous cell esophageal cancer tumors had higher incidence of KMT2D, PIK3CA, SETD2, NOTCH1, PTEN, RB1, FOXO3, BRCA1, and MSH6 than the other 2 disease types. All 3 tumor types had high levels of TP53 and mutations.

Table. Mutational Expression in Diffuse and Intestinal Gastric Cancers2

Protein expression screened by immunohistochemistry (IHC) testing also distinguished these tumors from each other in the analysis. All tumors had high expressions of EGFR, with squamous cell esophageal cancer running the highest. Squamous cell esophageal cancer also had higher expression of ERCC1, PD-L1, RRM1, TLE3, TOPO1, and TUBB3.There is also a statistically significant difference between diffuse and intestinal tumors based on NGS testing in the analysis (Table). “We recognize diffuse and intestinal [tumors] as different anatomically, but we don’t treat them differently clinically,” Marshall said. This distinction persists in IHC testing, where intestinal tumors have higher expressions of HER2, TOP2A, TS, RRM1, and cMET.

Biomarker interpretation is complicated by the variations in expression tendencies across tumor sites. For example, although HER2 expression in gastric cancer is common in proximal cancers and is less prevalent distally, in colorectal cancers, HER2 expression is more common in distal tumors and almost never seen in proximal tumors.

Another challenge is simply the nomenclature of mutational testing. When a patient’s tumor is assessed with IHC testing for MSI and dMMR status, clinicians are measuring the presence or absence of the key proteins MLH1, MLH6, PMS, and MSH2. The language is problematic, because the lab report indicates that the proteins are “present,” which is actually a normal result, but could be misinterpreted as meaning that the patient has MSI-high disease.

Also, if there are 1 or 2 proteins missing from the test, clinicians tend to reflexively order genetic testing with either NGS or a fragment analysis. If a patient tests positive for MSI-high disease, a physician is likely to proceed to germline genetic testing to see if it reveals an inherited cancer syndrome or somatic disease.

Pembrolizumab (Keytruda) is an approved immunotherapy for patients with any MSI-high or dMMR solid tumor; however, no established guidelines exist for MSI-high testing. “This is science that is evolving over time,” Marshall noted.

TMB and TML are also emerging as important biomarkers for immune therapy responsiveness. “The theory is that the more mutations one has, the more likely neoantigens will be available and checkpoint inhibition will work,” Marshall said. Some have decided to make TMB a standard biomarker test for immunotherapy checkpoint inhibitors, but MSI, TMB, and PD-L1 tests each have relevance based on biomarker expression across tumors. “I believe what’s going to happen over time is that there will be increasingly more than 1 assay to determine whether checkpoint inhibition will be [effective],” he said.

On a cautionary note, Marshall said that environmental factors may play a role in each patient’s disease, and even if the genetic profile looks the same across 2 tumors, that doesn’t necessarily mean identical treatment will succeed in both. Also, obtaining good-quality tissue samples is an important consideration. As soon as the surgeon cross-clamps the artery to the tumor, the clock is ticking to get an accurate sampling, he said. If it takes longer than 15 to 20 minutes for that tumor specimen to be fixed in formalin, then about 40% of the proteins and mRNA have changed.

It is also difficult to decide how to study upper GI and other cancers. Does each molecular cohort and each tumor location require a different trial? If the protocol focuses on a very rare subtype, recruiting may be challenging, or patients who could benefit from the trial may be excluded. “When you start to divide and conquer, you end up with not enough patients in these cohorts,” Marshall said.

Basket trial designs could be the answer to this problem, he said: “Our vision moving forward is to take this kind of [basket trial] structure, molecularly profile everybody, and offer them clinical trials based on their molecular characterization, not just their tumor location.”


  1. Anderson WF, Rabkin CS, Turner N, Fraumeni JF Jr, Rosenberg PS, Camargo MC. The changing face of noncardiac gastric cancer incidence among US non-Hispanic whites [published online January 19, 2018]. J Natl Cancer Inst. doi: 10.1093/jnci/djx262.
  2. Salem ME, Puccini A, Xiu J, et al. Comparative molecular analyses of esophageal squamous cell carcinoma, esophageal adenocarcinoma, and gastric adenocarcinoma [published online June 4, 2018]. Oncologist. doi: 10.1634/theoncologist.2018-0143.