elifeeLifeeLife2050-084XeLife Sciences Publications, Ltd10710310.7554/eLife.10710310.7554/eLife.107103.3version of recordResearch ArticleCancer BiologyEndoscopic liquid biopsies of gastric fluid in a large human patient cohort reveal DNA content as a candidate tumor biomarker in gastric cancerCadonáFrancine C12BartelliThais F1†PelosofAdriane G3SztokfiszClaudia Z3BuenoAdriana P4Batista do Carmo dos SantosLuana1BrancoGabriela P1Oliveira dos SantosGabriel4NunesWarley A4PintorFernanda A5Lie Senda de AbrantesLaís5DefelicibusAlexandre6Gonzaga Vaz CoelhoLuiz7LejaMarcis8InHaejin910LiSharon1112HochsterHoward913CoimbraFelipe JF14DrummondRodrigo6Tojal Da SilvaIsraelhttps://orcid.org/0000-0002-4687-14996ChokshiRavi J1115PasqualiniRenata1116ArapWadihwa116@rutgers.edu1117NunesDiana N11116Dias-NetoEmmanuelhttps://orcid.org/0000-0001-5670-8559eyd10@rutgers.edu11116https://ror.org/03025ga79Laboratory of Medical Genomics, A. C. Camargo Cancer CenterSão PauloBrazilhttps://ror.org/00wbge811Masters in Health and Life Sciences, Franciscana UniversitySanta MariaBrazilhttps://ror.org/03025ga79Endoscopy Section, A. C. Camargo Cancer CenterSão PauloBrazilhttps://ror.org/03025ga79Department of Pathology, A. C. Camargo Cancer CenterSão PauloBrazilhttps://ror.org/03025ga79Research Support Center, A. C. Camargo Cancer CenterSão PauloBrazilhttps://ror.org/03025ga79Laboratory of Computational Biology, A. C. Camargo Cancer CenterSão PauloBrazilhttps://ror.org/0176yjw32Postgraduation Program in Applied of Sciences of Adult Health & Alpha Gatroenterology Institute, Hospital das Clínicas, Faculty of Medicine, Federal University of Minas GeraisBelo HorizonteBrazilhttps://ror.org/05g3mes96University of Latvia, Institute of Clinical and Preventive Medicine and Center for Gastric Diseases GASTRORigaLatviahttps://ror.org/0060x3y55Rutgers Cancer Institute, New BrunswickNew BrunswickUnited Stateshttps://ror.org/029z02k15Division of Surgical Oncology, Department of Surgery, Rutgers Robert Wood Johnson Medical SchoolNew BrunswickUnited Stateshttps://ror.org/0060x3y55Rutgers Cancer InstituteNewarkUnited Stateshttps://ror.org/014ye1258Division of Hematology/Oncology, Department of Medicine, Rutgers New Jersey Medical SchoolNewarkUnited Stateshttps://ror.org/029z02k15Division of Medical Oncology, Department of Medicine, Rutgers Robert Wood Johnson Medical SchoolNew BrunswickUnited Stateshttps://ror.org/03025ga79Department of Abdominal Surgery, A. C. Camargo Cancer CenterSão PauloBrazilhttps://ror.org/014ye1258Division of Surgical Oncology, Department of Surgery, Rutgers New Jersey Medical SchoolNewarkUnited Stateshttps://ror.org/014ye1258Division of Cancer Biology, Department of Radiation Oncology, Rutgers New Jersey Medical SchoolNewarkUnited Stateshttps://ror.org/014ye1258Division of Hematology/Oncology, Department of Medicine, Rutgers New Jersey Medical SchoolNewarkUnited StatesYangYongliangReviewing Editorhttps://ror.org/03ns6aq57Shanghai University of Medicine and Health SciencesChinaLiuCaigangSenior Editorhttps://ror.org/032d4f246Shengjing Hospital of China Medical UniversityChina
Department of Clinical Cancer Prevention, The University of Texas M. D. Anderson Cancer Center, Houston, United States
Gastric cancer remains a diagnostic and therapeutic challenge worldwide. Improved prognostic biomarkers could aid treatment planning across surgical, neoadjuvant, and adjuvant settings. We evaluated a novel liquid-biopsy approach integrated with esophagogastroduodenoscopy (EGD) by analyzing gastric fluid DNA (gfDNA) from a large cohort (n=1056) to assess its diagnostic utility and prognostic value in gastric cancer. In this exploratory study, gfDNA concentration was measured in patients with normal gastric mucosa, peptic diseases, preneoplastic conditions, or cancer. Variables included sex, gastric fluid pH, proton-pump inhibitor use, tumor subtype, stage, and outcomes. gfDNA levels were significantly higher in gastric cancer than in all comparison groups (mean 26.86 ng/µL; 95% CI 20.05–33.79; p=3.61 × 10e-12) and as compared to non-malignant controls (mean 10.77 ng/µL; 95% CI 9.23–12.33; p=9.55 × 10e-13) and preneoplastic states (mean 10.10 ng/µL; 95% CI 7.59–12.60; p=1.10 × 10e-5). Advanced tumors (T3) exhibited higher gfDNA than earlier stages (T2 or below; mean 25.66 vs 15.12 ng/µL; p=5.97 × 10e-4). In a subset of gastric cancer patients, gfDNA >1.28 ng/µL associated with longer progression-free survival (p=0.009) and correlated with increased tumor-infiltrating immune cells (p=0.001); this association remained after adjusting for stage (p=0.014). Elevated gfDNA supports gastric cancer presence in the general human population and may inform disease management when combined with tissue biopsies. Importantly, gfDNA shows prognostic potential in established gastric cancer, where higher gfDNA content may paradoxically relate to better outcomes, potentially linked to immune-cell infiltration. These findings warrant further validation and integration with complementary diagnostic modalities to enhance accuracy and clinical utility.
gastric cancerprognosisliquid biopsygastric fluidsDNA levelsgastric fluidResearch organismHumanhttps://ror.org/02ddkpn78Fundação de Amparo à Pesquisa do Estado de São Paulo2014/26897-0Dias-NetoEmmanuelhttps://ror.org/046w98q07Ministério da SaúdePRONON/DECITDias-NetoEmmanuelLevy-LongenbaughDonor-Advised FundPasqualiniRenataArapWadihhttps://ror.org/02ddkpn78Fundação de Amparo à Pesquisa do Estado de São Paulo2018/14267-2Dias-NetoEmmanuelhttps://ror.org/02ddkpn78Fundação de Amparo à Pesquisa do Estado de São Paulo2018/02972-3Dias-NetoEmmanuelhttps://ror.org/046w98q07Ministério da SaúdeSIPAR 2500.035-167/2015-23Dias-NetoEmmanuelThe funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.Author impact statementElevated gfDNA supports gastric cancer diagnosis and may guide management when combined with tissue biopsies, and in established cancer, higher gfDNA may predict better outcomes, potentially indicating immune-cell infiltration.publishing-routeprcIntroduction
Esophagogastroduodenoscopy (EGD) is routinely performed to work up vague digestive complaints and is used to biopsy suspicious lesions and ultimately diagnose gastric cancer, a tumor that ranks as the fifth in terms of incidence and mortality in the world, with current estimates of ~660,000 annual deaths globally (Globocan/IARC, 2022). In areas of high incidence, EGD has been shown to contribute to a reduction in gastric cancer mortality (Hamashima et al., 2013; Matsumoto and Yoshida, 2014) and is also employed for presurgical gastric cancer staging and even as a treatment method (endoscopic resection) in early-stage disease (Kim and Jung, 2021), rendering it a universally used tool in this setting. During EGD, gastric fluids are collected and discarded, allowing a better evaluation of the gastric mucosa.
The study of DNA found in body-fluid samples (e.g. blood, urine, saliva, cerebrospinal fluid, peritoneal washes) can support the diagnosis and/or follow-up of cancer (Forshew et al., 2012). In a liquid biopsy approach, tumor-derived DNA permits the monitoring of specific mutations diluted in the fluids under investigation. As the gastric fluid is in direct contact with the stomach mucosa and gastric cancer lesions, liquid biopsies performed with gastric fluid are neither restricted to just the few selected areas sampled by tissue biopsy, nor by the minute amounts of tumor DNA that might be diluted in peripheral blood. Also, it may indeed allow a full representation of cells that reach the stomach, derived from the upper digestive tract, including immune cells, tumor cells, and the microbiota. In a prior feasibility study, we showed that gastric fluid incidentally collected from gastric cancer patients during EGD contains tumor-derived DNA, allowing the capture of tumor mutations (Pizzi et al., 2019). We have now collected gastric fluids from a large patient cohort (n=1056), obtained during diagnostic endoscopy, to investigate whether gastric fluid DNA (gfDNA) concentrations would vary according to disease diagnosis, including the presence of non-malignant, premalignant, or malignant lesions, gastric cancer subtype and stage, tumor location, and patient age or sex. Notably, we also evaluated whether gfDNA concentrations would have predictive and prognostic value on gastric cancer recurrence and patient outcomes.
ResultsPatient groups and diagnoses
Gastric fluids were collected and gfDNA was extracted from a total of 1056 subjects. After excluding 115 cases with a diagnosis of hepatic portal hypertension, partial or total gastrectomy, esophagectomy, or other non-gastric cancer malignancies (e.g. esophageal squamous cell carcinoma, duodenal adenocarcinoma, Hodgkin’s disease, well-differentiated neuroendocrine tumor, and metastatic breast cancer, among others), a total of 941 patients (89.1%) remained for gfDNA analysis. These patients were grouped according to the EGD findings as: normal EGD (n=10, 1.1%), non-malignant peptic diseases (n=596, 63.3%) including erosive gastritis, erosive and non-erosive esophagitis with or without hiatal hernia, infectious or inflammatory disorders, preneoplastic conditions (n=99, 10.5%) including atrophic gastritis, gastric intestinal metaplasia, dysplasia, Barrett’s esophagus, or gastric adenocarcinoma (n=236, 25.1%). The diagnostic group, age, sex, and body-mass index (BMI) of the participants are shown in Table 1.
Clinical and demographic attributes of participating subjects (n=941).
Non-gastric cancer*(n=705, 74.9%)
Gastric cancer(n=236, 25.1%)
Age
≤45
130 (18.4%)
34 (14.4%)
45–60
224 (31.8%)
83 (35.2%)
≥60
343 (48.7%)
119 (50.4%)
Missing data
8 (1.1%)
0
Sex
Male
330 (46.8%)
150 (63.6%)
Female
367 (52.1%)
86 (36.4%)
Missing data
8 (1.1%)
0
†BMI
Underweight
10 (1.4%)
9 (3.8%)
Normal weight
206 (29.2%)
91 (38.6%)
Overweight
310 (44%)
84 (35.6%)
Obese
177 (25.1%)
51 (21.6%)
Missing data
2 (0.3%)
1 (0.4%)
Includes all patients diagnosed as presenting normal mucosa, with peptic diseases or pre-neoplastic lesions.
Body-mass index (BMI): Underweight, BMI <18.5; Normal weight, BMI = 18.5–24.9; Overweight, BMI = 25–29.9; Obese, BMI ≥30.
Gastric fluid DNA concentrations according to age, pH, and body-mass index
No significant variations in gfDNA concentration were found by age, pH of gastric fluid, or BMI when examined by group amongst gastric cancer patients or non-cancer controls. Regarding age, subjects were clustered in the three groups (<45, 45–60, and >60 years old), and no significant differences were found between subjects with (p=0.26) or without (p=0.14) gastric cancer. Similarly, no difference was found by pH analysis of gastric fluid (grouped as pH <7, pH = 7, or pH >7); we found p-values of 0.60 and 0.90 for gastric cancer and non-cancer controls, respectively. For BMI analysis, subjects were grouped as underweight (BMI <18.5), normal (BMI between 18.5–24.9), overweight (BMI between 25–29.9), or obese (BMI ≥30). Again, no differences in gfDNA concentrations were found by BMI in individuals with gastric cancer (p=0.59) or non-cancer controls (p=0.25).
gfDNA concentrations and sex
Next, we evaluated whether gfDNA concentrations vary according to sex in patients with gastric cancer and subjects without cancer. For the non-cancer cohort, we observed elevated gfDNA concentrations in females versus males (p=6.8e–4), but no differences were seen between males and females in the cancer group (p=0.7; Figure 1). Therefore, when we compared gfDNA concentration between cancer versus non-cancer subjects, considering sex, we found striking differences for men and for women, respectively, p=7.98e–12 and p=1.19e–4 (Figure 1).
gfDNA concentrations observed in all subjects with available data, grouped according to sex, and presence of gastric cancer.
Non-cancer subjects have normal gastric mucosa or minor peptic diseases (excludes pre-neoplastic disease). Statistical comparisons were performed using the Kruskal–Wallis test, followed by pairwise Mann–Whitney U tests with Benjamini–Hochberg correction. P-values < 0.05 were considered significant. Error bars indicate median gfDNA concentration (ng/µL) with interquartile ranges.
As some studies have suggested that the use of proton pump inhibitors (PPI) is more frequent in women (Rückert-Eheberg et al., 2022), we evaluated possible links between gfDNA concentrations and the reported history of current, previous, or no PPI use. In general, we observed no impact of PPI on gfDNA concentrations. However, for the group of non-cancer subjects that used PPI (n=410), increased gfDNA concentrations were seen for women (n=206), as compared to men (n=204; p=1.51e–4), apparently reflecting the sex-related results shown in Figure 1, between males and females, unrelated to PPI use. Moreover, gfDNA concentrations were not different for subjects that reported current and previous administration, as well as no PPI administration, for females with or without cancer (n=55, n=6 and n=25; n=111, n=45 and n=206, respectively), as well as for males with or without cancer diagnosis (n=60, n=19 and n=62; n=75, n=51 and n=209, respectively, with corresponding p-values of p=0.92; p=0.07; p=0.24, and p=0.71).
Analysis of bacteria to human ratios in gfDNA
As the gfDNA origin is likely to be a sum of human DNA (including epithelial, stromal, and immune cells, tumor and non-tumor cells), combined with microbiota-derived DNA (transient, saliva-derived microorganisms, the resident microbiota of the stomach and other upper digestive tract structures), we performed a quantitative analysis of bacteria and human DNA for a subset of samples (n=180), based on human-to-bacteria DNA ratios, as recently described (de Albuquerque et al., 2022). The samples selected for the analysis of this subset analysis included all 10 controls without any pathologies detected, as well as samples from the peptic diseases group (n=51), preneoplastic conditions (n=55), and the gastric cancer group (n=64). This analysis revealed no significant differences in the ratios of human- or bacteria-derived DNA found in the gfDNA (Supplementary file 1).
gfDNA concentrations are increased in gastric cancer patients compared to non-cancer controls
We first determined the concentration of gfDNA from subjects with different diagnoses after EGD and found the following values for the groups: normal mucosa and peptic diseases - mean 10.77 ng/µL; 95% CI: 9.23–12.33; n=606; preneoplastic conditions - mean 10.10 ng/µL; 95% CI: 7.59–12.60; n=99 and gastric cancer - mean 26.86 ng/µL; 95% CI: 20.05–33.79; n=236. Notably, mean concentration of gfDNA was significantly different among the groups (p=3.61e–12). gfDNA concentration was higher in gastric cancer as compared to the combined group of normal EGD and peptic disease patients (p=9.55e–13), as well as to the preneoplastic group (p=1.10e–5). In contrast, gfDNA concentrations were not different between the non-cancer groups (p=0.89; Figure 2A).
gfDNA concentration according to disease diagnosis and tumor stage.
(A) gfDNA concentration (ng/µL) in patients with no endoscopic findings: Normal (N) or presenting minor peptic diseases (PD) – N+PD; preneoplastic conditions (PN), or gastric cancer (Cancer). (B) gfDNA concentrations (ng/µL) for early-stage disease patients – T0-T2 (T0 +Tis + T1+T2), as compared to more advanced disease stages T3 and T4. Statistical comparisons were performed using the Kruskal–Wallis test, followed by pairwise Mann–Whitney U tests with Benjamini–Hochberg correction. P-values < 0.05 were considered significant. Error bars indicate mean gfDNA concentration (ng/µL) and 95% confidence intervals.
For gastric cancer patients, we evaluated possible correlations between gfDNA concentrations and clinicopathological variables such as tumor stage and histopathological grades. In general, increments in gfDNA were found in subjects as the disease progresses (Figure 2B). Notably, the gfDNA mean concentration for early-stage disease (tumor stages - T0 +Tis + T1+T2, or T0-T2 n=65) was 15.12 ng/µL; 95% CI: 9.73–20.50, increasing in T3 (25.66 ng/µL; 95% CI: 19.46–31.85; n=91; p=5.97e–4) and T4 stages (38.12 ng/µL; 95% CI: 15.02–61.22; n=63; p=0.06). Other comparisons were not significant, including gfDNA concentrations for patients with localized (n=165) versus metastatic disease (n=71). gfDNA concentrations could not differentiate tumor histopathological grades (p>0.05), or Lauren gastric cancer subtypes: diffuse (n=97), intestinal (n=93), or mixed (n=25; p=0.28; data not shown).
gfDNA correlates with the presence of inflammatory cell infiltrates and gastric cancer diagnosis and survival
Our liquid biopsy results revealed increased gfDNA concentrations in gastric cancer patients as compared to non-cancer controls. To explore whether gfDNA concentrations could be used to support gastric cancer diagnosis, we analyzed the receiver operating characteristic (ROC) curve of gfDNA between gastric cancer versus non-cancer controls. We observed a statistically significant curve (area under the curve [AUC] = 0.66; p=5.23e–10), with a fair capability of detecting true positive gastric cancer patients by simply considering gfDNA concentrations (Figure 3A; Supplementary file 2), indicating its fair diagnostic support value. We also found a statistically significant curve of gfDNA between cancer/preneoplastic conditions and peptic diseases/normal (AUC = 0.61; p=6.86e–7; Figure 3B).
Analysis of receiver operating characteristics (ROC) and area under the curve (AUC) for gfDNA levels between cancer patients and non-cancer individuals.
(A) ROC curve of gfDNA between the cancer patients and non-cancer individuals. (B) ROC curve of gfDNA between cancer versus non-cancer and cancer + preneoplastic conditions versus peptic diseases + normal.
We next evaluated if gfDNA concentrations would correlate with disease prognosis. Well-annotated clinical records were collected for all subjects to determine gfDNA concentrations cutoff values that would split the patients into groups with contrasting overall survival through a log-rank test. By using this approach, a gfDNA cutoff of 1.28 ng/µL was set and gastric cancer patients presenting gfDNA concentrations above this threshold had better overall survival (n=163; p=0.019), even more significant after patients with metastatic tumors were not considered (n=148; p=0.014; Figure 4A). Finally, to further investigate the possible link between patient survival and gfDNA concentration, we examined a subset of gastric cancer patients (n=32), for which pathology slides were available, to investigate possible correlations between gfDNA concentrations and the intensity of inflammatory cell infiltrates (Figure 4B). For this, we managed to include eight samples with gfDNA below the threshold and 24 above it. Even with a small sample set, this result was also significant (p=0.001; Figure 4C and Supplementary file 3), suggesting a possible mechanism to support the correlation of gfDNA concentration and gastric cancer survival data (Figure 4A).
gfDNA analysis according to patient survival and infiltration of inflammatory cells.
(A) For a total of 148 non-metastatic GC patients, cutoff gfDNA concentrations determined by a log-rank test (1.28 ng/µL) discriminate survival free of tumor progression (p=0.009). (B) For a subset of 32 cases, for which representative biopsy H&E slides were available, inflammatory cell infiltrates in the gastric tumors were inspected by experienced pathologists and (C) have shown that more intense inflammatory infiltrates in tumors are characteristic of subjects with higher gfDNA concentrations (p=0.001). gfDNA value distributions were compared using the non‑parametric Mann–Whitney U test.P‑values < 0.05 were considered statistically significant. Error bars indicate mean gfDNA concentration (ng/µL) and 95% confidence intervals.
Discussion
Gastric cancer remains a leading cause of tumor mortality worldwide, a clinical feature attributable at least in part to its often-late stage at diagnosis and/or lack of prognostic factors to enable personalized disease management. However, with varying access to subspecialists, a role for gfDNA may help in the clinical management of patients. Early gastric cancer detection may often be delayed during EGD if biopsy samples are either taken too superficially or if the tissue is too scanty for meaningful pathological analysis. Thus, the translational application of a biomarker that provides support to gastric cancer diagnosis and prognosis would be useful, especially in a less specialized outpatient ambulatory or inpatient hospital settings.
In Leon et al., 1977, published a seminal paper showing that cancer patients had higher levels of cell-free DNA (cfDNA) in their blood compared to healthy individuals, suggesting its use as a potential cancer biomarker. Since then, numerous studies have confirmed these findings and have explored the potential of using cfDNA for cancer diagnosis, prognosis, and monitoring treatment response, including gastric cancer. Qian et al., 2017 analyzed cfDNA in serum samples of subjects with gastric dysfunctions, including gastric cancer, and found that cancer patients presented elevated cfDNA when compared to benign gastric disease or controls. However, the diluted amounts of tumor-derived/tumor-microenvironment DNA remain minute in the peripheral circulation. This prompted us to determine DNA concentrations in the gastric fluids of subjects undergoing EGD to investigate non-specific digestive symptoms. Patients included those with diverse peptic diseases (controls), pre-neoplastic lesions, as well as patients with confirmed gastric cancer. We reasoned that liquid biopsies of gastric fluid incidentally obtained during EGD would offer diagnostic support and would possibly offer clinically useful information. Whereas EGD is an invasive procedure, most of these patients will receive at least one diagnostic EGD during their initial workup and tumor staging, and gastric fluids collected at this moment can be useful. However, prior studies of gastric fluids have heretofore been limited to the evaluation of certain molecules, such as proteins, RNAs, or for the analysis of mutations in specific genes (Wu and Chung, 2013; Shao et al., 2014; Pizzi et al., 2019). Here we show that the simple evaluation of gfDNA concentration serves as a convenient and promising biomarker to support gastric cancer diagnosis and staging and correlates with immune cell infiltrates. In a large cohort, we show that increased DNA concentrations in gastric fluids may differentiate gastric cancer patients from individuals without cancer, independent of age, BMI, sex, PPI use, or Lauren’s histological subtypes. Increasing amounts of gfDNA were significantly associated with gastric cancer diagnosis, as well as advanced tumor stages.
Of note, there were several interesting observations from our results. Overall, we found that T4 tumors had gfDNA concentration similar to T3 tumors (T3: 25.66 ng/µL; 95% CI: 19.46–31.85; and T4: 38.12 ng/µL; 95% CI: 15.02–61.22; p=0.27) and the difference between T4 and earlier stages (T2 and below) was non-significant (p=0.065). One might speculate that this may be due to the higher variability of gfDNA concentration seen for T4 (Figure 2B) that could be intrinsic to the T4 stage, as some tumors are more invasive than others, and if the tumor grows through the gastric wall and the serosa (Fukuda et al., 2011), it might perhaps be shedding DNA into adjacent peritoneal structures instead of just the gastric cavity, leading to a reduction of tumor-derived DNA in the gastric cavity. This contrasts with T0-T3 tumors that appear to continuously and preferentially release DNA inside the gastric cavity. In agreement with this possibility, we found that advanced tumors have increased gfDNA concentration as compared to early stages, a finding that might suggest that the origin of the gfDNA could perhaps derive from the tumor, but also from the tumor environment, from immune cell infiltrates, and the microbiota. However, our quantitative analysis of bacterial and human DNA in the gfDNA for a set of samples (n=180), including gastric cancer samples (n=64), showed no clear trends towards an increased bacterial DNA content, rendering that possibility unlikely (Supplementary file 1).
We also observed an intriguing correlation between gfDNA concentration and survival, with gastric cancer patients presenting lower gfDNA concentrations (set at ≤1.28 ng/µL) associated with a worse prognosis. We envisaged that this observation could be associated with poorer immune responses (immunologically ‘cold tumors’), which was supported by the finding that patients with higher gfDNA concentrations (set at >1.28 ng/µL) more often presented with moderate-to-intense inflammatory infiltrates, whereas individuals with lower gfDNA usually presented minimal-to-moderate inflammatory infiltrates (p=0.001). Notably, two-thirds (16 out of 24) of the cases with increased gfDNA had moderate/intense/elevated immune cell infiltrates, and all cases within this classification of inflammatory infiltrates were in the group of increased gfDNA (Supplementary file 3). This suggests that the elevated gfDNA concentration observed in subjects with better outcomes may originate from two primary sources: an active population of infiltrating immune cells, together with the tumor cells eliminated by these immune infiltrates.The results presented here suggest wide implications for the translational application of liquid biopsies with gfDNA concentration analysis in the diagnosis, staging, and management of gastric cancer patients. The most obvious barrier to the translational application of liquid biopsies and gfDNA measurement in the routine clinical setting is the need for an EGD to access it. However, the use of gfDNA may hold the highest benefit by increasing the value of information obtained from the initial EGD, which is – as mentioned previously – a procedure that all patients with gastric cancer will eventually undergo multiple times during their disease workup. Indeed, the suction of gastric fluids during endoscopy is readily performed during EGD, when the gastric walls must be free of fluids for a better visualization of the mucosa. Thus, the convenient incorporation of this liquid biopsy approach during EGD is cost-effective and a minor additional step to the procedure. The benefit is that the continuous proximity of gastric fluids to the tumor site portends a much-reduced dilution of gfDNA, some of it being tumor-derived DNA, which allows a better representation of tumor-related DNA alterations. In a comparative sense, liquid biopsy of gastric fluid with gfDNA measurement is similar to obtaining DNA from peritoneal lavage of patients with endometrial cancer, but indeed much harder to obtain, in which peritoneal fluid has higher mutant allelic fractions as compared to plasma (Mayo-de-Las-Casas et al., 2020), similar to previous findings in gastric cancer in our early work (Pizzi et al., 2019).
The results presented here suggest wide implications for the translational application of liquid biopsies with gfDNA concentration analysis in the diagnosis, staging, and management of gastric cancer patients. The most obvious barrier to the translational application of liquid biopsies and gfDNA measurement in the routine clinical setting is the need for an EGD to access it. However, the use of gfDNA may hold the highest benefit by increasing the value of information obtained from the initial EGD, which is – as mentioned previously – a procedure that all patients with gastric cancer will eventually undergo multiple times during their disease workup. Indeed, the suction of gastric fluids during endoscopy is readily performed during EGD, when the gastric walls must be free of fluids for a better visualization of the mucosa. Thus, the convenient incorporation of this liquid biopsy approach during EGD is cost-effective and a minor additional step to the procedure. The benefit is that the continuous proximity of gastric fluids to the tumor site portends a much-reduced dilution of gfDNA, some of it being tumor-derived DNA, which allows a better representation of tumor-related DNA alterations. In a comparative sense, liquid biopsy of gastric fluid with gfDNA measurement is similar to obtaining DNA from peritoneal lavage of patients with endometrial cancer, but indeed much harder to obtain, in which peritoneal fluid has higher mutant allelic fractions as compared to plasma (Mayo-de-Las-Casas et al., 2020), similar to previous findings in gastric cancer in our early work (Pizzi et al., 2019).
Much interest has been given to developing new tools for detecting diagnostic, prognostic, and predictive markers in gastric cancer (Mayo-de-Las-Casas et al., 2020). The area under the receiver operating characteristic (AUROC) analysis for liquid biopsy of gastric fluid with gfDNA analysis yielded a discrimination capability of 0.66, which is comparable to serological tumor biomarkers such as CEA (0.68), CA72-4 (0.67), or CA19-9 (0.64) that are currently used in standard oncological practice (Tahara and Arisawa, 2015; Yu et al., 2016; Lin et al., 2020). Moreover, the monitoring of gfDNA concentration may also be valuable for a closer follow-up and better diagnostic accuracy of patients with potentially premalignant diseases and/or at higher genetic/familial risk of developing gastric cancer. One might also speculate that our findings could perhaps also apply to other upper gastrointestinal malignant tumors such as esophageal cancer and/or gastroesophageal junction neoplasia, and future studies should rigorously evaluate those open possibilities.
This study represents an initial evaluation of the prognostic impact of gfDNA concentration in human subjects diagnosed with gastric cancer, as compared to other conditions such as pre-malignant lesions and peptic diseases. The participants were referred to diagnostic EGD, and their distribution in the diagnostic groups reflects the real-life case numbers seen in a single institution, a dedicated cancer center. Before receiving sedation for the EGD, participants were invited to join the study. Therefore, samples were collected prospectively, before diagnosis, from all subjects and no attempts to balance the groups were made to avoid introducing any potential selection bias. Limitations also include the primer sets that allow a partial quantification of some bacterial groups (i.e. microbiome analysis), the limited number of pathology slides available from gastric cancer patients to investigate immune cell infiltrates (n=32), and the lack of a more comprehensive approach that would also allow a cell-type-specific deconvolution and a better explanation of the functional mechanisms that lead to increased gfDNA concentrations. The diagnostic potential of elevated gfDNA levels was not explored together with other known biomarkers such as CEA or CA19-9. Our data suggest that gfDNA is not only derived from tumor cells but may have expressive amounts of DNA derived from infiltrating immune cells. The precise determination of the presumed cell-of-origin of this increased gfDNA remains to be unequivocally demonstrated as it would have to rely on methods not used in this study, such as RNA-Seq and DNA-seq.
Moreover, the large patient cohort analyzed in this study was derived from a single specialized institution (a tertiary cancer center), and it is therefore unclear whether our findings might be broadly generalized; this open question should be addressed in further confirmatory studies from other hospital types and geographical regions.
Finally, for considering subjects with an established gastric cancer diagnosis, gfDNA has the potential to serve as a rapid, low-cost surrogate marker of the tumor immune microenvironment, whose characterization may have implications for predicting chemotherapy and immunotherapy response (Jiang et al., 2018; Jiang et al., 2019; Chen et al., 2022; Espinosa-Carrasco et al., 2024), and prognostic repercussions that, in the future, could be useful to treatment planning and risk-adapt therapeutic strategies. As we have gathered no data on gfDNA changes over time, conclusions on its utility for monitoring treatment response and/or predicting recurrence are currently limited but warrant further investigation.
Given that gastric cancer has currently no suitable tumor prognostic biomarker, one hopes that the initial results of this report will encourage the liquid biopsy of EGD-collected gastric fluids to evaluate gfDNA concentrations towards clinic-ready translational applications. Further gastric fluid studies may help revolutionize gastric cancer care, especially in places where specialties may not be readily available. Also, it may assist as a first-line exam that can be done to help detect cancer cases, with prognostic value. Although more research is needed, early data indicates this may be a way to provide access and availability for primary and recurrent gastric cancer.
Materials and methods
Reagent type (species) or resource
Designation
Source or reference
Identifiers
Additional information
Chemical compound, drug
PhenolChloroform
SigmaSigma
P4682472476
Used to remove proteins and lipids during DNA extraction.
Software, algorithm
R statistical environmentGraphPad
R Core TeamPrism
RRID:SCR_025679RRID:SCR_002798
Version 4.2.0, 2022Version 10.4.1 (532) Dec. 2024
Other
Proteinase-K
Bioline
BIO37037
Enzyme used for protein digestion during DNA extraction.
Patients
This single-center convenience study performed at the A. C. Camargo Cancer Center, São Paulo, Brazil, included all patients who provided informed consent and met the inclusion criteria of being at least 18 years old and having a clinical indication for diagnostic EGD. Samples were collected at diagnosis, during standard EGD, from February 2016 to March 2021. EGDs were only performed after referral from an independent physician (i.e. not related to the study) to assess abdominal symptoms, under standard sedation. Samples were obtained after either patients or their guardians signed a written informed consent form approved by the Clinical Research Committee, the Ethics Research Committee, and the Institutional Review Board (IRB; protocol #2134/15). Diagnostic and epidemiological data were collected as described (Bartelli et al., 2019). Samples and analyses were excluded if requested by the participants (even after the signature of a written informed consent), or if other unrelated diagnoses were found (e.g. previous gastrectomy or presence of other non-gastric cancer).
Gastric fluid collection, DNA extraction, and quantification
As usual and customary for upper endoscopic examination, all subjects were fasting for 8–12 hr. At the beginning of the procedures, gastric fluids that are routinely removed to allow better examination of the mucosa were collected in sterile plastic containers attached to the endoscope suction channel and kept on ice until pH measurement and neutralization. Samples were divided into aliquots and kept frozen at –20°C (each aliquot was thawed just once), until DNA was extracted from an equal volume of gastric fluid (800 µl) by proteinase-K digestion, followed by phenol-chloroform extraction and ethanol DNA precipitation. DNA was subsequently resuspended in 100 µl of sterile nuclease-free double-deionized water and quantified by Qubit fluorometry (Thermo Fisher). DNA concentrations found for each subject, as well as clinical/pathological data and patient demographics related to this article are freely available at: https://doi.org/10.5061/dryad.bzkh189qz.
Quantification of human- versus bacteria-derived DNA in gastric fluids
For a representative fraction of the liquid biopsies (n=180, 19.1%), including all 10 control subjects with no clinical findings (i.e. no pathologies detected on EGD) along with peptic diseases (n=51), premalignant lesions (n=55), and gastric cancer cases (n=64), we evaluated the proportion of human- versus bacteria-derived DNA in gastric fluids by using quantitative PCR (de Albuquerque et al., 2022).
Pathological analysis of inflammatory cell infiltrates in tissue biopsies
Tissue biopsies were available from a subset of liquid biopsies evaluated for the gfDNA content, allowing the investigation of inflammatory cell infiltrates. In such cases, pathological tissue analyses were done independently by two expert cancer pathologists, both specialized in upper gastrointestinal tract neoplasia, who were blinded to the gfDNA concentration results from the liquid biopsies. Discrepant cases for the immune infiltration levels (n=3) were revised and discussed, and a consensus was achieved. After the microscopic identification of tumor cells and the histological classification of gastric cancer, the presence and intensity of inflammatory cell infiltrates in the neoplastic lesions were also evaluated, as well as their patterns in the tumor stroma (e.g. diffuse, microscopic lymphoid aggregates, lymphoid aggregates with the formation of hyperplastic germinal centers). We have set the predominant cellular composition of the inflammatory infiltrates and quantified the infiltrate/tumor stroma ratios as follows: (i) minimal/rare – rare inflammatory cells in the tumor stroma; (ii) moderate – intermediate cellularity, no distortion of the epithelial component and without formation of luminal micro-abscesses, and (iii) intense/elevated – large amounts of inflammatory cells, making it difficult to visualize the stroma, with the formation of luminal micro-abscesses and/or epithelial thinning. We have also evaluated the presence of lymphoid aggregates, necrosis/ulcer, and neutrophilic infiltrates. Areas directly related to ulcer formation and/or granulation tissue were not considered in the polymorphonuclear component assessment.
Statistical analysis
No formal sample size calculation was performed as in this exploratory study we used convenience sampling, recruiting all consecutive eligible patients who met the inclusion criteria and consented to participate during the study period. Baseline gfDNA concentration values for patient groups are presented as mean ± standard deviation or as median and interquartile range, as appropriate. Analyses were performed in the R statistical environment (version 4.2.0) as described by the R Development Core Team, 2022. Data distribution was assessed using the Shapiro–Wilk test and visual inspection of Q–Q plots. Normally distributed data were analyzed using one-way ANOVA with Tukey’s HSD post hoc test, whereas non-normally distributed data were analyzed using the Kruskal–Wallis test with Dunn’s post hoc test. Parametric tests were applied to normally distributed data, and non-parametric alternatives were used otherwise. Optimal cutoffs for gfDNA concentrations to stratify survival groups were determined using maximally selected rank statistics (maxstat.test from the R package maxstat; Hothorn, 2017). Distributions of gfDNA concentrations among the four patient groups (normal, peptic ulcer disease, preneoplastic disease, and gastric cancer) were compared using the Kruskal–Wallis test, followed by pairwise Mann–Whitney U tests with Benjamini–Hochberg adjustment of p-values. ROC curves and the corresponding AUC were obtained using logistic regression analysis. Results were considered statistically significant when p≤0.05.
Additional informationCompeting interests
No competing interests declared
Reviewing editor, eLife
Author contributions
Data curation, Formal analysis, Investigation, Visualization, Writing – original draft, Writing – review and editing
Formal analysis, Investigation, Writing – review and editing
Investigation, Methodology, Writing – original draft, Writing – review and editing, Sample collection
Formal analysis, Investigation, Writing – review and editing, Sample collection
Data curation, Investigation, Methodology, Writing – review and editing, Pathology analysis
Formal analysis, Investigation, Sample and clinical data collection
Investigation, Methodology, Sample and clinical data collection
Investigation, Writing – review and editing, pathology analysis
Formal analysis, Investigation, Writing – review and editing, Pathology analysis
Investigation, Writing – review and editing, Clinical data collection
Investigation, Clinical data collection
Writing – review and editing, Statistical support
Writing – review and editing, Discussion of findings and clinical relevance
Writing – review and editing, Discussion of findings and clinical relevance
Writing – review and editing, Discussion of findings and clinical relevance
Writing – review and editing, Discussion of findings and clinical relevance
Writing – review and editing, Discussion of findings and clinical relevance
Investigation, Writing – review and editing, Discussion of findings and clinical relevance
Writing – review and editing, Statistical support
Data curation, Investigation, Writing – review and editing, Statistical support
Writing – review and editing, Discussion of findings and clinical relevance
Writing – review and editing, Discussion of findings and clinical relevance
Supervision, Writing – review and editing, Discussion of findings and clinical relevance
Conceptualization, Investigation, Writing – original draft, Project administration, Writing – review and editing, Discussion of findings and clinical relevance
Conceptualization, Formal analysis, Supervision, Funding acquisition, Investigation, Methodology, Writing – original draft, Project administration, Writing – review and editing
Ethics
Samples were obtained after either patients or their guardians signed a written informed consent form approved by the Clinical Research Committee, the Ethics Research Committee, and the Institutional Review Board (IRB) of the A.C.Camargo cancer Center (protocol #2134/15).
Additional files
Analysis of CT values for bacteria and human target genes.
Performance metrics of the ROC/AUC curves shown in <xref ref-type="fig" rid="fig3">Figure 3</xref>.
gfDNA concentration and intensity of inflammatory cell infiltrates.
Data availability
The datasets generated and/or analyzed during the current study are available either in this published article or in the following publicly accessible repository: https://doi.org/10.5061/dryad.bzkh189qz.
The following dataset was generated:
CadonáFCBartelliTFPelosofAGSztokfiszCZBuenoAPSantosLBBrancoGPSantosGONunesWAPintorFAAbrantesLLDefelicibusACoelhoLVLejaMInHLiSHochsterHCoimbraFJDrummondRSilvaIChokshiRPasqualiniRArapWNunesDNDias-NetoE2025Patient data, demographics, clinical data and DNA concentrationDryad Digital Repository10.5061/dryad.bzkh189qz
Acknowledgements
Authors acknowledge the support given by the institutional Tumor Bank from AC Camargo Cancer Center. We thank Dr. Christian Abnet and Dr. M Constanza Camargo (NCI, NIH, USA) for their critical review of this manuscript.
ReferencesBartelliTFSenda de AbrantesLLFreitasHCThomasAMSilvaJMAlbuquerqueGEAraújoLFBrancoGPde AmorimMGSerpaMSTakenakaIKTMSouzaDTMonçãoLOModaBSValierisRDefelicibusABorgesRDrummondRDAlvesFIASantosMNPBobrovnitchaiaIGElhaikECoelhoLGVKhayatADemachkiSAssumpçãoPPSantiagoKMTorrezanGTCarraroDMPeresSVCalsavaraVFBurbanoRNóbregaCRBaladãoGPPPereiraACCGattiCMFagundesMAAraújoMSMirandaTVBarbosaMSCardosoDMMCarneiroLCBritoAMRamosAFPLSilvaLLLPontesJCTiengoTArantesPESantanaVCordeiroMSant’AnaROAndradeHBAnaissiAKMSampaioSVAbdallahEAChinenLTDBraunACFloresBCTMelloCALClaroLCLSztokfiszCZAltamiranoCCCarterDRFJesusVHFRiechelmannRMedinaTGollobKJMartinsVRSetúbalJCPelosofAGCoimbraFJCosta-JrWLSilvaITNunesDNCuradoMPDias-NetoEGE4GAC group2019Genomics and epidemiology for gastric adenocarcinomas (GE4GAC): a Brazilian initiative to study gastric cancerApplied Cancer Research391410.1186/s41241-019-0081-4ChenYJiaKSunYZhangCLiYZhangLChenZZhangJHuYYuanJZhaoXLiYGongJDongBZhangXLiJShenL2022Predicting response to immunotherapy in gastric cancer via multi-dimensional analyses of the tumour immune microenvironmentNature Communications13485110.1038/s41467-022-32570-z35982052de AlbuquerqueGEModaBSSerpaMSBrancoGPDefelicibusATakenakaIKTMde AmorimMGMiolaECMartinsVCATorresKLBezerraSMClaroLCLPelosofAGSztokfiszCZAbrantesLLSCoimbraFJFKowalskiLPAlvesFAZequiSCUdekwuKISilvaITNunesDNBartelliTFDias-NetoE2022Evaluation of bacteria and fungi DNA abundance in human tissuesGenes1323710.3390/genes1302023735205282Espinosa-CarrascoGChiuEScrivoAZumboPDaveABetelDKangSWJangHJHellmannMDBurtBMLeeHSSchietingerA2024Intratumoral immune triads are required for immunotherapy-mediated elimination of solid tumorsCancer Cell19e01010.1016/j.ccell.2024.01.01038906155ForshewTMurtazaMParkinsonCGaleDTsuiDWYKaperFDawsonSJPiskorzAMJimenez-LinanMBentleyDHadfieldJMayAPCaldasCBrentonJDRosenfeldN2012Noninvasive identification and monitoring of cancer mutations by targeted deep sequencing of plasma DNAScience Translational Medicine4136ra6810.1126/scitranslmed.300372622649089FukudaNSugiyamaYWadaJ2011Prognostic factors of T4 gastric cancer patients undergoing potentially curative resectionWorld Journal of Gastroenterology171180118410.3748/wjg.v17.i9.118021448423Globocan/IARC2022Globocan 2020 – World fact sheethttps://gco.iarc.who.int/media/globocan/factsheets/cancers/7-stomach-fact-sheet.pdfJune 26, 2024HamashimaCOgoshiKOkamotoMShabanaMKishimotoTFukaoA2013A community-based, case-control study evaluating mortality reduction from gastric cancer by endoscopic screening in JapanPLOS ONE8e7908810.1371/journal.pone.007908824236091HothornT2017Maxstat: maximally selected rank statistics0.7-25R packagehttps://CRAN.R-project.org/package=maxstatJiangYZhangQHuYLiTYuJZhaoLYeGDengHMouTCaiSZhouZLiuHChenGLiGQiX2018ImmunoScore signature: A prognostic and predictive tool in gastric cancerAnnals of Surgery26750451310.1097/SLA.000000000000211628002059JiangYXieJHuangWChenHXiSHanZHuangLLinTZhaoL-YHuY-FYuJCaiS-RLiTLiG2019Tumor immune microenvironment and chemosensitivity signature for predicting response to chemotherapy in gastric cancerCancer Immunology Research72065207310.1158/2326-6066.CIR-19-031131615816KimGHJungHY2021Endoscopic resection of gastric cancerGastrointestinal Endoscopy Clinics of North America3156357910.1016/j.giec.2021.03.00834053639LeonSAShapiroBSklaroffDMYarosMJ1977Free DNA in the serum of cancer patients and the effect of therapyCancer Research3764665010.1158/0008-5472.CAN-77-037837366LinRLiCLiuZWuRLuJ2020Genome-wide DNA methylation profiling identifies epigenetic signatures of gastric cardiac intestinal metaplasiaJournal of Translational Medicine1829210.1186/s12967-020-02453-232736574MatsumotoSYoshidaY2014Efficacy of endoscopic screening in an isolated island: a case-control studyIndian Journal of Gastroenterology33464910.1007/s12664-013-0378-223996741Mayo-de-Las-CasasCVelascoASanchezDMartínez-BuenoAGarzón-IbáñezMGatiusSRuiz-MiróMGonzalez-TalladaXLlordellaITresserraFRodríguezSAldeguerERoman-CanalBBertran-AlamilloJGarcía-PeláezBRosellRMolina-VilaMAMatias-GuiuX2020Detection of somatic mutations in peritoneal lavages and plasma of endometrial cancer patients: A proof-of-concept studyInternational Journal of Cancer14727728410.1002/ijc.3287231953839PizziMPBartelliTFPelosofAGFreitasHCBegnamiMDde AbrantesLLSSztokfiszCValierisRKnebelFHCoelhoLGVda CostaWLJrCoimbraFJFda SilvaITde AmorimMGNunesDNDias-NetoE2019Identification of DNA mutations in gastric washes from gastric adenocarcinoma patients: Possible implications for liquid biopsies and patient follow-upInternational Journal of Cancer1451090109810.1002/ijc.3221730779121QianCJuSQiJZhaoJShenXJingRYuJLiLShiYZhangLWangZCongH2017Alu-based cell-free DNA: a novel biomarker for screening of gastric cancerOncotarget8540375404510.18632/oncotarget.1107928903321R Development Core Team2022R: a language and environment for statistical computingVienna, AustriaR Foundation for Statistical Computinghttps://www.R-project.orgRückert-EhebergI-MNoldeMAhnNTauscherMGerlachRGüntnerFGünterAMeisingerCLinseisenJAmannUBaumeisterS-E2022Who gets prescriptions for proton pump inhibitors and why? A drug-utilization study with claims data in Bavaria, Germany, 2010-2018European Journal of Clinical Pharmacology7865766710.1007/s00228-021-03257-z34877614ShaoYYeMJiangXSunWDingXLiuZYeGZhangXXiaoBGuoJ2014Gastric juice long noncoding RNA used as a tumor marker for screening gastric cancerCancer1203320332810.1002/cncr.2888224986041TaharaTArisawaT2015DNA methylation as a molecular biomarker in gastric cancerEpigenomics747548610.2217/epi.15.426077432WuWChungMCM2013The gastric fluid proteome as a potential source of gastric cancer biomarkersJournal of Proteomics9031310.1016/j.jprot.2013.04.03523665003YuJZhangSZhaoB2016Differences and correlation of serum CEA, CA19-9 and CA72-4 in gastric cancerMolecular and Clinical Oncology444144910.3892/mco.2015.7122699830110.7554/eLife.107103.3.sa0eLife AssessmentYangYongliangReviewing EditorShanghai University of Medicine and Health SciencesChinaSolidValuable
This valuable work substantially advances our understanding of prognostic value of total gfDNA in gastric cancer. The evidence supporting the conclusions is solid, supported by a large, well-classified patient cohort and controlled clinical variables. The work will be of broad interest to scientists and clinical pathologist working in the field of gastric cancer.
The study analyzes the gastric fluid DNA content identified as a potential biomarker for human gastric cancer. However, the study lacks overall logicality, and several key issues require improvement and clarification. In the opinion of this reviewer, some major revisions are needed:
(1) This manuscript lacks a comparison of gastric cancer patients' stages with PN and N+PD patients, especially T0-T2 patients.
(2) The comparison between gastric cancer stages seems only to reveal the difference between T3 patients and early-stage gastric cancer patients, which raises doubts about the authenticity of the previous differences between gastric cancer patients and normal patients, whether it is only due to the higher number of T3 patients.
(3) The prognosis evaluation is too simplistic, only considering staging factors, without taking into account other factors such as tumor pathology and the time from onset to tumor detection.
(4) The comparison between gfDNA and conventional pathological examination methods should be mentioned, reflecting advantages such as accuracy and patient comfort.
(5) There are many questions in the figures and tables. Please match the Title, Figure legends, Footnote, Alphabetic order, etc.
(6) The overall logicality of the manuscript is not rigorous enough, with few discussion factors, and cannot represent the conclusions drawn.
Comments on revisions:
The authors have addressed all concerns in the revision.
The authors aimed to evaluate whether total DNA concentration in gastric fluid (gfDNA) collected during routine endoscopy could serve as a diagnostic and prognostic biomarker for gastric cancer. Using a large cohort (n=941), they reported elevated gfDNA in gastric cancer patients, an unexpected association with improved survival, and a positive correlation with immune cell infiltration.
Strengths
The study benefits from a substantial sample size, clear patient stratification, and control of key clinical confounders. The method is simple and clinically feasible, with preliminary evidence linking gfDNA to immune infiltration.
Weaknesses
(1) While the study identifies gfDNA as a potential prognostic tool, the evidence remains preliminary. Unexplained survival associations and methodological gaps weaken support for the conclusions.
(2) The paradoxical association between high gfDNA and better survival lacks mechanistic validation. The authors acknowledge but do not experimentally distinguish tumor vs. immune-derived DNA, leaving the biological basis speculative.
(3) Pre-analytical variables were noted but not systematically analyzed for their impact on gfDNA stability.
Comments on revisions:
To enhance the completeness and credibility of this research, it is essential to clarify the biological origin of gastric fluid DNA and validate these preliminary findings through a prospective, longitudinal study design.
10.7554/eLife.107103.3.sa3Author responseCadonáFrancine CAuthorAC Camargo HospitalSao PauloBrazilBartelliThais FAuthorAC Camargo HospitalSao PauloBrazilPelosofAdriane GAuthorAC Camargo HospitalSao PauloBrazilSztokfiszClaudia ZAuthorAC Camargo HospitalSao PauloBrazilBuenoAdriana PAuthorAC Camargo HospitalSao PauloBrazilSantosLuana BCAuthorAC Camargo HospitalSao PauloBrazilBrancoGabriela PAuthorAC Camargo HospitalSao PauloBrazilSantosGabriel OAuthorAC Camargo HospitalSao PauloBrazilNunesWarley AAuthorAC Camargo HospitalSao PauloBrazilPintorFernanda AAuthorAC Camargo HospitalSao PauloBrazilAbrantesLaís LSAuthorAC Camargo HospitalSao PauloBrazilDefelicibusAlexandreAuthorAC Camargo HospitalSao PauloBrazilCoelhoLuiz Gonzaga VAuthorhttps://ror.org/035rpst33Hospital das Clínicas da Universidade Federal de Minas GeraisBelo HorizonteBrazilLejaMarcisAuthorUniversity of LatviaRigaLatviaInHaejinAuthorRutgers, The State University of New JerseyNewarkUnited StatesLiSharonAuthorRutgersNewarkUnited StatesHochsterHowardAuthorRutgers, The State University of New JerseyNewarkUnited StatesCoimbraFelipe JFAuthorAC Camargo HospitalSao PauloBrazilDrummondRodrigoAuthorAC Camargo HospitalSao PauloBrazilTojal Da SilvaIsraelAuthorAC Camargo Cancer CenterSao PauloBrazilChokshiRaviAuthorRutgers, The State University of New JerseyNewarkUnited StatesPasqualiniRenataAuthorRutgers Cancer Institute of New JerseyNewarkUnited StatesArapWadihAuthorRutgers Cancer Institute of New JerseyNewarkUnited StatesNunesDiana NAuthorRutgers, The State University of New JerseyNewarkUnited StatesDias-NetoEmmanuelAuthorRutgers, The State University of New JerseyNewarkUnited States
The following is the authors’ response to the original reviews.
Reviewer #1 (Public review):
“The study analyzes the gastric fluid DNA content identified as a potential biomarker for human gastric cancer. However, the study lacks overall logicality, and several key issues require improvement and clarification. In the opinion of this reviewer, some major revisions are needed:”
(1) “This manuscript lacks a comparison of gastric cancer patients' stages with PN and N+PD patients, especially T0-T2 patients.”
We are grateful for this astute remark. A comparison of gfDNA concentration among the diagnostic groups indicates a trend of increasing values as the diagnosis progresses toward malignancy. The observed values for the diagnostic groups are as follows:
Diagnosis group
Mean gfDNA and 95% Confidence Interval
Normal mucosa and peptic diseases (N=606)
10.77ng//muL; 95% CI: 9.23 to 12.33;
Preneoplastic conditions (N=99)
10.10ng//muL; 95% Cl: 7.59 to 12.60
T2 and below (N=65)
15.12ng//muL; 95% Cl: 9.73 to 20.50
T3 and above (N=154)
30.75ng//muL; 95% Cl: 14.24 to 36.62
The chart below presents the statistical analyses of the same diagnostic/tumor-stage groups (One-Way ANOVA followed by Tukey’s multiple comparison tests). It shows that gastric fluid gfDNA concentrations gradually increase with malignant progression. We observed that the initial tumor stages (T0 to T2) exhibit intermediate gfDNA levels, which in this group is significantly lower than in advanced disease (p = 0.0036), but not statistically different from non-neoplastic disease (p = 0.74).
(2) “The comparison between gastric cancer stages seems only to reveal the difference between T3 patients and early-stage gastric cancer patients, which raises doubts about the authenticity of the previous differences between gastric cancer patients and normal patients, whether it is only due to the higher number of T3 patients.”
We appreciate the attention to detail regarding the numbers analyzed in the manuscript. Importantly, the results are meaningful because the number of subjects in each group is comparable (T0-T2, N = 65; T3, N = 91; T4, N = 63). The mean gastric fluid gfDNA values (ng/µL) increase with disease stage (T0-T2: 15.12; T3-T4: 30.75), and both are higher than the mean gfDNA values observed in non-neoplastic disease (10.81 ng/µL for N+PD and 10.10 ng/µL for PN). These subject numbers in each diagnostic group accurately reflect real-world data from a tertiary cancer center.
(3) “The prognosis evaluation is too simplistic, only considering staging factors, without taking into account other factors such as tumor pathology and the time from onset to tumor detection.”
Histopathological analyses were performed throughout the study not only for the initial diagnosis of tissue biopsies, but also for the classification of Lauren’s subtypes, tumor staging, and the assessment of the presence and extent of immune cell infiltrates. Regarding the time of disease onset, this variable is inherently unknown--by definition--at the time of a diagnostic EGD. While the prognosis definition is indeed straightforward, we believe that a simple, cost-effective, and practical approach is advantageous for patients across diverse clinical settings and is more likely to be effectively integrated into routine EGD practice.
(4) “The comparison between gfDNA and conventional pathological examination methods should be mentioned, reflecting advantages such as accuracy and patient comfort. “
We wish to reinforce that EGD, along with conventional histopathology, remains the gold standard for gastric cancer evaluation. EGD under sedation is routinely performed for diagnosis, and the collection of gastric fluids for gfDNA evaluation does not affect patient comfort. Thus, while gfDNA analysis was evidently not intended as a diagnostic EGD and biopsy replacement, it may provide added prognostic value to this exam.
(5) “There are many questions in the figures and tables. Please match the Title, Figure legends, Footnote, Alphabetic order, etc. “
We are grateful for these comments and apologize for the clerical oversight. All figures, tables, titles and figure legends have now been double-checked.
(6) “The overall logicality of the manuscript is not rigorous enough, with few discussion factors, and cannot represent the conclusions drawn. “
We assume that the unusual wording remark regarding “overall logicality” pertains to the rationale and/or reasoning of this investigational study. Our working hypothesis was that during neoplastic disease progression, tumor cells continuously proliferate and, depending on various factors, attract immune cell infiltrates. Consequently, both tumor cells and immune cells (as well as tumor-derived DNA) are released into the fluids surrounding the tumor at its various locations, including blood, urine, saliva, gastric fluids, and others. Thus, increases in DNA levels within some of these fluids have been documented and are clinically meaningful. The concurrent observation of elevated gastric fluid gfDNA levels and immune cell infiltration supports the hypothesis that increased gfDNA—which may originate not only from tumor cells but also from immune cells—could be associated with better prognosis, as suggested by this study of a large real-world patient cohort.
In summary, we thank Reviewer #1 for his time and effort in a constructive critique of our work.
Reviewer #2 (Public review):
Summary:
“The authors investigated whether the total DNA concentration in gastric fluid (gfDNA), collected via routine esophagogastroduodenoscopy (EGD), could serve as a diagnostic and prognostic biomarker for gastric cancer. In a large patient cohort (initial n=1,056; analyzed n=941), they found that gfDNA levels were significantly higher in gastric cancer patients compared to non-cancer, gastritis, and precancerous lesion groups. Unexpectedly, higher gfDNA concentrations were also significantly associated with better survival prognosis and positively correlated with immune cell infiltration. The authors proposed that gfDNA may reflect both tumor burden and immune activity, potentially serving as a cost-effective and convenient liquid biopsy tool to assist in gastric cancer diagnosis, staging, and follow-up.”
Strengths:
“This study is supported by a robust sample size (n=941) with clear patient classification, enabling reliable statistical analysis. It employs a simple, low-threshold method for measuring total gfDNA, making it suitable for large-scale clinical use. Clinical confounders, including age, sex, BMI, gastric fluid pH, and PPI use, were systematically controlled. The findings demonstrate both diagnostic and prognostic value of gfDNA, as its concentration can help distinguish gastric cancer patients and correlates with tumor progression and survival. Additionally, preliminary mechanistic data reveal a significant association between elevated gfDNA levels and increased immune cell infiltration in tumors (p=0.001).”
Reviewer #2 has conceptually grasped the overall rationale of the study quite well, and we are grateful for their assessment and comprehensive summary of our findings.
Weaknesses:
(1) “The study has several notable weaknesses. The association between high gfDNA levels and better survival contradicts conventional expectations and raises concerns about the biological interpretation of the findings.“
We agree that this would be the case if the gfDNA was derived solely from tumor cells. However, the findings presented here suggest that a fraction of this DNA would be indeed derived from infiltrating immune cells. The precise determination of the origin of this increased gfDNA remains to be achieved in future follow-up studies, and these are planned to be evaluated soon, by applying DNA- and RNA-sequencing methodologies and deconvolution analyses.
(2) “The diagnostic performance of gfDNA alone was only moderate, and the study did not explore potential improvements through combination with established biomarkers. Methodological limitations include a lack of control for pre-analytical variables, the absence of longitudinal data, and imbalanced group sizes, which may affect the robustness and generalizability of the results.“
Reviewer #2 is correct that this investigational study was not designed to assess the diagnostic potential of gfDNA. Instead, its primary contribution is to provide useful prognostic information. In this regard, we have not yet explored combining gfDNA with other clinically well-established diagnostic biomarkers. We do acknowledge this current limitation as a logical follow-up that must be investigated in the near future.
Moreover, we collected a substantial number of pre-analytical variables within the limitations of a study involving over 1,000 subjects. Longitudinal samples and data were not analyzed here, as our aim was to evaluate prognostic value at diagnosis. Although the groups are imbalanced, this accurately reflects the real-world population of a large endoscopy center within a dedicated cancer facility. Subjects were invited to participate and enter the study before sedation for the diagnostic EGD procedure; thus, samples were collected prospectively from all consenting individuals.
Finally, to maintain a large, unbiased cohort, we did not attempt to balance the groups, allowing analysis of samples and data from all patients with compatible diagnoses (please see Results: Patient groups and diagnoses).
(3) “Additionally, key methodological details were insufficiently reported, and the ROC analysis lacked comprehensive performance metrics, limiting the study's clinical applicability.“
We are grateful for this useful suggestion. In the current version, each ROC curve (Supplementary Figures 1A and 1B) now includes the top 10 gfDNA thresholds, along with their corresponding sensitivity and specificity values (please see Suppl. Table 1). The thresholds are ordered from-best-to-worst based on the classic Youden’s J statistic, as follows:
Youden Index = specificity + sensitivity – 1 [Youden WJ. Index for rating diagnostic tests. Cancer 3:32-35, 1950. PMID: 15405679]. We have made an effort to provide all the key methodological details requested, but we would be glad to add further information upon specific request.
Reviewer #1 (Recommendations for the authors):
The authors should pay attention to ensuring uniformity in the format of all cited references, such as the number of authors for each reference, the journal names, publication years, volume numbers, and page number formats, to the best extent possible.
Thank you for pointing this inconsistency. All cited references have now been revisited and adjusted properly. We apologize for this clerical oversight.
Reviewer #2 (Recommendations for the authors):
(1) “High gfDNA levels were surprisingly linked to better survival, which conflicts with the conventional understanding of cfDNA as a tumor burden marker. Was any qualitative analysis performed to distinguish DNA derived from immune cells versus tumor cells?“
Tumor-derived DNA is certainly present in gfDNA, as our group has unequivocally demonstrated in a previous publication [Pizzi M. P., et al. (2019) Identification of DNA mutations in gastric washes from gastric adenocarcinoma patients: Possible implications for liquid biopsies and patient follow-up Int J Cancer 145:1090–1097. DOI: 10.1002/ijc.32114]. However, in the present manuscript, our data suggest that gfDNA may also contain DNA derived from infiltrating immune cells. This may also be the case for other malignancies, and qualitative deconvolution studies could provide more informative information. To achieve this, DNA sequencing and RNA-Seq analyses may offer relevant evidence. Our study should be viewed as an original and preliminary analysis that may encourage such quantitative and qualitative studies in biofluids from cancer patients. Currently, this is a simple approach (which might be its essential beauty), but we hope to investigate this aspect further in future studies.
(2) “The ROC curve AUC was 0.66, indicating only moderate discrimination ability. Did the authors consider combining gfDNA with markers such as CEA or CA19-9 to improve diagnostic accuracy?“
This is indeed a logical idea, which shall certainly be explored in planned follow-up studies.
(3) “DNA concentration could be influenced by non-biological factors, including gastric fluid pH, sampling location, time delay, or freeze-thaw cycles. Were these operational variables assessed for their effect on data stability?“
We appreciate the rigor of the evaluation. Yes, information regarding gastric fluid pH was collected. All samples were collected from the stomach during EGD procedure. Samples were divided in aliquots and were thawed only once. This information is now provided in the updated manuscript text.
(4) “This cross-sectional study lacks data on gfDNA changes over time, limiting conclusions on its utility for monitoring treatment response or predicting recurrence.“
Again, temporal evaluation is another excellent point, and it will be the subject of future analyses. In this exploratory study, samples were collected at diagnosis, at a single point. We have not obtained serial samples, as participants received appropriate therapy soon following diagnosis.
(5) The normal endoscopy group included only 10 patients, the precancerous lesion group 99 patients, while the gastritis group had 596 patients. Such uneven sample sizes may affect statistical reliability and generalizability. Has weighted analysis or optimized sampling been considered for future studies?“
Yes, in future studies this analysis will be considered, probably by employing stratified random sampling with relevant patient attributes recorded.
(6) “The SciScore was only 2 points, indicating that key methodological details such as inclusion/exclusion criteria, randomization, sex variables, and power calculation were not clearly described. It is recommended that these basic research elements be supplemented in the Methods section. “
This was an exploratory research, the first of its kind, to evaluate prognostic potential of gfDNA in the context of gastric cancer. Patients were not included if they did not sign the informed consent or excluded if they withdrew after consenting. Other exclusion criteria included diagnoses of conditions such as previous gastrectomy or esophagectomy, or the presence of non-gastric malignancies. Randomization and power analyses were not applicable, as no prior data were available regarding gfDNA concentration values or its diagnostic/prognostic potential. All subjects, regardless of sex, were invited to participate without discrimination or selection.
(7) “Although a ROC curve was provided in the supplementary materials (Supplementary Figure 1), only the curve and AUC value were shown without sensitivity, specificity, predictive values, or cutoff thresholds. The authors are advised to provide a full ROC performance assessment to strengthen the study's clinical relevance.
These data are now given alongside the ROC curves in the Supplementary Information section, specifically in Supplementary Figure 1 and in the newly added Supplementary Table 1.
We thank Reviewer #2 for an insightful and positive overall assessment of our work.