The cohort of sixty-three GEAC patients was primarily composed of males (87.3%) and patients aged between 60 and 69 years (38.1%). Most patients were white (93.7%) and had stage 4 disease (63.5%) or stage 3 disease (12.7%). Forty-one biopsies (65.1%) were from a primary tumor and twenty-one patients (33.3%) had a metastatic site biopsied. The esophagus was the most common tumor location (58.7%) followed by the stomach (23.8%), and gastroesophageal junction (17.5%). Most of the patients had an overweight BMI (71.4%) and a positive PD-L1 IHC status (61.9%). Eighteen patients in the cohort were treated with an immune checkpoint inhibitor (286%). Characteristics of the clinical cohort are summarized in Table 1.
Table 1
Demographic and Disease Characteristics of the Study Cohort
Characteristics | n (%) (N = 63) |
Age at Initial Diagnosis | |
30–39 | 5 (7.9%) |
40–49 | 4 (6.3%) |
50–59 | 12 (19%) |
60–69 | 24 (38.1%) |
70–79 | 16 (25.4%) |
≥ 80 | 2 (3.2%) |
Gender | |
Male | 55 (87.3%) |
Female | 8 (12.7%) |
Race | |
White | 59 (93.7%) |
Black | 2 (3.2%) |
Asian | 1 (1.6%) |
American Indian | 1 (1.6%) |
BMI | |
< 25 | 18 (28.6%) |
≥ 25 | 45 (71.4%) |
Stage | |
1 | 2 (3.1%) |
2 | 6 (9.5%) |
3 | 8 (12.7%) |
4 | 40 (63.5%) |
No Data | 7 (11.1%) |
Tumor Location | |
Stomach | 15 (23.8%) |
Esophagus | 37 (58.7%) |
Gastroesophageal Junction | 11 (17.5%) |
Primary or Metastatic | |
Primary | 41 (65.1%) |
Metastatic | 21 (33.3%) |
No Data | 1 (1.6%) |
PD-L1 IHC Status (CPS ≥ 1) | |
High | 39 (61.9%) |
Low | 14 (22.2%) |
No Data | 10 (15.9%) |
Immunotherapy Received | |
No | 45 (71.4%) |
Yes | 18 (28.6%) |
Demographics of the patient population in the study cohort including age, gender, race, and BMI are listed above. Characteristics of the disease including tumor location, stage of cancer, primary/metastatic, PD-L1 IHC status, and ICI treatment status are also listed. All values are given as total numbers and percentages. *N = number, BMI = body mass index, PD-L1 = Programmed death ligand-1, IHC = Immunohistochemistry, CPS = Combined positive score |
Additionally, we identified 329 GEAC patients from the TCGA database to validate the findings in the clinical cohort identified at our institution. The TCGA cohort was primarily composed of male patients (70.5%) between the ages of 70 to 79 years (32.2%). Most patients were white (74.5%) and had stage 2 (37.1%) or stage 3 disease (36.8%). (Table S1 of Online Resource 1).
The prevalence of CTA genes in the primary cohort is given in Fig. 1. The majority of genes range from a high of 58% for MAGEA3 to 19% for SSX2. MLANA, primarily a melanoma-associated gene, is an exception with a prevalence of only 3%. Apart from MAGEC2 (32%) and GAGE13 (21%), the MAGE and GAGE family of genes exhibit higher expression levels (38% − 58% and 38% − 49% respectively), while BAGE (24%), NY-ESO-1(27%), XAGE1B (33%) and LAGE1A (35%) show the opposite trend. In the TCGA cohort, MLANA was found the be the most expressed (95%) with GAGE12J being the least expressed (9%). Similar to the primary cohort, the MAGE family of genes was found to exhibit higher expression levels, ranging from 57% for MAGEA3 to 35% for MAGEC2. NY.ESO.1 (23%), BAGE (26%), and LAGE1A (28%) were again found to have lower expression levels. The prevalence of CTA genes in the TCGA cohort is given in Figure S1 of Online Resource 1.
We performed Pearson correlation among the seventeen CTA genes in the primary cohort and eleven genes in the TCGA cohort to identify co-expression clusters as potential targets of multimodal treatment. The correlation matrix as well as the network plot for the primary cohort are given in Fig. 2. We found a single cluster of genes as well as two genes separated from the others. The cluster includes MAGEA4, MAGEA3, NY.ESO.1, MAGEC2, MAGEA1, MAGEA10, MAGEA12, GAGE13, BAGE, LAGE1A, GAGE1.GAGE12L.GAGE12F, SSX2, GAGE12J, GAGE2C.GAGE2A.GAGE2E. XAGE1B is also part of the cluster but slightly offset. Both Gage10 and MLANA are separate from the cluster. The R2 Pearson correlation coefficient values in this cluster range from a low of 0.27 between XAGE1B and MAGEA4 to a high of 0.73 between the MAGE group of genes and the GAGE group of genes. Additionally, NY.ESO.1 had a high correlation with MAGEA1, MAGEA10, LAGE1A, MAGEA12, and GAGE1.GAGE12l.GAGE12F. In the primary cohort, MLANA was only correlated with MAGEA4, and MAGEA10 while GAGE10 was not co-expressed with any other CTA gene. The TCGA cohort showed similar results, displaying a single cluster of CTA genes with MLANA being separate (Fig. 3). The cluster in the TCGA cohort consisted of BAGE, GAGE12J, MAGEA12, NY.ESO.1, LAGE1A, MAGEA10, MAGEA1, MAGEA4, MAGEC2, AND MAGEA3. The R2 Pearson correlation coefficient values in this cluster range from a low of 0.33 between BAGE and MAGEC2 to a high of 0.77 between MAGEA3 and MAGEA2. Additionally, NY.ESO.1 showed high co-expression with MAGEA12, MAGEA3, LAGE1A, and GAGE12J. MLANA was only correlated with BAGE (R2 = 0.11).
We also performed survival analysis using the Kaplan-Meier method in the primary cohort as well as the TCGA cohort. In the primary cohort, we found that patients positive for MAGEC2 had significantly better survival than patients negative for MAGEC2 (positive vs. negative median survival in months, p-value; 31 months vs. 21 months, p = 0.034). We did not find any other significant trends with the other CTA genes. In the TCGA cohort, we found that patients with positive for MAGEA4 had significantly worse survival than patients negative for MAGEA4 (25 months vs. 32 months, p = 0.032). Additionally, patients positive for MAGEA1 had significantly worse survival than patients negative for MAGEA1 (23 months vs. 31 months, p = 0.013). One reason for the reversal of trends seen between the primary cohort and the TCGA cohort could be since the TCGA database was compiled prior to the introduction of immunotherapy. Therefore, we repeated the survival analysis in the primary cohort, splitting patients into subgroups depending on ICI treatment status.
Here, we found several significant results. In the non-ICI treated subgroup, MAGEC2-positive patients had worse overall survival than MAGEC2-negative patients (9 months vs. 30 months, p = 0.013). Similarly, GAGE2C.GAGE2A.GAGE2E-positive patients fared worse than GAGE2C.GAGE2A.GAGE2E negative patients (15 months vs. 29 months, p = 0.05). In the ICI-treated subgroup, GAGE13-positive patients tended to have better survival than GAGE13-negative patients, though the trend was not significant (72 months vs 13 months, p = 0.077). Similarly, SSX2 positive patients trended to better survival than SSX2 negative patients in the ICI-treated group (p = 0.068). MAGEA1 positive patients (34 months vs. 14.5 months, p < 0.01), MAGEA12 positive patients (34 months vs. 18.5 months, p = 0.043), MAGEA3 positive patients (34 months vs. 14.5 months, p = 0.028), MAGEA4 positive patients (72 months vs. 13 months, p < 0.01), and NY.ESO.1 positive patients (72 months vs. 18.5 months, p = 0.027) all had significantly higher overall survival than those negative for the corresponding CTA genes in the ICI-treated subgroup. The Kaplan Meier plots for the significant survival analyses are given in Figures S2, S3, and S4 of Online Resource 1.