Tumor Microenvironment Prognostic Risk and Its Association With MUC5AC in Ampullary Carcinoma

This study was approved by the Institutional Review Board of Incheon St Mary’s Hospital (OC13SISI0162; Incheon, Republic of Korea). A total of 64 surgically resected primary AC cases were collected. The requirement for patient consent was waived because of the retrospectively obtained and anonymized data of this study. Carcinomas in which the epicenter was in the ampulla were included, whereas carcinomas extending from the duodenum, extrahepatic CBD, and pancreas were excluded. Cases with neoadjuvant chemotherapy were excluded from this study.

Clinical data included patient age and sex, operation date, most recent follow-up date, chemotherapy, radiotherapy, recurrence date, and survival status of patients. Pathologic data included tumor size and growth pattern (polypoid, nodular, or infiltrative); histologic subtype and grade; site-specific classification; pancreatic, duodenal, and CBD invasions; nodal metastasis; perineural and lymphovascular invasion; margin status; and stage grouping based on the 8th edition of the AJCC cancer staging system.³  Histologic subtypes and tumor grades followed the 5th edition of the World Health Organization (WHO) classification.²  ACs were further categorized into 4 categories according to the site-specific classification proposed by Adsay et al¹¹: (1) “intraampullary,” carcinomas arising in intraampullary papillary-tubular neoplasms; (2) “ampullary-ductal,” tumors forming sclerotic thickening of the walls of the CBD and/or pancreatic duct; (3) “periampullary-duodenal,” exophytic tumors growing into the duodenal lumen and/or eccentrically encasing the ampullary orifice; and (4) “ampullary-NOS,” ampullary tumors lacking any of the specific characteristics of the other categories.¹¹  Histologic and immunohistochemical assessment was performed using an Olympus BX-53 microscope (Olympus, Tokyo, Japan) by an experienced GI pathologist blinded to clinicopathologic information.

BD, defined as a single cell or clusters of as many as 4 tumor cells, was counted in 1 hot spot of a ×20 objective field (area, 0.785 mm²) on hematoxylin-eosin–stained slides and was graded as BD1 (0–4), BD2 (5–9), or BD3 (≥10), according to the ITBCC criteria.²⁹  Stromal TIL densities were calculated using a ×20 objective and scored as the percentage of the stromal area occupied by TILs over the total intratumoral stromal areas, as recommended by the ITWG.^27,28  Stromal TILs were classified as 0%, 1% (≤1%), 5% (2%–5%), 10% (6%–10%), or in additional 10% increments, and the mean stromal TIL density was calculated using 10 randomly selected fields, as previously reported.²⁷  Receiver operating characteristic curve analyses were performed to maximize the sensitivity and specificity of BD and TIL in predicting overall survival (OS), as described in previous studies.^23,27 

Staining for MUC1 (Ma695, Leica Biosystems Inc), MUC2 (CCP58, Leica Biosystems Inc), MUC5AC (CLH2, Dako, Carpinteria, California), CDX2 (EPR2764Y, Roche Diagnostics, Indianapolis, Indiana), and CK20 (KS20.8, Dako) was performed using a Benchmark XT immunostainer (Ventana Medical System, Tucson, Arizona) following the manufacturers’ protocol. The percentages of cytoplasmic (MUC1, MUC2, MUC5AC, and CK20) and nuclear (CDX2) labeling were scored.

We classified ACs into intestinal, pancreatobiliary, and ambiguous immunophenotypes according to the schema proposed by Ang and colleagues.¹²  More than 25% of tumor labeling for each antibody was considered positive.¹²  “Intestinal type” was defined as (1) CK20⁺ or CDX2⁺ or MUC2⁺ and MUC1⁻, or (2) CK20⁺, CDX2⁺, and MUC2⁺, regardless of MUC1 result; and “pancreatobiliary type” was defined as MUC1⁺, CDX2⁻, and MUC2⁻, irrespective of CK20 result. Tumors not fitting one of the above-described categories were regarded as “ambiguous type.”¹²  In addition, we analyzed MUC5AC immunoreactivity with the cutoff of greater than 20%, as previously described by Xue et al.¹⁴ 

Immunohistochemical staining was performed for MLH1 (clone M1, Roche Diagnostics), MSH2 (clone G219-1129, Cell Marque, Rocklin, California), MSH6 (clone 44, Cell Marque), and PMS2 (clone EP51, Dako). Loss of protein expression was defined as complete absence of nuclear staining of tumor cells.²⁷  For microsatellite instability (MSI) analysis, 5 quasi-monomorphic mononucleotide repeats (BAT25, BAT26, NR21, NR24, and NR27) were amplified in a single multiplex polymerase chain reaction. According to the National Cancer Institute (NCI) guidelines, tumors were classified as MSI-low (instability at a single mononucleotide locus), MSI-high (instability at ≥2 mononucleotide loci), and microsatellite stable (no instability at any of the loci tested).²⁷ 

The SPSS Statistics for Windows (version 28.0, IBM, Armonk, New York) and Harrell rms R package (version 3.6.3; https://cran.r-project.org/web/packages/rms/index.html) were used for the analysis.^22,23  Harrell rms is characterized by its ability to easily plot regression effects of any predictors and save a summary of the distribution of predictors with the fit (http://www.rdocumentation.org/packages/rms/versions/6.2-0/topics/rmsOverview). Categoric variables were assessed using the χ² and/or Fisher exact test, and continuous variables were evaluated using simple analysis of variance. Survival curves were plotted using the Kaplan-Meier method, and the statistical significances were tested using the log-rank test and the Cox proportional hazards model. OS and recurrence-free survival (RFS) were estimated from the date of surgery to the date of event (death or last follow-up of the patients in OS; recurrence of cancer in RFS). The relative contribution (%) of each factor to OS and RFS was estimated using χ² based on multivariable Cox regression models.^22,23  Multivariate relationships were analyzed by fitting logistic regression models with calculated hazard ratio (HR) and 95% CI. A P value of <.05 was considered statistically significant.

Most cases (63 of 64; 98%) received pancreaticoduodenectomies, including pylorus-preserving pancreaticoduodenectomies (17 cases) or classic Whipple operations (46 cases). Ampullectomy was performed in 1 case (2%). The follow-up period after surgical resection ranged from 1.1 to 215.8 months (mean, 48.7 months). During follow-up, 44% (28 of 64) of patients showed cancer recurrence; the median time to recurrence was 35.2 months.

The baseline clinicopathologic findings are described in Table 1. The mean patient age was 65.4 ± 10.4 years (range, 39–85 years), and approximately half of the patients were female. The tumor size ranged from 0.4 to 6.0 cm (mean, 2.5 ± 1.2 cm). Most tumors (51 of 64; 80%) showed an infiltrative growth pattern, whereas polypoid and nodular patterns were observed in 12 cases (12 of 64; 19%) and 1 case (1 of 64; 1%), respectively. Tubular adenocarcinomas were commonly found in 58 cases (58 of 64; 92%), followed by 3 mucinous carcinomas (5%) and 1 each of poorly cohesive cell carcinoma (1%), squamous carcinoma (1%), and mixed neuroendocrine-nonendocrine neoplasm (MiNEN; 1%). Regarding histologic grades, 42 (42 of 64; 66%) ACs were moderately differentiated tumors, 16 (25%) were well-differentiated ones, and 6 (9%) were poorly differentiated ones. Among the 64 ACs, 21 tumors (33%) were classified as ampullary-ductal type, 16 (25%) as intraampullary type, and 4 (6%) as periampullary-duodenal type; the other 23 cases (36%) were categorized into ampullary-NOS. ACs involved the duodenum in 52 cases (52 of 64; 81%), extending into the submucosa (10 of 64; 16%), proper muscle layer (9 of 64; 14%), subserosa (32 of 64; 50%), and serosa (1 of 64; 1%) of the duodenum. Pancreatic invasion was observed in half of the ACs, and 12 cases (12 of 64; 18%) showed an involved tumor size of 0.5 cm or less. Tumor size exceeding 0.5 cm in pancreatic invasion was observed in 10 cases (10 of 64; 16%), and peripancreatic soft tissue involvement was also observed in 10 cases (10 of 64; 16%). CBD involvement was found in 19 cases (19 of 64; 30%), lymphovascular invasion in 30 cases (30 of 64; 47%), and perineural invasion in 24 cases (24 of 64; 38%). Resection margins were clear except for 1 case (1 of 64; 2%). Peritumoral adenomas were observed in 29 cases (29 of 64; 45%). Regarding the AJCC staging scheme, 9 (9 of 64; 14%) cases were T1a, 13 cases (20%) were T1b, 11 cases (17%) were T2, 4 cases (6%) were T3, and 27 cases (43%) were T3b; no T4 cases were observed. Lymph node metastases were identified in 28 cases, including 23 N1 (23 of 64; 36%) and 5 N2 (5 of 64; 8%). Distant metastases were only found in 3 cases (3 of 64; 5%). Tumors were classified into stages IA (8 of 64; 12%), IB (14; 22%), IIA (2; 3%), IIB (12; 19%), IIIA (22; 34%), IIIB (3; 5%), and IV (3; 5%). Additional chemotherapy and/or radiotherapy was performed in 18 patients (28%), including chemotherapy alone in 9 patients, radiotherapy alone in 1 patient, and chemoradiation therapy in 8 patients.

Immunohistochemically, tumors were positive for CDX2 in 34 cases (34 of 57; 60%), CK20 in 23 cases (23 of 58; 40%), MUC1 in 27 cases (27 of 58; 47%), and MUC2 in 7 cases (7 of 58; 12%). Using the criteria of Ang et al,¹²  there were 26 (26 of 58; 44%) intestinal, 16 (16 of 58; 28%) pancreatobiliary, and 16 (16 of 58; 28%) ambiguous types. MUC5AC was positive in 25 cases (25 of 57; 44%).¹² 

BD was variably observed; tumors were categorized as BD1 in 17 cases (17 of 64; 27%), BD2 in 13 cases (20%), and BD3 in 34 cases (53%), based on the ITBCC scoring system for BD. The mean stromal TIL density was 23.1% ± 21.2% (range, 0%–90.0%). Using receiver operating characteristic curve analysis, BD3 and stromal TIL density greater than 5% were defined as constituting high levels, and high levels of BD and stromal TIL were observed in 34 (34 of 64; 53%) and 45 (45 of 64; 70%) cases (Figure 1). The area under the curve (AUC) at the BD3 cutoff for BD was 0.66 (95% CI, 0.53–0.78), which indicated fair sensitivity in discriminating OS. Stromal TIL showed poor accuracy in predicting OS, with an AUC of 0.59 (95% CI, 0.46–0.71). Of the 64 patients, 58 were available for MSI analysis: All 56 cases (56 of 58; 97%) with retained expression of the 4 mismatch repair proteins had microsatellite stability as expected, whereas 2 cases (2 of 58; 3%) exhibiting PMS2⁻ (n = 1) and MSH6⁻/MSH2⁻ (n = 1) were revealed as MSI. The case with PMS2⁻ was MSI-low, and the one with MSH6⁻/MSH2⁻ was MSI-high.

The association between clinicopathologic factors and survival is listed in Table 2. In univariate analysis for OS, margin positivity (P < .001), higher tumor grade (P = .05), higher stage grouping (P = .05), high levels of BD (P = .01) and low levels of TILs (P = .03) were associated with worse OS. Univariate analysis for RFS revealed that higher tumor grade (P = .04), duodenal invasion (P = .03), lymphovascular invasion (P = .01), margin positivity (P < .001), nodal metastasis (P < .001), higher stage grouping (P = .01), chemotherapy (P = .03), high levels of BD (P = .01), and low levels of TILs (P = .02) were related to worse RFS. MUC1 immunopositivity tended to be related to worse RFS, but this was statistically borderline (P = .06). MUC5AC expression and mismatch repair gene status did not correlate with OS or RFS.

We combined BD and TIL, which individually showed prognostic value, to categorize cases into 4 groups: BD^Low/TIL^High (26 of 64; 41%), BD^Low/TIL^Low (4; 6%), BD^High/TIL^High (19; 30%), and BD^High/TIL^Low (15; 23%). Figure 2, A, shows the survival analyses of the BD and TIL groups for OS. Patient survival rates were significantly related to the 4 BD/TIL groups (P = .02). The 5-year survival rates (YSR) of the BD^Low/TIL^High, BD^Low/TIL^Low, BD^High/TIL^High, and BD^High/TIL^Low groups were 96.0%, 75.0%, 56.0%, and 50.9%, respectively. In pairwise comparisons, patients with BD^Low/TIL^High showed higher survival rates than those with BD^High/TIL^High (P = .05) and BD^High/TIL^Low (P = .001). However, there was no significant difference in the survival rate between patients with BD^Low/TIL^High and those with BD^Low/TIL^Low (P = .32), between patients with BD^Low/TIL^Low and those with BD^High/TIL^High (P = .80), and between patients with BD^High/TIL^High and those with BD^Low/TIL^High (P = .15). In further analysis, we grouped BD^Low/TIL^Low and BD^High/TIL^High together and generated a collective TME prognostic risk index: low-risk (BD^Low/TIL^High; 26 of 64; 41%), intermediate-risk (BD^Low/TIL^Low or BD^High/TIL^High; 23; 36%), and high-risk (BD^High/TIL^Low; 15; 23%). This TME-based prognostic risk index significantly discriminated patient survival (P = .008; Figure 2, B). The 5-YSRs for patients with low, intermediate, and high risk were 96.0%, 62.5%, and 50.9%, respectively. By pairwise comparisons, patients with low risk had significantly higher survival rates than those with high risk (P = .001). However, there was no significant difference in survival rate between the low- and intermediate-risk groups (P = .07) and between the intermediate- and high-risk groups (P = .13).

Survival analyses of the combined BD and TIL groups for RFS are shown in Figure 2, C. The survival times of patients for RFS were significantly associated with the 4 BD/TIL groups (P = .02). All patients with BD^Low/TIL^High were alive at the end of the study. The median survival times of BD^Low/TIL^Low, BD^High/TIL^High, and BD^High/TIL^Low groups were 62.0, 57.5, and 29.3 months, respectively. Pairwise comparisons revealed that patients with BD^Low/TIL^High had longer survival times than those with BD^High/TIL^Low (P = .001). However, no significant differences in survival times were seen between patients with BD^Low/TIL^High and those with BD^Low/TIL^Low (P = .27), between patients with BD^Low/TIL^Low and those with BD^High/TIL^High (P = .75), and between patients with BD^High/TIL^High and those with BD^High/TIL^Low (P = .10). We then performed survival analysis for RFS of the TME prognostic risk index and found significant differences in RFS (P = .006; Figure 2, D). As mentioned above, all patients with low risk were alive at the end of the study. The median survival times for patients with intermediate risk and high risk were 57.5 and 29.3 months, respectively. In pairwise comparisons, patients with low risk had significantly longer survival times than those with high risk (P = .001). However, there was no significant difference in survival times between the low- and intermediate-risk groups (P = .09) and between the intermediate- and high-risk groups (P = .07).

We compared the relative contributions of BD, TIL, and the TME prognostic risk index to OS and RFS. In multivariable models with BD, the largest contributor to OS was BD (46.1%), followed by margin status (31.6%), stage grouping (19.5%), and tumor grade (2.8%; Figure 3, A). The contribution of each TIL (Figure 3, B) or the TME prognostic risk index (Figure 3, C) to OS also ranked first. When analyzing the relative weights for the contribution of BD, TIL, and the TME prognostic risk index, the TME risk index showed the best predictive value for OS (53.9 in TME risk index versus 46.1 in BD versus 42.2 in TIL; Figure 3, D).

In multivariable models for RFS with BD, the top contributor was margin status (35.9%), followed by chemotherapy (17.7%), BD (16.9%), lymphovascular invasion (14.1%), nodal metastasis (11.2%), tumor grade (3.4%), and stage grouping (0.8%; Figure 4, A). In models with TIL, TIL ranked third (15.3%) behind margin status (36.5%) and chemotherapy (18.1%; Figure 4, B). The contribution of the TME prognostic risk index to RFS was increased, ranking second (24.8%) behind margin status (32.4%; Figure 4, C). In the relative weight analysis of the contribution of BD, TIL, and the TME prognostic risk index, the predictability of the TME risk index for RFS was highest (24.8 for TME risk index versus 16.9 for BD versus 15.3 for TIL; Figure 4, D).

In multivariate analysis, higher TME prognostic risk was an independent prognosticator for both OS (P = .003) and RFS (P = .03; Table 3). In addition, high stage IV was related to worse OS (P = .007), whereas lymphovascular invasion was associated with shorter survival time for RFS (P = .05).

The prognostic impact of the TME risk index for predicting OS was investigated with respect to site-specific classification, immunophenotypic or histologic subtype, and stage (Figure 5).

With regard to immunophenotypic or histologic subtypes, the TME risk index significantly stratified the survival of patients with either the nonintestinal immunophenotype or the tubular type of AC. In the intestinal immunophenotype (n = 26), the OS rate of patients tended to be adversely related to TME risk index, but no statistical significance was identified (P = .67; Figure 5, A). For the nonintestinal immunophenotype (n = 32), OS rates tended to be different according to TME risk (P = .05; Figure 5, B). The 3-YSRs of patients with low, intermediate, and high risk were 100%, 80.2%, and 51.4%, respectively. In terms of histologic type, the survival rate of patients with the tubular type of ACs (n = 58) was affected by TME risk (P = .03; Figure 5, C). The 5-YSRs of patients with low, intermediate, and high risk were 95.5%, 61.2%, and 55.7%, respectively. In the nontubular type of ACs (n = 6), no difference in OS rate was found (P = .22; Figure 5, D). TME risk index also significantly stratified the survival of patients in the lower stage group. In the lower stage group (stages I to III, n = 61), the OS rates for patients were significantly different according to TME risk (P = .004; Figure 5, E). The 5-YSRs for the low-, intermediate- and high-risk groups were 95.8%, 68.6%, and 50.9%, respectively. In the high stage group (stage IV, n = 3), no statistical significance was identified (P = .23; Figure 5, F). In terms of site-specific classification, the survival rate of patients was not affected by TME risk.

The relationship between TME prognostic risk and clinicopathologic factors is presented in Table 4. Higher TME prognostic risk was more commonly found in tumors with poor differentiation (P = .03) and lymphovascular invasion (P = .05). Interestingly, tumors with higher TME prognostic risk more frequently expressed MUC5AC (P = .02).

We report the clinical utility of assessing TME risk via BD and TIL to predict the prognosis for patients with ACs, including OS and RFS, and the association of TME prognostic risk index with MUC5AC expression.

In this study, we dichotomized BD and TIL according to cutoffs optimized for correlation with OS. The AUC value of BD at the BD3 cutoff was larger (0.66; 95% CI, 0.53–0.78) than that of the BD2 (0.62; 95% CI, 0.49–0.74), so we defined BD3 as high level. When using a cutoff point of BD3, the sensitivity was 78.6% and specificity was 54.0%. We also analyzed the AUC value of TIL to maximize the sensitivity and specificity in predicting OS. When using a cutoff point of 5% for TILs, an optimal sensitivity of 50.0% and specificity of 76.0% could be obtained. In a recent SIAC study, the frequency of high BD (BD3) was reported to be 67.4% (155 of 230), which is relatively higher than that (53%; 34 of 64) of AC in this study.²⁶  This difference may be due to the higher number of advanced-stage SIAC cases than in AC. Although 50.7% (108 of 213) of tumors were high-stage (stages III and IV) disease in the SIAC study,²⁶  in this study, 44% (28 of 64) were high-stage ACs. However, the cutoff level of stromal TIL in this study was lower than that in the SIAC study (5% versus 20%),²⁷  because the ampullary mucosa has fewer immunocompetent cells than the mucosa of the small intestine.

BD is a key component of the TME and contributes to the invasiveness of tumors via epithelial-mesenchymal transition.²⁴  BD is a robust prognostic biomarker in malignancies, particularly in CRCs. The ITBCC system standardized the grading and evaluation of BD in 2016 and has since been validated in many CRC studies.²⁴  Studies of other carcinomas of the GI tract followed the ITBCC grading system, with a cutoff of either BD2 or BD3 used to define a high level of BD.^24,26  However, a few studies applied a different cutoff (≥3 buds), similar to a previous study in AC by Ohike and colleagues.⁹  Ohike et al⁹  identified the predictive role of high BD count (≥3 buds) for shorter OS of 244 patients with AC. Although we applied the ITBCC grading system for BD, we found the same worse prognostic effect of BD in ACs for both OS and RFS. Hence, BD was confirmed as a significant prognostic factor in ACs.

In the TME, the innate immune system interacts with BD, which is referred to as the attacker–defender model.³⁰  CD8⁺ T cells and FOXP3⁺ regulatory T cells often surround budding tumor cells and stromal macrophages can engulf them, but BD can escape from the immune system via loss of MHC class I molecule expression on the cell surface.^24,31  Recent studies of CRC that address the interaction of BD with immune cells have shown that a high density of CD8⁺ T cells at the tumor center was associated with a favorable outcome,³²  and combining BD and CD8⁺ T-cell count in a ratio improved the prediction of OS.³³  Other studies have evaluated the prognostic impact of a combined factor of BD and intratumoral stromal TIL, independent of immune cell lineage.^20,23  Similar to our study, Lang-Schwarz et al²⁰  identified a shorter survival time of the BD^High/TIL^Low group compared with that of other groups (BD^Low/TIL^High, BD^High/TIL^High, and BD^Low/TIL^Low) through evaluation of BD and stromal TIL together in CRC cases. In SIAC, the combination of BD and TIL was found as a powerful prognostic predictor, regardless of tumor location, staging, and mismatch repair gene status.²³  In this study, we found that TME risk index was significantly related to the OS of patients with nonintestinal immunophenotype (P = .05), tubular adenocarcinoma (P = .03), and low-stage disease of ACs (P = .004). Although not statistically significant, there was a trend for patient survival to be influenced by the TME risk index in patients with the intestinal immunophenotype (26 cases), nontubular carcinoma (6), and high-stage disease (3) of ACs. Therefore, further studies including more cases are needed to clearly define the prognostic impact of the TME index in AC.

MUC5AC is a heavily glycosylated gel-forming mucin expressed in the normal tissues from the gastric pits, gallbladder, and bronchial tract.³⁴  The prognostic relevance of aberrant MUC5AC expression in GI malignancies, particularly in pancreatic ductal adenocarcinomas, has been studied, but the results were inconclusive.^35–38  In Crohn disease–associated SIAC, MUC5AC was reported to predict worse survival.³⁹  Most immunohistochemical studies for MUC5AC in GI malignancies used CLH2 monoclonal antibodies and showed poor outcomes when the cutoff level of MUC5AC expression was set high.^14,34,37,38  In previous studies on ACs, MUC5AC expression was almost entirely evaluated using the CLH2 antibody, and the relationship of MUC5AC expression to survival has been reported with various cutoffs (Table 5).^{14,38,40–53}  Only 29% (2 of 7) of these studies, including studies by Xue et al¹⁴  and Aloysius et al,³⁸  described adverse effects of MUC5AC on prognosis. In the study by Aloysius et al,³⁸  31 cases of AC were included in the overall cohort of periampullary carcinomas (n = 104), but the survival effect of MUC5AC was not discriminated by specific disease. Xue et al¹⁴  defined MUC5AC as an independent worse prognosticator in ACs (n = 136) using a high threshold (>20% cutoff). However, Perkins et al⁵³  found that MUC5AC in AC (n = 91) was not related to OS and RFS despite using the same cutoff. This discrepancy may be caused by different immunohistochemical analyses: Perkins et al⁵³  assessed MUC5AC immunoreactivity using tissue microarray, whereas Xue et al¹⁴  conducted immunohistochemical analysis with full tumor sections. In this study, we performed full-section slide staining with an identical cutoff of more than 20% for MUC5AC. However, similar to the results of Perkins et al,⁵³  we did not find any prognostic significance of MUC5AC for OS and RFS. We observed that MUC5AC was more commonly expressed in nonintestinal immunophenotypes than in the intestinal type of AC (P = .03), similar to previous studies of AC.^42,48  Further studies are needed to clearly define the prognostic impact of MUC5AC in AC.

Aberrantly expressed MUC5AC was related to tumor progression and aggressive behavior, as mainly described in pancreatobiliary tract cancers,³⁴  but its mechanism is not clear. MUC5AC regulates cell-to-cell and cell-stroma interactions¹⁴  and leads to enhanced cell invasion and migration by interacting with integrins and matrix metalloproteinase 3 or by enabling disruption of the E-cadherin/β-catenin axis.^15–17  In addition, MUC5AC protects tumor cells from the host immune system and modulates tumor cell responsiveness by inhibiting apoptosis and decreasing the production of chemokines.^54,55  We therefore speculate that MUC5AC may be closely involved in the tumor-host interaction. This may be the reason why MUC5AC was more commonly expressed in AC with a higher TME prognostic risk (high invasiveness and low immune response) in the present study. Therefore, MUC5AC may modulate the interaction between tumor cells and immunity toward enhancing invasiveness in the TME.

We investigated the roles of TME risk index and MUC5AC for additional treatment. MUC5AC provides resistance to chemotherapeutic regimens, such as gemcitabine and 5-fluorouracil (5-FU), by promoting E-cadherin/β-catenin axis activation or CD44/β-catenin/p53/p21 signaling.^17,56  Hence, in this study, we examined the influence of MUC5AC on the response to therapeutic agents. The AC patients in this study, regardless of histologic subtype, were treated with chemotherapy, and 94% (16 of 17) of patients received gemcitabine-based or 5-FU–based chemotherapy. When we analyzed recurrence rate of patients receiving chemotherapeutic agents, MUC5AC was not significantly related to an increased recurrence rate (P = .61). The prognostic impact of the TME risk index for predicting RFS was analyzed according to additional chemotherapy and/or radiotherapy. As a result, the TME risk index significantly stratified the survival of patients without chemotherapy and/or radiotherapy treatment (n = 46; P = .03; data not shown). Unfortunately, the RFS of patients with additional chemotherapy and/or radiotherapy was not affected by TME risk index (n = 18; P = .18; data not shown).

Our cohort included 1 case of MiNEN. A 2.0-cm tumor of ampullary-NOS type had components of moderately differentiated adenocarcinoma and large cell neuroendocrine carcinoma (Supplemental Figure 1, see the supplemental digital content at https://meridian.allenpress.com/aplm in the September 2023 table of contents). This tumor showed BD^Low/TIL^Low and was categorized in the intermediate risk group of the TME prognostic risk index. Future studies evaluating more MiNEN cases would allow for the characterization of TME risk of MiNEN in AC compared with that of conventional adenocarcinoma.

The novelty of this study is the simple application of TME-based prognostic index through standardized assessments of BD count and TIL density during routine pathologic practice. TME prognostic risk index displayed better predictive ability for both OS and RFS than BD or TIL alone, and its association with MUC5AC may suggest a regulatory mechanism of MUC5AC that enhances the invasiveness of tumor cells against immunity. However, our study had some limitations. The underlying mechanisms of MUC5AC in immune interactions with tumor cells were not precisely known. In addition, a small number of cases were included because of a single-center study of AC with low prevalence. Although MUC5AC was not related to OS or RFS in univariate analysis, its close association with TME risk index might be due to the small number of cases. Further multi-institutional studies with greater numbers of ACs are needed to elucidate the value of our TME index and the regulatory mechanisms of MUC5AC.

In conclusion, our simple TME-based prognostic index using the combination of histologic BD and TIL assessment strengthened risk stratification for patients with AC compared with either BD or TIL assessment alone. This TME prognostic risk index effectively predicts the OS and RFS in patients with ACs and consistently stratifies the OS of patients with ACs of nonintestinal immunophenotypes, tubular adenocarcinomas, and lower stage diseases. Therefore, the TME prognostic risk index is a strong and useful prognosticator for patients with AC. In addition, MUC5AC in the TME may modulate the interaction between tumor cells and immunity toward enhancing invasiveness.