Combination immunohistochemistry for SMAD4 and Runt-related transcription factor 3 may identify a favorable prognostic subgroup of pancreatic ductal adenocarcinomas

Purposes SMAD4/DPC4 mutations have been associated with aggressive behavior in pancreatic ductal adenocarcinomas (PDAC), and it has recently been suggested that RUNX3 expression combined with SMAD4 status may predict the metastatic potential of PDACs. We evaluated the prognostic utility of SMAD4/RUNX3 status in human PDACs by immunohistochemistry. Materials and Methods Immunohistochemical stains were performed for SMAD4 and RUNX3 on 210 surgically resected PDACs, and the results were correlated with the clinicopathological features. Results Loss of SMAD4 expression was associated with poor overall survival (OS) (p = 0.015) and progression-free survival (PFS) (p = 0.044). Nuclear RUNX3 expression was associated with decreased OS (p = 0.010) and PFS (p = 0.009), and more frequent in poorly differentiated PDACs (p = 0.037). On combining RUNX3/SMAD4 status, RUNX3-/SMAD4+ PDACs demonstrated longer OS (p = 0.008, median time; RUNX3-/SMAD4+ 34 months, others 17 months) and PFS (p = 0.009, median time; RUNX3-/SMAD4+ 29 months, others 8 months) compared to RUNX3+/SMAD4+ and SMAD4- groups; RUNX3-/SMAD4+ was a significant independent predictive factor for both OS [p = 0.025, HR 1.842 (95% CI 1.079-3.143)] and PFS [p = 0.020, HR 1.850 (95% CI 1.100-3.113)]. Conclusions SMAD4-positivity with RUNX3-negativity was a significant independent predictive factor for favorable OS and PFS in PDAC. This is the first and large clinicopathological study of RUNX3/SMAD4 expression status in human PDAC. Combination immunohistochemistry for SMAD4 and RUNX3 may help identify a favorable prognostic subgroup of PDAC.


INTRODUCTION
Pancreas cancer is the fourth most common cause of cancer-related deaths in industrialized countries, with an overall 5-year survival rate of about 8% [1]. The majority of patients are diagnosed at an advanced stage with distant metastasis or invasion of adjacent organs, and less than 20% of patients are amenable to curative resection [1].

Research Paper
Oncotarget 76700 www.impactjournals.com/oncotarget landscape of pancreatic ductal adenocarcinomas (PDACs) is extremely vast. For example, recent large-scale genomic analyses on PDACs have identified various subtypes of these cancers highlighting their heterogeneity, e.g. the "Collisson class" (classical, quasimesenchymal and exocrine-like subtypes), "Moffitt tumor class" (basallike versus classical), "Moffitt stromal class" (activated stroma, normal stroma) and the subtypes recently defined by Bailey et al. (squamous, pancreatic progenitor, immunogenic and aberrantly differentiated endocrine exocrine (ADEX) subtypes) [4][5][6]. Thus, targeting the specific molecular subtypes would ideally help improve the outcomes of PDAC patients, in the way that selected patients with breast and lung cancers have greatly benefitted from targeted therapy over the past decades.
While such high-throughput genomic analyses have yielded promising results, it is also important that inexpensive and easily applicable tests (e.g. immunohistochemical assays) are developed for predicting therapeutic response and prognosis. In this regard, Runtrelated transcription factor 3 (RUNX3) is an interesting candidate biomarker for PDAC, as it has recently been demonstrated to play a role as a "metastatic switch" in PDACs in genetically engineered mouse models: RUNX3 was suggested to be an important determinant of the ability of PDAC cells to metastasize to distant sites or to proliferate locally [7,8]. Moreover, RUNX3 expression was shown to be influenced by SMAD4 status in the same study: heterozygous deletion of SMAD4 (+/−) attenuated RUNX3 levels compared to intact SMAD4 (+/+) and homozygously-deleted SMAD4 (−/−) [8].
If the combination of RUNX3 and SMAD4 status could predict the behavior of PDAC, analyzing RUNX3 and SMAD4 protein expression status by immunohistochemistry in resected or biopsied PDAC tissues would greatly benefit pancreatic cancer patients by helping to guide treatment plans. For example, PDACs with increased metastatic potential could benefit from systemic chemotherapy, while the more locally proliferative PDACs would be candidates for more aggressive local treatment, such as surgery or radiotherapy. Therefore, in this study, we performed an immunohistochemical analysis of RUNX3 and SMAD4 expression in our cohort of 210 surgically resected PDACs, and compared the clinicopathological features of PDACs according to RUNX3/SMAD4 expression status.
When the clinicopathological features were compared according to SMAD4 and RUNX3 expression status (Table 2), we found SMAD4 loss was associated with increased tumor size (p = 0.030), higher pN stage (AJCC 8th ed.) (p = 0.035), poor histological differentiation (p = 0.051) and higher pT stage (AJCC 8th ed.) (p = 0.054), although the latter two parameters were marginally significant. When the disease recurrence patterns were compared, PDACs with SMAD4 loss tended to result more frequently in distant metastasis compared to SMAD4-intact PDACs (p = 0.058), while there was no difference in the frequency of local recurrence according to SMAD4 expression status. RUNX3 expression was associated with tumor size (p = 0.039) and poor histological differentiation (p = 0.033).
On combining the expression status of the two markers, PDACs with RUNX3-/SMAD4+ profiles showed less frequent distant metastasis (p = 0.017) and tended to be smaller in size (p = 0.092) compared to the other PDACs which showed SMAD4 loss and/or RUNX3 expression (Table 3).
When the survival was analyzed according to the immunohistochemical stain results, loss of SMAD4 expression was significantly associated with poor overall survival (median survival 16 months versus 25 months, p = 0.015) and progression-free survival (median survival 8 months versus 11 months, p = 0.044) compared to PDACs with intact SMAD4 expression. RUNX3 expression was significantly associated with decreased overall survival (median survival 16 months versus 23 months, p = 0.010) and progression-free survival (median survival 8 months versus 12 months, p = 0.009) compared to RUNX3negative PDACs. When the RUNX3/SMAD4 status was combined, RUNX3-/SMAD4+ tumors were associated with remarkably longer overall survival (median 34 months versus 17 months, p = 0.008) and progressionfree survivals (median 29 months versus 8 months, p = 0.009) compared to other groups. Interestingly, among the PDACs with no RUNX3 expression, those with SMAD4 loss demonstrated similarly decreased overall and progression-free survivals as RUNX3-expressing PDACs.

Comparison of immunohistochemical status between primary and metastatic lesions
Of the 210 patients studied, 121 (57.6%) subsequently developed distant metastases. Biopsies or resected tissues from metastatic sites were available for 14/121 patients and the whole tissue sections of the 14 metastatic tumors were subjected to SMAD4 and RUNX3 immunohistochemistry. The liver was the most common site of distant metastasis (n = 9) ( Table 6).
Identical RUNX3/SMAD4 profiles were seen in the primary and metastatic tumors of 5/14 (35.7%) cases. While 11/14 (78.6%) of the cases showed concordant SMAD4 expression status between the matched primary and metastatic tumors, only 6/14 (42.9%) cases showed concordant RUNX3 expression status. Interestingly, out of the 14 matched cases, the RUNX3-/SMAD4+ profile was seen in none of the primary PDACs and also in none of their subsequent metastatic tumors. In addition, only 1/14 (7.1%) of the metastatic tumors had intact SMAD4 expression, and all 8 (100%) cases with discordant RUNX3 status were RUNX3+ in the primary tumor and RUNX3-negative in the matched metastatic tumors.

DISCUSSION
The purpose of this study was to evaluate the expression status of SMAD4 and RUNX3 in PDACs, and to explore the utility of the SMAD4/RUNX3 marker combination assay in predicting PDAC behavior. We demonstrate in this retrospective study of 210 PDACs that the combination of intact SMAD4 and RUNX3 negativity may help to identify a subset of PDAC patients with favorable prognosis.    [13][14][15][16][17]. Loss of SMAD4 protein expression has been demonstrated to be a sensitive and specific surrogate marker for SMAD4 alteration, independent of whether SMAD4 inactivation occurred by homozygous deletions or by loss of heterozygosity, and the presence of intact SMAD4 protein expression corresponds to tumors with at least one wildtype allele (i.e. SMAD4(+/+) or SMAD4(+/−)) [11,18]. Therefore, PDACs with intact SMAD4 protein expression would comprise a mixed population of SMAD4(+/+) and SMAD4(+/−) PDACs, and hence the discrepancies in the clinicopathological studies on the prognostic significance of SMAD4 immunohistochemistry are not surprising. In addition, PDACs with SMAD4 loss tended to show more frequent distant metastasis on follow-up compared to SMAD4-intact tumors, while there was no difference in the frequency of local progression according to SMAD4 status, supporting the previous literature that suggested correlations between SMAD4 status and disease progression patterns in PDAC [19,20].
The role of RUNX3 in cancer is still not wellestablished; it has been described to function as a tumor suppressor in most malignancies, including cancers of the breast, stomach, liver, lung and prostate, [21][22][23][24][25] but tumor promoting roles of RUNX3 have been suggested for some tumors such as ovarian cancers, melanomas and PDACs [8,[26][27][28]. An association between RUNX3 protein expression and decreased overall survival has also been demonstrated in PDAC patients, although in a small number of cases [8]. A recent genetically engineered mouse model study by Whittle et al. demonstrated an interesting relationship between SMAD4 and RUNX3 status and the biological behavior (metastasis versus primary tumor growth) of PDACs according to SMAD4/ RUNX3 status [7,8,29]. In brief, KPC (SMAD4(+/+)) and KPDDC (SMAD4(−/−)) tumors showed increased RUNX3 expression levels and a high metastatic disease burden, and the high expression of RUNX3 was associated with increased metastatic potential of PDACs independent of epithelial-mesenchymal transition [8,30]. On the other hand, SMAD4 haploinsufficient KPDC mice (SMAD4(+/−)) lacked RUNX3 expression, had a significantly lower metastatic burden, but showed an increased proliferative activity of the primary tumor compared to KPC tumors. We did not find differences in the proliferative activity (Ki-67 labeling index) of PDACs according to SMAD4/RUNX3 status in this study (data not shown), but we did find that PDACs with RUNX3 Oncotarget 76707 www.impactjournals.com/oncotarget expression and/or SMAD4 loss were more likely to metastasize to distant sites and follow an aggressive clinical course compared to PDACs that did not show these immunophenotypes. Therefore, combining the SMAD4 and RUNX3 protein expression status would provide a more accurate prognostic marker compared to SMAD4 status alone.
It is also interesting that when the immunohistochemical profiles of the primary and metastatic tumors were compared in the separate analysis of 14 primary versus matched metastatic PDACs, SMAD4 labeling status was concordant in the majority of the cases, while the RUNX3 expression status were more frequently discordant. All primary tumors showed RUNX3 expression and/or SMAD4 loss; RUNX3-/SMAD4+ status was not seen in any of the primary or metastatic tumors in this analysis. In addition, all RUNX3-discordant cases showed conversion from RUNX3-positivity in the primary tumor to RUNX3-negativity in the matched metastatic tumors. Attenuation of RUNX3 levels once the metastatic niche is established, switching the metastatic tumor to the locally proliferative state, would be an attractive explanation for this observation; however, experimental evidence would be required to support this. This is to our knowledge the first and the largest single institutional clinicopathological study demonstrating RUNX3/SMAD4 combination Oncotarget 76708 www.impactjournals.com/oncotarget immunohistochemistry to be a useful predictor of both overall and disease-free survivals in PDAC patients. Although further validation in independent cohorts would be required, application of RUNX3/ SMAD4 immunohistochemistry on biopsies or resected PDAC tissues may not only be a prognostic marker but also help guide treatment plans in PDAC patients; RUNX3−/SMAD4+ PDACs are less likely to metastasize to distant sites and may therefore be candidates for more aggressive local surgery and radiotherapy, whereas PDACs with RUNX3 expression and/or SMAD4 loss profiles may benefit more from systemic chemotherapy.

Patient selection and clinicopathological analysis
This retrospective analysis was approved by the Institutional Review Board of Seoul National University Bundang Hospital (IRB #B-1611-369-302). The study subjects comprised 210 consecutive cases of PDACs resected between June 2003 and January 2013 at Seoul National University Bundang Hospital, Seongnam, South Korea. Clinicopathological data were retrieved by reviewing electronic medical records, pathology reports and glass slides, and included patient sex, initial treatment, age at operation, tumor size, histologic differentiation, serum tumor marker levels (CEA, CA19-9), and pathological T and N stages according to American Joint Committee on Cancer (AJCC) 7th edition and the new 8th edition. Follow up data was also retrieved from the electronic medical records, including the status at last follow up, and occurrence of local recurrence and/or distant metastasis during the follow up period. The dates of local recurrence and distant metastasis were defined as when new lesions appeared on imaging studies or when tumor marker levels were elevated. The date of disease progression was defined as when the event has occurred postoperatively such as local recurrence, distant metastasis or death.

Immunohistochemistry
Two mm-core tissue microarrays were constructed from 210 PDACs and their matched non-neoplastic pancreatic tissues (Superbiochips Laboratories, Seoul, Korea), and 4 μm-thick tissue sections obtained from the microarray blocks were subjected to immunohistochemical staining for SMAD4 (1:100, rabbit monoclonal, Abcam, Cambridge, UK) and RUNX3 (1:100, rabbit monoclonal, Cell signaling technology, Denver, USA). In brief, tissue sections were deparaffinized in xylene, rehydrated in graded alcohol, and antigen retrieval was performed using citrate buffer (pH 6.0). Incubation with primary antibodies was performed for 1 hour, and then with secondary antibodies (EnVision Detection System, Dako) for 30 minutes. Counterstaining was performed using Mayer's hematoxylin.
The labeled slides were interpreted by 2 pathologists (YL and HK). SMAD4 was expressed in the cytoplasm and nuclei of normal pancreatic duct epithelia, acinar cells, lymphocytes, stromal fibroblasts and endocrine cells, and served as positive controls. Expression in < 20% of tumor cell nuclei was defined as SMAD4 loss ("SMAD4-"). RUNX3 was not expressed in normal pancreatic ductal epithelial cells or acinar cells, but was expressed in the lymphocytes. RUNX3 expression status was evaluated in terms of both intensity and distribution of staining. The staining intensity was defined as follows: "0", no expression; "1", weaker intensity compared to staining of lymphocytes; "2", strong intensity similar to that seen in lymphocytes. For distribution, the percentage of positively stained tumor cell nuclei was estimated in 5% increments. The staining distribution was relatively homogeneous throughout the tumor. The presence of weak or strong RUNX3 positivity in 5% or more of tumor cells were regarded as positive ("RUNX3+") for subsequent analysis.
For cases where both primary pancreatic tumors and metastatic tumors were available for examination, whole tissue sections were subjected to the same immunohistochemical stains for RUNX3 and SMAD4, and the expression status of both markers were compared between the primary and matched metastatic tumors. These cases were from the same cohort of 210 PDACs.

Statistical analysis
All statistical analyses were performed using SPSS 19.0K (SPSS Korea, Seoul, South Korea). Chi-square tests, Fisher exact tests and Student t-tests were performed as deemed appropriate. Univariate survival analyses for overall and progression-free survivals were performed by the Kaplan-Meier method and log-rank test. The Cox regression models were used for multivariate analysis. Statistical significance was defined as p < 0.05.