Stromal Annexin A2 expression is predictive of decreased survival in pancreatic cancer

Pancreatic ductal adenocarcinoma (PDA) is renowned for high rates of metastasis and poor survival. Its notoriously dense fibrotic stroma contributes to chemoresistance. Stromal signaling in PDA is recognized for its multiple roles in regulating tumor invasion and metastasis. However, no stromal biomarker which can predict survival in PDA exists. Annexin A2 (AnxA2) was formerly identified as a metastasis-associated protein in PDA and tumoral overexpression is associated with poor survival. In this study, we examined AnxA2 expression in the tumor microenvironment in a preclinical model of PDA which suggests its role in tumor colonization. We injected wild-type (KPC) and AnxA2 knockout (KPCA) pancreatic cells into C57BL/GJ (B6) and AnxA2 knockout (KO) mice using the hemi-spleen model and observed their survival. We performed quantitative immunohistochemistry examining stromal AnxA2 expression in 56 patients who had surgically resected PDA and correlated expression with clinical outcomes. B6 mice injected with KPC cells demonstrated decreased median survival compared to those injected with KPCA cells (90 days vs. not reached, p < 0.0001) whereas there was no survival difference in the AnxA2 KO mice (p = 0.63). In patient specimens, we found that high stromal AnxA2 expression (≥80th percentile) was associated with significantly reduced disease-free survival (p = 0.002) and overall survival (p < 0.001). Using multivariate Cox models, there were no significant associations between other clinical covariates apart from high stromal AnxA2 expression. This study highlights the role that stromal AnxA2 expression plays as a prognostic marker in PDA and its potential as a predictive biomarker for survival outcomes in PDA.


INTRODUCTION
Pancreatic ductal adenocarcinoma (PDA) is associated with high rates of metastasis, resistance to cytotoxic chemotherapeutic agents and poor 5-year survival rates [1].The majority of patients (80%) present with inoperable disease and even those who undergo extensive surgical resection typically recur in 80% of cases and die within 5 years of recurrence [2].
PDA is renowned for its dense fibrous stroma, which contains significant cellular heterogeneity and contributes to chemoresistance [3,4].Emerging evidence

Research Paper
suggests that stromal signaling regulates growth, invasion and metastasis in PDA [5].There are increasing efforts focusing on targeting the stromal components of PDA that are known to be aberrantly regulated (e.g., hedgehog inhibition and hyaluronan) [6] which have resulted in the increased efficacy of some chemotherapeutics [7].Specifically, depleting tumor-associated stromal tissue by inhibiting hedgehog signaling in a genetic mouse model of PDA resulted in a transient increase in intratumoral vascular density and intratumoral concentration of gemcitabine [7].However, inhibiting hedgehog signaling remains controversial because the depletion of stromal cells through hedgehog dependent and independent mechanisms results in the promotion of PDA growth and metastasis in some mouse models [8,9].An additional stromal targeting agent, PEGPH20, which is a recombinant human hyaluronidase enzyme that degrades hyaluronan (HA) in the tumor microenvironment (TME), has also been shown to increase drug perfusion in the TME.HA is an extracellular matrix component that contributes to the formation of the dense and fibrous connective tissue that is present in PDA, which results in increased intratumoral pressure in the TME of PDA [10].Therefore, targeting the breakdown of HA with PEGPH20 has resulted in increased perfusion of gemcitabine into the TME, which resulted in improved survival outcomes [11].
A phase 2 clinical trial involving GVAX (allogeneic granulocyte-macrophage colony stimulating factorsecreting tumor vaccine) identified Annexin A2 (AnxA2) from the sera of patients with PDA, as a metastasisassociated protein.Patients who developed post-vaccination immune responses had improved recurrence-free survival with neoadjuvant and adjuvant GVAX in conjunction with surgical resection [12,13].AnxA2 is a calcium-dependent phospholipid binding protein that plays key roles in cellular structure and function including cell division and movement [14,15].It is overexpressed in numerous cancer types and is associated with poor prognostic rates in lung and colorectal cancer [16,17].AnxA2 forms heterotetramers on the cell surface that bind to tissuetype plasminogen activator to mediate the conversion of plasminogen to plasmin, resulting in extracellular matrix degradation and proteolytic activation of inactive proteases [18,19].Phosphorylation of Tyr 23 of AnxA2 is responsible for the cell-surface localization of AnxA2 and mediating the invasive and metastatic function of PDA cells [12].
AnxA2 has also been shown to regulate the secretion of a class III semaphorins (Sema3D) from PDA cells [20].It controls the interaction between Sema3D and its receptor plexin D1 (PlxnD1) on the tumor cell surface, thereby promoting tumor invasion and metastasis.AnxA2 has been reported to have 5-15 fold higher levels of expression in pancreatic tumors compared to normal pancreatic cells [21].AnxA2 overexpression has been associated with resistance to gemcitabine in PDA, and patients with high levels of AnxA2 expression in their tumors have shorter disease-free survival rates and worse overall survival than those with low expression levels [22].Given the important role of stroma in PDA, we wished to examine the association between stromal expression of AnxA2 and clinical outcomes.
Understanding the complexity of the stromal compartment may provide further insights into the development of PDA, which could identify potential biomarkers for predicting survival for patients with PDA.
Using C57BL/6J (B6) mice, mice receiving KPCA + AnxA2 cells (functionally resembling KPC cells) had decreased survival in the hemi-spleen model compared to mice receiving KPCA GFP + cells (Figure 1A, Supplementary Figure 1, median survival 90 days vs. not reached, p < 0.0001).Using the same model in AnxA2 knockout (AnxA2 KO) mice, there was no significant difference in survival when KPCA GFP + or KPCA + AnxA2 cells were used (Figure 1B, Supplementary Figure 1, median survival not reached, p = 0.63).These data suggest that AnxA2 expression in the stroma and TME is important for pancreatic tumor homing in the liver (Supplementary Figure 2).

Clinical characteristics of cohort of patients with resected pancreatic tumors
The clinical characteristics of the patient cohort are summarized in Table 1.The median age of the 56 patients was 62 years (range 43-83).There were more male patients (57.1%) and the majority had positive lymph node status (87.5%).Most tumors had low/intermediate grade differentiation (60%).The majority of the tumors demonstrated perineural invasion (90.7%) and vascular invasion (56.6%), and most patients in our cohort had negative surgical margins (60.7%).The mean stromal AnxA2 score was 0.82 (standard deviation, 0.55; range, 0.02-2.27).

AnxA2 stromal expression is correlated with reduced disease-free survival
The preclinical study showed that stromal AnxA2 expression is important for the homing of the tumors in the liver with subsequent survival benefit.Next we wanted to determine the correlation between AnxA2 expression in the stroma of human PDA and DFS.
We used immunohistochemical staining to quantify stromal AnxA2 expression (Figure 2). Figure 3A shows the Kaplan-Meier curve of DFS for all 56 patients.The estimated median DFS was 16.8 months (95% CI 12.3-25.0months).Moreover, the progression-free probabilities at 1 and 2 years are 0.63 (95% CI 0.52-0.78)and 0.36 (95% CI 0.25-0.52),respectively.To explore the association between DFS and stromal AnxA2 stromal expression, we fitted a Cox model with penalized smoothing splines, with a degree of freedom of 4, to allow for a nonlinear association between log-hazard of DFS and AnxA2 score.Figure 3B shows the ratio of the estimated hazard function, along with 95% CI, for patients with different values of AnxA2 stromal expression relative to a stromal AnxA2 score of 1.This suggests that the effect was approximately constant for stromal AnxA2 scores smaller than 1 and that the effect increased with AnxA2 scores greater than 1.The lower limit of the 95% CI exceeds 1 for AnxA2 scores greater than 1.5, suggesting that a cutoff value between 1 and 1.5 in AnxA2 score may be used to identify patients at a higher risk of disease progression.When a cutoff value of 1.29 is used (corresponding to the 80th percentile of the observed AnxA2 scores), there was a significant difference in DFS between the high and low AnxA2 expression groups (10.3 vs. 19.0months, p = 0.002) (Figure 3C).
The univariate Cox model estimates a hazard ratio of 2.98 (95% CI 1.45-6.14)comparing patients with and without high AnxA2 stromal expression score (≥1.29).We also used a multivariate Cox model to evaluate the adjusted effect of AnxA2 expression on DFS, adjusting for potential clinical and pathological risk factors, and found that the high AnxA2 stromal expression score (≥1.29) remained significantly associated with DFS (p < 0.001) (Table 2).

AnxA2 stromal expression is correlated with reduced overall survival
We evaluated the association between stromal AnxA2 expression scores with overall survival (OS) in our cohort.The median follow-up time for overall survival was 25.4 months (range 4.1-112.5 months).The Kaplan-Meier method estimates a median OS of 25.4 months (95% CI 23.7-32.4months), and the 1-year and 2-year survival probabilities are estimated to be 0.84 (95% CI 0.75-0.94)and 0.57 (95% CI 0.45-0.71),respectively (Figure 4A).As before, we used Cox models with penalized smoothing splines to explore the effect of stromal AnxA2 expression on survival.The hazard ratio relative to stromal AnxA2 1 is shown in Figure 4B.Similar to Figure 3B for DFS, we observe a constant effect for stromal AnxA2 expression score less than 1 and an increasing trend in the risk of death for score greater than 1.When we use the same cutoff value of 1.29 (corresponding to the 80th percentile) for AnxA2 expression score, there is a significant difference in OS between high and low AnxA2 expression groups (14.0 vs. 30.2months, p < 0.001, Figure 4C).
We then used multivariate Cox models to evaluate the adjusted effect of stromal AnxA2 expression on OS.Apart from the cutoff value of AnxA2 stromal expression, there was no significant association found between covariates and OS (Table 3).

DISCUSSION
We have shown that stromal expression of AnxA2 is associated with poor survival in patients with metastatic PDA.In addition, we have shown that AnxA2 plays an important role in the homing of PDA cells at distant metastatic sites.Specifically, when KPCA cells with restored AnxA2 function were administered in the hemispleen model, decreased survival was observed in mice expressing AnxA2 in the TME.This indicates that AnxA2 expression is necessary in both the TME and on tumor cells [12,20].
In our cohort of patients who had undergone resection of pancreatic primary tumors, high stromal expression of AnxA2 was associated with a significantly reduced disease-free and overall survival.Consistent with our and other's published studies delineating the role of AnxA2 in PDA neoplastic cells, Takano et al previously demonstrated that high expression of AnxA2 on the tumor cell surface (defined as >30% of tumor cells with positive staining) was correlated with rapid recurrence after adjuvant gemcitabine chemotherapy including decreased disease-free and overall survival in patients with resected PDA [22].Here, we used automated analysis of stromal AnxA2 expression to distinguish high vs. low expression values and determined that high expression (defined as ≥80th percentile) was a factor that influenced survival outcomes in our cohort.This highlights the potential role of AnxA2 in cancer associated fibroblasts and the potential value for stromal AnxA2 to serve as a biomarker in patients with PDA.
The complex dynamic between components of the TME can be seen in the reverse Warburg effect where epithelial tumor cells are said to induce metabolic effects in neighboring stromal fibroblasts [28].The notoriously dense fibrotic stroma in PDA results in a relatively hypovascular environment that poses a challenge to drug delivery.Several studies show concordance between tumoral and stromal expression of tumor markers suggesting that AnxA2 expression in both tumor cells and stromal cells might play a significant role in PDA metastasis [29,30].Previous studies have shown the role that AnxA2 plays in regulating the secretion of axon guidance molecules in PDA, but further studies are necessary to further delineate the specific roles of the stromal compartment in the AnxA2-PlxnD1-Sema3D pathway [20].
We have previously shown that AnxA2 expression is localized on the tumor cell surface and that cell surface expression increases as pancreatic lesions progress from precursor lesions (pancreatic intra-epithelial neoplasms) to invasive PDA [12].This study demonstrated cytoplasmic staining of stromal fibroblasts in the PDA tumor microenvironment.AnxA2 is known to have key intracellular roles involving exocytosis, endocytosis and membrane trafficking [31][32][33][34].Intracellular AnxA2 has also been shown to induce the expression of anti-apoptotic genes (including IL-6) leading to chemo-resistance in PDA cell lines [35].It is plausible that the increased cytoplasmic expression of AnxA2 in fibroblasts enhances the secretory processes of axon guidance molecules, thereby assisting the processes of invasion and metastasis.

CONCLUSIONS
In summary, this study shows that stromal AnxA2 expression may serve as a prognostic marker of survival in patients with PDA.It also provides further evidence of the importance of stromal components in the TME in mediating metastasis of PDA.

Mouse models of PDA
Animals used in experiments were maintained according to the Animal Care and Use Committee guidelines of Johns Hopkins University.Cox models with penalized polynomial smoothing splines were used to explore the nonlinear association between log-hazard and stromal AnxA2 expression.All analyses were performed using R Statistical Software (http://www.R-project.org;R Foundation for Statistical Computing, Vienna, Austria).

Figure 1 :
Figure 1: Knock-down of AnxA2 improves survival in the hemi-splenectomy KPC model.Kaplan-Meier survival analyses of C57BL/6J (B6) and Annexin A2 knockout (AnxA2 KO) mice injected with KPCA GFP + (AnxA2 deficient) or KPCA +AnxA2 (AnxA2 re-expressed) cells.(A) Liver metastases were generated in the hemi-spleen model in B6 where KPCA + AnxA2 cells had shorter survival than KPCA GFP + cells (p < 0.001).(B)In liver metastases generated in AnxA2 KO mice, there was no significant survival difference between either cell line (p = 0.63).10 mice per group were used in each experiment.Survival curves were compared using log rank test.

Figure 2 :
Figure 2: Immunohistochemistry of stromal AnxA2 expression in pancreatic cancer.Immunohistochemical staining of AnxA2 expression shows that the subcellular location of AnxA2 in stromal cells is in the cytoplasm of stromal fibroblasts.Representative images of tumor specimen with high stromal AnxA2 expression at 10× magnification (A) and 20× magnification (B).Representative image of tumor specimen with low stromal AnxA2 expression at 10× magnification (C) and 20× magnification (D).Red arrows indicate tumor and green arrows indicate stroma.

Figure 3 :
Figure 3: Effect of stromal AnxA2 score on disease-free survival in pancreatic cancer.We correlated the disease-free survival rates of the patients in our cohort with stromal AnxA2 expression.(A) Kaplan-Meier curve, along with pointwise 95% confidence interval, showing that the median DFS in this cohort is 16.8 months (95% CI 12.3-25.0months).(B) Hazard ratio is estimated by fitting a Cox model with penalized smoothing spline of stromal AnxA2 score as the covariate.The ratio is calculated by dividing the estimated hazard for a given stromal AnxA2 score by that for the stromal AnxA2 score of 1.0.(C) Kaplan-Meier curves show that patients with high stromal AnxA2 expression (score > 1.29) have decreased DFS (p < 0.001).

Figure 4 :
Figure 4: Effect of stromal AnxA2 score on overall survival.We correlated the overall survival rates of the patients in our cohort with stromal AnxA2 expression.(A) Kaplan-Meier curve showing that the median OS in this cohort is 25.4 months (range, 23.7-32.4months).(B) We used a Cox model with penalized smoothing splines of stromal AnxA2 score as covariates to evaluate log hazard at different covariate values.This plot shows the ratio of the hazard function for patients with different values of stromal AnxA2 score relative to that for patients with stromal AnxA2 score of 1.0.(C) Kaplan-Meier curves show that high stromal AnxA2 expression (>1.29) is associated with decreased DFS (p < 0.0001).

Table 2 : Multivariate Cox regression analysis of factors related to disease-free survival
Multivariate Cox modelling showing that clinical characteristics, except for high stromal AnxA2 expression, are not significantly associated with DFS.HR: hazard ratio, CI: confidence interval.