Impact of histological response after neoadjuvant therapy on podocalyxin as a prognostic marker in pancreatic cancer

Podocalyxin overexpression associates with poor survival in pancreatic cancer (PDAC). We investigated whether podocalyxin expression correlates with treatment response or survival in neoadjuvant-treated PDAC. Through immunohistochemistry, we evaluated podocalyxin expression in 88 neoadjuvant and 143 upfront surgery patients using two antibodies. We developed a six-tier grading scheme for neoadjuvant responses evaluating the remaining tumor cells in surgical specimens. Strong podocalyxin immunopositivity associated with poor survival in the patients responding poorly to the neoadjuvant treatment (HR 4.16, 95% CI 1.56–11.01, p = 0.004), although neoadjuvant patients exhibited generally low podocalyxin expression (p = 0.017). Strong podocalyxin expression associated with perineural invasion (p = 0.003) and lack of radiation (p = 0.036). Two patients exhibited a complete neoadjuvant response, while a strong neoadjuvant response (≤ 5% of residual tumor cells) significantly associated with lower stage, pT-class and grade, less spread to the regional lymph nodes, less perineural invasion, and podocalyxin negativity (p < 0.05, respectively). A strong response predicted better survival (HR 0.28, 95% CI 0.09–0.94, p = 0.039). In conclusion, strong podocalyxin expression associates with poor survival among poorly responding neoadjuvant patients. A good response associates with podocalyxin negativity. A strong response associates with better outcome.


Methods
We identified patients surgically treated for PDAC at Helsinki University Hospital between 2000 and 2016, finding 90 patients who received NAT and 144 propensity scored patients matched for age, sex, and time of surgery who underwent upfront surgery (US). Survival data and cause of death were collected from patient records, Statistics Finland, and the Finnish Population Registry. Patients who died of immediate postoperative complications were excluded (n=3).
Tumor tissue microarray and immunohistochemistry. Formalin-fixed and paraffin-embedded postpancreatectomy specimens were collected from hematoxylin-eosin stained samples and re-evaluated by experienced pathologists to confirm diagnosis and determine well-represented areas of PDAC for the tumor tissue microarray (TMA) preparation. Six 1.0-mm cores were drilled from each tumor with a semiautomatic tissue microarrayer (TMA Grand Master, TMA Control 3.0, 3D Histech, Hungary).
To identify and stain podocalyxin, two different antibodies that recognize two different amino acid residues were used. The monoclonal antibody (mAb) recognizes amino acid residues 189-192 of podocalyxin. It was created through the immunization of mice using undifferentiated human embryonic stem cells (hES), stem cell line SA167 (Cellartis, Gothenburg, Sweden, www. celle ctis-biore search. com), and establishing mAb production against hES by conventional hybridoma technology. This process has been previously described in detail 35 . The polyclonal antibody (pAb) (HPA 2110, Atlas Antibodies, Stockholm, Sweden) recognizes amino acid residues 278-415. Both mAb and pAb have been validated 47,48 and used in PDAC and podocalyxin studies before 26,28,35 .
TMA blocks were cut into 4-µm sections, deparaffinized in xylene and rehydrated through a gradually decreasing concentration of ethanol to distilled water. Slides were treated in a PreTreatment module (Lab Vision Corp., Fremont, and Ca, USA) in Tris-Hcl (pH 8.5) buffer for 20 min at 98°C for antigen retrieval. Slides were immunostained with pAb and mAb in an Autostainer 480 (Lab Vision Corp., Fremont, CA, USA) by the Dako REAL EnVision Detection system, Peroxidase/DAB+, rabbit/mouse (Dako, Glostrup, Denmark). Tissues were incubated with the mAb (1:500 = 2.4 mg/ml) and with the pAb (1:250 = 0.1 mg/ml) at room temperature for one hour. A sample of podocalyxin-positive colon cancer tissue served as the positive control.
All TMA samples were evaluated by two independent, blinded investigators (AE and JH). The cytoplasmic staining intensity, in the PDAC cells, of podocalyxin stained with mAb and pAb was scored from 0 to 3: 0, negative; 1, weakly positive; 2, moderately positive; and 3, strongly positive (Fig. 1). The highest score across all six TMA samples was considered representative. In the case of differing scores, consensus was reached through re-evaluation.
Statistical analysis. For analytical purposes, a strong response to NAT was defined as a class 0 or 1 response, corresponding to ≤5% RTCs. In the survival analysis, non-responders were defined as a class 4 or 5 response corresponding to >50% RTCs, and responders as a class 0 to 3 response, ≤50% RTCs. Podocalyxin staining scores were combined and grouped as follows: 1, strong staining for both pAb and mAb (both scored 3); 2, either exhibiting strong staining (either scored 3); and 3, both staining weakly (both scored 0-2), corresponding to strong, moderate, and weak, respectively.

Results
Response to neoadjuvant therapy. Among 88 NAT patients, gemcitabine-cisplatin was the most used NAT (n = 43, 48.9%), either with (n = 26, 30%) or without radiation therapy. All NATs used are described in Supplementary Table 1.
Among NAT patients, 2 (2.3%) had no viable residual tumor cells remaining and, thus, a complete response to NAT. A class 5 response with more than 90% of the RTCs remaining was identified in 39 (44.3%) patients, making it the most common response class. The distribution of all responses appears in Fig. 2.
Podocalyxin expression. NAT patients exhibited a weaker (p = 0.049) and less strong podocalyxin immunoexpression compared to US patients using mAb (p = 0.017). In the combined mAb and pAb podocalyxin expression category we identified a statistical difference between the two groups so that the NAT patients had weak podocalyxin immunoexpression (p = 0.032) and US patients had moderate podocalyxin immunoexpression (p = 0.005). Using pAb staining, we detected no significant difference between the groups. Table 2 summarizes the podocalyxin expression distribution.
In the DSS multivariate analysis among non-responders, a strong podocalyxin expression (HR 6.175, 95% CI 2.057-18.544, p = 0.001) associated with a poor survival. Other variables including age, stage, tumor size, adjuvant therapy, or perivascular invasion did not associate with DSS (Supplementary Table 2). In the PFS multivariate analysis of non-responders, a strong podocalyxin expression associated with a poor survival (HR 4.06, 95% CI 1.33-12.37 p = 0.014; Supplementary Table 3). We could not carry out a DSS or PFS multivariate analysis among responders given the small number of patients.

Discussion
We demonstrate that podocalyxin overexpression in PDAC serves as an independent prognostic marker for poor prognosis in patients treated with NAT only if no significant response to NAT was recorded, whereby more than 50% of the viable residual tumor cells appeared in the post-pancreatectomy specimen. When the response to NAT was moderate to complete, with 50% or less viable residual tumor cells, no association between podocalyxin expression and DSS was found. In addition, podocalyxin expression associated with poor clinicopathological factors in the NAT group. In general, NAT patients exhibited a weaker podocalyxin expression compared to US patients using mAb. We also demonstrated that podocalyxin overexpression serves as an independent prognostic factor for poor progression-free survival among NAT patients only if the response to NAT was insubstantial, such that more than 50% of the viable residual tumor cells could be seen post-pancreatectomy.  www.nature.com/scientificreports/ Podocalyxin expression and its association with a poor survival in PDAC was described previously [26][27][28][29][30][31] . For example, Saukkonen et al. (2015) showed that strong podocalyxin expression independently served as a prognostic factor for a poor prognosis in PDAC in a study that used the same antibodies. However, patients receiving NAT were excluded and, to our knowledge, our study represents the first report describing the expression of podocalyxin in neoadjuvant-treated PDAC patients.
While podocalyxin serves as an independent marker of poor prognosis in treatment-naïve PDAC, the underlying mechanisms are not well understood. The upregulation of podocalyxin appears necessary for the epithelial-mesenchymal transition, characterized by migratory and invasive behavior and the involvement metastatic events in cancer [41][42][43] . Podocalyxin has been shown to promote cancer cell proliferation, prevent apoptosis, promote tumorigenesis and participate in cancer cell renewal 40,41,49,50 . Some studies suggest that metastatic, circulating PDAC cells employ podocalyxin as a functional E-and L-selectin ligand promoting cell adhesion to vascular endothelia in metastasis 27,29 . However, the causal relationship between a weaker podocalyxin expression and a better NAT response remains unclear. In PDAC, a preoperative tumor biopsy is difficult to obtain; thus, the preoperative podocalyxin expression is unknown. A comparative analysis of podocalyxin expression or the tumor cell density evaluation of the NAT response between the preoperative and postoperative tumor samples could be performed. Thus, we cannot draw a definitive conclusion regarding whether a weaker podocalyxin expression leads to a better NAT response or, alternatively, if a better NAT response leads to a weaker podocalyxin expression. Among poorly responding NAT patients, a strong podocalyxin expression associated with poor survival, similar to observations for treatment-naïve PDAC. Furthermore, in these patients, podocalyxin could be used as a survival marker. Among strong NAT responders, podocalyxin expression becomes generally weaker. Hence, podocalyxin can be used as a response marker, but is of no use as a survival marker in strong responders.
Furthermore, we show that podocalyxin serves as an independent prognostic factor of poor progressionfree survival in US patients. This supports findings from an earlier study of podocalyxin and DSS among US patients 28 . Here, we detected a difference in survival, both DSS and PFS, between strong and weak podocalyxin expression, but not between moderate and weak podocalyxin expression, and these results agree with previous findings. Saukkonen et al. showed that strong podocalyxin expression compared to weak podocalyxin expression was associated with poor survival 28 .
In addition, we found that only 2 (2.3%) patients exhibited a complete response to NAT. Chatterjee at al. described a similar complete response rate of 2.7% 16 while White at al. described a complete response rate of 6% 14 . We detected a significant association between the NAT response and survival between groups with 5% or fewer and groups with more than 5% residual tumor cells. We found no difference, however, in survival between groups with ≤10% and >10% RTCs, the groups with ≤50% and >50% RTCs, or the groups with ≤90% and >90% of RTCs. Thus, when using NAT, improvement in survival may only occur if the response to NAT is nearly complete, such that no more than 5% of residual tumor cells appear in the surgical specimen. A strong response to NAT was also associated with a better stage, T-class, lymph node ratio, grade, perineural invasion, and negative podocalyxin expression. Patients treated with NAT radiotherapy significantly more often exhibited a strong response to NAT, supporting its use as a part of NAT. In the study of Chatterjee et al. patients treated with neoadjuvant chemotherapy instead of chemoradiation had more residual tumor cells in the post-pancreatectomy specimen 16 .
We designed a six-tier scheme to grade the response to NAT based on the percentage of viable residual tumor cells observed in the post-pancreatectomy specimen. This scheme mirrors previously established schemes. For example, Ishikawa et al. (1989) proposed a grading system with three categories based on the percentage of severely degenerative cancer cells: less than one-third, one-third to two-thirds, and more than two-thirds 11 . Evans et al. (1992) proposed a four-tiered system based on assessing the percentage of tumor cell destruction: I, destruction of <10% tumor cells; IIa, destruction of 10% to 50% of tumor cells; IIb, destruction of 51% to 90% of  Table 3. Association between patient characteristics and podocalyxin using the polyclonal (pAb) and the monoclonal antibody (mAb) in the neoadjuvant group.   www.nature.com/scientificreports/ found no difference between Evans grade I, IIa, and IIb or CAP grades 2 and 3. Thus, a three-tier grading system has been proposed: no residual carcinoma, <5% residual carcinoma, and >5% residual carcinoma 16 . By contrast, in the study by White et al. (2005), patients with a large residual tumor (class 5) experienced a shorter survival time than those with a moderate or good response to NAT 13 . In our progression-free survival analysis, we detected a difference between groups with 5% or less and more than 5% residual tumor cells, observing an additional difference between groups with 10% or less and more than 10% residual tumor cells, but not between the other groups. Chatterjee et al. recommended using a three-tier grading system to group responses to complete, <5%, and >5% residual tumor cells. Our results regarding the progression-free survival identified a difference between groups with <10% and >10% residual tumor cells, suggesting that perhaps additional groups are warranted. and monoclonal antibodies (mAb), HES9 combined among non-responders. A categorization for podocalyxin expression with three groups was created as follows: 1, strong staining for both pAb and mAb (both scored 3); 2, either exhibiting strong staining (either scored 3); and 3, both staining weakly (both scored 0-2), corresponding to strong, moderate, and weak, respectively. Non-responders were defined as a class 4 or 5 response corresponding to > 50% residual tumor cells. Kaplan-Meier analysis. Log-rank test was used for statistical significance. (b) The association of progression-free survival (PFS) and podocalyxin immunoexpression based on polyclonal (pAb) and monoclonal antibodies (mAb) combined among non-responders. A categorization for podocalyxin expression with three groups was created as follows: 1, strong staining for both pAb and mAb (both scored 3); 2, either exhibiting strong staining (either scored 3); and 3, both staining weakly (both scored 0-2), corresponding to strong, moderate, and weak, respectively. Non-responders were defined as a class 4 or 5 response corresponding to > 50% residual tumor cells. Death from any cause (overall survival, OS) was used as a competing event. Aalen-Johansen analysis. Gray's test was used for statistical significance. (c) The association of progression-free survival (PFS) and podocalyxin immunoexpression with polyclonal antibody HPA2110, (pAb) in the upfront surgery group. Podocalyxin expressions were grouped as follows: 1, strong staining for pAb (scored 3); and 2, moderate, weak or negative staining for pAb (scored 0-2) corresponding to strong and weak, respectively. Upfront surgery patients included for the analysis. Aalen-Johansen analysis. Death from any cause (overall survival, OS) was used as a competing event. Grays' test was used for statistical significance. (d) The association of progression-free survival (PFS) and podocalyxin immunoexpression with monoclonal antibody HES9, (mAb) in the upfront surgery group. Podocalyxin expressions were grouped as follows: 1, strong staining for mAb (scored 3); and 2, moderate, weak or negative staining for mAb (scored 0-2) corresponding to strong and weak, respectively. Upfront surgery patients included for the analysis. Aalen-Johansen analysis. Death from any cause (overall survival, OS) was used as a competing event. Grays' test was used for statistical significance. www.nature.com/scientificreports/ One strength to our study lies in its relatively large, well-described cohort of patients with histologically determined PDAC. We excluded other tumors of the pancreas or cancers originating from the bile ducts. The NAT and US groups were matched for age, sex, and time of surgery. Moreover, our data on the survival and follow-up are reliable and precise. To evaluate the neoadjuvant response, we designed the six-tier scheme adapting it from previously designed schemes. Our scheme follows the recommendations of the latest tumor response consensus statements from the Amsterdam International Consensus Meeting 19 . Furthermore, multiple representative samples for each patient were reviewed to evaluate the NAT response. Given these strengths, we acknowledge limitations to our study. The TMA samples are relatively small (1-mm diameter) and, thus, the representativeness of the cancer and podocalyxin staining may have suffered. Specifically, podocalyxin may be unevenly distributed in the tumor. These limitations were minimized by taking several (6) TMA samples per tumor and by scoring the podocalyxin intensity using the strongest intensity visible in the tumor. An experienced pathologist determined the TMA drilling spots on the specimen. Furthermore, the podocalyxin intensity scoring was determined by two independent investigators in order to minimize the effect of subjectivity. This study was conducted to clarify the relationship between podocalyxin staining and NAT, considering the response to NAT. Due to the retrospective nature of our study, implementation of NAT varies between patients. This study featured a limited number of patients with a good or strong response to NAT and, thus, we could not perform a multivariate survival analysis for those patients.

Conclusions
To our knowledge, this is the first report to show that podocalyxin serves as an independent predictor of a poor prognosis in neoadjuvant-treated PDAC patients, but only if the response to NAT is not significant. The causal relationship between a better NAT response, decreased podocalyxin expression, and an inability of podocalyxin to serve as a prognostic marker in NAT patients who respond well remains unclear. A complete or nearly complete response to NAT is rare, whereby those patients seem to benefit from NAT in terms of survival. . NAT responses were grouped as follows: a strong response as class 0 or 1 response, corresponding to ≤ 5% residual tumor cells (RTCs) and a poor response as class 2-5 response corresponding to > 5% RTCs. Kaplan-Meier analysis. Log-rank test was used for statistical significance. (b) The association between NAT response and progression-free survival (PFS). NAT responses were grouped as follows: a strong response as class 0 or 1 response, corresponding to ≤ 5% residual tumor cells (RTCs) and a poor response as class 2-5 response corresponding to > 5% RTCs Death from any cause (overall survival, OS) was used as a competing event. Aalen-Johansen analysis. Gray's test was used for statistical significance.