Specific Subtypes of Carcinoma-Associated Fibroblasts Are Correlated with Worse Survival in Resectable Pancreatic Ductal Adenocarcinoma

Simple Summary Despite massive research efforts, the mortality of pancreatic ductal adenocarcinoma is still high. In recent years, the tumor microenvironment has been revealed to play a key role in carcinogenesis. Therefore, our study aimed to further characterize the family of cancer-associated fibroblasts. We conducted stainings of four common fibroblast markers in a tissue microarray of 321 patients. Here, we describe three subgroups of expression patterns of these markers which are associated with worse survival. Following further basic research, this could lead to new targeted treatment options for patients with pancreatic ductal adenocarcinoma. Abstract Purpose: The pancreatic ductal adenocarcinoma (PDAC) is one of the most lethal cancer entities. Effective therapy options are still lacking. The tumor microenvironment possibly bears further treatment possibilities. This study aimed to describe the expression patterns of four established carcinoma-associated fibroblast (CAFs) markers and their correlation in PDAC tissue samples. Methods: This project included 321 patients with PDAC who underwent surgery with a curative intent in one of the PANCALYZE study centers. Immunohistochemical stainings for FAP, PDGFR, periostin, and SMA were performed. The expression patterns of each marker were divided into low- and high-expressing CAFs and correlated with patients’ survival. Results: Tumors showing SMAhigh-, PeriostinhighSMAhigh-, or PeriostinhighSMAlowPDGFRlowFAPhigh-positive CAFs demonstrated significantly worse survival. Additionally, a high expression of SMA in PDAC tissue samples was shown to be an independent risk factor for worse survival. Conclusion: This project identified three subgroups of PDAC with different expression patterns of CAF markers which showed significantly worse survival. This could be the base for the further characterization of the fibroblast subgroups in PDAC and contribute to the development of new targeted therapy options against CAFs.


Background
Due to its prolonged life span, as well as its risk factor increase, pancreatic ductal adenocarcinoma (PDAC) is progressing in its importance as a tumor entity [1]. Since effective therapy regimes are lacking, PDAC continues to have a high mortality rate. In 2017, pancreatic cancer caused more deaths than breast cancer in the European Union [2]. The wide variety of risk factors supports the hypothesis that PDAC is not a homogenous tumor entity but should rather be considered as a heterogenous disease based on a great number of known pathomechanisms. A greater knowledge of biomarkers for PDAC in order to predict the patients' survival or facilitate treatment decisions is needed. lung, pancreas, and head and neck cancer [28]. However, attempts to further target FAPpositive tumors in mice with T cells engineered with specific chimeric antigen receptors against FAP led to bone marrow toxicity and cachexia [29]. Hence, this underlines the fact that further characterization of fibroblast subgroups is needed.
The aim of this study is to describe this composition with the already established fibroblast markers FAP, PDGFR, periostin, and SMA in a big patient population in order to form further subgroups.

Patients and Tumor Samples
We included 321 patients who were operated on according to the German S3 guidelines for PDAC between 2013 and 2020 in one of the PANCALYZE study centers [30]. Formalin-fixed and paraffin-embedded samples were transferred by each study center to the University Hospital of Cologne. Tissue microarrays (TMAs) were assembled as described before [31]. Trained technicians created two 1.2 mm cylinders of each tumor sample with a semi-automated punch and transferred them to a paraffin block. For further staining, the tissue microarray was cut into 4 µm thick slides.
Written informed consent was obtained from each of the included patients. Data were collected prospectively according to the PANCALYZE study protocol and analyzed retrospectively [32]. Overall survival was defined as the time between the operative resection and the patients' death or loss of follow up. Clinicians and pathologists of the study center captured the included clinicopathologic values following the 7th edition of the Union for International Cancer Control. The study was approved by the local ethic committees and was conducted in accordance with the Declaration of Helsinki.

Immunohistochemistry (IHC) and Analysis
Immunohistochemical stainings with antibodies against SMA, FAP, PDGFR, and periostin were performed following the manufacturer's recommendations (Supplementary Table S1). Control stainings were conducted as mentioned in Supplementary Table S1. Stainings were performed automatically with the Leica BOND-MAX automated system (Leica Biosystems, Wetzlar, Germany). The obtained stainings were scanned with the Aperio GT 450 DX (Leica Biosystems, Wetzlar, Germany) and later analyzed digitally via QuPath v0.3.2 in a previously published manner [33]. The cell detection was performed under the following settings: setup parameters: detection image optical density sum, requested pixel size 1 µm; nucleus parameters: background radius 8 µm, median filter radius 0 µm, sigma 2 µm, minimum area 12 µm 2 , maximum area 400 µm 2 ; intensity parameters: threshold 0.1, max background intensity 2; and cell parameters: cell expansion 5 µm, cell nucleus included. Both entire TMA samples were analyzed. The average value was calculated from these two samples for each patient. The total population was divided into a group with a high expression and a group with a low expression of each marker. The cutoff was defined as the median for FAP, PDGFR, and periostin. We defined the 45th percentile as the cutoff for SMA. Values lower or equal to the cutoff were defined as low.

Statistical Analysis
p-Values below 0.05 were considered statistically significant. All analyses were performed with IBM SPSS Statistics (version 28.0.1.1). Analyses of qualitative markers were performed with the chi-square test. Survival analyses were carried out with Kaplan-Meier curves. Interdependence of clinicopathologic values and survival was analyzed with the univariate and multivariate Cox regression analyses.

Results
A total of 321 patients with pancreatic ductal adenocarcinoma were included in this study. All patients were operated on with curative intention. Patient characteristics are shown in Table 1 Fibroblasts are an important part of the tumor microenvironment. Therefore, we performed immunohistochemical stainings with the common fibroblast markers SMA, FAP, PDGFR, and periostin. Representative images of each marker are shown in Supplementary Figure S1. Full tissue section validation in twenty tumor samples was performed to demonstrate the homogeneity of the marker expression. The four analyzed markers showed a homogenous expression.
We divided our patient cohort into low expression and high expression for each marker (SMA: n (low) = 145, n (high) = 176; FAP: n (low) = 161, n (high) = 160; PDGFR: n (low) = 161, n (high) = 160; periostin: n (low) = 161, n (high) = 160). After analyzing the stainings digitally, we looked for the interdependences between the different fibroblast markers and the overall survival. Here, we could show that a higher number of infiltrating SMA-positive cells correlates with a worse overall survival (p = 0.029, Figure 1). On the contrary, neither FAP nor PDGFR nor periostin stainings seemed to have an impact on survival when analyzed for the total population (FAP: p = 0.208; PDGFR: p = 0.237; periostin: Fibroblasts are an important part of the tumor microenvironment. Therefore, we performed immunohistochemical stainings with the common fibroblast markers SMA, FAP, PDGFR, and periostin. Representative images of each marker are shown in Supplementary Figure S1. Full tissue section validation in twenty tumor samples was performed to demonstrate the homogeneity of the marker expression. The four analyzed markers showed a homogenous expression. We divided our patient cohort into low expression and high expression for each marker (SMA: n (low) = 145, n (high) = 176; FAP: n (low) = 161, n (high) = 160; PDGFR: n (low) = 161, n (high) = 160; periostin: n (low) = 161, n (high) = 160). After analyzing the stainings digitally, we looked for the interdependences between the different fibroblast markers and the overall survival. Here, we could show that a higher number of infiltrating SMApositive cells correlates with a worse overall survival (p = 0.029, Figure 1). On the contrary, neither FAP nor PDGFR nor periostin stainings seemed to have an impact on survival when analyzed for the total population (FAP: p = 0.208; PDGFR: p = 0.237; periostin: p = 0.295). We performed a Cox regression analysis to correct our results for effect modifiers. Here, overexpression of SMA was proven to be an independent risk factor for worse overall survival (HR: 1.389, CI: 1.019-1.893, p = 0.038, Supplementary Table S2 and Table 2). Additionally, clinicopathological values such as T and N status are independent risk factors for worse overall survival (T status: p = 0.007, N status: p < 0.001, Table 2). We performed a Cox regression analysis to correct our results for effect modifiers. Here, overexpression of SMA was proven to be an independent risk factor for worse overall survival (HR: 1.389, CI: 1.019-1.893, p = 0.038, Supplementary Table S2 and Table 2). Additionally, clinicopathological values such as T and N status are independent risk factors for worse overall survival (T status: p = 0.007, N status: p < 0.001, Table 2). As we develop a deeper understanding of the tumor microenvironment, it is becoming clear that each cell type-such as macrophages or fibroblasts-is a heterogenous cell group playing various roles and expressing different markers. Therefore, we built further subgroups for different expression patterns of the four examined fibroblast markers.
Here, patients with Periostin high SMA high -expressing tumors were shown to have a significantly worse survival compared to those with Periostin high SMA low tumors (n (low) = 56, n (high) = 103, p = 0.030, Figure 2A). Tumor samples with Periostin high FAP low PDGFR low SMA high just showed a trend for a worse overall survival (n (low) = 15, n (high) = 16, p = 0.083, Figure 2B). The Periostin high SMA low PDGFR low FAP high expression pattern was revealed to be an additional subgroup with a worse overall survival (n (low) = 15, n (high) = 9, p = 0.045, Figure 2C). All other combinations did not show an impact on the overall survival. No significant differences in patient characteristics between the groups Periostin high SMA high and Periostin high SMA low or Periostin high SMA low PDGFR low FAP high and Periostin high SMA low PDGFR low FAP low could be detected (Table 3). Representative consecutive pictures of each significant marker combination are depicted in Figure 3.        Similar analyses were conducted for the treatment-naïve subcohort. Here, we could observe similar results for the survival analyses as we showed for the whole patient cohort above (Figure 4). Similar analyses were conducted for the treatment-naïve subcohort. Here, we could observe similar results for the survival analyses as we showed for the whole patient cohort above (Figure 4). Additionally, we could detect a significantly higher infiltration of PDGFR high fibroblasts in neoadjuvant-treated patients (Table 4). In contrast, significantly fewer FAP high expression patterns could be found in neoadjuvant-treated patients. No differences could be detected for SMA and periostin.  Additionally, we could detect a significantly higher infiltration of PDGFR high fibroblasts in neoadjuvant-treated patients (Table 4). In contrast, significantly fewer FAP high expression patterns could be found in neoadjuvant-treated patients. No differences could be detected for SMA and periostin. To compare the impact on patients' survival of the three described subgroups of CAFs, we compared the median overall survival rate of the latter. Here, the subgroup with Periostin high SMA high (we called them my-p-CAFs) showed the worst overall survival (SMA high : 20 ± 2.3 months, Periostin high SMA high : 15 ± 2.35 months, and Periostin high SMA low PDGFR low FAP high : 18 ± 4.85 months).
Recapitulated, this project described three subgroups of PDAC demonstrating worse patient survival-SMA high , Periostin high SMA high , and Periostin high SMA low PDGFR low FAP high . A high SMA expression was shown to be an independent risk factor for worse overall survival.

Discussion
This study investigated the role of four previously established fibroblast markers-FAP, PDGFR, periostin, and SMA-in a large study population of 321 patients with pancreatic ductal adenocarcinoma who underwent a surgical resection with a curative intension in one of the PANCALYZE study centers. The focus of this work was to characterize clinically relevant subgroups of CAFs by using a combination of these four listed markers. The importance of defining further subgroups of fibroblasts in patients with PDAC lies in the currently still limited treatment options. The use of FOLFIRINOX improved survival significantly [34]. However, the median disease-free survival is still 21.4 months [35]. Immunotherapy, which showed promising results in other malignancies, has not led to a breakthrough for patients with PDAC yet [36]. A depletion of FAP-expressing fibroblasts in in vivo experiments for pancreatic cancer showed synergisms with anti-PD-L1 therapy and led to tumor suppression [37].
We described that a higher infiltration of SMA-positive my-CAFs correlated with a worse overall survival in our study population. Additionally, a my-CAF-rich tumor stroma is an independent risk factor for a worse survival in patients with PDAC. This finding is supported by the literature [17]. Although many studies investigated different fibroblast markers in PDAC, only a few of them defined fibroblasts with a combination of these fibroblast markers. Öhlund et al. described two subgroups of fibroblasts. The FAP + SMA high fibroblasts were located proximal to, and the IL6 + SMA low more distant from, the neoplastic cells [38]. In this study, we described Periostin high SMA high CAFs (we called them my-p-CAFs), which showed a significantly worse survival in our study population. Additionally, we could define a subtype of CAFs, which is characterized by Periostin high SMA low PDGFR low FAP high , with a significantly worse overall survival compared to that of Periostin high SMA low PDGFR low FAP low . One should highlight that the sample sizes in these latter subgroups were small. However, thanks to the study design and the big study population, it was possible to form these subgroups. These findings are supported by several studies which characterized the impact on patients' survival of fibroblast markers separately [17,23,26,39]. In a software-based manner, we could define as the first group, to the best of our knowledge, these CAF subgroups with a prognostic impact.
One of the limitations of this study was the inability to perform multiple stainings on a single tissue section, which could have provided more comprehensive insights into the underlying mechanisms, as the stainings described in this study were performed on consecutive sections. However, a direct overlay to describe the expression pattern of each cell was not technically possible. In future studies, immunofluorescent, sequential immunohistochemistry technics or multiplex immunohistochemistry could overcome this limitation [40]. Multiplex immunohistochemistry may offer a convenient method for evaluating the expression of multiple markers in individual cells while also providing spatial information. Once thoroughly implemented, this technique has the potential for use in clinical routine as it does not require any additional devices with high maintenance costs or extensive training of technician staff [41]. It was shown that multiplex immunohistochemistry is a more sensitive technique. However, a higher unspecific background staining could be detected, and a more thoughtful selection of antibodies must be kept in mind in order to avoid possible interactions [42].
One general limitation of the immunohistochemical analyses of the described markers was the lacking interpretation standardization of the staining. On the one hand, stainings are either analyzed qualitatively via optical microscope, semiquantitatively, or-as in this study-digitally via evaluation programs. On the other hand, no standardized thresholds are used in the literature for the four analyzed markers. After careful consideration, we decided to use cutoff values around the median. Given the absence of universally accepted cutoff values, we believe that this approach strikes a practical compromise. It is important to acknowledge that there is no consensus on this matter in the literature, as reported cutoff values for SMA in other studies range from 17% and 65% of the patient cohort [43][44][45]. This poses a challenge when comparing findings across studies. Therefore, it is crucial for the research community to address this issue and work towards achieving greater harmonization in this regard in the future. Additionally, using fresh tumor tissue or building organoids, followed by building single-cell suspensions with cell sorting via flow cytometry, could help to further investigate the role of each fibroblast subgroup regarding, among other things, angiogenic abilities and chemotherapy escape mechanisms. In breast cancer, four fibroblast subgroups could be identified by using several immunohistochemical stainings for similar markers as this study described [46]. Further investigations could show that these different subgroups not only occur in different kinds of breast cancer types but also show variations in immune cell infiltration and altered possibilities of the cell crosstalk [46].
Comparing the median overall survival of the three described CAF subgroups, we could show that my-p-CAFs (Periostin high SMA high ) relate to the worst overall survival. In recent years, the role of fibroblasts in polarization-dependent tumor supporting or suppressing has been actively discussed [47]. Further characterization of the tumorsupporting fibroblasts subgroups could lead to new treatment options. A targeted inhibition of the tumor-supporting fibroblasts could possibly facilitate the propagation of the tumor-suppressing fibroblasts and, therefore, have an antitumor effect.
Taken all together, we were able to describe three subgroups of fibroblasts which showed a significantly worse survival. Especially, the fibroblast subtypes with Periostin high SMA high and Periostin high SMA low PDGFR low FAP high could bear the possibility of developing novel targeted therapy options for inhibition of these fibroblast subtypes without encountering severe side effects.

Conclusions
In this study, we were able to further describe the family of carcinoma-associated fibroblasts by using four widely established fibroblast markers. Three subgroups with specific expression patterns of these markers-SMA high , Periostin high SMA high , and Periostin high SMA low PDGFR low FAP high -demonstrated significantly worse overall survival. This, after further profound fundamental research, could lead to new personalized treatment targets, which could expand the current treatment options for patients with PDAC.  Data Availability Statement: The datasets generated and analyzed during the current study are available from the corresponding author upon reasonable request.