Stromal hyaluronan accumulation is associated with low immune response and poor prognosis in pancreatic cancer

Hyaluronan (HA) accumulation has been associated with poor survival in various cancers, but the mechanisms for this phenomenon are still unclear. The aim of this study was to investigate the prognostic significance of stromal HA accumulation and its association with host immune response in pancreatic ductal adenocarcinoma (PDAC). The study material consisted of 101 radically treated patients for PDAC from a single geographical area. HA staining was evaluated using a HA-specific probe, and the patterns of CD3, CD8, CD73 and PD-L1 expression were evaluated using immunohistochemistry. HA staining intensity of tumour stromal areas was assessed digitally using QuPath. CD3- and CD8-based immune cell score (ICS) was determined. High-level stromal HA expression was significantly associated with poor disease-specific survival (p = 0.037) and overall survival (p = 0.013) In multivariate analysis, high-level stromal HA expression was an independent negative prognostic factor together with histopathological grade, TNM stage, CD73 positivity in tumour cells and low ICS. Moreover, high-level stromal HA expression was associated with low ICS (p = 0.017). In conclusion, stromal HA accumulation is associated with poor survival and low immune response in PDAC.


Scientific Reports
| (2021) 11:12216 | https://doi.org/10.1038/s41598-021-91796-x www.nature.com/scientificreports/ influences immune responses 8 . A complex regulation system controls HA metabolism, mainly dependent on HA-producing synthases and degrading hyaluronidases. The activation of the cell surface HA receptors such as CD44 9 and RHAMM 10 modulate cell proliferation, aggregation, migration and angiogenesis 11 and may also be involved in the HA-induced epithelial-mesenchymal transition 12 and stem cell functions 13 . HA has been shown to be overexpressed in most human malignancies [14][15][16][17][18][19][20][21][22][23] . Several studies have indicated that hyaluronan accumulation in the tumour cells and/or peritumoral stroma is related to tumour progression and poor survival in many cancer types [23][24][25][26][27][28][29][30][31] . However, the mechanisms underlying the association between accumulation of HA, host immune response and poor survival remain unclear, especially in PDAC. The association between a strong immune response and better survival is well established in various cancers [32][33][34][35][36] . We have previously introduced a T-lymphocyte-based immune cell score (ICS) as a strong favourable prognostic factor in PDAC 37 . Extensive alterations occur in the complex PDAC microenvironment during the tumorigenesis. Multiple mechanisms, such as overexpression of the immunosuppressive molecules CD73 and PD-L1, may lead to immune suppression [38][39][40][41] . There is some evidence showing that HA plays a role in immune response regulation 42,43 . According to our hypothesis this might be one of the key factors explaining the association between HA accumulation and low survival among cancer patients 25 .
The aim of the present study was to examine the prognostic role of stromal HA accumulation and its relation to immune cell infiltration and the immune-suppressing molecules CD73 and PD-L1 in PDAC.

Methods
Patients. From 2000 to 2016, a total of 129 patients with PDAC were operated on with curative intent in the Central Hospital of Central Finland, Jyväskylä, Finland. Patients with locally inoperable tumour, peritoneal carcinosis or distant metastases were excluded, resulting in the 101 patients with stage IA-IIB disease. Detailed information on patient and tumour characteristics, surgical treatment and complications, oncological treatment and follow-up were collected prospectively, updated and confirmed by a review of patient records review. None of the included patients received neoadjuvant chemotherapy before surgery.
Histopathological examination. All histopathological tumour specimens were reviewed by an experienced gastrointestinal histopathologist (JB). Tumour staging was done according to the 7th edition of the UICC/ AJCC TNM categories 44 . The grading was performed according to the WHO classification of tumours 2010 45 . Tumour sampling, HA assay, and immunohistochemistry. Tissue microarray (TMA) blocks were constructed as described previously, from formalin-fixed paraffin-embedded primary PDAC patient tumour samples. Two tissue cores 0.6 mm in diameter were taken both from the core of the tumour and the invasive margin from representative tumour blocks. Sections of 2 µm thickness were used for immunohistochemical (IHC) analyses 37 .
Signal visualization for all IHC was done by diaminobenzidine and sections were counterstained with haematoxylin.
In general, HA staining was clearly seen in all specimen both in stroma and in tumour epithelium. Stained TMA sections were scanned using an Aperio digital slide scanner (Leica Biosystems), followed by analysis using QuPath v 0.1.2 as described below.
ICS was determined using the TMA technique as described earlier 37 . Briefly, ICS describes the immune response represented by the density of CD3 and CD8 positive immune cells in the tumour centre and at the invasive margin. PD-L1 expression was evaluated by estimating the proportion of PD-L1 positivity on the tumour cell surface as described earlier 39 . In addition, we analyzed also the proportion of PD-L1 expression in stromal cells using the 5% staining proportion as a cutoff.
Quantitative evaluation of HA staining. HA was evaluated using QuPath v 0.1.2 46 . First, the stain vectors and background values were estimated using the Estimate stain vectors command to facilitate stain separation with the color deconvolution method. Simple tissue detection command was used to delineate the tissue area from the white background. This area was manually edited with the brush tool to exclude tumour epithelial regions. SLIC superpixel segmentation was used to divide the area into superpixels (neighboring groups of pixels sharing similar characteristics). DAB intensity was calculated for each superpixel, and the data were exported at individual superpixel level. R statistical programming language version 3.5.2 (R Foundation for Statistical Computing, Vienna, Austria) was used to summarize the mean intensity for each case. The distribution of DAB intensities was similar in different TMAs suggesting that the assay had performed uniformly (Fig. S.1 www.nature.com/scientificreports/ from tumour centre and invasive margin had a strong positive correlation (Spearman's rho = 0.69), and average HA intensity of all available tumour regions was therefore used in the main analyses. Samples were divided into two groups based on the mean intensity value: high and low stromal HA expression. To determine cut-off values for HA expression with optimal sensitivity and specificity, we used receiver operating characteristic (ROC) curve drawn in relation to disease-specific 3-year mortality.
Statistical analyses. The chi-square test was used when analysing the associations between HA and clinical and histopathological variables, CD73 positivity and PD-L1 positivity in tumour cells. The estimates for hazard ratios for overall survival (OS) and disease specific survival (DSS) were calculated using univariate and multivariate Cox proportional hazards regression model. Only variables with p < 0.05 in univariate analysis were entered into the multivariate analysis despite the a-priori determined confounder, tumour stage (p = 0.117). All statistical tests were two-sided. A p value less than 0.05 was considered significant. The statistical analyses were performed and

Results
Patient demographics. A total of 101 of PDAC patients were included in this study. The distribution of key clinicopathological variables among these patients is shown in Table 1.

Associations between stromal HA expression and other histopathological variables. Stromal
HA accumulation appeared to associate with low ICS (p = 0.017). The associations between stromal HA expression and other clinical and histopathological variables, cell-specific CD73 positivity and PD-L1 positivity in tumour cells were also assessed and are shown in Table 2. Stromal HA accumulation was not associated with   (Fig. 5).
In the multivariate analysis, stromal HA accumulation was an independent negative prognostic factor together with histopathological grade, TNM stage, CD73 positivity in tumour cells and low ICS (Table 3).

Discussion
In the present study, using a larger, consecutive patient series from a single geographical area of Northern Finland without apparent selection bias, we showed the role of stromal HA accumulation as an independent prognostic factor for poor survival in pancreatic cancer. We also found an association between the HA accumulation and low immune response as judged by the tumour-infiltrating T-cell densities.
While the number of patients in the present work was higher than in previous studies on HA in PDAC 6,24 , an even larger material would probably have allowed a connection between T-cell score and hyaluronan stronger than that now established. The small tissue cores turned out to be quite acceptable for the analysis since in preliminary tests the HA intensities were strongly correlated between different cores of the same tumour even between the tumour centre and invasive margin. The computer-assisted evaluation of HA staining adopted for  www.nature.com/scientificreports/ the present work was felt easier than manual scoring of sometimes minor differences in intensity. It can also be recommended for future studies due to its independence of personal variation between evaluators. Stromal HA accumulation in malignancies originating from non-stratified epithelium is associated with a poor survival in a number of solid tumours 7 . Indeed, given the large desmoplastic stroma in PDAC, a major role of HA was expected in the progression of this cancer. The idea was further supported for example by a fact that a drug specifically reducing HA synthesis inhibits human PDAC cell growth in vitro and in mice in vivo 49 . However, clinical trials combining enzymatic removal of HA and cytostatic drugs have been disappointing 50,51 suggesting that it is not just the content of HA that enhances malignant growth. Rather, activated synthesis and concurrent degradation of HA probably provide an environment supporting cancer spreading 7 . This becomes understandable by considering the two opposite influences of HA on cell migration. By its swelling pressure HA gel creates free space for cells to move in, while at the same time blocks attachment to adjacent cells and matrix proteins.

PD-L1 in tumour cells
Indeed, the cell surface hyaluronidase TMEM2 is an independent negative prognostic factor in PDAC 52 , demonstrating the importance of HA degradation in PDAC progression. TMEM2 associates to integrins and clears HA to facilitate cancer cell adhesion and migration 53 . Besides facilitating migration in HA-rich matrix the fragments created by hyaluronidase act as a signal that amplify inflammation.
Increasing numbers of studies have shown the impact of HA on the host immune response: It is suggested to protect tumour cells against immune attack by forming peri-cellular coats 54 . Moreover, HA accumulation seems to facilitate tumour-associated macrophage infiltration and their differentiation into the pro-tumoral M2 phenotype 25,42 with an immunosuppressive effect preventing antitumour immunity by T-cells. Interestingly, in the present study we demonstrate an inverse correlation between T-cell-based ICS and stromal HA accumulation in PDAC. This supports the idea that an HA-rich extracellular matrix not only acts as a shield between T-cells and tumour cells, but also prevents T-cell infiltration in the whole tumour microenvironment. We have previously published a paper showing that CD73 positivity in PDAC cells is a prognostic factor in PDAC independently of ICS and hypothesized that CD73 suppresses immune response by impacting on the activity of the tumour infiltrating lymphocytes rather than their number 39 . This might also explain why stromal HA expression associates with ICS but not with CD73 or PD-L1.
Recently, different phenotypes of ECM-producing cancer-associated fibroblasts (CAFs) have been described, including inflammatory CAFs and myofibroblastic CAFs 55 . Inflammatory CAFs are supposed to be tumour-promoting via immune suppression 56 . In future, it would be reasonable to find out if the HA accumulation associates with the polarization of CAFs, since this would further give some insight into the potential mechanism behind the association between HA and immune status. One possible link between CAF polarisation and HA synthesis is the STAT3-signaling pathway, since the inhibition of STAT3-pathway has been shown to downregulate HA-synthesis 57 and, on the other hand, to promote differentiation of CAFs into myCAFs 56 .
As far as we know, the association of HA on T-cell immune response has not been studied earlier in PDAC but the present finding clearly warrants further expansion of the studies to obtain a more detailed view of the interactions between HA and lymphocytes in this disease with such a bleak prognosis. In future studies, information of physical properties (for example molecular mass) of HA molecules is also needed, since there are data indicating that molecular weight affects the biological functions of HA molecules 58 .
In conclusion, our study indicates that stromal HA accumulation may be associated with low T cell densities in the PDAC microenvironment, but still represents an adverse prognostic parameter independent of T