Role of stromal PD-L1 expression in colorectal liver metastasis

Background and Aim The outcomes of immune checkpoint blockade for colorectal cancer (CRC) treatment are unsatisfactory. Furthermore, the efficacy of immune checkpoint blockade for liver metastasis of various cancer is poor. Here, we investigated the relationship between stromal programmed death-ligand 1 (PD-L1) expression and the prognosis of patients with colorectal cancer liver metastasis (CRLM). Methods The present study enrolled 84 CRLM patients who underwent surgery (R0) for CRC. Immunohistochemistry was performed to analyze stromal PD-L1 expression in CRLM. Results Stromal PD-L1 was expressed in 52.3% of CRLM samples, which was associated with fewer not optimally resectable metastases (p = 0.04). Stromal PD-L1 also tended to associate with a lower tumor grade (p = 0.08). Stromal PD-L1-positive patients had longer overall survival (p = 0.003). Multivariate analysis identified stromal PD-L1 expression (p = 0.008) and poorer differentiation (p < 0.001) as independent prognostic indicators. Furthermore, stromal PD-L1 expression was correlated to a high number of tumor-infiltrating lymphocytes (TILs). Stromal PD-L1– and low TIL groups had shorter OS than stromal PD-L1 + and high TIL groups (46.6% vs. 81.8%, p = 0.05) Stromal PD-L1-positive patients had longer disease-free survival (DFS) (p = 0.03) and time to surgical failure (p = 0.001). Interestingly, stromal PD-L1 expression was positively related to the desmoplastic subtype (p = 0.0002) and inversely related to the replacement subtype of the histological growth pattern (p = 0.008). Conclusions Stromal PD-L1 expression may be a significant prognostic marker for CRLM. Supplementary Information The online version contains supplementary material available at 10.1186/s12885-024-11869-8.


Introduction
Colorectal cancer (CRC) is steadily increasing and has the second highest incidence among all cancers worldwide [1].Approximately 25% of CRC patients have liver metastasis at the time of diagnosis.Moreover, approximately 50% of patients develop liver metastasis during disease progression [2].Therefore, treatment of CRC efficacy of immune blockade for liver metastasis of various cancers including CRC is unsatisfactory [9][10][11].We previously reported the characteristics of immune markers in CRLM including PD-1, PD-L1, tumor-associated macrophages and indoleamine-pyrrole 2,3-dioxygenase [12].Expression of these immunosuppressive proteins correlates to favorable survival and diminished tumor aggressiveness of CRLM.PD-L1 expression in CRLM is significantly correlated to better differentiation and low incidence of lymph node metastasis, which contribute to better overall survival.
Stromal cells in the tumor microenvironment have been the focus of recent studies [13,14].PD-L1 is primarily detected on tumor cells and overexpressed in various tumor cell types but also expressed in stromal cells such as lymphocytes, macrophages, dendritic cells, and fibroblasts [15].Regarding PD-L1 expression in stromal cells, there have been some studies of small cohorts, but its role in prognosis is controversial, even in the primary site of CRC [16][17][18].Only one previous study analyzed the expression patterns of stromal PD-L1 in the primary site and CRLM [19].Stromal PD-L1 expression was higher than in the primary site in patients with synchronous metastasis, but its prognostic value was not elucidated.Furthermore, one study of stromal PD-L1 expression focused on a limited number of patients with colorectal liver oligometastasis defined as no more than five liver metastases [20].Therefore, we evaluated the influence of stromal PD-L1 expression on CRLM and its prognostic value.

Patients
The present study enrolled 84 patients with CRLM who underwent surgery (R0) between 1995 and 2014 at Tokushima University Hospital.The classifications of liver metastasis, such as H-stage and grade, were defined in accordance with the Japanese Classification of Colorectal Carcinoma, Second English Edition [21].Briefly, H-stage was classified by the number and maximum diameter of liver tumors.Grade classifications were determined by H-stage, mesenteric lymph node metastasis, and extrahepatic metastasis.The definition of non-optimally resectable was determined by previous reports as follows: single nodule (> 5 cm); multiple nodules (> 4 cm) and/or bilobar lesions; synchronous liver metastasis [22].This study was reviewed and approved by the Institutional Review Board of the University of Tokushima Graduate School (Approval no.2395, authorized in 2015).Written informed consent for inclusion was obtained from each patient.All methods were carried out in accordance with Declaration of Helsink.

Histological assessment
Hematoxylin and eosin-stained tissue sections were used to evaluate tumor-infiltrating lymphocytes (TILs) that were quantitated in accordance with the International TIL Working Group 2014 Guidelines [23].The cutoff value used to determine high and low TIL groups was defined in accordance with published data [24].
The histological characteristics of tumor growth pattern in CRC was evaluated with hematoxylin and eosinstained tissue sections.Growth patterns classified as pushing, replacement, and desmoplastic [25].The replacement (invasive) subtype is difficult to clearly distinguish at the tumor border.Tumor cells directly contact the liver parenchyma and replace hepatocytes and the hepatic sinusoidal structure.The desmoplastic subtype showed a tumor separates from the surrounding liver parenchyma by desmoplastic stroma formation [43,44].The pushing subtype showed expansive growth patterns.The tumor has clear border but no fibrotic tissue intervenes.

Statistical analysis
JMP 8.0.1 software (SAS, Cary, NC, USA) was used for statistical analysis.The statistical method to evaluate the relationship between two groups was the chi-squared test.Continuous variables are presented as the median and were compared by the Mann-Whitney test.Survival was calculated by Kaplan-Meier analysis and compared using the log-rank test.Multivariable Cox regression analysis was performed for variables with p < 0.05 in the univariate analysis.Time to surgical failure (TSF) was defined as the time from the initial curative surgery to an unresectable recurrence.Overall survival (OS) was defined as the time from curative surgery to death [30].Because CRLM has unique biological characteristics, in which the first recurrence after an initial hepatic resection does not reflect surgical failure, TSF is a suitable endpoint for CRLM [30,31].P < 0.05 was considered statistically significant.

Pathological findings
PD-L1 was mainly expressed by round mononuclear cells in the stroma (Fig. 1a), indicating that most positive cells were lymphocytes.There were few PD-L1-positive cells, which appeared large and round, suggesting macrophages.PD-L1 was infrequently expressed in fibroblasts that were shaped as spindles.To reveal the characteristics of stromal PD-L1-positive cells, immunohistochemistry was performed in serial sections using CD3, CD4, and CD8 as T cell markers, CD68 as a macrophage marker, and αSMA as a fibroblast marker (Fig. 2).PD-L1-positive cells were mainly positive for CD8 and some were positive for CD68.

Correlations between stromal PD-L1 expression and clinicopathological characteristics
Clinicopathological characteristics in accordance with stromal PD-L1 expression are shown in Table 1.Stromal PD-L1 expression in CRLM tended to associate with a lower tumor grade (p = 0.08).Furthermore, there were significantly fewer instances of non-optimally resectable metastases in stromal PD-L1 + patients compared with stromal PD-L1-patients (p = 0.04).Primary tumor characteristics did not significantly differ in accordance with stromal PD-L1 expression.Stromal PD-L1 expression was significantly correlated to tumor-specific PD-L1 expression in CRLM (p = 0.05).

Correlation between stromal PD-L1 expression and histological findings
Stromal PD-L1 expression tended to correlate to a high number of TILs (Table 1, p = 0.14), although there was no significant difference.There was no correlation between tumor PD-L1 expression and TILs (p = 0.34).Interestingly, stromal PD-L1 expression was associated with the histological growth pattern of CRLM (p = 0.0009, Table 1).Stromal PD-L1 expression was inversely related to the replacement subtype (Fig. 3a, p = 0.008) and positively related to the desmoplastic type (Fig. 3b, p = 0.0002).Clinicopathological characteristics in accordance with the histological growth pattern are shown in S1 Table (replacement subtype) and S2 Table (desmoplastic subtype).The characteristics did not significantly differ in accordance with the histological growth pattern except for stromal PD-L1 expression.

Discussion
Evidence indicates that the tumor stroma, which comprises the tumor microenvironment, is required for tumor growth and progression [32].In the host microenvironment, the liver contains phenotypically distinct stromal cells including macrophages, fibroblasts, lymphocytes, and dendritic cells.These cells interact in a complex manner mediated by cytokines and chemokines [33].The tumor stroma promotes tumor cell proliferation and dissemination through various mechanisms.Tumor stroma remodels the extracellular matrix and recruits inflammatory cells [34,35].Furthermore, the tumor stroma has been implicated in the prognostic outcomes of CRC patients [35].However, the prognostic value of stromal PD-L1 expression in CRC [16,17] and tumor PD-L1 expression [6,36] is highly debated.Only one study reported the stromal PD-L1 expression in colorectal liver oligometastasis [20].Unlike our study results, PD-L1 expression was correlate to poor prognosis in patients with liver oligometastasis.Because PD-1/PD-L1 signaling attenuates host immunity and maintains peripheral tolerance [37], the association of the immunosuppressive ligand PD-L1 with a better prognosis of CRLM may seem counterintuitive.This apparent contradiction may be explained if stromal PD-L1 expression acts as an adaptive anti-tumor response to tumor antigens mediated by an activated immune escape pathway.This possibility is consistent with our findings that stromal PD-L1 expression tended to correlate to a high number of TILs.For example, CRLM patients with CD8 + TILs had better OS than patients with CD8-TILs [38,39].Furthermore, a high CD8+/CD4 + ratio and low FOXP3/CD8 ratio correlate to longer survival of stromal PD-L1 + patients, but not that of stromal PD-L1 − patients with esophageal cancer [40].These findings support the conclusion that immune-mediated and tumor-intrinsic oncogenic activation controls stromal PD-L1 expression.Furthermore, patients with PD-L1+/TILs + have longer OS and TSF but shorter DFS in this study.One possible reason could be a tolerability of chemotherapy.The previous report showed TILs were negatively associated with frailty [41].The frailty increased the risk of chemotherapy.In this point, the patients with PD-L1+/TILs + could show the good tolerance for treatment without frail.Future analysis will be required to determine the impact of frailty on the survival in patients with PD-L1+/TILs+.
Here, we found that stromal PD-L1 was mainly expressed by round mononuclear cells in the stroma, indicating lymphocytes as its source.Representative images of IHC staining revealed PD-L1-positive cells were mainly showed positive for CD8.Conversely, there were few PD-L1-positive cells showed positive for CD3, CD4 and αSMA.TIL subsets, including CD19, CD20, and FOXP3, should be investigated and the balance of TILs between immune-reactive and immune-tolerant should be determined in future studies.In this study, stromal PD-L1 expression was correlated to a stronger influence on the prognosis of CRLM compared with tumor PD-L1 expression and the number of TILs.Moreover, CRLM patients with stromal PD-L1-and low TILs had the lowest OS and TSF rates.Furthermore, PD-L1-and high TIL groups tended to have longer DFS than stromal PD-L1-and low TIL groups.TILs may contribute to the different characteristics of stromal PD-L1-positive cells.Therefore, it will be necessary to simultaneously evaluate PD-L1 expression in tumors and stromal cells as well as the proportion of TIL subsets in CRLM.Furthermore, recent study revealed that peripheral PD-1/PD-L1 expression in circulating T lymphocytes had a significant consistency with PD-L1 expression in immune cells in breast cancer [42].For the clinical application, the relationship between stromal PD-L1 and PD-L1 expression in circulating T lymphocytes should be investigated in future study.It could provide an alternative choice of tissue biopsy to detect the stromal PD-L1 expression for the patients with CRLM.
In this study, we also found that stromal PD-L1 expression in CRLM was inversely related to the replacement subtype and indicated a better prognosis.The main histological characteristics of tumor growth in CRC include pushing, replacement, and desmoplastic [43].Patients with the replacement subtype have a worse prognosis after curative liver resection compared with patients with the desmoplastic subtype [44,45].The different histological patterns of CRLM are associated with different types of tumor vascularization [43].The replacement subtype shows a non-angiogenic growth pattern in contrast to the desmoplastic subtype [46].Because vascular co-option from the normal liver is highly efficient, the replacement   subtype shows minimal hypoxia [47].Furthermore, angiotropism resembling the co-opted capillary bed contributes cancer cell motility and invasion during replacement growth [48].This may be a reason that tumors with the replacement subtype show aggressive characteristics resulting in a poor prognosis.The other reason for the difference between replacement and desmoplastic subtypes may be the immune response to metastatic tumors.A previous study revealed that CD8 + TILs in desmoplastic subtypes are associated with longer survival [49].Furthermore, a tumor with the replacement subtype has reduced infiltration of CD8 + immune cells [50].In this study, stromal PD-L1 expression tended to correlate to high TILs, although it was not significantly different.Taken together, these findings suggest that patients with stromal PD-L1 expression and a favorable prognosis have the ability to perform less replacement growth, possibly through immune infiltration.Thus, to confirm and clarify this association, further studies are required to delineate the role and influence of stromal cells to the histological characteristics of tumor growth.
This study has certain limitations.Firstly, this was a retrospective single-center study of a limited number of patients.The correlation between TILs and stromal PD-L1 will be evident with cohort study.Moreover, future research should analyze TIL subsets and the histological characteristics of tumor growth.Secondly, this study employed immunohistochemistry as the only method to evaluate protein expression in CRLM.Our result should be confirmed through a prospective study by flow cytometry analysis.

Conclusions
Our present study demonstrates the strong influence of stromal PD-L1 expression on the prognostic outcomes of CRLM patients.The expression of stromal PD-L1 served Synch/meta: synchronous/metachronous; diff./undiff.: differentiated histological type/undifferentiated histological type as a better independent prognostic factor for OS.Patients with stromal PD-L1-negative expression and low TILs had the worst OS.These results contribute to a better understanding of the interactions between a tumor and its microenvironment, and will thereby enable prediction of the prognoses of CRLM patients.

Fig. 4
Fig. 4 Prognostic value of stromal PD-L1 expression in CRLM Kaplan-Meier survival analysis of CRLM patients with stromal PD-L1 expression: (a) Overall survival.(b) Disease-free survival.(c) Time to surgical failure.Kaplan-Meier survival analysis of CRLM patients with stromal PD-L1 expression combined with TILs: (d) Overall survival.(e) Disease-free survival.(f) Time to surgical failure

Fig. 3
Fig. 3 Correlation Between stromal PD-L1 and histological growth patterns.Correlation between stromal PD-L1 expression and the (a) replacement subtype and (b) desmoplastic type

Table 1
Clinicopathological characteristics in accordance with stromal PD-L1 expression

Table 2
Univariate and multivariate analyses of clinicopathological factors associated with overall survival after hepatectomy