Targeting the Tumor Immune Microenvironment Could Become a Potential Therapeutic Modality for Aggressive Pituitary Adenoma

Object: This study aimed to explore the relationship between the aggressiveness and immune cell infiltration in pituitary adenoma (PA) and to provide the basis for immuno-targeting therapies. Methods: One hundred and three patients with PA who underwent surgery at a single institution were retrospectively identified. The infiltration of macrophages and T-lymphocytes was quantitatively assessed. Results: The number of CD68+ macrophages was positively correlated with Knosp (p = 0.003) and MMP-9 expression grades (p = 0.00). The infiltration of CD163+ macrophages differed among Knosp (p = 0.022) and MMP-9 grades (p = 0.04). CD8+ tumor-infiltrating lymphocytes (TILs) were also positively associated with Knosp (p = 0.002) and MMP-9 grades (p = 0.01). Interestingly, MGMT expression was positively correlated with MMP-9 staining extent (p = 0.000). The quantities of CD8+ TILs (p = 0.016), CD68+ macrophages (p = 0.000), and CD163+ macrophages (p = 0.043) were negatively associated with MGMT expression levels. The number of CD68+ macrophages in the PD-L1 negative group was significantly more than that in the PD-L1 positive group (p = 0.01). The rate of PD-L1 positivity was positively correlated with the Ki-67 index (p = 0.046) and p53 expression (p = 0.029). Conclusion: Targeted therapy for macrophages and CD8+ TILs could be a helpful treatment in the future for aggressive PA. Anti-PD-L1 therapy may better respond to PAs with higher Ki-67 and p53 expression and more infiltrating CD68+ macrophages. Multiple treatment modalities, especially combined with immunotherapy could become a novel therapeutic strategy for aggressive PA.


Introduction
Pituitary adenoma (PA) accounts for approximately 15% of intracranial tumors and represents the second-most common primary brain tumor in humans [1]. Most PAs are non-aggressive, benign tumors that grow slowly in the sellar and suprasellar regions [2]. However, about 35% of PAs are aggressive adenomas that infiltrate the adjacent sphenoid sinus, cavernous sinus (CS), and internal carotid arteries [3]. In addition, third ventricular compression may occur, which leads to hydrocephalus [4]. Due to their invasive nature, the goal of complete removal of PA is difficult to achieve, which further increases the risk of tumor recurrence [4]. As a result, local compression symptoms and neuroendocrine symptoms caused by aggressive PAs remain after traditional treatment. Hence, a multidisciplinary approach including surgery, radiotherapy, chemotherapy, and immunotherapy is
The stained sections were viewed at low magnification to identify 5-10 "hot-spot" areas with the highest density of stained immune cells. The number of selected "hot-spots" mainly depended on the size of the specimen under 400× microscopic examination. For every region of interest, the number of stained immune cells was counted. The average of the highest three values was used as the final value. The method is a comprehensive reference to previous studies [9,24].

Statistical Analysis
The R language and GraphPad Prism software 9.0 (GraphPad Software, San Diego, CA, USA) were used to analyze and map the data. The Student's t-test was conducted on two-group comparisons. Spearman correlations and regression analyses were mainly used to analyze the relationship between the factors. For the comparison of multiple groups, a Brain Sci. 2023, 13, 164 5 of 15 one-way ANOVA was performed. The Chi-square test was used for categorical variables, when needed. The Kaplan-Meier method was used to evaluate the overall survival (OS) and progression-free survival (PFS). Univariate and multivariate analyses were performed with the Cox proportional hazards model. A p value < 0.05 was deemed significant.

Markers of Invasiveness and Aggressive Infiltration
CD68+ macrophage infiltration surrounding tumor blood vessels was often observed (Figure 1c). There was significant differences in the distribution of CD68+ (p = 0.016, Figure 1c,d) and CD163+ macrophage (p = 0.022, Figure 1e,f), but not in CD8+ TILs (p = 0.377, Figure 1a,b). Among them, post hoc analysis indicated that the number of CD163+ macrophages in tumors of Knosp grade 0 was lower than those with Knosp grade 1 and grade 3. The CD68+ cells infiltrating PAs were positively correlated with tumor volume (p = 0.03, r = 0.21) and Knosp grade (p = 0.003, r = 0.29, Figure 2). Similar to the results described above, CD8+ TILs were positively associated with Knosp grade (p = 0.002, r = 0.30, Figure 2). The different distribution of CD8+ TILs between the non-aggressive PA and aggressive PA at ×400 magnification (a) and the Knosp grades (b); the different distribution of CD68+ macrophages surrounding the vessels between the non-aggressive PA and aggressive PA at ×400 magnification (c) and the Knosp grades (d); the different distribution of the density of infiltrating CD163+ macrophages between the non-aggressive PA and aggressive PA at ×400 magnification (e) and the Knosp grades (f). The different distribution of CD8+ TILs between the non-aggressive PA and aggressive PA at ×400 magnification (a) and the Knosp grades (b); the different distribution of CD68+ macrophages surrounding the vessels between the non-aggressive PA and aggressive PA at ×400 magnification (c) and the Knosp grades (d); the different distribution of the density of infiltrating CD163+ macrophages between the non-aggressive PA and aggressive PA at ×400 magnification (e) and the Knosp grades (f).
The final scores of MMP-9 staining were positively related to Knosp grade (r = 0.22, p = 0.03) ( Figure 2); however, statistical significance was not found in the expression of MMP-9 between the Knosp grades (p = 0.127; Figure 3e,f). In addition, they were not related to the Ki-67 index (p = 0.76) and tumor volumes (p = 0.33). Interestingly, the final scores of MMP-9 immunoreactivity were positively correlated with the number of CD8+ TILs (r = 0.26, p = 0.01) and CD68+ macrophages (r = 0.32, p = 0.00) beyond CD163+ macrophages (p = 0.38). The numbers of CD163+ macrophages infiltrating PAs were highly variable with statistically significant differences among the MMP-9 scores (p = 0.04) and  The different distribution of CD8+ TILs between the non-aggressive PA and aggressive PA at ×400 magnification (a) and the Knosp grades (b); the different distribution of CD68+ macrophages surrounding the vessels between the non-aggressive PA and aggressive PA at ×400 magnification (c) and the Knosp grades (d); the different distribution of the density of infiltrating CD163+ macrophages between the non-aggressive PA and aggressive PA at ×400 magnification (e) and the Knosp grades (f).

Figure 2. Correlation heatmap (Spearman correlation) of immune cells and pro-invasive factors.
Only statistically significant differences are displayed.
The final scores of MMP-9 staining were positively related to Knosp grade (r = 0.22, p = 0.03) ( Figure 2); however, statistical significance was not found in the expression of MMP-9 between the Knosp grades (p = 0.127; Figure 3e,f). In addition, they were not related to the Ki-67 index (p = 0.76) and tumor volumes (p = 0.33). Interestingly, the final scores of MMP-9 The numbers of CD163+ macrophages infiltrating PAs were highly variable with statistically significant differences among the MMP-9 scores (p = 0.04) and were higher in the moderately MMP-9 positive group than in the weakly positive group (p = 0.006). It is interesting to note that the expression of MMP-9 was positively correlated with tumor recurrence (r = 0.23, p = 0.03), and patients with MMP-9 positivity had a 2.63-fold increased risk for relapse (B = 0.97, p = 0.03). It is interesting to note that the expression of MMP-9 was positively correlated with tumor recurrence (r = 0.23, p = 0.03), and patients with MMP-9 positivity had a 2.63-fold increased risk for relapse (B = 0.97, p = 0.03).

MGMT and Immune Checkpoint
There were four PAs with low MGMT expression (3.88%), 50 with intermediate (48.54%), and 49 with high MGMT expression (47.57%) on immunostaining. No significant MGMT expression between the Knosp grades was observed in this study (p = 0.782; Figure 3a,b).
A statistically significant difference in MGMT expression was found between primary and recurrent PAs (Chi-square = 6.646, p = 0.036, 5.06% vs. 0.00%). The distribution of MGMT expression did not differ among the pathological subtypes (Chi-square = 6.552, p = 0.886).

Classifications and Apoplexy of Pituitary Adenoma
There was no difference in the number of CD8+ TILs (p = 0.38) or CD68+ (p = 0.97) and CD163+ macrophages (p = 0.29) among the pathological classifications of tumors. The thyrotroph adenoma group was excluded since only one patient was in the group.

Survival Analysis
Ninety-three patients completed follow-ups, with a mean follow-up time of 30.56 ± 6.53 months. Two patients with PAs died from bleeding because of surgery complications, and two patients died of other causes. The average PFS was 29.26 ± 8.35 months (range, <1-35 months), and nine patients (9/93, 9.68%) had tumor recurrence during the follow-up. Only one patient with Knosp grade 0 received complete resection of the tumor among nine patients with tumor recurrence; seven patients with CS invasion had an incomplete tumor removal.

Classifications and Apoplexy of Pituitary Adenoma
There was no difference in the number of CD8+ TILs (p = 0.38) or CD68+ (p = 0.97) and CD163+ macrophages (p = 0.29) among the pathological classifications of tumors. The thyrotroph adenoma group was excluded since only one patient was in the group.

Survival Analysis
Ninety-three patients completed follow-ups, with a mean follow-up time of 30.56 ± 6.53 months. Two patients with PAs died from bleeding because of surgery complications, and two patients died of other causes. The average PFS was 29.26 ± 8.35 months (range, <1-35 months), and nine patients (9/93, 9.68%) had tumor recurrence during the follow-up. Only one patient with Knosp grade 0 received complete resection of the tumor among nine patients with tumor recurrence; seven patients with CS invasion had an incomplete tumor removal.
In the multivariate analyses, no independent risk factors were found for OS (Table 2). An analysis of possible risk factors for PA recurrence using multiple Cox regression showed that recurrent tumors were significantly more likely to recur (p = 0.04, Table 3). Univariate Cox regression analysis showed patients with MMP-9 positive expression (p = 0.04) and incomplete resection (p = 0.03) had a significantly increased risk of recurrence (Table 3). Kaplan-Meier curve analysis showed that patients with incomplete resection of the tumor had low PFS rates (p = 0.0079, Figure 6a). Patients with MMP-9 positive expression tended to have shorter PFS than MMP-9 negative patients (p = 0.021, Figure 6b). In the multivariate analyses, no independent risk factors were found for OS (Table 2). An analysis of possible risk factors for PA recurrence using multiple Cox regression showed that recurrent tumors were significantly more likely to recur (p = 0.04, Table 3). Univariate Cox regression analysis showed patients with MMP-9 positive expression (p = 0.04) and incomplete resection (p = 0.03) had a significantly increased risk of recurrence (Table 3). Kaplan-Meier curve analysis showed that patients with incomplete resection of the tumor had low PFS rates (p = 0.0079, Figure 6a). Patients with MMP-9 positive expression tended to have shorter PFS than MMP-9 negative patients (p = 0.021, Figure 6b).

Discussion
This study demonstrated that the number of immune cells infiltrating PAs and MGMT expression were correlated with tumor aggressiveness based on imaging and molecular features. In addition, the positive rate of PD-L1 was positively associated with the Ki-67 index and p53 expression. Therefore, the above results reveal that there will be a novel and promising therapeutic approach for the management of aggressive PA.
Our findings confirmed that infiltrating TAMs and M2 polarization in PA were correlated with Knosp grades and MMP-9 expression. However, studies investigating M2 polarized macrophages in PA are rare. Macrophages become polarized based on changes in the environment, creating the different macrophage subtypes, M1 and M2. As is wellknown, M1 macrophages generally have antitumor properties and enhanced antigenpresenting ability. In contrast, M2 macrophages show reduced antimicrobial and tumo-ricidal activity [28]. A study by Marques et al. [29]. showed that M2 macrophages could promote neo-angiogenesis in pituitary neuroendocrine tumors (PitNET).The number of CD68+ macrophages infiltrating PAs was positively correlated with tumor size and Knosp grade, which indicates tumor aggressiveness [24]. TAMs could enwrap tumor-associated blood vessels and create an environment conducive to tumor progression by promoting tumor angiogenesis and secreting growth factors [30]. MMP-9, as a member of the zincdependent endopeptidase family, could promote angiogenesis and cancer invasion by degrading type IV collagen of basal membrane near the tumor cell and extracellular matrix (ECM) [31]. There are few studies on this topic, and further work is required to validate the specific mechanism.
It has been reported that the number of CD68+ macrophages in sparsely granulated GH-secreting adenomas is greater than in ACTH adenomas, and nonfunctional adenomas show more infiltrating CD68+ macrophages than ACTH adenomas [24]. Interestingly, only GH-secreting adenomas had the highest number of CD163+ macrophage infiltration, which may support the evidence that GH adenomas tend to have an aggressive behavior [32].
Controversies exist regarding the associations between CD8+ TILs and PA invasiveness or aggressiveness. It has been reported that T-lymphocytes (CD8, CD4, FOXP3) recruited by pituitary neuroendocrine tumor-derived chemokines determine the aggressive behavior [33]. A tendency for higher invasiveness or aggressiveness in PAs with more infiltration of CD8+ TILs has been observed [9,34]. However, it was noted that the number of CD8+ lymphocytes was positively correlated with Knosp grades in our study.
In addition, the number of CD8+ TILs was positively correlated with the expression of MMP-9. MMP-9 plays a critical role in CD8+ T-cell infiltration into tissues and exerts a regulatory role on CD8+ T-cell activation [35,36]. Our results revealed that PAs with aggressive behavior and/or MMP-9 expression could present a microenvironment highly infiltrated by CD8+ TILs, which may exert a specific effect on the invasion of PAs. Whether CD8+ TILs could encourage an immunosuppressive phenotype or not warrants further study.
There was a trend suggesting a higher positive rate of PD-L1 in aggressive PA and non-functional PA [9,34]. However, the present results only confirmed that a higher expression of PD-L1 occurred in null cell PAs than the other PAs. In addition, it was reported that PD-L1 expression is positively associated with increased numbers of CD8+ TILs [34]. However, this result was not seen in this study. Interestingly, CD68+ macrophages were more infiltrated in the positive PD-L1 group. This result suggests that CD68+ macrophages may promote the expression of PD-L1, further encouraging the formation of an immunosuppressive microenvironment. In addition, it seems that patients with positive p53 expression and a high Ki-67 index would benefit most from anti-PD-L1 therapy. Although some studies on glioblastomas showed a correlation between PD-L1 and MGMT expression [37,38], a similar result was not found in this study.
A previous study demonstrated that MMP-9 has potential as a marker for invasion but not for recurrence [39]. Survival analysis showed that patients with positive MMP-9 expression had worse PFS and a higher risk of tumor recurrence. In addition, incomplete resection was a significant independent predictor of recurrence of tumor in multivariate analyses, and patients with incomplete removal of the tumor had a significantly shorter PFS in this study.
The lower the MGMT expression, the better the response to the TMZ treatment in PA [40]. Our results suggested that there may be some interaction between the aggressiveness of the tumor and MGMT expression. A previous study showed that low-to-moderate MGMT immunoexpression occurred significantly more often in aggressive PAs [41]. However, although it was not significantly different between aggressive and non-aggressive PAs in the imaging data, the level of MGMT immunopositivity was positively associated with the MMP-9 staining extent. This result may further exemplify that aggressive PAs could be a suitable candidate for TMZ therapy, but it is necessary to provide more evidence to document this. In addition, the expression of MGMT did show linear relationships with infiltrating immune cells, which may provide evidence that it is possible to use immunotherapy in combination with TMZ to treat aggressive PA.
Together, our results demonstrate that the tumor immune microenvironment serves a vital role in the invasion of PA. However, the effect of immunotherapy on pituitary adenomas has only been reported in some cases, and the effect is uncertain. One patient with ACTH-secreting PA progressed rapidly after four cycles of anti-PD-L1 (pembrolizumab) [42]. Clinical trials showed that the therapeutic effect of a single immune checkpoint inhibitor is not satisfactory for GBM, and attention has shifted focus to immunotherapies combined with other treatments. This novel approach might be more beneficial for patients with GBM [43]. There are currently only two clinical trials of combined immunotherapy for pituitary tumors in clinicaltrials.gov accessed on 2 November 2022. Combined immunotherapy (anti-PD-L1 and CTLA-4) for aggressive pituitary adenomas is still in phase 2 (Memorial Sloan Kettering cancer center, United States, NCT04042753). Another clinical trial of combined immunotherapy for rare pituitary tumors is still recruiting patients (National Cancer Institute, United States, NCT02834013). At present, there are no animal experiments or clinical studies on macrophage-targeting treatment for PA. In addition, the number of cases with a low expression of MGMT is relatively small. Unfortunately, a 2017 review reports that only approximately 42% of pituitary tumors show a radiological response to TMZ treatment [44].
It was noteworthy that in 2022, the WHO emphasized the various cell types and their subtypes and classifies new subtypes of the PitNET based on the tumor cell lineage, cell type, and related characteristics. Therefore, the new classification criteria may bring variations in immune cell infiltration and expression of MGMT, MMP-9, and PD-L1 between the different categories in this study. However, despite the controversies in the field, there is no novel definition of aggressive PA/PitNET in the 2022 WHO classification of pituitary tumors. Nevertheless, some subtypes of PitNETs, such as Crooke cell and immature PIT1-lineage tumors, may have features of aggressive behaviors. It is believed that the diagnosis of "aggressive PitNETs" in the future may also be mainly based on cell type, tumor lineage, and clinical features, and the cure of the disease might benefit primarily from new molecular or immune drugs.
In summary, we confirmed that there are significant differences in the infiltration of immune cells (CD8+ TILs, CD68+ and CD163+ macrophages) and the expression of PD-L1 and MGMT between aggressive and non-aggressive PA. Based on our results and combined with the experience of glioma immunotherapy, targeting macrophages/CD8+ TILs with anti-PD-L1 or other types of immune checkpoint inhibitors may provide a breakthrough in immunotherapy for aggressive PA. Whether immunotherapy based on TMZ treatment may benefit patients is expected to be confirmed by further preclinical and clinical trials.

Limitations
There were limitations to our study that should be acknowledged. First, the main limitation of this study is the low reproducibility of the immunohistochemical methods. However, immunohistochemistry is still the most important method for the diagnosis of PAs, based on the 2017 WHO classification of tumors of the pituitary gland [22]. Second, the conclusion that these data speak in favor of anti-lymphocytic or anti-macrophagic approaches in treatment is based on circumstantial evidence. Despite certain limitations of this retrospective study, the observed interesting clinical features have some enlightening significance and require further investigation.

Conclusions
Targeted therapy for immune cells could emerge as a potential treatment for aggressive PA. In PAs with a higher Ki-67 index, p53 expression, and more CD68+ macrophage infiltration, PD-L1 inhibitors may be more effective. In addition, the incomplete removal of the tumor, recurrent PA, and positive MMP-9 expression were confirmed as increased risk factors for recurrence of PA following surgery. Combined immunotherapy could open new chapters in the treatment of aggressive PA.
Author Contributions: S.H., X.Q. and C.Y. designed the study; Z.Y., X.T., Y.Y. and C.Y. collected data and revised the manuscript for important intellectual content; K.Y., X.Q. and L.Z. performed the histological examination of the samples; Z.Y., C.Y., X.T. and N.L. analyzed the data and Z.Y. wrote the article. All authors have read and approved the final manuscript. Informed Consent Statement: Informed consent was obtained from all subjects involved in the study. Written informed consent has been obtained from the patient(s) to publish this paper.

Data Availability Statement:
The clinical datasets generated during and/or analyzed during the current study are available from the corresponding authors on reasonable request.