In silico analysis of the immunological landscape of pituitary adenomas

Abstract

Purpose

Immunotherapy has gained traction in the treatment of solid tumors but the immunological landscape of pituitary adenomas is not well defined. We sought to investigate the immunological composition in pituitary adenomas using RNA deconvolution (CIBERSORTx) on an existing gene expression dataset for pituitary adenomas.

Methods

We applied an established computational approach (CIBERSORTx) on 134 pituitary adenomas from a previously published gene expression dataset to infer the proportions of 22 subsets of immune cells. We investigated associations between each immune cell type and tumor subtype.

Results

We found that the majority of infiltrating immune cells within pituitary adenomas were comprised of M2 macrophages followed by resting CD4⁺ memory T cells and mast cells. Silent pituitary tumors have higher M2 macrophage fractions when compared to other subtypes. In contrast, Cushing pituitary tumors, both overt and subclinical cases, had higher CD8⁺ T cells fractions than GH tumors, prolactinomas, hyperthyroid tumors, and silent tumors.

Conclusions

RNA deconvolution of the immune infiltrates of pituitary adenomas using CIBERSORTx suggests that most pituitary adenomas comprise of M2 macrophages, but each adenoma subtype has a unique immune landscape. This may have implications in targeting each adenoma subtype with different immunotherapies.

Introduction

Pituitary adenomas are common tumors in the brain that affects more than 10,000 people in the United States annually [1]. Although surgical resection remains the mainstay of most symptomatic cases with the option of adjuvant radiotherapy, there remain cases that are not amenable to surgical resection or harbor more aggressive phenotypes. Immunotherapy has revolutionized the treatment of solid tumors in the last decade but its implementation requires an understanding of the immunological landscape within the tumor microenvironment [2]. For pituitary adenomas, there is a paucity of reports exploring the immune cell composition within these tumors. One study by Lu et al. demonstrated that infiltration of CD68⁺ macrophages positively correlates with pituitary adenoma size and invasiveness, while CD4⁺ and CD8⁺ T cell infiltration was relatively scant [3]. Another report by Mei et al. demonstrated differential expression of immune inhibitory ligand programmed cell death −1 ligand (PD-L1) across pituitary adenoma subtypes, with significantly higher expression in functioning adenomas compared to non-functioning adenomas [4]. However, there was no correlation with tumor invasiveness [4]. Classification of the immune landscape in tumors depends on both PD-L1 expression and the presence of T cell infiltration, referred to as Tumor Immunity in MicroEnvironment (TIME) Classification, which have important implications for the success of anti-PD therapy [2]. Overall, our understanding of the TIME in pituitary adenomas is very limited.

Methods

We aimed to investigate the immunological landscape of pituitary adenomas using an in silico approach (CIBERSORTx) developed by Newman et al. to provide an estimation of the relative abundances of immune cell types in a mixed cell population, using previously published open-source gene expression data [5–7]. This dataset consists of 134 patients with pituitary adenomas [7]. Briefly, the gene expression dataset was generated using RNA-seq by Neou et al. and the database was downloaded from Arrayexpress (https://www.ebi.ac.uk/arrayexpress) under accession number E-MTAB-7768 [6, 7]. Then, the gene expression data was uploaded to CIBERSORTx’s online platform (https://cibersortx.stanford.edu). A validated leukocyte gene signature matrix (LM22) was used to identify 22 human hematopoietic cell subsets in bulk tissues, including tumors [5]. We utilized Bulk-mode batch correction and 1000 permutations for the analysis in relative mode. Quantile normalization was disabled as the dataset was generated from RNA-seq. The filter criteria of each sample is set as the CIBERSORT calculation of P < 0.05 (statistical significance of the deconvolution result across all cell subsets to ensure adequate goodness-of-fit).

All statistics were performed using Graphpad Prism Version 8.3. Statistics are presented as mean ± SEM. One-way or two-way ANOVA was used to compare the means of different immune cells in different tumor subtypes by analyzing false-discovery rate using the two-stage step-up method of Benjamini, Krieger and Yekutieli. A q-value ≤ 0.05 was considered significant.

Results

RNA deconvolution for immune cell subset identification

We analyzed 134 patients with pituitary adenomas in the above manner using CIBERSORTx. After elimination of samples that did not meet the P < 0.05 cutoff, there were 125 cases (Table 1). The most common classifications of the cases were silent/non-functional (n = 50), followed by acromegaly/growth hormone (GH) tumors (n = 24), Cushing ACTH-secreting tumor (overt: n = 18, subclinical n = 6), prolactinomas (n = 15), thyroid hormone-secreting (n = 6), “Mixed Acromegaly and prolactinoma” (n = 5), and one case of gonadotrophin-secreting tumor (excluded from analysis due to single case number). Each gender is similarly represented (63 male, 62 female).

Table 1.

Patient cohort’s demographics and tumor subtypes/clinical presentations

Tumor subtype	n	Age (years)	Male (%)
Acromegaly	24	48.3 ± 2.6	50.0
Cushing
Overt	18	43.4 ± 3.5	16.7
Subclinical	6	44.6 ± 4.0	33.3
Gonadotrophin	1	41.2
Hyperprolactinemia	15	38.0 ± 4.2	60.0
Hyperthyroidism	6	62.9 ± 4.8	33.3
Mixed Acromegaly and prolactinoma	5	45.3 ± 4.8	40.0
Silent	50	57.2 ± 2.0	64.0

Overall, we noticed that the majority of infiltrating immune cells within pituitary adenomas were comprised of M2 macrophages followed by resting CD4⁺ memory T cells and mast cells (Fig. 1a and b). The most dominant T cell phenotype was the CD4⁺ memory resting phenotype (Fig. 1a). M2 macrophages far outnumbered M1 macrophages in all pituitary adenoma subtypes (Fig. 1b).

Fig. 1.

In silico dissection of immunological landscape in pituitary adenomas. a Heatmap of the relative proportion of myeloid cells. b Heatmap of the relative proportion of lymphoid cells. c Relative proportion of M2 macrophages in various types of pituitary tumors. d Relative proportion of CD8⁺ T cells in various types of pituitary tumors

Silent pituitary tumors tend to have higher M2 macrophage fractions than most other subtypes, including Cushing tumors (q < 0.01), GH-secreting tumors (q < 0.01), prolactinomas (q = 0.036), and Mixed Acromegaly and prolactinoma (q < 0.01) (Fig. 1c). Cushing pituitary tumors, both overt and subclinical cases, had higher CD8⁺ T cells than GH tumors (q < 0.0001), prolactinomas (q = 0.001), hyperthyroid tumors (q < 0.0001) and silent tumors (q < 0.0001) (Fig. 1d).

Discussion

Using this digital dissection of the tumor immunological landscape from existing gene expression data, we have demonstrated distinct immune profiles within different subtypes of pituitary adenomas. Importantly, we corroborated the findings by Lu et al. that these tumors are highly infiltrated by myeloid cells [3]. A recent study by Marques et al. identified increased CD163:HLA-DR macrophages and a reduced CD8:CD4 T cell ratio in pituitary adenomas when compared to normal pituitary [8]. Sato et al. demonstrated that there was increased tumor-associated macrophages in more aggressive non-functional pituitary adenomas [9–11]. Our in silico data further supports that these macrophages are of the immunosuppressive, stromal-supportive M2 subtype. Kemeny et al. demonstrated increased CD3⁺ T cell infiltration in both ACTH and GH-secreting tumors, but did not define them by CD4 or CD8 expression [12]. In the current study, we found that T cell infiltration mostly consisted of resting memory CD4⁺ cells with the exception of ACTH-secreting tumors that exhibit higher CD8⁺ T cell infiltration. Our data suggests that these tumors may harbor completely different immunobiology depending on the subtype.

Tumor Immunity in MicroEnvironment (TIME) classification depends on both PD-L1 expression and the presence of T cell infiltration [2]. Our data suggests that most pituitary adenomas are Type 1 or 4 tumors based on TIME classification given scant T cell infiltration, especially CD8⁺ cytotoxic T cells, and variable PD-L1 expression [4, 12, 13]. Type 1 or 4 tumors are often referred to as “cold” tumors and account for most non-responders to anti-PD (PD-1 or PD-L1) therapy in solid tumors [14]. The one exception in our data is the ACTH-secreting (Cushing) pituitary adenomas that have higher CD8⁺ T cell infiltration. This is in contrast to the report by Lu et al. that described higher CD8⁺ cell infiltration in GH-secreting compared to ACTH-secreting tumors [3]. This may be explained, in part, by sample size as the current dataset has three times as many samples for these adenoma subtypes. Our current data indicating higher CD8⁺ T cells in ACTH-secreting/Cushing pituitary adenomas would suggest that these adenomas may be more amenable to anti-PD therapies. In cases where there is scant immune infiltration in other kinds of pituitary adenomas, the focus should be on increasing immune cell recruitment into the tumor microenvironment.

A limitation of our current approach is that in silico data processing is inferred from bulk RNA data. Its use precludes the study of cyto-architecture and the spatial relationship between immune cells and tumor cells. However, the use of this digital technique has successfully recapitulated flow cytometry and immunohistochemistry data and found to have significant implications in patient survival [15–19]. Although some of the pituitary tumor subtypes are low in numbers, this current paper is, to our knowledge, the first to characterize potential differences in immune cell subtypes among different types of pituitary adenomas using an unbiased computational approach. It provides robust and quick hypothesis-generating data using an existing gene expression database.

In conclusion, our in silico dissection of the immune infiltrates of pituitary adenomas using CIBERSORTx suggests that most pituitary adenomas comprise of M2 macrophages, but each adenoma subtype has a unique immune landscape. This calls for further investigations in the TIME of these tumors in order to develop rational immunotherapeutic strategies.