CD163 Dominance Within Macrophages/Microglia Reflects a Favorable Prognosis in Patients With Glioblastoma

ABSTRACT

Glioblastoma is the most common primary malignant central nervous system (CNS) tumor; however, its microenvironment, including tumor‐associated microglia/macrophages (TAMs), is not fully understood. Although ionized calcium‐binding adaptor molecule 1 (IBA‐1) is expressed in various microglial phenotypes in the healthy human brain, CD163 is a marker of activated/phagocytic microglia. Tissues from 34 patients with glioblastoma were analyzed, and 17.6% of cases were CD163‐dominant. CD163‐dominant patients showed better overall survival than IBA‐1‐dominant patients (p = 0.019). CD163 dominance was identified as an independent prognostic factor for overall survival (hazard ratio [HR], 0.17; p = 0.011). Compared with the IBA‐1‐dominant group, CD163‐dominant tumors showed more CD4‐positive cell infiltration (p = 0.005), fewer ameboid TAMs (p = 0.021), and fewer non‐microvascular proliferation (non‐MVP) vessels (p = 0.012). No significant differences were found in patient characteristics, such as age, sex, tumor location, or extent of resection. The CD163‐to‐IBA‐1 ratio of TAMs is a significant independent prognostic factor in glioblastoma, suggesting that the activation status of these cells and their interactions with the vascular endothelium and T cells influence tumor progression. These findings highlight TAMs as potential therapeutic targets for glioblastoma.

1. Introduction

Glioblastoma is the most common primary malignant tumor of the central nervous system (CNS) and is defined as a diffuse astrocytic glioma that is isocitrate dehydrogenase (IDH)‐wildtype and H3‐wildtype with high‐grade features [1]. The complex interaction between glioblastoma cells and their microenvironment has been discussed.

Microglia are the resident macrophages in the CNS and play a crucial role in maintaining brain homeostasis. The morphological state of the microglia depends on their activation status and can be classified as ramified (resting), activated, or phagocytic [2]. Activated/phagocytic microglia respond to acute destructive conditions, such as infarction or infection. Although no specific markers exist for each microglial subtype in the human brain, activated or phagocytic microglia tend to express CD163, whereas ionized calcium‐binding adaptor molecule 1 (IBA‐1) is present in all microglial cells regardless of their activation status in the healthy human brain [3].

The tumor microenvironment has been investigated in various cancers, and immune checkpoint inhibitors have shown clinical efficacy. However, the glioblastoma microenvironment is not fully understood. A major challenge in studying glioblastomas is the difficulty in precisely identifying glioma‐infiltrating TAMs owing to their heterogeneity [4]. In this study, we evaluated the status of TAMs within the glioblastoma microenvironment by assessing the ratio of CD163‐positive to IBA‐1‐positive cells and their morphology using immunohistochemical staining.

2. Materials and Methods

2.1. Clinical Sample and Histology

In this study, tissues from 34 patients with glioblastoma were obtained through surgery between 2011 and 2015 and stored in the Department of Pathology, Kurume University. The integrated diagnosis was re‐evaluated by two experienced neuropathologists based on the latest WHO criteria [1], including molecular analysis, as previously reported [5, 6]. We used a tissue microarray (TMA) containing two or more cores from each of the 34 specimens in a formalin‐fixed, paraffin‐embedded block, with each sample placed in a 2.5 mm diameter space.

The research ethics committee of Kurume University approved the use of patient materials and clinical information (approval no. 25102), and the study was conducted in accordance with the Declaration of Helsinki.

2.2. Evaluation of Immunohistochemical Staining of CD3, CD4, CD8, CD163, and IBA‐1

Slide‐mounted formalin‐fixed, paraffin‐embedded tissue sections (2.5 μm in thickness) from the TMA were prepared as previously described [5, 6]. For immunohistochemical analysis, the sections were stained with antibodies against CD3 (clone F7.2.38, Dako, Japan Co., Kyoto, Japan; 1:50 dilution), CD4 (clone SP35, VENTANA; 1:10 dilution), CD8 (clone C8/144B, DAKO; 1:50 dilution), CD163 (clone 10D6, Leica; 1:200 dilution), and IBA‐1 (clone E404W, Cell Signaling Technology; 1:1000 dilution). All cells included within each TMA core were counted and the median was recorded.

The phenotypes of the TAMs were assessed in each area using IBA‐1 staining and were classified by morphology by two neuropathologists (T.F. and M.M.) (Figure 1). Representative images of glioblastomas with hematoxylin‐and‐eosin staining are shown in (Figure 1A,E). As there were few ramified TAMs (Figure 1D), ameboid and activated TAMs (Figure 1H) were counted. All CD163‐positive TAMs (Figure 1B,F) and IBA‐1–positive TAMs (Figure 1C,G) were counted. If the total number of CD163‐positive TAMs exceeded the number of IBA‐1–positive TAMs, the patient was categorized as CD163‐dominant (Figure 1A–C). The following parameters were evaluated in each TMA core: microvascular proliferation (MVP) (Figure 2A), non‐MVP vessels (Figure 2B), and CD4‐positive T cells (Figure 2C,D). CD3‐positive (Figure 2E) and CD8‐positive cells (Figure 2F) were also evaluated. MVP (Figure 2A) and non‐MVP vessels (Figure 2B) were counted as described in our previous study [7]. Six cases (17.6%) were categorized as CD163‐dominant.

FIGURE 1.

Representative images of CD163‐dominant specimens (A–C) and IBA‐1–dominant specimens (E–G). Ramified TAM (D, arrowhead) and ameboid TAM (H, arrowhead) were identified (IBA‐1, scale bar = 20 μm). IBA‐1–positive TAMs (C) were fewer than CD163‐positive TAMs (B), whereas IBA‐1–positive TAMs (G) were more than CD163‐positive TAMs (F) (A, E, hematoxylin and eosin [HE] stain; B, F, CD163 stain; C, D, G, H, IBA‐1 stain).

FIGURE 2.

Representative images of MVP vessel (A), a non‐MVP vessel (B), and CD4‐positive cells (C, D). The IBA‐1‐dominant tumor had fewer non‐MVP vessels (mean, 33.5; median, 19.5) than the CD163‐dominant tumors (mean, 16.5; median, 8.3) (G). The IBA1‐dominant tumor had fewer ameboid TAMs (mean, 4.3; median, 4.2) than the CD163‐dominant tumors (mean, 17.8; median, 24.2) (H). The CD163‐dominant tumor had more CD4‐positive cells (mean, 9.1, median, 25.5, D) than the IBA‐1–dominant tumors (mean, 0; median, 15.8, C) (I). Some CD4‐positive cells were not only positive for CD3 (E) but also positive for CD8 (F). (A–C, HE stain; D, E, CD4 stain; scale bar = 50 μm).

3. Results

3.1. Clinicopathological Characteristics of Patients With Glioblastoma

The clinicopathological characteristics of the 34 patients are summarized in Table 1. The median age was 66.9 years (range, 32–86). A total of 22 males and 12 females were included in the study. The follow‐up period ranged from 2.1 to 51.4 months, with a median of 19.4 months. Eleven tumors (32.3%) were located in the frontal lobe; seven (21%) in the parietal lobe; 13 (38%) in the temporal lobe, and three (8.8%) in other areas (temporoparietal lobe, cerebellum, and thalamus). The median Karnofsky performance status (KPS) score was 70 (range, 30–90). Twenty‐two patients (64.7%) underwent gross total resection (GTR). IBA‐1‐positive TAMs were the most frequent inflammatory cells (mean, 211.3). CD8‐positive T cells (mean, 8.0) infiltrated more than CD4‐positive T cells (mean, 3.4); however, some populations were considered double‐positive T cells. Ameboid TAMs (mean, 20.7) were more abundant than activated TAMs (mean, 9.5) (Figures 1A,B and S1).

TABLE 1.

Patient characteristics (n = 34).

Characteristics	CD163 > IBA‐1 (n = 6)	CD163 < IBA‐1 (n = 28)	p
Median age, years (range)	65.5 (37–76)	67.5 (32–86)	0.79
Sex, n (%)
Male	4 (66.7)	18 (64.3)	0.91
Localization			0.09 a
Frontal	1	10
Temporal	2	1
Parietal	0	7
Others	3	10
Mean KPS at preoperation (range)	90 (30–90)	80 (60–90)	0.41
Extent of resection, n (%)
Gross total resection	3 (50.0)	19 (67.9)	0.16
CD4‐positive cells infiltration	15.8	25.5	0.01
MVP vessels	15.4	17.9	0.59
Non‐MVP vessels	19.5	8.3	0.01
Phenotype of TAMs
Activated	8.7 (0–25.3)	9.9 (0–44.5)	1.00
Ameboid	24.2 (0–130)	4.2 (2–6)	0.02 b

Abbreviation: KPS, Karnofsky performance status.

^a Fisher's exact test.

^b Wilcoxon rank‐sum test.

3.2. Clinicopathological Comparison Between CD163‐Dominant and IBA‐1–Dominant Patients

CD163‐dominant glioblastomas showed more CD4‐positive cell infiltration (p = 0.005), more thrombus formation (p = 0.011), fewer ameboid TAMs (p = 0.021), and fewer non‐MVP vessels (p = 0.012) than IBA‐1–dominant cases (Figure 2G,H). There were no significant differences in age, sex, tumor location, KPS, MVP vessels, or surgical procedures between the two groups (Table 1). Genetically, there were no significant differences in epidermal growth factor receptor (EGFR) amplification, methylation status of O ⁶ ‐methylguanine methyltransferase (MGMT) promoter, and telomerase reverse transcriptase (TERT) promoter mutations (data not shown).

3.3. Overall Survival (OS) Curves in Patients With Glioblastoma According to CD163 and IBA‐1 Expression

CD163‐dominant patients had better OS than IBA‐1–dominant patients (log‐rank test, p = 0.0193; Figure 3). Univariate analysis was performed using degree of KPS > 80 (hazard ratio [HR], 0.49; 95% confidence interval [CI], 0.21–1.12; p = 0.09), age < 50 (HR, 0.72; 95% CI, 0.22–2.45; p = 0.60), GTR (HR, 0.57; 95% CI, 0.25–1.30; p = 0.18), CD4 > average (HR, 0.65; 95% CI, 0.22–1.91; p = 0.43), and CD163‐dominant (HR, 0.25; 95% CI, 0.07–0.86; p = 0.03) (Table 2). Multivariate analysis using these variables indicated that CD163 > IBA‐1 was an independent prognostic factor (HR, 0.16; 95% CI, 0.04–0.66; p = 0.01, Table 2).

FIGURE 3.

Kaplan–Meier curves of overall survival in glioblastoma patients. CD163 > IBA‐1‐patients had significantly better overall survival than those with CD163 < IBA‐1 patients (p = 0.0193*).

TABLE 2.

Prognostic factors affecting the OS of patients with glioblastoma.

Variable	Univariate			Multivariate
	HR	95% CI	p	HR	95% CI	p
KPS > 80	0.49	0.21–1.12	0.09	0.37	0.13–1.05	0.06
Age < 50	0.72	0.22–2.45	0.60	0.23	0.05–0.97	0.04
Gross total resection	0.57	0.25–1.30	0.18	0.66	0.26–1.67	0.38
CD4 > average	0.65	0.22–1.91	0.43	1.01	0.28–3.66	0.98
CD163 > IBA‐1	0.25	0.07–0.86	0.03	0.16	0.04–0.66	0.01

4. Discussion

Patients with CD163‐dominant tumors had better OS than those with IBA‐1–dominant tumors, and CD163‐dominance was an independent prognostic factor. CD163‐dominant glioblastomas show more CD4‐positive cell infiltration, more thrombus formation, and fewer non‐MVP vessels than IBA‐1–dominant cases. In IBA‐1–dominant cases, ameboid TAMs were predominant.

Microglia account for 5%–12% of the cells in the CNS [8], originate from mesodermal cells, and begin to be recognized as ameboid microglia at 13 and 18 weeks of gestation. TAMs are the most abundant microenvironmental cells in glioblastoma [9]. However, the diversity and origin of TAMs in glioblastomas remain controversial. Macrophages originating from the bone marrow and TAMs in the brain may have different biological potentials, as traditional bone marrow transplantation approaches fail to replace TAMs in the brain [10]. IBA‐1‐positive cells are more closely associated with CD276‐expressing glioma stem cells (GSCs) than CD163‐positive cells [11]. The prevalence of ameboid phenotypes in CD163‐dominant tumors may reflect an activated and highly phagocytic state of TAMs rather than a uniformly tumor‐promoting phenotype. Activated microglia can acquire amoeboid morphology while simultaneously exhibiting enhanced antigen presentation and immune regulatory functions [12]. Thus, morphological activation alone may not necessarily indicate protumoral activity in glioblastoma.

IBA‐1 and CD163 are established TAM cell markers [13]. More CD163 infiltration [14] and less IBA‐1 infiltration have been reported as poor prognostic factors, respectively [15]. Interestingly, the prognostic significance of CD163‐positive TAMs remains controversial. Zeiner et al. reported that increased infiltration of CD163‐positive and CD206‐positive glioma‐associated microglia/macrophages within the vital tumor core was associated with prolonged OS in patients with IDH‐wildtype glioblastoma [16]. Their transcriptomic analyses demonstrated the simultaneous expression of both pro‐inflammatory and anti‐inflammatory genes, suggesting that CD163‐positive TAMs may reflect phagocytic and antigen‐presenting functions rather than a purely tumor‐promoting M2 phenotype. Therefore, a higher proportion of CD163‐positive TAMs may not necessarily indicate an immunosuppressive microenvironment. However, the majority of IBA‐1 and CD163 in TAMs have not been reported in glioblastoma. Notably, our data were not associated with prognosis using single immunohistochemical staining for CD163 or IBA‐1, suggesting that the balance between these two phenotypes was related to prognosis. Macrophages have been classified as classically activated macrophages (M1 cells) and anti‐inflammatory macrophages (M2 cells) [4]; the latter are positive for CD163, while the former are not [17]. The activated form of M1 plays an antitumor role, whereas M2 promotes tumor growth [4, 18]. The M1/M2 dichotomy does not fully capture the complex nature of TAMs in glioblastoma, as individual TAMs often express both M1 and M2 genes and these phenotypes change depending on their environment and treatments [3, 19]. Recent comprehensive reviews integrating single‐cell and spatial transcriptomic studies have further emphasized that glioblastoma‐associated TAMs occupy a spectrum of activation states that do not correspond to conventional M1 or M2 polarization [12]. Instead, TAM phenotypes are dynamically shaped by local microenvironmental factors (hypoxia, vascular niches, tumor architecture, and interactions with glioma cells). Moreover, CD163 expression is not restricted to monocyte‐derived macrophages, as activated resident microglia can also upregulate CD163 and become morphologically indistinguishable from infiltrating macrophages in glioblastoma. These findings highlight the limitations of interpreting CD163 positivity solely as a marker of immunosuppressive M2 macrophages. A major challenge in studying glioblastomas is the difficulty in precisely identifying TAMs. This is largely because of the lack of a specific immunohistochemical marker that can distinguish resident microglia from peripheral macrophages [20]. Previous studies that relied on immunohistochemical evaluations may have underestimated the importance of morphological assessment (such as classifying cells as ramified, phagocytic, or activated), leading to potentially ambiguous results regarding the exact composition and function of TAMs in glioblastoma.

Regarding the relationship between CD4‐positive cell infiltration and prognosis in glioblastoma, a higher infiltration of CD4‐ or CD8‐positive cells alone did not lead to a better prognosis. A combination of high CD4‐positive cells and low CD8‐positive cells is an independent poor prognostic factor for both progression‐free survival (PFS) and OS in patients with glioblastoma [21, 22]. An increased proportion of CD4‐ and FoxP3‐positive cells is associated with tumor recurrence and reduced survival [23]. As CD4‐positive cells are critical regulators of the functional state of infiltrating CD8‐positive T cells, TAMs can be stimulated by CD4‐positive T cells through MHC Class II, the balance between CD4‐ and CD8‐positive T cells, and their potential relationship with microglia in the context of glioblastoma immunotherapy [24]. Therefore, the prognosis of glioblastoma is determined by the immune balance at the tumor site. CD4‐positive and CD8‐positive T cells were often observed in glioblastoma (Figure 2D,F,G), as previously reported [9]. Innate lymphocytes and lymphocytes delivered from the bone marrow may be mixed in various situations. Innate lymphocytes have unique surface markers for peripheral T‐cells [25], and the diseased brain triggers the activation of both the innate immune response and peripheral lymphocytes. For example, in Parkinson's disease, the protein alpha‐synuclein functions as a damage‐associated molecular pattern, activating immune receptors on microglia and macrophages but failing to clear the protein. This triggers a pro‐inflammatory response, releasing cytokines that further damage neurons. The innate immune response is accompanied by an adaptive immune response involving CD4‐positive and CD8‐positive T cells [26]. Therefore, immune cell functions around lymphocytes and TAMs may be different in glioblastomas than in other organs.

The role of microglia in angiogenesis has also been previously investigated. Activated IBA‐1‐positive microglia are associated with the vasculature under neurotoxic condition [27]. TAMs in a hypoxic state, characterized by the upregulation of hypoxia response genes (ADM, BNIP3, and CSTB), produce a substance called adrenomedullin, making tumor vessels abnormally leaky by destabilizing the connections between endothelial cells [28]. CD163‐positive cells promote glioma tumor vessels and their infiltration is correlated with patient outcome [29]. TAMs and angiogenesis are correlated in glioblastomas and are thought to influence drug delivery and migration of tumor cells. As brain endothelial cells regulate lymphocyte migration through the blood–brain barrier via adhesion molecules such as Intercellular Adhesion Molecule‐1 (ICAM‐1) [30], TAMs have the potential to indirectly influence endothelium–lymphocyte interactions. High angiogenesis correlates with increased TAM infiltration and poor survival in patients with glioblastoma [31]. As mentioned above, owing to the ambiguity of M1/M2 polarization, TAMs have only recently been classified by gene expression profiling, and their immunophenotypes have not been investigated [32]. We found that CD163‐dominant tumors were associated with lower neoangiogenesis, suggesting normal drug delivery in these tumors. Recent spatial analyses have demonstrated that distinct TAM subsets preferentially localize to hypoxic and angiogenic niches within glioblastoma and that hypoxia is a major determinant of TAM functional states [12]. Because TAM‐mediated angiogenesis is strongly influenced by local microenvironmental cues, differences in the relative abundance of CD163‐positive and IBA‐1–positive TAM populations may reflect distinct vascular microenvironments rather than simple M1/M2 polarization states. Further studies are required to understand the association between CD163/IBA‐1‐positive TAMs and angiogenesis in glioblastoma.

The population of CD163‐ and IBA‐1–positive TAMs is an independent prognostic factor for glioblastoma. It has been suggested that TAMs influence tumor growth by interacting with both T cells and the vascular endothelium. Our findings support the emerging concept that the biological significance of TAMs in glioblastoma is determined not only by their abundance but also by their functional state, morphology, and relative composition within the tumor microenvironment. In particular, CD163‐dominance may represent a distinct TAM ecosystem that cannot be adequately explained by the conventional M1/M2 paradigm. Targeting the activation status of TAMs may be a key strategy for glioblastoma therapy.

As a supplementary analysis, we evaluated the prognostic significance of CD163 and AIF1, coding IBA‐1, expression using the The Cancer Genome Atlas (TCGA) glioblastoma dataset. In contrast to our immunohistochemical findings, high CD163 mRNA expression was associated with significantly worse OS, whereas AIF1 expression and the CD163/AIF1 ratio were not prognostic (Figure S2). This apparent discrepancy between transcriptomic and histopathological analyses may reflect the fundamentally different biological information captured by these methodologies. Bulk RNA sequencing measures average gene expression across all cells within a tumor and lacks information regarding cellular morphology, spatial distribution, and local microenvironmental context. In contrast, our histological approach assessed the relative predominance and morphology of CD163‐positive and IBA‐1–positive TAM populations. Furthermore, recent studies have shown that glioblastoma cells can acquire myeloid‐like transcriptional programs and express genes traditionally regarded as microglial or macrophage markers [16]. Thus, elevated CD163 or AIF1 transcript levels may not exclusively represent TAM infiltration.

Importantly, prognosis in our cohort was associated not with the absolute abundance of CD163‐positive or IBA‐1–positive cells, but with their relative predominance and accompanying morphological features. These spatial and cellular characteristics cannot be captured by bulk transcriptomic analyses. Therefore, the discrepancy between TCGA‐based and immunohistochemical results may indicate that the functional state and tissue organization of TAMs provides prognostic information beyond what is obtainable from gene‐expression profiling alone.

This study has several limitations. First, the sample size was relatively small (n = 34), which may have limited the statistical power of the analyses and the generalizability of the findings. Validation in larger, independent cohorts is therefore warranted. Second, immunohistochemical evaluation was performed using TMA rather than whole‐slide sections. Given the substantial spatial heterogeneity of glioblastoma and its tumor microenvironment, the sampled cores may not fully represent the entire spectrum of TAM distribution, morphology, vascular changes, and immune‐cell infiltration present within individual tumors. Future studies using whole‐slide analysis and spatially resolved approaches will be necessary to confirm our observations.

Funding

This work was supported by Japan Society for the Promotion of Science (JP23K14490).

Ethics Statement

The study protocol was approved by the Research Ethics Committee of Kurume University (Approval No. 25102), and the study was conducted in accordance with the Declaration of Helsinki.

Consent

The requirement for informed consent was waived by the Research Ethics Committee of Kurume University because this was a retrospective study using archived tissue specimens and anonymized clinical data.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Figure S1: Total numbers of CD3‐positive, CD4‐positive, and CD8‐positive T cells; IBA‐1‐positive and CD163‐positive TAMs, and ameboid and activated TAMs.

Figure S2: TCGA‐GBM dataset analysis for validation. High CD163 mRNA expression was associated with significantly worse overall survival (A), whereas AIF1 expression (B) and the CD163/AIF1 ratio (C) were not prognostic.

Data S1: neup70071‐sup‐0003‐Supinfo.docx.

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.