The study protocol was approved by the institutional review boards of The Jikei University School of Medicine [permission number: 35–181(11,810)] and University of Tsukuba [permission number: R05-124]. We implemented an opt-out method that was approved by the institutional review board to acquire consent for this retrospective study. Our institutional homepage has a notice that provides the necessary information.

This study included two different cohorts. First, we extracted data on ten tumors with documented stiffness according to the surgical records of patients with PCNSL. These records were included based on neurosurgeon descriptions clearly mentioning tumor stiffness. This cohort consisted of cases from the Departments of Neurosurgery at The Jikei University Hospital, The Jikei University Katsushika Medical Center, The Jikei University Kashiwa Hospital, and University of Tsukuba treated between January 2015 and June 2023. Next, 43 tumors were extracted from surgical records of consecutive PCNSL cases without “descriptions of stiffness”. Patient data, such as age, sex, surgical procedure, treatment, and overall survival were obtained from electronic charts and surgical records. Cases with secondary CNS lymphoma, histology other than DLBCL, immune human immunodeficiency virus (HIV)-related lymphoma, preoperative steroid administration, and lack of tumor sample for analysis were excluded from the present study.

All surgical specimens were fixed in 10% neutral-buffered formalin solution (Muto Pure Chemicals, Tokyo, Japan) or 10% formalin (Mitsubishi Gas Chemical, Tokyo, Japan), processed and embedded in paraffin, creating formalin-fixed paraffin-embedded (FFPE) sections. FFPE tissues were thinned to 4 μm for staining with hematoxylin and eosin (HE) and Masson’s trichrome (MT). All selected PCNSL cases were reviewed by two pathologists (MM and NF) and classified into germinal center B (GCB) or non-GCB subtypes according to expression of CD10, Bcl6, and MUM1 determined by immunohistochemistry [11]. FFPE tissues were also thinned to 7.5 μm for silver-impregnation using Watanabe’s method, to identify reticular fiber as black structure and collagen formation as red structure. We scored reticular fibers from 0 to + 3 according to the following criteria (excluding the perivascular reticulin network): 0, no cross-linking at all; + 1, slight cross-linking; + 2, moderate cross-linking; and + 3, prominent aggregation with collagenous stroma deposition (Fig. 1a). Cases scoring + 3 for reticular fibrosis were taken as the fibrosis group (FG). All other cases were allocated to the control group (CG).

Definitions for histopathological categorization. a Reticular fiber expression was scored as the degree of reticular fiber cross-linking and aggregation. b, c Both αSMA-positive spindle cells and GFAP-positive cells were classified into four levels according to density. Original magnifications: a × 100; b, c × 200. Scale bars: a 200 μm; b, c 100 μm

Immunohistochemical analysis using antibodies for CD20, CD3, CD4, CD8, programmed cell death 1 (PD-1), forkhead box P3 (Foxp3), CD68, CD163, α-smooth muscle actin (αSMA), and glial fibrillary acidic protein (GFAP) were performed on 4-μm FFPE sections. Details of each antibody are presented in Supplemental Table 1. Automated immunohistochemistry (IHC) with the CD20, CD3, CD4, CD8, PD1, CD68, CD163, αSMA, and GFAP were performed in a BenchMark-ULTRA autostainer or Discovery-ultra autostainer (Ventana Medical Systems, Tucson, AZ, USA; division of Roche Diagnostics, Basel, Switzerland). In brief, slides were deparaffinized using EZ prep (catalog no. 950–102; Roche) followed by antigen retrieval (cell conditioning solution [CC1], catalog no. 951–124; Roche) at the necessary temperature and time presented in Supplemental Table 1. After retrieval, slides were blocked for peroxidase for 4 min. IHC was performed with incubation for each antibody at the required temperature and time. An Ultra-View DAB Universal kit (catalog no. 760–500; Roche) was used according to the recommendations from the manufacturer for visualization of the bound primary antibody. Slides were then counterstained with hematoxylin II (catalog no. 790–2208; Roche) for 8 min followed by bluing reagent (catalog no. 760–2037; Roche) for 4 min, prior to dehydration in a graded alcohol series. For IHC with Foxp3, slides were deparaffinized followed by antigen retrieval in citrate buffer (pH 6.0) in an autoclave. Slides were incubated with anti-Foxp3 antibody (Supplemental Table 1) at 4 °C overnight.

To assess CD3, CD4, CD8, Foxp3, PD-1, CD68, and CD163 expressions, stained sections were surveyed under a low-power field (× 40). In each specimen, three hot spots were selected in the tumor center and three in the tumor margins, making a total of six spots. Positive cells in these areas were counted in three high-power fields (× 400, total 0.71 mm2) by one neurosurgeon and one pathologist (JT and MM), as previously described [12, 13]. For quantitative evaluation of αSMA and GFAP, stained sections were screened in a low-power field (× 40) and a middle-power field (× 200) showing the densest spots in the tumor center, then margins were assessed in 4 categories (0 to 3 + , as shown in Fig. 1b, c) by two pathologists (MM and NF). Expression of αSMA was qualified using the following criteria (Fig. 1b): 0, αSMA positivity detected only in the vessel wall; + 1, αSMA-positive spindle cells scattered in small numbers outside the vessel wall; + 2, αSMA-positive spindle cells moderately distributed with partial cross-linking; and + 3, αSMA-positive spindle cells distributed at high density. We used αSMA as a cancer-associated fibroblast (CAF) marker [14, 15] and categorized the expression of αSMA focusing exclusively on spindle-shaped cells as shown in supplemental Fig. 1. GFAP evaluation (Fig. 1c) used the following criteria: 0, no or few GFAP-positive cells; + 1, a few GFAP-positive gemistocytic astrocytes; + 2, moderate number of gemistocytic or fibrillary astrocytes; + 3, a large number of gemistocytic or fibrillary astrocytes with dense distribution. If these findings were only very localized in the specimen, cases were classified into a lower category. We used GFAP to distinguish gliosis. Cells were counted using Fiji software (version 2.0.0-re-69/1.52p) [16]. All photos were captured from the identical microscopic field. This ensured that similar locations were chosen across the specimens for CD3, CD4, CD8, Foxp3, PD-1, CD68, CD163, αSMA, and GFAP.

Continuous data are expressed as the median and interquartile range (IQR), and categorical data as numbers and percentages. To compare characteristics between the two groups, the Mann–Whitney U test (MWU) and paired-t-tests were used, as appropriate. Correlation analysis was used to measure the strength of the linear relationship between two variables. To compare the survival time between the two groups, a log-rank test was used. Cases with missing data were omitted, and the remaining available data were analyzed. Statistical analyses were performed using STATA version 18 (StataCorp LLC, College Station, TX, USA) or GraphPad Prism version 10 (GraphPad Software, Boston, MA, USA). All p-values were two-sided, with values of p < 0.05 considered significant.