GSEA in MTX-resistant PCNSL cells and clinical samples

In this study, to determine gene signatures in MTX-resistant PCNSL cells and PCNSL patients with poor prognoses, we conducted GSEA with NGS in PCNSL cell lines HKBML and TK and their MTX-resistant cells (HKBML-MTX and TK-MTX), and PCNSL tumor tissues. In rank scores, 200 upregulated genes in HKBML-MTX and TK-MTX, compared with HKBML and TK, respectively, were enriched into leading edges of PCNSL with poor prognoses. The top enrichment scores (ESs) were 0.479 and 0.534 in HKBML-MTX and TK-MTX, respectively (Fig. 1a,b, Suppl. Table S1). While, 200 downregulated genes in HKBML-MTX and TK-MTX, compared with HKBML and TK, respectively, were also enriched into leading edges of PCNSL with poor prognoses. The top ESs were 0.324 and 0.260 in HKBML-MTX and TK-MTX, respectively (Fig. 1c,d, Suppl. Table S1). These results indicate that the upregulated genes in HKBML-MTX and TK-MTX are strongly enriched in PCNSL with poor prognoses and also denote that a part of downregulated genes in HKBML-MTX and TK-MTX are inversely correlated with PCNSL with good prognoses, suggestive of malignancies in PCNSL.

Figure 1

Gene set enrichment analysis (GSEA) in MTX-resistant PCNSL cells. (a,b) GSEA for upregulated genes of MTX-resistant PCNSL cells in PCNSL with poor and good prognoses. The upregulated genes in (a) HKBML-MTX and (b) TK-MTX, compared with HKBML and TK, respectively. (c,d) GSEA for downregulated genes of MTX-resistant PCNSL cells in poor and good prognoses. The downregulated genes in (c) HKBML-MTX and (d) TK-MTX, compared with HKBML and TK, respectively. ES; enrichment score. (e,f) Venn diagram of genes in MTX-resistant PCNSL cells, compared with control cells. Genes associated with high enrichment score in GSEA were selected. Numbers of genes were shown in Venn diagram. (e) The numbers of upregulated genes in HKBML-MTX and TK-MTX, compared with control cells. (f) The numbers of downregulated genes in HKBML-MTX and TK-MTX, compared with control cells. (g,h) Lists of cell-type-specifically and commonly expressed genes in HKBML-MTX and TK-MTX, compared with control cells, associated with ESs in GSEA. (g) Upregulated genes. (h) Downregulated genes.

Rank score-associated genes of MTX-resistant PCNSL cells in GSEA

We next examined rank scored genes in GSEA, which was shown in Venn diagrams (Fig. 1e,f). The 39 genes were found in the upregulated gene sets in TK-MTX and HKBML-MTX (Fig. 1e,g, Suppl. Table S2). The upregulated genes in HKBML-MTX, compared with HKBML, included nicotinamide adenine dinucleotide (NAD)H dehydrogenase 1α subcomplex subunit 1 (NDUFA1) and 11 (NDUFA11) for mitochondrial electron transport, solute carrier family 25 member 5 (SLC25A5) for ADP/ATP exchange, protein tyrosine phosphatase receptor type C-associated protein (PTPRCAP) for protein phosphatase, cyclin-dependent kinases regulatory subunit 2 (CKS2) for cell cycle, and cluster of differentiation (CD) 79B (CD79B) as B lymphocyte antigen receptors, in addition to high mobility group nucleosomal binding domain 2 (HMGN2) and small nuclear ribonucleoprotein-associated proteins B and B’ (SNRPB). The upregulated genes in TK-MTX, compared with TK, included CD70 as tumor necrosis factor (TNF) ligands, Epstein-Barr virus-induced gene 3 (EBI3) for activation of T, B, and myeloid cells, interferon γ-inducible protein 30 (IFI30) for major histocompatibility complex (MHC) class II-restricted antigen processing, cytochrome b-245 α chain (CYBA) for superoxide production and phagocytosis, and cytochrome c oxidase subunit 5B, mitochondrial (COX5B) as a subunit of complex IV in mitochondrial electron transport chain, and β-galactoside-binding lectin (LGALS1) for suppression of Th1 and Th17 helper T cells, in addition to fatty acid-binding protein 5 (FABP5) and ferritin heavy chain 1 (FTH1). The commonly upregulated genes in both HKBML-MTX and TK-MTX included thymosin β-10 (TMSB10) for organization of cytoskeleton, and macrophage migration inhibitory factor (MIF) for suppression of anti-inflammatory effects. Interestingly, of the downregulated genes, antisense RNAs of forkhead box transcription factor D2 (FOXD2-AS1) and myosin light chain kinase (MYLK-AS1) were included in both HKBML-MTX and TK-MTX, implying that the expression of FOXD2 and MYLK allow. While, only the ST3 β-galactoside α-2,3-sialyltransferase 4 (ST3GAL4) gene as sialyltransferase for β-galactoside was common in the downregulated gene sets in both TK-MTX and HKBML-MTX (Fig. 1f,h, Suppl. Table S2). These results suggest that the upregulated or downregulated genes in HKBML-MTX and TK-MTX could be potential prognostic factors and estimation of molecular pathways for targeted therapy in PCNSL.

GSEA-assisted genes as prognostic marker candidates in PCNSL

As described above, the rank score-associated genes successfully promoted gene marker candidates in PCNSL. Then, we further examined gene marker candidates along expression changes in the MTX-resistant PCNSL cells. In the GSEA-assisted genes, the 21 genes were differentially expressed with >1.5-fold (p

Figure 2

Identification of cell-type-dependent marker candidates for PCNSL with poor survivals and MTX resistances. (a-c) Differential expression in (a) PCNSL with poor survivals, (b) HKBML-MTX, and (c) TK-MTX. (d) Clustering of differential expression marker candidates. Green-black-red as low-median-high expression. (e–l) Expression in PCNSL samples divided by prognoses in box-whisker plots. (e) FOXD2-AS1 and (f) MMP19 as gene marker candidates for PCNSL with poor survivals. (g) FABP5 and (h) CD70 as gene marker candidates for HKBML-MTX. (i) CLCN2, (j) HOXB9, (k) INE1, and (l) LRP5L as gene marker candidates for TK- MTX. FPKM; fragments per kilobase of exon model per million reads mapped.

Table 1 Characterization of prognosis marker candidates in MTX-resistant PCNSL with poor prognoses.

Differential expression of GSEA-assisted genes in PCNSL

The above-described results proposed a possibility that GSEA-assisted genes in the MTX-resistant-PCNSL cells were useful for prognosis prediction in PCNSL. We further examined their potentials as prognostic markers in PCNSL. Each clustering was divided PCNSL patients into three clusters, named clusters 1, 2, and 3 in each GSEA-assisted gene category in the MTX-resistant-PCNSL cells (Fig. 3a–d), followed by Kaplan-Meier survival analyses. Survival curves were clearly divided in the clustering of downregulated genes in HKBML-MTX (Table 2, Fig. 3f, Suppl. Fig. S2b) and TK-MTX (Table 2, Fig. 3h, Suppl. Fig. S3b), but not in the clustering of upregulated genes in HKBML-MTX (Table 2, Fig. 3e, Suppl. Fig. S2a) and TK-MTX (Table 2, Fig. 3g, Suppl. Fig. S3a). The GSEA-assisted downregulated genes expressed >3.0-fold (p

Figure 3

Differential expression of MTX-resistant PCNSL cells-agitated genes in PCNSL with poor prognoses. (a–d) Clustering of gene expression of MTX-resistant PCNSL cells in PCNSL specimens. GSEA-assisted upregulated genes in (a) HKBML-MTX and (c) TK-MTX. GSEA-assisted downregulated genes in (b) HKBML-MTX and (d) TK-MTX. Green-black-red as low-median-high expression. (e–h) Survival distribution of PCNSL patients divided into clusters. Clusters 1, 2, and 3 in each panel of (e–h), correspond to those in (a–d). Kaplan-Meier survival curves were evaluated with log-rank tests. OS; overall survival.

Table 2 Survival analyses for the MTX-resistant cells-promoted genes in PCNSL.

Gene ontology search for the GSEA-assisted genes in MTX-resistant PCNSL cells

Gene ontologies (GOs) of the aforementioned genes were surveyed (p

Molecular function and signaling pathways in MTX-resistant PCNSL cells

In public data sets including immunologic signatures, oncogenic signatures, computational gene set for cancer modules, and curated gene set for canonical pathways, we further analyzed the GSEA-assisted gene signatures proposed in this study (Suppl. Table S5). The HKBML-MTX-specifically upregulated gene set was correlated to formation of the ternary complex and subsequently 43 S complex (Suppl. Fig. S4a), a cancer module (Suppl. Fig. S4b), thymocytes compared with CD4 T cell in adult blood (Suppl. Fig. S4c), and neutrophil compared with B cell (Suppl. Fig. S4d). The comprehensive results in the GSEA-associated genes in HKBML-MTX suggest that HKBML-type poor prognosis marker candidates in PCNSL would be ribonucleoprotein complex formation, T cell maturation and activation, and cancers including neutrophils. The commonly upregulated gene set in HKBML-MTX and TK-MTX was correlated with a cancer module (Suppl. Fig. S4e), and thymocytes compared with CD4 T cell in adult blood (Suppl. Fig. S4f). Th1 compared with Th17 was detected in TK-MTX-specific cell differentiation (Suppl. Fig. S4g). Considering these results, the GSEA-assisted genes in TK-MTX and HKBML-MTX are similar in gene function. Furthermore, TK-type PCNSL with poor prognoses would also be involved in Th1 maturation from naïve CD4 T cells.

Estimated molecular networks in the MTX-resistant PCNSL cells and PCNSL tissues

Finally, to validate gene function and network in the MTX-resistant PCNSL cells, we investigated protein–protein interactions (PPIs) in the MTX-resistant PCNSL cells and PCNSL tissues (Suppl. Table S6). Here, the 200 upregulated genes in each HKBML-MTX and TK-MTX, and the 300 upregulated genes in PCNSL with poor prognoses were examined. Of these, 58.5–73.6% hit in the database, and 41–59% were found in clusters. Average numbers of edge per node were 37.7–45.7. Of the seed genes as hubs, peroxiredoxin-1 (PRDX1) was found in HKBML-MTX and TK-MTX. Nascent-polypeptide-associated complex α polypeptide (NACA) was found in TK-MTX and PCNSL with poor prognoses. These results suggest that the MTX-resistant PCNSL cells share a hub of oxidative stress and redox; in addition, TK-MTX and PCNSL with poor prognoses share a hub of ribosome complex and transcription. In HKBML-MTX, PPI hubs for immune system, oxidative phosphorylation, cytoskeleton, and ribosomal protein were detected (Fig. 4a), whereas in TK-MTX, PPI hubs for immune system, oxidative phosphorylation, ribosomal protein, and glycosylation were detected (Fig. 4b). Further, immune system, RNA-binding protein, mitochondria respiratory chain, and cell proliferation were clustered in HKBML-MTX-specific PPI networks (Fig. 4c). While, immune system, proteasome, ATP production with cytochrome c and metal chaperone, redox, and glycoprotein were clustered in TK-MTX-specific PPI networks (Fig. 4d). These results indicate that common PPI networks in both cells are ribosomal protein, immune system, and oxidative phosphorylation, whereas cytoskeleton and glycosylation are unique in HKBML-MTX and TK-MTX, respectively. The distinct PPI networks also included cell proliferation, RNA-binding protein, and NAD + /NADH dehydrogenase in HKBML-MTX, and proteasome, redox, and ATP synthesis with metal chaperone in TK-MTX. In the PCNSL with poor prognoses, PPI hubs with complex networks were involved in immune system, ribosomal protein, mitochondrial protein, and histone, which overlapped PPIs in HKBML-MTX and TK-MTX (Suppl. Fig. S5). These results also suggest that PCNSL includes different cell-types, whereas details for protein expression and interaction should await future studies.

Figure 4

Protein-protein interaction (PPI) networks in MTX-resistant PCNSL cells. Networks were estimated by MCODE and STRING. (a,b) PPI networks on upregulated genes in (a) HKBML-MTX and (b) TK-MTX, compared with control cells. (c,d) PPI networks on cell-type-dependent upregulated genes in (c) HKBML and (d) TK.