Primary central nervous system lymphoma (PCNSL) is an isolated type of lymphoma of the central nervous system and has a dismal prognosis despite intensive chemotherapy. Recent genomic analyses have identified highly recurrent mutations of MYD88 and CD79B in immunocompetent PCNSL, whereas LMP1 activation is commonly observed in Epstein–Barr virus (EBV)-positive PCNSL. However, a lack of clinically representative preclinical models has hampered our understanding of the pathogenic mechanisms by which genetic aberrations drive PCNSL disease phenotypes. Here, we establish a panel of 12 orthotopic, patient-derived xenograft (PDX) models from both immunocompetent and EBV-positive PCNSL and secondary CNSL biopsy specimens. PDXs faithfully retained their phenotypic, metabolic, and genetic features, with 100% concordance of MYD88 and CD79B mutations present in PCNSL in immunocompetent patients. These models revealed a convergent functional dependency upon a deregulated RelA/p65-hexokinase 2 signaling axis, codriven by either mutated MYD88/CD79B or LMP1 with Pin1 overactivation in immunocompetent PCNSL and EBV-positive PCNSL, respectively. Notably, distinct molecular alterations used by immunocompetent and EBV-positive PCNSL converged to deregulate RelA/p65 expression and to drive glycolysis, which is critical for intracerebral tumor progression and FDG-PET imaging characteristics. Genetic and pharmacologic inhibition of this key signaling axis potently suppressed PCNSL growth in vitro and in vivo. These patient-derived models offer a platform for predicting clinical chemotherapeutics efficacy and provide critical insights into PCNSL pathogenic mechanisms, accelerating therapeutic discovery for this aggressive disease.

Significance:

A set of clinically relevant CNSL xenografts identifies a hyperactive RelA/p65-hexokinase 2 signaling axis as a driver of progression and potential therapeutic target for treatment and provides a foundational preclinical platform.

Comprehensive genomic analyses have uncovered genetic heterogeneity in diffuse large B-cell lymphomas (DLBCL; refs. 1–3). Among distinct genetic subtypes in DLBCLs, the prognosis of those with MYD88L265P and CD79B concomitant mutations are unfavorable (1, 2). Therefore, novel therapeutic strategies are particularly needed for these tumors. Primary central nervous system lymphoma (PCNSL) is a highly aggressive subtype of extranodal non-Hodgkin's lymphoma, which is confined to the central nervous system (CNS). The vast majority of PCNSLs are DLBCL, belonging to the activated B-cell like (ABC)/nongerminal center B-cell like (non-GCB) subtype, and have an activated B-cell immunophenotype resembling exit of B cells from GC (4–6). Although treatment responses are observed in a subset of patients with PCNSL receiving standard-of-care high-dose methotrexate (HD-MTX)–based regimens, the overall outcome remains unsatisfactory with a 5-year survival rate of 30% to 50% (7). Thus, better understanding of PCNSL pathogenesis is needed to spur the development of novel treatment strategies that can improve prognosis for patients with this disease.

We have previously reported that immunocompetent PCNSLs are hallmarked by a high prevalence of mutations in MYD88 (76%) and CD79B (83%; ref. 8). Although variable mutation frequency has been reported, several studies demonstrated similar findings (9), indicating that these genomic alterations are highly common in PCNSL and considerably more frequent than in systemic DLBCL (10, 11). These mutations are frequently seen in ABC/non-GCB, but have also been identified in the GCB subtype of PCNSL (4, 5), suggesting that these mutations may be pivotal for establishing and maintaining lymphomagenesis within the CNS. In addition, an inhibitor of Bruton's tyrosine kinase (BTK), which links the B-cell receptor and toll-like receptor signaling pathways, has shown durable radiographic responses in a subset of patients with PCNSL (4, 12–14), suggesting a potential functional role for the BTK and downstream NF-κB pathway in PCNSL. Contrarily, in immunocompromised individuals with PCNSL, MYD88 or CD79B alterations are not common; PCNSL in patients with HIV/AIDS, organ transplantation, or chronic immunosuppressive therapy is associated with decreased CD4+ T cells and Epstein–Barr virus (EBV) infection (15, 16). Despite the two distinct etiologic pathways, immunocompetent and immunocompromised, in the development of PCNSL, the histologic and clinical features of PCNSL have been traditionally studied uniformly.

Although comprehensive genomic analyses have uncovered the somatic genetic landscape of PCNSL, the functional contribution of these genomic alterations to CNS lymphomagenesis and tumor progression in PCNSL is largely unexplored, partly due to the paucity of preclinical models of PCNSL (17). Although patient-derived xenograft (PDX) models from patients with systemic DLBCL have been established (18, 19), the rarity of PCNSL and the use of small needle biopsy specimens for diagnosis have slowed our biological understanding of this disease. In this study, we have established 12 novel orthotopic PDX models from patients with immunocompetent and EBV-positive PCNSL and secondary central nervous system lymphoma (SCNSL) and have performed comprehensive histologic, genomic, and metabolic investigations. These represent, to the best of our knowledge, the largest series of CNSL PDX models, recapitulating the critical features of PCNSL tumors and providing a platform to understand the functional impact of PCNSL genetics. Herein, we find that aberrant expression of the RelA/p65 hexokinase 2 (HK2) axis is critical for intracerebral tumor progression, and deactivation of this signaling pathway results in potent antitumor effects in PCNSL.

This study is conducted in accordance with the Declaration of Helsinki and was approved by the Institutional Review Board at Yokohama City University (YCU, Yokohama, Japan, A200900004 and B190700012) and National Cancer Center (Tokyo, Japan, 2013–042). Written-informed consent was obtained from all patients included in this study. All mouse experiments were approved by Institutional Animal Care and Use Committee at YCU (FA19–024). Detailed methods are described in Supplementary document.

Animal study

To evaluate if gene knockdown prolongs orthotopically tumor formation, EBV-positive-HKBML or -YML15 cells were infected with lentivirus containing human nonsilencing (NS), MYD88 (#2), PIN1(#1, #4), LMP1(#1), GLUT1(#1), HK2 (#1, Sigma Aldrich), and then 1 × 105 cells were orthotopically implanted into the right striatum of 6- to 9-week-old female SCID Beige mouse. To examine if intravascular tumor cell injection could form xenograft formation, 3 × 106 of PCNSL xenograft cells were injected from mouse tail vein. Autopsy was performed at 60 days after tail vein implantation. To test responsiveness of HD-MTX in PCNSL orthotopic xenograft model, 1 × 105 YML16 and YML15 cells were intracerebrally implanted. Seven days after implantation, mice were randomly assigned for subcutaneous treatment with MTX (400 mg/kg, every 1 week, twice) or vehicle treatment. Twelve hours after treatment, mice were treated with leucovorin (25 mg/kg) every 4 hours for 3 times as previously described (20). Mice were sacrificed when neurologic deficits or general conditions reached the criteria for euthanasia.

Statistical analysis

Statistical analysis was performed with JMP Pro15.0.0 software and GraphPad Prism (ver. 8.4.2). For parametric analysis, two-tailed t tests or one-way ANOVA were used. Two-tailed Fisher exact tests were used for analysis of frequencies of nominal data. To assess the association of genomic alterations in patients with xenograft tumors, Pearson's correlation coefficient was utilized. Survival analysis was performed using the Kaplan–Meier method using the log-rank test to compare the arms. Data were expressed as mean ± SEM. P < 0.05 was considered statistically significant.

Orthotopic PDXs of CNSL accurately recapitulated phenotypic features of primary tumors

To generate orthotopic xenografts, we obtained fresh surgical specimens from 14 patients with PCNSL and 1 patient with SCNSL (Supplementary Fig. S1A). Supplementary Table S1 summarizes the clinical characteristics of this cohort, representative of a typical PCNSL population. All PCNSL and SCNSL cells were confirmed to be viable prior to implantation. Following their implantation into immunocompromised mouse brains, development of neurological symptoms and xenograft formation were observed in 12 of 15 (80%) cases, including two EBV-positive PCNSL cases (Fig. 1A). The median survival time of mice with successfully implanted xenografts was 54.5 days (Supplementary Table S2). To assess the number of tumor cells required for successful orthotopic xenograft engraftment, we implanted varying numbers of PCNSL xenograft cells into mouse brains and found that 5 × 102 cells were sufficient to develop lethal xenografts. Median survival time was dependent on the number of cells that were injected (Fig. 1B; Supplementary Fig. S1B). In most cases, the histologic phenotype of xenografts was characterized by diffuse infiltration throughout both hemispheres; however, some xenografts showed more localization (Fig. 1A and C; Supplementary Fig. S1C). CD20-positive lymphoma cells diffusely infiltrated the brain in a time-dependent manner (Supplementary Fig. S1D). Serial transplantation induced second-generation PDXs expressing CD20 in 10 of 12 (83.3%) models (Supplementary Table S2). Orthotopic xenografts were also established from frozen cells in nine out of nine (100%) cases (Supplementary Table S1), indicating a potential for their broad use in future preclinical investigations. The time required to develop lethality in each CNSL PDX mouse was mostly constant, with only YML16 more rapidly developing tumor formation subsequent to a second passage in vivo (Supplementary Table S2). Hematoxylin and eosin staining of PDXs revealed histologic characteristics of DLBCL with positive expression of CD20 and CD79a in all patients with PCNSL and xenograft models. We noted that the histologic tumor architecture of the 12 xenograft-forming PCNSL patients preferentially exhibited either a sheet-like (SL) growth pattern (ten cases) or perivascular (PV) growth pattern (two cases) with 91.7% (11/12 or 12 of 12) consistency in the corresponding first-generation PDX models (passage1: sc1), and these features were completely recapitulated at multiple passages in vivo (e.g., passage2: sc2; Fig. 1C and D; Supplementary Fig. S1C and S1E; Supplementary Table S2). Cell proliferation based on Ki-67 index in xenografts was typically very high (40%–90%) and similar to that seen in corresponding primary tumors. Histologic subtyping based on expression of CD10, Bcl-6, and MUM1 (21) determined that among the xenograft-forming patient tumors, nine cases were of the ABC/non-GCB subtype, and three cases were of the GCB subtype. In all cases, the immunostaining pattern in the xenograft was consistent with the primary tumor (Fig. 1D; Supplementary Fig. S1E; Supplementary Table S3), and these subtypes were also retained at serial passages in vivo (Supplementary Fig. S1F). Two patients with PCNSL (YML1 and YML17) developed systemic lymphoma at recurrence. On the other hand, in all intracerebrally implanted PDX mice, no CD20+ neoplasm was identified in systemic organs except the CNS (Fig. 1E). To examine whether PCNSL cells had the potential to metastasize to other organs via the blood stream, 3 × 106 xenograft cells were injected into the tail vein of mice. After 60 days, only one YML16 (1 of 11, 9.1%) developed tumors in the brain but not in other organs (Fig. 1F and G; Supplementary Fig. S1G). The intracerebral xenograft establishment rate was significantly higher with orthotopic implantation method than that seen with intravascular injection method (P < 0.0001). Attempts to propagate long-term in vitro cultures from freshly dissociated xenografts were successful from two xenograft cells, YML5 and YML15, which were EBV-positive PCNSLs (Supplementary Fig. S1H), but long-term culture propagation was not successful in any cells derived from immunocompetent patients. Collectively, these results demonstrate the feasibility of establishing PDX models from immunocompetent and EBV-positive patients, which are broadly representative of the brain tropism and key phenotypic features of CNSL.

Figure. 1.

Establishment of 12 CNSL orthotopic PDX models. A, Overview of CD20 staining for all 12 generated orthotopic xenograft models. Bars, 500 μm. B, Kaplan–Meier curves showing survival differences of YML11 (top) and YML16 (bottom) cell–implanted mice at indicated cell numbers. C, Overview of H&E staining for CNSL xenografted brains at indicated passage numbers in vivo. Insets, CD20 IHC. Bars, 50 μm. D, Hematoxylin and eosin (H&E) staining and IHC for indicated markers in patient tumor cells (top) and corresponding PDX cells (bottom). Bars, 50 μm. E–G, CD20 IHC of indicated tissue in YML16 orthotopically implanted mice (E), mouse brain 60 days after intravascular injection (F), and indicated tissue in YML16 xenografted mice after intravascular injection (G).

Figure. 1.

Establishment of 12 CNSL orthotopic PDX models. A, Overview of CD20 staining for all 12 generated orthotopic xenograft models. Bars, 500 μm. B, Kaplan–Meier curves showing survival differences of YML11 (top) and YML16 (bottom) cell–implanted mice at indicated cell numbers. C, Overview of H&E staining for CNSL xenografted brains at indicated passage numbers in vivo. Insets, CD20 IHC. Bars, 50 μm. D, Hematoxylin and eosin (H&E) staining and IHC for indicated markers in patient tumor cells (top) and corresponding PDX cells (bottom). Bars, 50 μm. E–G, CD20 IHC of indicated tissue in YML16 orthotopically implanted mice (E), mouse brain 60 days after intravascular injection (F), and indicated tissue in YML16 xenografted mice after intravascular injection (G).

Close modal

Subclonal heterogeneity is aberrant somatic hypermutation independent and recurrent MYD88 and CD79B mutations are constitutively retained in immunocompetent PCNSL xenograft models

To evaluate genome-wide genetic differences between immunocompetent primary tumors and their corresponding orthotopic xenografts, whole-exome sequencing (WES) was performed for eight matched PCNSL samples from patients and with paired peripheral blood mononuclear cells (Supplementary Table S4). WES revealed nonsynonymous mutations (ns-mts) in each patient sample to be 158.9 ± 13.9 (mean ± SEM, Fig. 2A; Supplementary Fig. S2A). We found that 93.3% ± 1.3% and 91.2% ± 1.0% (mean ± SEM) of those ns-mts were shared in-common within the patient and xenograft tumors groups, respectively. Profiles of single-nucleotide variants (SNV) within the 96-trinucleotides relevant to the patient tumors, which were enriched with C>T transitions (50.6% ± 2.2%, mean ± SEM), were similar to the profiles obtained for their corresponding xenografts (49.3% ± 1.6%, P = 0.62; Supplementary Fig. S2B). We found that variant allele frequencies (VAF) of the shared ns-mts were significantly higher than those of the private ns-mts in 6 of 8 patients (75%) and 8 of 8 (100%) xenograft cases, respectively (Supplementary Fig. S2C). We also assessed WES to determine driver genes in PCNSL. WES detected 30 recurrent representative genes (Fig. 2B), and MutSigCV analysis and/or MutPanning analysis identified 3 significantly mutated genes (MYD88, CD79B, and HIST1H1E) as driver candidates (Supplementary Tables S5 and S6). We further performed targeted multiplex PCR-based next-generation sequencing in 10 patients and xenograft tumors (Supplementary Fig. S2D and Supplementary Table S7). MYD88L265P and CD79B SNVs were found in 8 of 10 (80%) and 6 out of 10 (60%) patients and tumors, respectively; and all CD79B-mutant tumors co-occurred with MYD88L265P (Fig. 2B; Supplementary Table S8; Supplementary Fig. S2D). In CD79B-mutant PCNSL, four of six (66.7%) SNVs were located at Y196. Interestingly, MYD88 and CD79B SNVs were serially retained in the respective xenografts, even after serial transplantation up to the 13th generation (Fig. 2C and D; Supplementary Fig. S2E and S2F; Supplementary Table S2), suggesting that these mutations might play a functional role in PCNSL xenograft growth and maintenance. According to a recent study, MYD88L265P can be detected from cell-free DNA in the serum and CSF of patients with PCNSL (14, 22). To examine whether MYD88L265P levels could be detectable from in cell-free DNA xenograft models, we extracted serum cell–free DNA from xenografted mice. Droplet digital PCR was able to detect MYD88L265P in serum cell–free DNA from PCNSL xenografts with mutated MYD88 (six of six samples, 100%), whereas no MYD88 SNV was identified in the serum of a MYD88 wild-type YMG30 glioblastoma xenograft used as a control (Fig. 2E; Supplementary Fig. S2G; Supplementary Table S9). CARD11 ns-mts were found in four out of ten (40%) patient tumors. Among them, the private SNV CARD11 was identified in YML3 xenografts and YML11 patient tumors, and these observations were also confirmed by data from the BAM files (Fig. 2B; Supplementary Table S8; Supplementary Fig. S2D and S2H).

Figure. 2.

Genomic landscape and subclonal diversity of immunocompetent PCNSL in matched pairs of patients and xenografts. A, Venn diagram indicating shared and private nonsynonymous mutations (ns-mts) between patients (pt) and xenografts (sc) at indicated in vivo passage number. B, Landscape of ns-mts in PCNSL patient and corresponding xenograft cells. C and D, Sanger sequencing and pyrosequencing, indicating MYD88 (L265P, arrow; C) and CD79BY196 single nucleotide variants (arrow; D) in serially passaged xenografts. E, Droplet digital PCR identifying MYD88L265P SNVs in xenograft cell-free DNA harvested from orthotopic xenografts. F, Log R ratio for assessing CNA in patients (pt) and corresponding xenograft tumors (sc) at indicated in vivo passage. G, Phylogenetic trees in patient and xenograft tumors at the indicated in vivo passage. Bars, 20 ns-mts. H, Phylogeny tracing of PCNSL patient tumor and xenograft cells. Phylogenetic reconstruction of CNAs and ns-mts. Each color represents a distinct clone within the tumor. I, Number of aSHM targets (red) and other target ns-mts (blue) in truncal clone and subclone.

Figure. 2.

Genomic landscape and subclonal diversity of immunocompetent PCNSL in matched pairs of patients and xenografts. A, Venn diagram indicating shared and private nonsynonymous mutations (ns-mts) between patients (pt) and xenografts (sc) at indicated in vivo passage number. B, Landscape of ns-mts in PCNSL patient and corresponding xenograft cells. C and D, Sanger sequencing and pyrosequencing, indicating MYD88 (L265P, arrow; C) and CD79BY196 single nucleotide variants (arrow; D) in serially passaged xenografts. E, Droplet digital PCR identifying MYD88L265P SNVs in xenograft cell-free DNA harvested from orthotopic xenografts. F, Log R ratio for assessing CNA in patients (pt) and corresponding xenograft tumors (sc) at indicated in vivo passage. G, Phylogenetic trees in patient and xenograft tumors at the indicated in vivo passage. Bars, 20 ns-mts. H, Phylogeny tracing of PCNSL patient tumor and xenograft cells. Phylogenetic reconstruction of CNAs and ns-mts. Each color represents a distinct clone within the tumor. I, Number of aSHM targets (red) and other target ns-mts (blue) in truncal clone and subclone.

Close modal

We also assessed copy-number alterations (CNA) of immunocompetent PCNSL patients and xenograft samples and found a positive correlation between the two (r = 0.91 ± 0.01, mean ± SEM; Fig. 2F; Supplementary Fig. S2I). To assess previously reported CNAs in PCNSL (23–25), we performed WES and multiplex ligation-dependent probe amplification assay. We found that 5 of 8 (62.5%) patients harbored 6q21–23 deletion, including the NF-κB pathway inhibiting TNFAIP3 (6q23.3) and 7 of 10 patients with PCNSL (70%) harbored 9p21.3 deletion (CDKN2A), whereas 9p24.1 amplification (PD-L1) and 12q amplification were absent. These CNAs were completely retained in the xenografts (Supplementary Fig. S2J; Supplementary Tables S10 and S11). We also assessed the evolution of subclonal heterogeneity in xenografts. Phylogenetic tree analysis revealed that MYD88L265P and CD79BY196 mutations arise as early genomic alterations (Fig. 2G). On the other hand, although truncal clones and subclones identified in individual cases were minimally divergent in terms of somatic mutations, clonal compositions changed dynamically during the establishment and passage of xenografts (Fig. 2H). Aberrant somatic hypermutation (aSHM) can target genes that, when mutated, are capable of causing genome instability in systemic DLBCL as well as PCNSL (4, 5, 26), and are therefore a potential source of oncogenic mutations in PCNSL and genetically heterogeneous clones. We therefore assessed the effect of aSHM in truncal clones and subclones. Using 76 aSHM target genes, which we had previously identified from patients with PCNSL (5), we found that aSHM target gene alterations were not significantly increased in subclones (Fig. 2I). We further focused on PIM1, BTG2, and HIST1H1E, as they are most affected by aSHM and are commonly observed to be altered in PCNSL (4, 5). In our cohort, only 2.3% (±1.1% mean ± SEM), 0%, and 0% of additional mutations in the PIM1, BTG2, and HIST1H1E genes, respectively, which included percentages below the cutoff threshold, were identified at first or second passage in vivo (Supplementary Fig. S2K–S2M). These results indicate that aSHM does not occur robustly during orthotopic xenograft formation. Thus, we conclude that the heterogeneous genomic evolution in PCNSL identified by WES was probably not induced by ongoing aSHM.

Excessive glycolysis in CNSL is recapitulated in xenograft cells

PET imaging of the brain reveals that PCNSL is associated with high 18F-fluorodeoxyglucose (FDG) uptake (27, 28), which indicates increased glucose transport and glycolysis. We assessed FDG uptake in patients with PCNSL, who were imaged at Yokohama City University Hospital. In line with prior reports, FDG-standardized uptake values (SUV) were significantly higher in lymphoma lesions (SUVmax = 14.5 ± 0.8) when compared with those seen in the contralateral white matter (SUVmax = 3.5 ± 0.1, P < 0.0001, Fig. 3A). FDG-PET also demonstrated high FDG uptake in tumors of patients with CNSL that gave rise to xenograft models (mean SUVmax = 17.8 ± 2.1; Fig. 3B; Supplementary Fig. S3A; Supplementary Table S1). We found that glucose transporter 1 (Glut-1) and HK2 expression was highly correlated with FDG uptake. Lactate dehydrogenase A levels were also elevated in tumors of patients with CNSL when compared with those seen in the normal brain cortex (Fig. 3C; Supplementary Fig. S3B). We also tested if animal FDG-PET could detect high FDG uptake in PDXs. As expected, FDG-PET and MR imaging demonstrated a higher FDG uptake in YML12 PDXs (SUVmax = 17.3) when compared with the contralateral normal brain (P < 0.0001, Fig. 3D and E; Supplementary Fig. S3C). Significant increase in FDG uptake was observed in multiple PDXs when compared with the sham-treated mouse brain (Fig. 3F). Immunohistochemistry demonstrated that the expression of proteins in the glycolysis pathway was higher in CNSL xenografts than that seen in control brain tissues (Fig. 3G; Supplementary Fig. S3D), likely indicative of excessive glycolysis in CNSL xenografts. To evaluate selective cytotoxicity by inhibiting glycolysis, WZB117 (GLUT1 inhibitor) and HK2 inhibitor (2-deoxyglucose, 2-DG) were tested. CNSL xenograft cells were significantly more sensitive to these compounds than normal human astrocytes (Fig. 3H; Supplementary Fig. S3E). In addition, GLUT1 and HEXOKINASE 2 (HK2) knockdown suppressed cell viability and prolonged overall survival in PCNSL orthotopic xenografts (Fig. 3I,M; Supplementary Fig. S3F–S3J).

Figure. 3.

Excessive glycolysis plays a crucial role in PCNSL. A, SUV max of FDG in PCNSL patient tumors and contralateral normal brain (NB). B, FDG-PET images of the indicated patients with PCNSL. C, IHC for Glut-1, HK2, and lactate dehydrogenase A (LDHA) in patients with PCNSL tumors and normal brain. D, Coronal view of a mouse brain with the YML12 xenograft. Hematoxylin and eosin staining, T2-weighted MRI, FDG-PET image, and merged image of MRI and FDG-PET. E, SUV max and SUV mean of FDG-PET images in YML12 xenograft and sham brain. *, P < 0.05 between tumor and sham brain. F, FDG-PET images for YML9, YML15, YML16 xenografts, and sham brain. SUV max and SUV mean of the xenograft brains and sham brain. *, P < 0.05 between tumor and sham brain. G, IHC for Glut-1, HK2, and LDHA expression in the indicated PCNSL xenograft tumors and sham brain. H, Relative cell viability of WZB117 and 2-DG–treated PCNSL xenograft cells and immortalized normal human astrocyte cells. *, P < 0.05 for difference between DMSO versus WZB117 and 2-DG. I, Western blotting for GLUT1 expression in nonsilencing (NS) and GLUT1 shRNA-transduced PCNSL xenograft cells. J, Relative cell viability of NS and GLUT1 shRNA-transduced PCNSL xenograft cells at the indicated time points. *, P < 0.05 between NS and GLUT1 shRNA-transduced xenograft cells. K, Western blot of HK2 expression in NS and HK2 shRNA-transduced PCNSL xenograft cells. L, Relative cell viability of NS and HK2 shRNA-transduced PCNSL xenograft cells at indicated time points. *, P < 0.05 between NS and HK2 shRNA-transduced xenograft cells. M, Kaplan–Meier curves indicating the survival difference between NS and GLUT1 shRNA (#1; top) and HK2 shRNA (#1; bottom)-transduced YML15 cells. Data represent the mean ± SEM. Bars, 50 μm.

Figure. 3.

Excessive glycolysis plays a crucial role in PCNSL. A, SUV max of FDG in PCNSL patient tumors and contralateral normal brain (NB). B, FDG-PET images of the indicated patients with PCNSL. C, IHC for Glut-1, HK2, and lactate dehydrogenase A (LDHA) in patients with PCNSL tumors and normal brain. D, Coronal view of a mouse brain with the YML12 xenograft. Hematoxylin and eosin staining, T2-weighted MRI, FDG-PET image, and merged image of MRI and FDG-PET. E, SUV max and SUV mean of FDG-PET images in YML12 xenograft and sham brain. *, P < 0.05 between tumor and sham brain. F, FDG-PET images for YML9, YML15, YML16 xenografts, and sham brain. SUV max and SUV mean of the xenograft brains and sham brain. *, P < 0.05 between tumor and sham brain. G, IHC for Glut-1, HK2, and LDHA expression in the indicated PCNSL xenograft tumors and sham brain. H, Relative cell viability of WZB117 and 2-DG–treated PCNSL xenograft cells and immortalized normal human astrocyte cells. *, P < 0.05 for difference between DMSO versus WZB117 and 2-DG. I, Western blotting for GLUT1 expression in nonsilencing (NS) and GLUT1 shRNA-transduced PCNSL xenograft cells. J, Relative cell viability of NS and GLUT1 shRNA-transduced PCNSL xenograft cells at the indicated time points. *, P < 0.05 between NS and GLUT1 shRNA-transduced xenograft cells. K, Western blot of HK2 expression in NS and HK2 shRNA-transduced PCNSL xenograft cells. L, Relative cell viability of NS and HK2 shRNA-transduced PCNSL xenograft cells at indicated time points. *, P < 0.05 between NS and HK2 shRNA-transduced xenograft cells. M, Kaplan–Meier curves indicating the survival difference between NS and GLUT1 shRNA (#1; top) and HK2 shRNA (#1; bottom)-transduced YML15 cells. Data represent the mean ± SEM. Bars, 50 μm.

Close modal

Intracerebral PCNSL progression is mediated by genetic alterations in the NF-κB canonical pathway

Previous studies have demonstrated that mutations in MYD88 and CD79B activate the NF-κB canonical pathway in systemic DLBCL (10, 11). Therefore, we hypothesized that these genomic alterations could signal via the NF-κB canonical pathway to induce glycolysis in PCNSL. As expected, Western blot analysis demonstrated that the expression of proteins in glycolysis and the NF-κB canonical pathway was higher in xenografts with mutated MYD88 and CD79B than that observed in normal brain tissues (Fig. 4A; Supplementary Fig. S4A). To test if inhibition of NF-κB canonical pathway was associated with concomitant inhibition of glycolysis in PCNSL, we introduced knockdowns of MYD88 and CD79B in MYD88L265P and CD79B mutant cells and HKBML cells (MYD88 and CD79B wild-type). We observed a decrease in the expression levels of phospho-p65, p65, and glycolytic proteins (particularly HK2) in these cells (Fig. 4B,D; Supplementary Fig. S4B–S4D). We also found that knocking down MYD88 and CD79B inhibited cell growth in vitro, suggesting that these mutant genes are strong drivers of tumor progression (Fig. 4E; Supplementary Fig. S4E and S4F). In contrast, knockdown of MYD88 and CD79B reversed 2-DG sensitivity in vitro (Fig. 4F). We also confirmed that MYD88 knockdown in HKBML PCNSL cells extended overall survival of xenografted mice (Fig. 4G). IHC demonstrated a decrease in the expression levels of phospho-p65, p65, and HK2 in MYD88 knockdown cells when compared with those seen in control cells (Fig. 4H). NF-κB inhibitor also (BAY11–7082) decreased the expression of phospho-p65 and HK2, and suppressed cell viability in PCNSL xenograft cells when compared with control cells (Fig. 4I and J; Supplementary Fig. S4G and S4H). To determine whether changes in RelA/p65 expression directly affect glycolysis, we created RELA knockdown cells. We observed that knockdown of RELA suppressed RelA/p65 and HK2, and cell growth in vitro (Fig. 4K,M). Taken together, these findings indicate that NF-κB RelA/p65-HK2 axis plays an important mechanistic role in tumor growth during immunocompetent PCNSL.

Figure. 4.

NF-κB canonical pathway gene alterations promote xenograft formation in immunocompetent PCNSL. A, Western blotting showing indicated protein levels in the PCNSL xenograft tumors and normal brain (NB). B and C, Western blot analysis of indicated protein expression in nonsilencing (NS) and MYD88 (B) and CD79B (C) shRNA-transduced PCNSL xenograft cells. D, Enzyme-linked immunosorbent assay indicating semiquantitative p65 value in NS, MYD88, and CD79B shRNA-transduced cells. *, P < 0.05 between NS versus MYD88 and CD79B. E, Relative cell viability of NS, MYD88, and CD79B shRNA-transduced xenograft cells. P < 0.05 for difference between NS versus MYD88 and CD79B. F, Relative cell viability of NS, MYD88, and CD79B shRNA-transduced YML16 and HKBML cells following 2-DG treatment. * and **, P <0.05 between NS and MYD88 (*) or CD79B (**). G, Kaplan–Meier curves indicating survival difference between NS and MYD88 shRNA-transduced HKBML cells. H, IHC of mouse brain implanted with NS and MYD88 shRNA-transduced HKBML cells at day 21. Bars, 50 μm. I, Western blotting showing indicated proteins in PCNSL xenograft cells after 12 hours of DMSO and 10 μmol/L of BAY11–7082 treatment. J, Relative cell viability of BAY11–7082–treated xenograft cells and normal human astrocytes at day 3. *, P < 0.05 for difference between DMSO versus BAY11–7082. K, Western blot analysis of indicated proteins in NS and RELA shRNA-transduced PCNSL cells. L, ELISA for semiquantitative p65 value in NS, RELA shRNA-transduced cells. *, P <0.05 between NS and RELA. M, Relative cell viability of NS and RELA shRNA-transduced PCNSL cells at the indicated time points. * and **, P < 0.05 between NS and RELA#1, RELA#2, or RELA#3. Data represent the mean ± SEM.

Figure. 4.

NF-κB canonical pathway gene alterations promote xenograft formation in immunocompetent PCNSL. A, Western blotting showing indicated protein levels in the PCNSL xenograft tumors and normal brain (NB). B and C, Western blot analysis of indicated protein expression in nonsilencing (NS) and MYD88 (B) and CD79B (C) shRNA-transduced PCNSL xenograft cells. D, Enzyme-linked immunosorbent assay indicating semiquantitative p65 value in NS, MYD88, and CD79B shRNA-transduced cells. *, P < 0.05 between NS versus MYD88 and CD79B. E, Relative cell viability of NS, MYD88, and CD79B shRNA-transduced xenograft cells. P < 0.05 for difference between NS versus MYD88 and CD79B. F, Relative cell viability of NS, MYD88, and CD79B shRNA-transduced YML16 and HKBML cells following 2-DG treatment. * and **, P <0.05 between NS and MYD88 (*) or CD79B (**). G, Kaplan–Meier curves indicating survival difference between NS and MYD88 shRNA-transduced HKBML cells. H, IHC of mouse brain implanted with NS and MYD88 shRNA-transduced HKBML cells at day 21. Bars, 50 μm. I, Western blotting showing indicated proteins in PCNSL xenograft cells after 12 hours of DMSO and 10 μmol/L of BAY11–7082 treatment. J, Relative cell viability of BAY11–7082–treated xenograft cells and normal human astrocytes at day 3. *, P < 0.05 for difference between DMSO versus BAY11–7082. K, Western blot analysis of indicated proteins in NS and RELA shRNA-transduced PCNSL cells. L, ELISA for semiquantitative p65 value in NS, RELA shRNA-transduced cells. *, P <0.05 between NS and RELA. M, Relative cell viability of NS and RELA shRNA-transduced PCNSL cells at the indicated time points. * and **, P < 0.05 between NS and RELA#1, RELA#2, or RELA#3. Data represent the mean ± SEM.

Close modal

Pin1 activation induces RelA/p65 stability and enhances tumor progression in immunocompetent PCNSL

To assess whether in vitro culturing would diminish xenograft formation, we implanted cultured cells at two different stages: at day zero and day four. We did not see any difference in survival rates in these cells and were also able to confirm the retention of phospho-p65 at day 4 (Supplementary Fig. S5A–S5C), indicating that xenografts grown in vitro (for a short period) did not show any formation defects. In contrast, immunocompetent YML7 and YML13 (MYD88 and CD79B mutant) confirmed cell viability, but did not generate PDXs (Fig. 5A; Supplementary Fig. S5D and S5E; Supplementary Table S7). Therefore, we propose that additional mechanisms that enhance RelA/p65 expression may be crucial for orthotopic xenograft formation. Because the p65 stability is regulated by peptidyl-prolyl isomerase Pin1, which enhances oncogenic activity (29, 30), we assessed whether Pin1 expression levels are different between xenograft-forming and nonforming patient cells. The expression levels of Pin1 were similar to those of phospho-p65 and p65 in cells of patients with PCNSL. Pin1 was highly expressed in xenograft-forming tumors (Fig. 5B and C). We also found that these expressions levels were relatively high in xenograft cells when compared with sham-treated mouse brains (Fig. 5D; Supplementary Fig. S5F). Pin1 is also known to promote the Warburg effect (31). Interestingly, YML16 xenografts, which exhibited slow growth in their initial growth period (sc1), became aggressive upon serial in vivo passages (Fig. 5E; Supplementary Table S2). Along with increased aggressiveness, expression levels of Pin1, phospho-p65, and p65 as well as HK2 protein were also elevated in YML16sc2 (passage 2) cells when compared with YML16sc1 cells (Fig. 5F and G). Cell viability assay demonstrated that YML16sc2 cells were more sensitive to 2-DG than YML16sc1 cells (Fig. 5H). PIN1 knockdown decreased the expressions levels of phopsho-p65, p65, and glycolytic proteins, inhibited cell growth in vitro, and delayed orthotopic xenograft formation in vivo (Fig. 5I,L; Supplementary Fig. S5G). Similarly, Pin1 inhibitor Juglone decreased the expression levels of Pin1 and phospho-p65, and suppressed cell viability in PCNSL xenograft cells (Supplementary Fig. S5H and S5I). These findings suggest that genetic alterations in the NF-κB pathway such as mutated MYD88 and CD79B genes and Pin1 co-operatively enhance the RelA/p65-HK2 axis, which promotes tumor progression in PCNSL.

Figure. 5.

Pin1 complementary activates the NF-κB canonical pathway and promotes tumor progression in immunocompetent PCNSL. A,MYD88, CD79B, and CARD11 SNVs (blue) in xenograft-forming and nonxenograft-forming PCNSL patient tumors. B, Western blotting for protein expression of Pin1 and the NF-κB canonical signaling pathway in xenograft-forming (YML1R, YML11, and YML12) and nonforming PCNSL tumors (YML13). C, IHC for indicated proteins in immunocompetent xenograft-forming PCNSL patient tumors and xenograft nonforming tumors. D, IHC for indicated proteins in immunocompetent PCNSL xenograft tumors and sham-treated mice brain. E, Hematoxylin and eosin staining of YML16 indicated passage xenograft tumors. F and G, Western blotting (F) and IHC (G) of Pin1, NF-κB canonical pathway, and glycolytic proteins in first-generation (sc1) and second-generation (sc2) YML16 xenograft tumors. H, Relative cell viability of 2-DG–treated YML16sc1 and YML16sc2 cells at day 3. *, P < 0.05 for difference between YML16sc1 versus YML16sc2 cells. I, Western blotting showing Pin1, NF-κB canonical pathway, and glycolytic protein expressions in nonsilencing (NS) and PIN1 shRNA-transduced PCNSL cells. J, Enzyme-linked immunosorbent assay for semiquantitative p65 value in NS and PIN1 shRNA-transduced cells. *, P < 0.05 between NS and PIN1 shRNA-transduced xenograft cells. K, Relative cell viability of NS and PIN1#1 shRNA-transduced PCNSL cells at the indicated time points. * and **, P < 0.05 between NS and PIN1#1 (*), or PIN1#2 (**). L, Kaplan–Meier curves indicating survival difference between NS and PIN1 shRNA (#1)-transduced HKBML cells. Data represent the mean ± SEM.

Figure. 5.

Pin1 complementary activates the NF-κB canonical pathway and promotes tumor progression in immunocompetent PCNSL. A,MYD88, CD79B, and CARD11 SNVs (blue) in xenograft-forming and nonxenograft-forming PCNSL patient tumors. B, Western blotting for protein expression of Pin1 and the NF-κB canonical signaling pathway in xenograft-forming (YML1R, YML11, and YML12) and nonforming PCNSL tumors (YML13). C, IHC for indicated proteins in immunocompetent xenograft-forming PCNSL patient tumors and xenograft nonforming tumors. D, IHC for indicated proteins in immunocompetent PCNSL xenograft tumors and sham-treated mice brain. E, Hematoxylin and eosin staining of YML16 indicated passage xenograft tumors. F and G, Western blotting (F) and IHC (G) of Pin1, NF-κB canonical pathway, and glycolytic proteins in first-generation (sc1) and second-generation (sc2) YML16 xenograft tumors. H, Relative cell viability of 2-DG–treated YML16sc1 and YML16sc2 cells at day 3. *, P < 0.05 for difference between YML16sc1 versus YML16sc2 cells. I, Western blotting showing Pin1, NF-κB canonical pathway, and glycolytic protein expressions in nonsilencing (NS) and PIN1 shRNA-transduced PCNSL cells. J, Enzyme-linked immunosorbent assay for semiquantitative p65 value in NS and PIN1 shRNA-transduced cells. *, P < 0.05 between NS and PIN1 shRNA-transduced xenograft cells. K, Relative cell viability of NS and PIN1#1 shRNA-transduced PCNSL cells at the indicated time points. * and **, P < 0.05 between NS and PIN1#1 (*), or PIN1#2 (**). L, Kaplan–Meier curves indicating survival difference between NS and PIN1 shRNA (#1)-transduced HKBML cells. Data represent the mean ± SEM.

Close modal

LMP1 and Pin1 promote formation of EBV-positive PCNSL xenografts

In EBV-positive YML15, CNAs were strongly correlated with the corresponding xenografts (r = 0.966, Supplementary Fig. S6A). In addition, the number of ns-mts (40 ns-mts) was relatively lower, and MYD88/CD79B and aSHM target gene alterations were not identified in YML15 (Supplementary Tables S4 and S8), implying that they have a distinct genomic profile that is not similar to that of immunocompetent PCNSL. Among them, 80% of ns-mts in the patient tumor were also found in their corresponding xenografts, and shared VAFs were significantly higher than the VAFs of private ns-mts (Supplementary Fig. S6B). Similar to immunocompetent PCNSL, phylogenetic tree analysis of YML15 also revealed subclonal heterogeneity (Supplementary Fig. S6C). It has been reported that latency membrane protein-1 (LMP1) is a major EBV-encoded oncogene that activates the NF-κB pathway in EBV-positive lymphoma (32). Consistent with this, LMP1 was expressed in our EBV-positive, MYD88/CD79B wild-type patient PCNSL cells (YML5, YML15, and YML14), but was not detected in EBV-negative YML16 cells (Fig. 6A). The expression levels of phospho-p65 and Pin1 were relatively higher in xenograft-forming YML15 and YML5 tumors than in xenograft nonforming YML14 tumors, findings similar to those seen in immunocompetent PCNSL tumors (Fig. 6A and B; Supplementary Fig. S6D). Phospho-p65, Pin1, and glycolytic proteins were highly expressed in EBV-positive PCNSL xenografts (Figs. 3G, 6C and D; Supplementary Fig. S3D), recapitulating the RelA/p65-HK2 axis activation seen in immunocompetent PCNSL xenografts. We additionally observed that knockdown of LMP1 decreased protein expression in the NF-κB canonical pathway and glycolysis, suppressed cell viability, and extensively prolonged overall survival of mice orthotopically xenografted with EBV-positive PCNSL cells (Fig. 6E,H). Furthermore, BAY11–7082 suppressed protein levels in the NF-κB canonical pathway and glycolytic pathways and cell viability (Fig. 6I; Supplementary Fig. S6E). And, PIN1 knockdown and Pin1 inhibitor (Juglone) decreased cell viability and expression levels of phospho-p65 and glycolytic proteins, (Fig. 6J and K; Supplementary Fig. S6F and S6G). PIN1 knockdown also extended overall survival in vivo (Fig. 6L). We observed a similar decrease in cell viability and in expression levels of phospho-p65 and HK2 in RELA knockdown PCNSL cells (Fig. 6M,O). Collectively, these findings suggest that activation of the RelA/p65-HK2 signaling axis, codriven by LMP1 and Pin1, is critical for tumor progression in EBV-positive PCNSL, identifying a mechanistic convergence in these different etiologies for PCNSL.

Figure. 6.

NF-κB canonical pathway deregulation promotes xenograft formation in EBV-positive PCNSL. A, Western blotting of indicated proteins in cells of patients with PCNSL. B, IHC for indicated proteins in patient tumors and the normal brain (NB). C, Western blotting showing indicated protein levels in the EBV-positive PCNSL xenograft tumors and NB. D, IHC of indicated proteins in EBV-positive PCNSL xenograft and sham-treated brain. E, Western blotting of indicated proteins in nonsilencing (NS) and LMP1 shRNA-transduced PCNSL xenograft cells. F, Enzyme-linked immunosorbent assay for semiquantitative p65 value in NS and LMP1 shRNA-transduced cells. *, P < 0.05 between NS and LMP1. G, Relative cell viability of NS and LMP1 shRNA-transduced cells. * and **, P < 0.05 between NS versus LMP1#1 (*) and LMP1#2 (**). H, Kaplan–Meier curves showing survival difference between NS and LMP1 shRNA-transduced YML15 cells. I and J, Western blotting for indicated protein expression in PCNSL xenograft cells treated with DMSO and 10 μmol/L of BAY11–7082 for 12 hours (I), and NS and PIN1 shRNA-transduced cells (J). K, Relative cell viability of NS and PIN1 shRNA-transduced cells. Asterisks, P < 0.05 between NS and PIN1. L, Kaplan–Meier curves showing survival difference between NS and PIN1 shRNA-transduced YML15 cells. M, Western blotting for protein expression in NS and RELA shRNA-transduced YML15 cells. N, Enzyme-linked immunosorbent assay for semiquantitative p65 value in NS and RELA shRNA-transduced cells. *, P < 0.05 between NS and RELA. O, Relative cell viability of NS and RELA shRNA-transduced cells at the indicated time points. Asterisks, P < 0.05 between NS and RELA. Data represent the mean ± SEM.

Figure. 6.

NF-κB canonical pathway deregulation promotes xenograft formation in EBV-positive PCNSL. A, Western blotting of indicated proteins in cells of patients with PCNSL. B, IHC for indicated proteins in patient tumors and the normal brain (NB). C, Western blotting showing indicated protein levels in the EBV-positive PCNSL xenograft tumors and NB. D, IHC of indicated proteins in EBV-positive PCNSL xenograft and sham-treated brain. E, Western blotting of indicated proteins in nonsilencing (NS) and LMP1 shRNA-transduced PCNSL xenograft cells. F, Enzyme-linked immunosorbent assay for semiquantitative p65 value in NS and LMP1 shRNA-transduced cells. *, P < 0.05 between NS and LMP1. G, Relative cell viability of NS and LMP1 shRNA-transduced cells. * and **, P < 0.05 between NS versus LMP1#1 (*) and LMP1#2 (**). H, Kaplan–Meier curves showing survival difference between NS and LMP1 shRNA-transduced YML15 cells. I and J, Western blotting for indicated protein expression in PCNSL xenograft cells treated with DMSO and 10 μmol/L of BAY11–7082 for 12 hours (I), and NS and PIN1 shRNA-transduced cells (J). K, Relative cell viability of NS and PIN1 shRNA-transduced cells. Asterisks, P < 0.05 between NS and PIN1. L, Kaplan–Meier curves showing survival difference between NS and PIN1 shRNA-transduced YML15 cells. M, Western blotting for protein expression in NS and RELA shRNA-transduced YML15 cells. N, Enzyme-linked immunosorbent assay for semiquantitative p65 value in NS and RELA shRNA-transduced cells. *, P < 0.05 between NS and RELA. O, Relative cell viability of NS and RELA shRNA-transduced cells at the indicated time points. Asterisks, P < 0.05 between NS and RELA. Data represent the mean ± SEM.

Close modal

RelA/p65-HK2 axis deactivation is correlated with response to chemotherapy in CNSL xenograft models

We examined whether our patient-derived cell models phenocopied the primary tumor response to the clinically effective chemotherapeutic MTX. After 2 cycles of HD-MTX monotherapy, 7 of 11 patients responded radiographically to the therapy (complete response + partial response; Fig. 7A; Supplementary Fig. S7A). The response of cultured PCNSL xenograft cells to MTX was positively correlated with the radiographic response observed in the corresponding patients (P = 0.02, Fig. 7B and C; Supplementary Fig. S7B; Supplementary Table S12). As our CNSL xenograft cells grown in vitro mirrored the clinical response of patients to chemotherapy, their therapeutic testing in vitro has the potential for predictive value. Although HD-MTX is a standard chemotherapy regimen, biomarkers of sensitivity to HD-MTX remain unclear (33). As MTX suppresses NF-κB canonical pathway activation in human T lymphocytes (34), we hypothesized that HD-MTX may inactivate the RelA/p65-HK2 axis in MTX-sensitive PCNSL cells. Indeed, we found that MTX suppressed the expression of phospho-p65 and HK2 in MTX-sensitive cells, but not in MTX-resistant cells, in a time- and dose-dependent manner (Fig. 7D and E). Decrease in cell viability and expression of phospho-p65 were restored by exogenous leucovorin, indicating that MTX-induced inactivation of phospho-p65 is a readout of on-target drug effect (Fig. 7F and G). Consistent with clinical responsiveness, HD-MTX treatment prolonged overall survival in YML16, but not in YML15. Pharmacodynamic analysis revealed that HD-MTX treatment decreased expression levels of phospho-p65 and HK2 and increased rate of terminal deoxynucleotidyl transferase–mediated dUTP nick end labeling-positive cells when compared with vehicle treatment in the YML16 model. This effect was not seen in the YML15 model, in vivo (Fig. 7H,K; Supplementary Fig. S7C and S7D), suggesting inhibition of the NF-κB canonical pathway and glycolysis could serve as a biomarker for sensitivity to HD-MTX. Taken together, our results indicate that responsiveness to HD-MTX is tightly correlated with the inhibition of the NF-κB canonical pathway and excessive glycolysis in CNSL.

Figure. 7.

Chemotherapeutic vulnerability is mediated by NF-κB canonical pathway deactivation in CNSL. A, Contrast-enhanced MR images of patients with PCNSL, prior (left) and post (after 2 cycles) HD-MTX treatment (right). B, Relative cell viability of CNSL xenograft cells 3 days after HD-MTX treatment. DMSO, control. *, P <0.05 for the difference between DMSO and MTX. C, Summary of sensitivity for HD-MTX treatment in PCNSL patients and xenografts. D and E, Western blotting for protein expression in the NF-κB canonical pathway and glycolysis in HD-MTX–sensitive cells (YML3, YML11, YML12, YML16, and HKBML) and -resistant cells (YML2, YML5, and YML15) in a dose- (D) and time-dependent manner (E). F, Relative cell viability of PCNSL xenograft cells treated with DMSO, 100 μmol/L MTX, 100 μmol/L of leucovorin (LV), and a combination of MTX and leucovorin on day 2. *, P < 0.05. G, Western blotting of phospho-p65 expression in PCNSL xenograft cells at the indicated treatment for 12 hours. H, Kaplan–Meier curves showing survival difference of YML16 orthotopically implanted mice treated with vehicle and HD-MTX. I, IHC for indicated protein expression in vehicle and HD-MTX–treated mouse brain in YML16 xenografts. J, Kaplan–Meier curves showing differences in survival rates of YML15 orthotopically implanted mice treated with vehicle control and HD-MTX. K, IHC for indicated protein expression in vehicle and HD-MTX–treated mice brain in YML15 xenografts. L, Schematic illustration of NF-κB canonical pathway activation to promote tumor progression in PCNSL. Data represent the mean ± SEM. Bars, 50 μm.

Figure. 7.

Chemotherapeutic vulnerability is mediated by NF-κB canonical pathway deactivation in CNSL. A, Contrast-enhanced MR images of patients with PCNSL, prior (left) and post (after 2 cycles) HD-MTX treatment (right). B, Relative cell viability of CNSL xenograft cells 3 days after HD-MTX treatment. DMSO, control. *, P <0.05 for the difference between DMSO and MTX. C, Summary of sensitivity for HD-MTX treatment in PCNSL patients and xenografts. D and E, Western blotting for protein expression in the NF-κB canonical pathway and glycolysis in HD-MTX–sensitive cells (YML3, YML11, YML12, YML16, and HKBML) and -resistant cells (YML2, YML5, and YML15) in a dose- (D) and time-dependent manner (E). F, Relative cell viability of PCNSL xenograft cells treated with DMSO, 100 μmol/L MTX, 100 μmol/L of leucovorin (LV), and a combination of MTX and leucovorin on day 2. *, P < 0.05. G, Western blotting of phospho-p65 expression in PCNSL xenograft cells at the indicated treatment for 12 hours. H, Kaplan–Meier curves showing survival difference of YML16 orthotopically implanted mice treated with vehicle and HD-MTX. I, IHC for indicated protein expression in vehicle and HD-MTX–treated mouse brain in YML16 xenografts. J, Kaplan–Meier curves showing differences in survival rates of YML15 orthotopically implanted mice treated with vehicle control and HD-MTX. K, IHC for indicated protein expression in vehicle and HD-MTX–treated mice brain in YML15 xenografts. L, Schematic illustration of NF-κB canonical pathway activation to promote tumor progression in PCNSL. Data represent the mean ± SEM. Bars, 50 μm.

Close modal

By using newly established xenograft models, we present convincing experimental evidence demonstrating that immunocompetent and EBV-positive PCNSL use distinct signaling pathways that converge to activate the RelA/p65-HK2 pathway. This signaling axis proves crucial for intracerebral tumor progression and therefore represents a potent therapeutic target in PCNSL (Fig. 7L). These data underscore the pivotal role of the canonical NF-κB pathway and subsequent aberrant glycolysis in PCNSL progression.

Our PDX models of CNSL robustly recapitulate patient tumor characteristics, allowing us to investigate tumor biology and therapeutic vulnerability in these tumors. In our experience, we found that xenograft formation was more successful with orthotopic injection than with intravascular injection. However, we did not find any CD20-positive tumor cells in non-CNS organs in both intracerebral and intravascular injection models. A recent study states that lymphoma cells frequently enter the normal CNS, but tumor formation is rare, whereas age-related gliosis promotes CNSL through NF-κB–mediated CCL19 activation (35). The present study used young adult mice that may have critical barriers diminishing the entry of CNSL cells into the CNS. Strikingly however, our orthotopic xenograft models displayed high affinity for CNSL cells within the CNS environment, capturing the characteristic phenotypic features of CNSL such as diffuse infiltration of the CNS with perivascular growth pattern, rare incidents of recurrence outside the CNS, and active glycolysis, as indicated by FDG-PET imaging (27, 33).

Our work also confirmed the identification of recurrent mutations in MYD88 and CD79B genes as candidate driver mutations, as these early genomic alterations were constitutively retained in xenografts over serial passages. Our findings also validate the key role of signaling alterations in the NF-κB pathway during xenograft formation in immunocompetent PCNSL. Experimental MYD88 and CD79B knockdown downregulated the RelA/p65-HK2 axis and reduced cell viability, and knockdown of MYD88 delayed xenograft formation. Knockdown of GLUT-1 and HK2 also suppressed cell viability and xenograft formation, confirming the crucial role of these gene alterations and subsequent aberrant glycolysis in tumor progression during PCNSL. In line with the findings that the majority of immunocompetent PCNSL cases show MYD88 and CD79B mutations (5, 8, 23, 24), this working model also could, in part, explain the elevated frequency of NF-κB–related genetic alterations in PCNSL when compared with systemic lymphomas (10, 11).

WES indicated that subclonal heterogeneity evolved during orthotopic xenograft formation in immunocompetent PCNSL, which was not primarily mediated by aSHM. This observation, along with the accurate retention of histologic subtypes and lack of extra-CNS development, indicates that the GC reaction, including aSHM, is inactive during PCNSL orthotopic xenograft formation. Therefore, combined with the finding of the MYD88L265P mutation in PCNSL cells and peripheral blood mononuclear cells (5), our data further support the proposal that PCNSL precursor cells complete the GC reaction outside the CNS before adapting to it, and the subclone is then selected during intracerebral tumor progression. For instance, the CARD11 mutation, which mediates resistance to BTK inhibitor (4), was newly identified in YML3 xenograft cells, but was lost in YML11 patient tumor cells, indicating subclonal selection during progression of the xenograft tumor. Thus, these findings indicate that our PCNSL xenografts not only recapitulate histologic and molecular subtypes of PCNSL, but also can probably reproduce the inter- and intratumoral genomic fidelity and heterogeneity seen in patients as well.

Importantly, we also found that LMP1 knockdown suppressed the RelA/p65-HK2 axis and delayed xenograft formation, indicating a pivotal role of NF-κB canonical pathway and glycolysis during tumor progression in EBV-positive PCNSL. On the other hand, two immunocompetent PCNSLs that harbored MYD88 and CD79B comutations and one EBV-positive PCNSL did not form xenografts, and phosho-p65 and Pin1 were weakly expressed in these tumors. In addition, knockdown of PIN1 in xenografts also prolonged mouse survival in vivo. Pin1 is known to regulate the conformational transformation of the phospho-serine/threonine-proline motif and facilitates multiple cancer driving pathways, including the NF-κB canonical pathway, by enhancing nuclear accumulation of phospho-p65 (31, 36). These findings indicate that MYD88/CD79B mutations and Pin1 activation, or LMP1 and Pin1 activation, converge on the RelA/p65-HK2 signaling in immunocompetent and EBV-positive PCNSL, respectively. Therefore, our results strengthen the hypothesis that after adapting within the CNS, lymphoma cells with hyperactive NF-κB canonical pathway become uniquely competent to complete engraftment within the CNS.

From a clinically relevant perspective, this represents the first study to establish CNSL PDX models accurately mirroring the patient tumor's sensitivity to HD-MTX. We anticipate that this translational approach will be useful for guiding therapeutic strategies in patients with CNSL. For instance, we found that inactivation of the NF-κB canonical pathway, as indicated by downregulation of phospho-p65, with suppressed HK2 expression levels, has the potential to serve as a biomarker of HD-MTX sensitivity. In our work here, similar cytotoxic and signaling effects were also observed in PCNSL xenograft cells after BAY11–7082 (NF-κB inhibitor) and Juglone (Pin1 inhibitor) treatment, suggesting that inactivation of the NF-κB canonical pathway and downstream glycolytic metabolism is a strong indicator of responsiveness to chemotherapeutic agents. Thus, our CNSL xenograft models could foster future efforts to further understand the contribution of genetic alterations to lymphomagenesis and tumor progression, and to develop more effective therapeutic agents targeting these pathways.

M. Natsumeda reports grants from JSPS during the conduct of the study and grants from JSPS outside the submitted work. I. Aoki reports grants from AMED and grants from JSPS during the conduct of the study. M. Natsumeda and Quantum and Radiological Sciences and Technology are members of collaborative research framework MRI Alliance that consists of Astellas, Daiichi-sankyo, Kowa, Kyowa-Kirin, Canon Medical, and Braizon. D.P. Cahill reports personal fees from Merck, personal fees from Lilly, and personal fees from Boston Pharmaceuticals outside the submitted work. A.S. Chi reports other compensation from Mirati Therapeutics outside the submitted work; in addition, A.S. Chi has a patent for WO 2020/073031 A1 issued. T.T. Batchelor reports personal fees from Genomicare outside the submitted work. M. Nagane reports grants and personal fees from ONO Pharmaceutical, personal fees from Novocure, grants and personal fees from Chugai Pharmaceutical, grants and personal fees from Nippon Kayaku, grants and personal fees from Daiichi-Sankyo, grants and personal fees from AbbVie, personal fees from RIEMSER, personal fees and non-financial support from Bristol-Myers-Squibb, grants and personal fees from Eisai, grants and personal fees from MSD, grants from Pfizer, grants from Astellas, grants from Shionogi, grants from BAYER, and grants from Otsuka outside the submitted work. No potential conflicts of interest were disclosed by the other authors.

K. Tateishi: Conceptualization, resources, data curation, software, formal analysis, supervision, funding acquisition, validation, investigation, methodology, writing-original draft, writing-review and editing. Y. Miyake: Formal analysis, funding acquisition, investigation, methodology. M. Kawazu: Conceptualization, resources, data curation, formal analysis, supervision, funding acquisition, methodology, writing-original draft, writing-review and editing. N. Sasaki: Investigation, methodology. T. Nakamura: Resources, data curation, funding acquisition, writing-review and editing. J. Sasame: Investigation. Y. Yoshii: Resources, investigation. T. Ueno: Data curation, software, formal analysis. A. Miyake: Investigation, methodology. J. Watanabe: Investigation. Y. Matsushita: Investigation. N. Shiba: Validation, investigation. N. Udaka: Validation, investigation. K. Ohki: Investigation. A.L. Fink: Investigation, methodology. S.S. Tummala: Investigation, methodology. M. Natsumeda: Conceptualization, funding acquisition, methodology, writing-review and editing. N. Ikegaya: Resources, investigation. M. Nishi: Investigation, methodology. M. Ohtake: Funding acquisition, writing-review and editing. R. Miyazaki: Resources. J. Suenaga: Resources, writing-review and editing. H. Murata: Resources. I. Aoki: Investigation, methodology. J.J. Miller: Supervision. Y. Fujii: Supervision, funding acquisition. A. Ryo: Resources, supervision. S. Yamanaka: Investigation, methodology. H. Mano: Supervision, funding acquisition, writing-review and editing. D.P. Cahill: Conceptualization, supervision, writing-original draft, writing-review and editing. H. Wakimoto: Conceptualization, supervision, writing-original draft, writing-review and editing. A.S. Chi: Conceptualization, writing-original draft, writing-review and editing. T.T. Batchelor: Conceptualization, supervision, writing-review and editing. M. Nagane: Conceptualization, supervision, funding acquisition, writing-original draft, writing-review and editing. K. Ichimura: Conceptualization, data curation, supervision, writing-original draft, writing-review and editing. T. Yamamoto: Resources, supervision, writing-original draft, project administration.

The authors thank Dr. Sayaka Shibata, Dr. Nobuhiro Nitta, Dr. Takeyoshi Eda (MS), Emi Hirata, Yuki Hoshino, Fukiko Hihara, and Yasuko Tanaka for technical and administrative assistance. This work was supported by Grant-Aid for Scientific Research (16K10765 and 19K09488 to K. Tateishi; 19K18398 to Y. Miyake; 18K16565 to T. Nakamura; 17K16632 and 19K09476 to M. Natsumeda, 18K16592 to M. Ohtake; 19K18418 to J. Watanabe; 16H05442 and 20H03795 to M. Nagane), the Japan Agency for Medical Research and Development (AMED, JPcm0106502 to M. Kawazu), Yokohama City University research grant “KAMOME Project,” Princess Takamatsu Cancer Research Fund, Takeda Science Foundation, The Uehara Memorial Foundation, The Collaborative Research Project of Brain Research Institute, Niigata University, Alumni Association of Faculty of Medicine, Kagawa University, The Ichiro Kanehara Foundation, Japan Cancer Society, and Japanese Brain Foundation (to K. Tateishi).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

1.
Schmitz
R
,
Wright
GW
,
Huang
DW
,
Johnson
CA
,
Phelan
JD
,
Wang
JQ
, et al
Genetics and pathogenesis of diffuse large B-Cell lymphoma
.
N Engl J Med
2018
;
378
:
1396
407
.
2.
Chapuy
B
,
Stewart
C
,
Dunford
AJ
,
Kim
J
,
Kamburov
A
,
Redd
RA
, et al
Molecular subtypes of diffuse large B cell lymphoma are associated with distinct pathogenic mechanisms and outcomes
.
Nat Med
2018
;
24
:
679
90
.
3.
Reddy
A
,
Zhang
J
,
Davis
NS
,
Moffitt
AB
,
Love
CL
,
Waldrop
A
, et al
Genetic and functional drivers of diffuse large B Cell lymphoma
.
Cell
2017
;
171
:
481
94
.
4.
Grommes
C
,
Pastore
A
,
Palaskas
N
,
Tang
SS
,
Campos
C
,
Schartz
D
, et al
Ibrutinib unmasks critical role of Bruton tyrosine kinase in primary CNS lymphoma
.
Cancer Discov
2017
;
7
:
1018
29
.
5.
Fukumura
K
,
Kawazu
M
,
Kojima
S
,
Ueno
T
,
Sai
E
,
Soda
M
, et al
Genomic characterization of primary central nervous system lymphoma
.
Acta Neuropathol
2016
;
131
:
865
75
.
6.
Camilleri-Broet
S
,
Criniere
E
,
Broet
P
,
Delwail
V
,
Mokhtari
K
,
Moreau
A
, et al
A uniform activated B-cell-like immunophenotype might explain the poor prognosis of primary central nervous system lymphomas: analysis of 83 cases
.
Blood
2006
;
107
:
190
6
.
7.
Grommes
C
,
Rubenstein
JL
,
DeAngelis
LM
,
Ferreri
AJM
,
Batchelor
TT
. 
Comprehensive approach to diagnosis and treatment of newly diagnosed primary CNS lymphoma
.
Neuro Oncol
2019
;
21
:
296
305
.
8.
Nakamura
T
,
Tateishi
K
,
Niwa
T
,
Matsushita
Y
,
Tamura
K
,
Kinoshita
M
, et al
Recurrent mutations of CD79B and MYD88 are the hallmark of primary central nervous system lymphomas
.
Neuropathol Appl Neurobiol
2016
;
42
:
279
90
.
9.
Ho
KG
,
Grommes
C
. 
Molecular profiling of primary central nervous system lymphomas - predictive and prognostic value?
Curr Opin Neurol
2019
;
32
:
886
94
.
10.
Ngo
VN
,
Young
RM
,
Schmitz
R
,
Jhavar
S
,
Xiao
W
,
Lim
KH
, et al
Oncogenically active MYD88 mutations in human lymphoma
.
Nature
2011
;
470
:
115
9
.
11.
Davis
RE
,
Ngo
VN
,
Lenz
G
,
Tolar
P
,
Young
RM
,
Romesser
PB
, et al
Chronic active B-cell-receptor signalling in diffuse large B-cell lymphoma
.
Nature
2010
;
463
:
88
92
.
12.
Narita
Y
,
Nagane
M
,
Mishima
K
,
Terui
Y
,
Arakawa
Y
,
Yonezawa
H
, et al
Phase 1/2 study of tirabrutinib, a second-generation bruton's tyrosine kinase inhibitor, in relapsed/refractory primary central nervous system lymphoma
.
Neuro Oncol
2020
:
noaa145
.
13.
Soussain
C
,
Choquet
S
,
Blonski
M
,
Leclercq
D
,
Houillier
C
,
Rezai
K
, et al
Ibrutinib monotherapy for relapse or refractory primary CNS lymphoma and primary vitreoretinal lymphoma: Final analysis of the phase II ‘proof-of-concept’ iLOC study by the Lymphoma study association (LYSA) and the French oculo-cerebral lymphoma (LOC) network
.
Eur J Cancer
2019
;
117
:
121
30
.
14.
Grommes
C
,
Tang
SS
,
Wolfe
J
,
Kaley
TJ
,
Daras
M
,
Pentsova
EI
, et al
Phase 1b trial of an ibrutinib-based combination therapy in recurrent/refractory CNS lymphoma
.
Blood
2019
;
133
:
436
45
.
15.
Kleinschmidt-DeMasters
BK
,
Damek
DM
,
Lillehei
KO
,
Dogan
A
,
Giannini
C
. 
Epstein Barr virus-associated primary CNS lymphomas in elderly patients on immunosuppressive medications
.
J Neuropathol Exp Neurol
2008
;
67
:
1103
11
.
16.
Biggar
RJ
,
Chaturvedi
AK
,
Goedert
JJ
,
Engels
EA
,
Study
HACM
. 
AIDS-related cancer and severity of immunosuppression in persons with AIDS
.
J Natl Cancer Inst
2007
;
99
:
962
72
.
17.
Pouzoulet
F
,
Alentorn
A
,
Royer-Perron
L
,
Assayag
F
,
Mokhtari
K
,
Labiod
D
, et al
Primary CNS lymphoma patient-derived orthotopic xenograft model capture the biological and molecular characteristics of the disease
.
Blood Cells Mol Dis
2018
;
75
:
1
10
.
18.
Zhang
L
,
Nomie
K
,
Zhang
H
,
Bell
T
,
Pham
L
,
Kadri
S
, et al
B-Cell lymphoma patient-derived xenograft models enable drug discovery and are a platform for personalized therapy
.
Clin Cancer Res
2017
;
23
:
4212
23
.
19.
Chapuy
B
,
Cheng
H
,
Watahiki
A
,
Ducar
MD
,
Tan
Y
,
Chen
L
, et al
Diffuse large B-cell lymphoma patient-derived xenograft models capture the molecular and biological heterogeneity of the disease
.
Blood
2016
;
127
:
2203
13
.
20.
Sirotnak
FM
,
Moccio
DM
,
Dorick
DM
. 
Optimization of high-dose methotrexate with leucovorin rescue therapy in the L1210 leukemia and sarcoma 180 murine tumor models
.
Cancer Res
1978
;
38
:
345
53
.
21.
Hans
CP
,
Weisenburger
DD
,
Greiner
TC
,
Gascoyne
RD
,
Delabie
J
,
Ott
G
, et al
Confirmation of the molecular classification of diffuse large B-cell lymphoma by immunohistochemistry using a tissue microarray
.
Blood
2004
;
103
:
275
82
.
22.
Watanabe
J
,
Natsumeda
M
,
Kanemaru
Y
,
Okada
M
,
Oishi
M
,
Kakita
A
, et al
Comparison of circulating tumor DNA between body fluids in patients with primary central nervous system lymphoma
.
Leuk Lymphoma
2019
;
60
:
3587
9
.
23.
Nayyar
N
,
White
MD
,
Gill
CM
,
Lastrapes
M
,
Bertalan
M
,
Kaplan
A
, et al
MYD88 L265P mutation and CDKN2A loss are early mutational events in primary central nervous system diffuse large B-cell lymphomas
.
Blood Adv
2019
;
3
:
375
83
.
24.
Chapuy
B
,
Roemer
MG
,
Stewart
C
,
Tan
Y
,
Abo
RP
,
Zhang
L
, et al
Targetable genetic features of primary testicular and primary central nervous system lymphomas
.
Blood
2016
;
127
:
869
81
.
25.
Braggio
E
,
McPhail
ER
,
Macon
W
,
Lopes
MB
,
Schiff
D
,
Law
M
, et al
Primary central nervous system lymphomas: a validation study of array-based comparative genomic hybridization in formalin-fixed paraffin-embedded tumor specimens
.
Clin Cancer Res
2011
;
17
:
4245
53
.
26.
Khodabakhshi
AH
,
Morin
RD
,
Fejes
AP
,
Mungall
AJ
,
Mungall
KL
,
Bolger-Munro
M
, et al
Recurrent targets of aberrant somatic hypermutation in lymphoma
.
Oncotarget
2012
;
3
:
1308
19
.
27.
Kawai
N
,
Miyake
K
,
Yamamoto
Y
,
Nishiyama
Y
,
Tamiya
T
. 
18F-FDG PET in the diagnosis and treatment of primary central nervous system lymphoma
.
Biomed Res Int
2013
;
2013
:
247152
.
28.
Go
JL
,
Lee
SC
,
Kim
PE
. 
Imaging of primary central nervous system lymphoma
.
Neurosurg Focus
2006
;
21
:
E4
.
29.
Fan
G
,
Fan
Y
,
Gupta
N
,
Matsuura
I
,
Liu
F
,
Zhou
XZ
, et al
Peptidyl-prolyl isomerase Pin1 markedly enhances the oncogenic activity of the rel proteins in the nuclear factor-kappaB family
.
Cancer Res
2009
;
69
:
4589
97
.
30.
Ryo
A
,
Suizu
F
,
Yoshida
Y
,
Perrem
K
,
Liou
YC
,
Wulf
G
, et al
Regulation of NF-kappaB signaling by Pin1-dependent prolyl isomerization and ubiquitin-mediated proteolysis of p65/RelA
.
Mol Cell
2003
;
12
:
1413
26
.
31.
Lu
Z
,
Hunter
T
. 
Prolyl isomerase Pin1 in cancer
.
Cell Res
2014
;
24
:
1033
49
.
32.
Ok
CY
,
Papathomas
TG
,
Medeiros
LJ
,
Young
KH
. 
EBV-positive diffuse large B-cell lymphoma of the elderly
.
Blood
2013
;
122
:
328
40
.
33.
Grommes
C
,
DeAngelis
LM
. 
Primary CNS lymphoma
.
J Clin Oncol
2017
;
35
:
2410
8
.
34.
Majumdar
S
,
Aggarwal
BB.
. 
Methotrexate suppresses NF-kappaB activation through inhibition of IkappaBalpha phosphorylation and degradation
.
J Immunol
2001
;
167
:
2911
20
.
35.
O'Connor
T
,
Zhou
X
,
Kosla
J
,
Adili
A
,
Garcia Beccaria
M
,
Kotsiliti
E
, et al
Age-related gliosis promotes central nervous system lymphoma through CCL19-mediated tumor cell retention
.
Cancer Cell
2019
;
36
:
250
67
.
36.
Chen
Y
,
Wu
YR
,
Yang
HY
,
Li
XZ
,
Jie
MM
,
Hu
CJ
, et al
Prolyl isomerase Pin1: a promoter of cancer and a target for therapy
.
Cell Death Dis
2018
;
9
:
883
.