1. Biochemistry and Chemical Biology
  2. Human Biology and Medicine
Download icon

GPIHBP1 expression in gliomas promotes utilization of lipoprotein-derived nutrients

  1. Xuchen Hu
  2. Ken Matsumoto
  3. Rachel S Jung
  4. Thomas A Weston
  5. Patrick J Heizer
  6. Cuiwen He
  7. Norma P Sandoval
  8. Christopher M Allan
  9. Yiping Tu
  10. Harry V Vinters
  11. Linda M Liau
  12. Rochelle M Ellison
  13. Jazmin E Morales
  14. Lynn J Baufeld
  15. Nicholas A Bayley
  16. Liqun He
  17. Christer Betsholtz
  18. Anne P Beigneux
  19. David A Nathanson
  20. Holger Gerhardt
  21. Stephen G Young  Is a corresponding author
  22. Loren G Fong  Is a corresponding author
  23. Haibo Jiang  Is a corresponding author
  1. University of California, Los Angeles, United States
  2. VIB-KU Leuven Center for Cancer Biology (CCB), Belgium
  3. Uppsala University, Sweden
  4. Karolinska Institutet, Sweden
  5. Max Delbrück Center for Molecular Medicine, Germany
  6. University of Western Australia, Australia
Research Article
  • Cited 0
  • Views 683
  • Annotations
Cite this article as: eLife 2019;8:e47178 doi: 10.7554/eLife.47178

Abstract

GPIHBP1, a GPI-anchored protein of capillary endothelial cells, binds lipoprotein lipase (LPL) within the subendothelial spaces and shuttles it to the capillary lumen. GPIHBP1-bound LPL is essential for the margination of triglyceride-rich lipoproteins (TRLs) along capillaries, allowing the lipolytic processing of TRLs to proceed. In peripheral tissues, the intravascular processing of TRLs by the GPIHBP1–LPL complex is crucial for the generation of lipid nutrients for adjacent parenchymal cells. GPIHBP1 is absent from the capillaries of the brain, which uses glucose for fuel; however, GPIHBP1 is expressed in the capillaries of mouse and human gliomas. Importantly, the GPIHBP1 in glioma capillaries captures locally produced LPL. We use NanoSIMS imaging to show that TRLs marginate along glioma capillaries and that there is uptake of TRL-derived lipid nutrients by surrounding glioma cells. Thus, GPIHBP1 expression in gliomas facilitates TRL processing and provides a source of lipid nutrients for glioma cells.

https://doi.org/10.7554/eLife.47178.001

Introduction

GPIHBP1, a GPI-anchored protein of capillary endothelial cells, is required for lipoprotein lipase (LPL)–mediated processing of triglyceride-rich lipoproteins (TRLs) (Beigneux et al., 2007). The principal function of GPIHBP1 is to capture LPL within the interstitial spaces, where it is secreted by parenchymal cells, and then to shuttle this enzyme to the luminal surface of capillary endothelial cells (Davies et al., 2010). GPIHBP1 is a long-lived protein (Young et al., 2011; Olafsen et al., 2010) that moves bidirectionally across endothelial cells, with each trip to the abluminal plasma membrane representing an opportunity to capture LPL and bring it to the capillary lumen (Davies et al., 2012). When GPIHBP1 is absent or defective, LPL is stranded within the interstitial spaces, where it remains bound to sulfated proteoglycans near the surface of cells (Young et al., 2011; Davies et al., 2010; Allan et al., 2017a; Fong et al., 2016). The inability of LPL to reach the capillary lumen in the absence of GPIHBP1 expression profoundly impairs TRL processing, resulting in severe hypertriglyceridemia (chylomicronemia) (Beigneux et al., 2007; Davies et al., 2010; Goulbourne et al., 2014).

GPIHBP1 is expressed in the capillary endothelial cells of peripheral tissues, with particularly high levels of expression in heart and brown adipose tissue (Beigneux et al., 2007; Davies et al., 2010; Fong et al., 2016). Most of the LPL within those tissues is bound to GPIHBP1 on capillaries (Beigneux et al., 2007; Davies et al., 2010; Davies et al., 2012; Allan et al., 2017a; Fong et al., 2016; Allan et al., 2017b; Allan et al., 2016), and the processing of TRLs in these tissues is robust, generating fatty acid nutrients for nearby parenchymal cells (Fong et al., 2016; Jiang et al., 2014a; He et al., 2018a). By contrast, GPIHBP1 is absent from capillaries of the brain (Young et al., 2011; Davies et al., 2010; Olafsen et al., 2010), a tissue that depends on glucose for fuel (Mergenthaler et al., 2013). When wild-type mice are injected intravenously with a GPIHBP1-specific antibody, the antibody rapidly binds to GPIHBP1-expressing capillaries in peripheral tissues and disappears from the plasma (Davies et al., 2010; Olafsen et al., 2010). By contrast, there is no antibody binding to the capillaries of the brain (Davies et al., 2010; Olafsen et al., 2010).

For the lipolytic processing of TRLs to proceed, lipoproteins in the bloodstream must marginate along the luminal surface of capillaries (Goulbourne et al., 2014). TRL margination along capillaries depends on GPIHBP1, more specifically on GPIHBP1-bound LPL (Goulbourne et al., 2014). In GPIHBP1-deficient mice, TRLs never stop along heart capillaries and instead simply ‘flow on by’ in the bloodstream (Goulbourne et al., 2014). In wild-type mice, TRLs marginate along heart capillaries, but TRL margination is absent along capillaries of the brain (Goulbourne et al., 2014).

Even though GPIHBP1 is not found in brain capillaries, there is ample evidence for LPL expression within the brain (Ben-Zeev et al., 1990; Bessesen et al., 1993; Goldberg et al., 1989; Vilaró et al., 1990; Yacoub et al., 1990; Eckel and Robbins, 1984). Several groups have found LPL in the rat brain, specifically in neurons of the dentate gyrus and hippocampus, in pyramidal cells of the cortex, and in Purkinje cells of the cerebellum (Ben-Zeev et al., 1990; Bessesen et al., 1993; Goldberg et al., 1989; Vilaró et al., 1990; Eckel and Robbins, 1984). Using single-cell RNA sequencing, Zhang et al. (2014) found Lpl transcripts in the resident macrophages of the brain (microglia), with lower levels in astrocytes, neurons, and oligodendrocytes. Using the same approach, Vanlandewijck et al. (2018) found LPL expression in brain smooth muscle cells and in perivascular fibroblasts (at even higher levels than in microglial cells). Given the absence of GPIHBP1 expression in brain capillaries and the absence of TRL margination along brain capillaries, we have proposed that the LPL in the brain probably has an extravascular function, presumably to hydrolyze glycerolipids within the extracellular spaces (Young et al., 2011; Adeyo et al., 2012).

Despite the absence of GPIHBP1 expression in brain capillaries, we were curious about whether GPIHBP1 might be expressed in the capillaries of gliomas. Glioma capillaries are morphologically distinct from normal brain capillaries (Yuan et al., 1994; Hobbs et al., 1998; Monsky et al., 1999; Bullitt et al., 2005), and the blood–brain barrier is often defective (Zhang et al., 1992). Electron microscopy has suggested that glioblastoma capillaries resemble capillaries in peripheral tissues (Vaz et al., 1996).

If GPIHBP1 were to be expressed in glioma capillaries, it could be relevant to glioma metabolism. The GPIHBP1 might capture locally produced LPL, allowing for TRL margination and TRL processing, and thereby providing a source of lipid nutrients for glioma cells. Interestingly, Dong et al. (2017) documented LPL expression in gliomas. Also, several studies have raised the possibility that glioma cells use fatty acids for fuel (Lin et al., 2017; Guo et al., 2011; Guo et al., 2009a; Guo et al., 2009b; Guo et al., 2013) and that levels of free fatty acids are higher in gliomas than in normal brain tissue (Guo et al., 2013; Gopal et al., 1963).

In the current study, we sought to determine whether glioma capillaries express GPIHBP1 and, if so, whether it binds LPL and facilitates TRL margination and the lipolytic processing of TRLs. In our study, we took advantage of NanoSIMS imaging, a high-resolution mass spectrometry–based imaging modality that makes it possible to visualize TRL margination and TRL processing in tissue sections (He et al., 2018a; Jiang et al., 2014a; Jiang et al., 2014b; He et al., 2017a; He et al., 2017b; He et al., 2018b; He et al., 2018c). This imaging modality allowed us to visualize TRL margination in glioma capillaries as well as the entry of TRL-derived nutrients into tumor cells.

Results

GPIHBP1 is expressed in the endothelial cells of human gliomas

We sectioned 20 human gliomas (Table 1) and screened them for GPIHBP1 expression by confocal microscopy with three GPIHBP1-specific monoclonal antibodies (mAbs) — RF4, which binds to residues 27–44 downstream from GPIHBP1’s acidic domain (Kristensen et al., 2018); and RE3 and RG3, which both bind to GPIHBP1’s LU (Ly6/uPAR) domain (Hu et al., 2017). GPIHBP1 in capillary endothelial cells was detected in 14 of 20 gliomas (Table 1) and colocalized with von Willebrand factor, an endothelial cell marker (Figure 1). GPIHBP1 expression in glioma capillaries did not appear to correlate with glioma grade, 1p/19q co-deletions, or IDH1 mutations (Table 1). GPIHBP1 was not detectable in the capillaries of human brain specimens (Figure 1). The GPIHBP1 in glioma capillaries could be detected with all three GPIHBP1-specific mAbs (Figure 2A). To be confident in the specificity of the antibodies, we performed studies in which recombinant human GPIHBP1 was added to the GPIHBP1-specific mAbs before incubating the solution with the glioma sections. As expected, the presence of recombinant GPIHBP1 eliminated binding of the GPIHBP1-specific mAbs to glioma capillaries (Figure 2B). GPIHBP1 expression in glioma capillaries could also be detected by immunoperoxidase staining (Figure 1—figure supplement 1).

Figure 1 with 1 supplement see all
GPIHBP1 expression in the endothelial cells of several human gliomas.

Immunohistochemical studies on surgically resected gliomas (Gliomas 1, 5, 9; Table 1) and non-diseased human frontal lobe (n = 3), revealing GPIHBP1 expression in capillaries of gliomas but not in frontal lobe specimens. GPIHBP1 (detected with a combination of the mAbs RE3 and RF4, 10 μg/ml each [red]) colocalized with von Willebrand factor (vWF, a marker for endothelial cells [green]), but not with glial fibrillary acidic protein (GFAP, a marker for astroglial cells [magenta]). DNA was stained with DAPI (blue). Three sections of each tumor and normal brain were evaluated and representative images are shown. Scale bar, 50 μm.

https://doi.org/10.7554/eLife.47178.002
Detecting GPIHBP1 in capillaries of human glioma specimens with three different monoclonal antibodies (mAbs) against GPIHBP1.

(A) Confocal fluorescence microscopy studies on sections from glioma sample 1 (Table 1), demonstrating the detection of GPIHBP1 with three different human GPIHBP1–specific monoclonal antibodies (mAbs). Tissue sections were fixed with 3% PFA and then stained with mAbs against human GPIHBP1 (RF4, RE3, or RG3, 10 μg/ml [red]), an antibody against von Willebrand factor (vWF[green]), and an antibody against glial fibrillary acidic protein (GFAP [magenta]). All three GPIHBP1-specific mAbs detected GPIHBP1 in capillaries, colocalizing with von Willebrand factor. DNA was stained with DAPI (blue). Scale bar, 50 μm. (B) Immunofluorescence confocal microscopy studies on human glioma sample 5, performed with mAbs RF4 and RE3 (10 μg/ml) in the presence or absence of 50 μg of recombinant soluble human GPIHBP1 (hGPIHBP1). Adding recombinant hGPIHBP1 to the antibody incubation abolished binding of the GPIHBP1-specific mAbs to GPIHBP1 on glioma capillaries. Images show GPIHBP1 (red), vWF (green), GFAP (magenta), and DAPI (blue). Three sections of tumors were evaluated; representative images are shown. Scale bar, 50 μm.

https://doi.org/10.7554/eLife.47178.004
Table 1
Human glioma tumor specimens.

Expression of GPIHBP1 was assessed by immunohistochemistry with mAbs against human GPIHBP1 (RF4, RE3, RG3). Those conducting the studies were blinded to diagnoses. This table details the tumor diagnosis, location, 1p/19q co-deletion, and IDH1 mutation status, as well as the presence of GPIHBP1.

https://doi.org/10.7554/eLife.47178.005
Sample IDTissue diagnosisLocation1p/19q co-deletionIDH1 mutationGPIHBP1
1Glioblastoma (GBM)Right frontal, parietalNoNegativeYes
2GBMLeft temporalNoNegativeYes
3GBMRight occipitalNoNegativeYes
4GBMLeft frontalNoNegativeYes
5Oligodendroglioma Grade IILeft anterior temporal, left posterior temporalYesNegativeYes
6Oligoastrocytoma Grade IIIRight temporalNoNegativeYes
7GBM + oligodendroglial componentLeft frontalYesNegativeYes
8GBM + extensive oligodendroglial componentRight frontalNoNegativeYes
9Oligodendroglioma Grade IIILeft frontalYes+R132HYes
10Oligodendroglioma Grade IIILeft frontalYes+R132HYes
11OligoastrocytomaRight parietalNoNegativeYes
12Oligodendroglioma Grade IIIRight parietalYes+R132HYes
13Oligodendroglioma Grade IIIRight parietalYesNegativeYes
14Oligoastrocytoma Grade IIILeft temporalNo+R132HYes
15Oligoastrocytoma Grade IIIRight temporalNo+R132GNo
16Oligoastrocytoma Grade IIIRight frontalNo+R132HNo
17Oligodendroglioma Grade IIILeft frontalYesNegativeNo
18Oligodendroglioma Grade IIILeft frontalYes+R132HNo
19Oligodendroglioma Grade IIILeft temporalYesNegativeNo
20Oligodendroglioma Grade IIIRight temporalYes+R132HNo

GPIHBP1 is present in the capillary endothelial cells of mouse gliomas

To determine whether GPIHBP1 is expressed in a mouse model of glioblastoma, spheroids of syngeneic C57BL/6 mouse CT-2A glioma cells (Seyfried et al., 1992; Oh et al., 2014), modified to express a blue fluorescent protein (BFP) (Mathivet et al., 2017), were engrafted into the brains of mice harboring an endothelial cell–specific Pdgfb-iCreERT2 transgene (Claxton et al., 2008) and a ROSAmT/mG reporter allele (Muzumdar et al., 2007). ROSAmT/mG is a two-color fluorescent, membrane-targeted Cre-dependent reporter allele. In the absence of Cre, all cells express a membrane-localized tdTomato and fluoresce red. In the setting of Cre expression, cells express membrane-localized EGFP (rather than tdTomato) and fluoresce green. Before tumor implantation, mice were injected with tamoxifen to induce Pdgfb-driven Cre expression in endothelial cells; thus, the endothelial cells of the mice expressed EGFP and fluoresced green. Mice harboring gliomas (after three weeks of growth) were injected intravenously with an Alexa Fluor 647–conjugated antibody against mouse GPIHBP1 (11A12) (Beigneux et al., 2009). Mice were perfused with PBS and then perfusion-fixed with 2% PFA, and tumor sections were processed for confocal immunofluorescence microscopy. GPIHBP1 was detected in endothelial cells of the gliomas, colocalizing with EGFP (brain endothelial cells), but GPIHBP1 was absent from capillaries in the adjacent normal brain (Figure 3, Figure 3—figure supplement 1). Using transmission electron microscopy, we observed large and irregularly shaped capillaries in gliomas, with numerous villus-like structures on the luminal surface of endothelial cells (Figure 3—figure supplement 2), similar to findings reported for capillaries in human gliomas (Vaz et al., 1996; Coomber et al., 1987; Weller et al., 1977).

Figure 3 with 3 supplements see all
GPIHBP1 is expressed by capillary endothelial cells in mouse gliomas.

Confocal microscopy images of a BFP-tagged CT-2A glioma implanted in a ROSAmT/mG::Pdgfb-iCreERT2 mouse, revealing the expression of GPIHBP1 in capillary endothelial cells of the glioma but not those of normal brain. Tamoxifen was administered prior to implantation of the glioma spheroid to activate membrane-targeted EGFP in endothelial cells (green). After three weeks of glioma growth, mice were anesthetized and injected via the tail vein with an Alexa Fluor 647–labeled antibody against mouse GPIHBP1 (11A12; red). The mice were then perfused with PBS and perfusion-fixed with 2% PFA in PBS. Glioma and adjacent normal brain were harvested, and 200-μm-thick sections were imaged by confocal microscopy. GPIHBP1 was present on endothelial cells of the glioma (blue) but was absent from normal brain. High-magnification images of the boxed area are shown on the right. Three mice were evaluated; representative images are shown. Scale bar, 50 μm.

https://doi.org/10.7554/eLife.47178.006

The factors that regulate Gpihbp1 expression in the capillary endothelial cells of peripheral tissues and gliomas are incompletely understood. However, a recent study found that Gpihbp1 transcript levels in rat aortic endothelial cells are upregulated by vascular endothelial growth factor (VEGF) (Chiu et al., 2016), an angiogenic factor known to be expressed at high levels by glioma cells (Plate et al., 1994; Pietsch et al., 1997; Christov et al., 1998). We found that Gpihbp1 expression in the mouse brain endothelial cell line bEnd.3 is upregulated by recombinant VEGF (Figure 3—figure supplement 3).

GLUT1 is expressed in the capillaries of gliomas and normal brain

We used immunofluorescence microscopy to examine the expression of GPIHBP1 and GLUT1 (the main glucose transporter in brain capillaries [Maher et al., 1994; Pardridge et al., 1990]) in mouse gliomas and adjacent normal brain. GPIHBP1 expression was detected in gliomas but was absent in the normal brain. The signal for GLUT1 was strong in the endothelial cells of the normal brain and was easily detectable in the capillaries of gliomas (Figure 4, Figure 4—figure supplements 12). Consistent findings were observed in single-cell RNA-seq studies on vascular cells of gliomas (Ken Matsumoto, manuscript in preparation) and normal brain vascular cells (Vanlandewijck et al., 2018; He et al., 2018d). Endothelial cells of gliomas (identifed by high von Willebrand factor [vWF] expression) exhibit high expression of Gpihbp1 and somewhat lower levels of Glut1 expression (e.g., Endothelial cell cluster 5 in Figure 4—figure supplement 3). In normal brain, Glut1 was highly expressed in endothelial cells, whereas Gpihbp1 expression was absent (Figure 4—figure supplement 3). In Gpihbp1-deficient mice, GLUT1 expression was detectable in the capillaries of gliomas and normal brain (Figure 4—figure supplement 4).

Figure 4 with 4 supplements see all
Expression of GPIHBP1 and GLUT1 in the endothelial cells of mouse gliomas.

Immunohistochemical studies of a BFP-expressing CT-2A glioma (after three weeks of growth). Mice were injected via the tail vein with an Alexa Fluor 647–labeled antibody against mouse GPIHBP1 (11A12; green), then perfused with PBS and perfusion-fixed with 2% PFA. Glioma and adjacent normal brain tissue were harvested, then 200-μm thick sections cut, fixed with 4% PFA, and stained with an antibody against GLUT1 (red). GPIHBP1 was present in the capillaries of mouse gliomas (blue) but absent from the capillaries of the normal brain. High-magnification images in the boxed region are shown below. Three mice were evaluated; representative images are shown. Scale bar, 50 μm.

https://doi.org/10.7554/eLife.47178.010

LPL is present on GPIHBP1-expressing capillaries of mouse gliomas

Most of the LPL in peripheral tissues (e.g., heart or brown adipose tissue) is bound to GPIHBP1 on capillaries; consequently, LPL and GPIHBP1 colocalize in tissue sections (Young et al., 2011; Davies et al., 2010; Davies et al., 2012; Allan et al., 2017a; Fong et al., 2016; Allan et al., 2017b; Allan et al., 2016). We hypothesized that GPIHBP1-expressing endothelial cells of gliomas could capture LPL. Several observations prompted us to consider this hypothesis. First, as noted earlier, there is ample evidence for LPL expression in the brain (Ben-Zeev et al., 1990; Bessesen et al., 1993; Goldberg et al., 1989; Vilaró et al., 1990; Yacoub et al., 1990; Zhang et al., 2014), and it seemed reasonable that some of that LPL would reach high-affinity GPIHBP1-binding sites on endothelial cells. Second, gliomas contain large numbers of macrophages (F4/80-expressing cells; Figure 5—figure supplement 1), and macrophages are known to express LPL (Mahoney et al., 1982). We found that LPL could be detected in peritoneal macrophages from wild-type mice but not in macrophages harvested from Lpl–/– mice carrying a skeletal muscle–specific human LPL transgene (Lpl–/–MCK-hLPL) (Levak-Frank et al., 1995) (Figure 5—figure supplement 2). We also found that LPL could be detected in some of the macrophages in mouse gliomas and in normal brain of wild-type mice, but not in the brain of Lpl–/–MCK-hLPL mice (Figure 5—figure supplement 3). These findings were consistent with single-cell RNA-seq data from glioma and normal brain, in which Lpl transcripts were found in the macrophages of gliomas and microglia of normal brain (Figure 4—figure supplement 3). Lpl transcripts are not present in capillary endothelial cells. Third, the most highly upregulated fatty acid metabolism gene in human gliomas, compared to normal brain tissue, is LPL (Figure 5—figure supplement 4). The second most perturbed gene in gliomas is CD36, which encodes a putative fatty acid transporter (Figure 5—figure supplement 4).

To determine whether LPL is bound to GPIHBP1-expressing capillaries of gliomas, we performed immunohistochemical studies, taking advantage of an affinity-purified goat antibody against mouse LPL (Page et al., 2006). These studies revealed colocalization of GPIHBP1 and LPL in glioma capillaries (Figure 5, Figure 5—figure supplement 5). LPL was not present in the capillaries of the normal brain or in the capillaries of gliomas from Gpihbp1–/– mice (Figure 5, Figure 5—figure supplement 5). As expected, the binding of the goat LPL antibody to tissues of Lpl–/–MCK-hLPL mice was low (Figure 5, Figure 5—figure supplement 5), whereas mouse LPL was easily detectable in the heart capillaries of wild-type mice (colocalizing with GPIHBP1) (Figure 5—figure supplement 6). Consistent with earlier publications (Ben-Zeev et al., 1990; Vilaró et al., 1990), we observed a strong mouse LPL signal in the hippocampal neurons of wild-type mice but not of Lpl–/–MCK-hLPL mice (Figure 5—figure supplement 7). Of note, LPL was undetectable in ‘secondary antibody–only’ experiments (i.e., when the incubation of the primary antibody with tissue sections was omitted) (Figure 5, Figure 5—figure supplement 57).

Figure 5 with 8 supplements see all
Lipoprotein lipase (LPL) colocalizes with GPIHBP1 in glioma capillaries.

Confocal immunofluorescence microscopy studies on glioma and normal brain from wild-type and Gpihbp1–/– mice, along with the brain from an Lpl–/– mouse carrying a skeletal muscle–specific human LPL transgene (MCK). Glioma and brain sections (10-μm-thick) were fixed with 3% PFA and then stained with a mAb against mouse GPIHBP1 (11A12; green), a goat antibody against mouse LPL (red), and a rabbit antibody against CD31 (white). LPL colocalizes with GPIHBP1 and CD31 in the capillaries of gliomas; GPIHBP1 and LPL were absent from normal brain capillaries and from glioma capillaries in Gpihbp1–/– mice. DNA was stained with DAPI (blue). No LPL was detected in the capillaries of Lpl-deficient mice (MCK) or when the incubation with primary antibodies was omitted (Secondary Only). Staining of all tissue sections was performed simultaneously, and all images were recorded with identical microscopy settings. Three mice per genotype were evaluated; representative images are shown. Scale bar, 50 μm.

https://doi.org/10.7554/eLife.47178.015

There is little reason to suspect that the expression of LPL influences the expression of GPIHBP1 in capillaries. The overexpression of human LPL in the skeletal muscle of Lpl–/–MCK-hLPL mice did not alter levels of Gpihbp1 expression (Figure 5—figure supplement 8).

Margination of TRLs along glioma capillaries and uptake of TRL-derived nutrients in glioma cells

Given the presence of GPIHBP1-bound LPL on glioma capillaries, we suspected that we might find evidence of TRL margination and processing in gliomas. To test this idea, TRLs that were heavily labeled with deuterated lipids ([2H]TRLs) (He et al., 2018a) were injected intravenously into mice harboring CT-2A gliomas (after three weeks of glioma growth). After allowing the [2H]TRLs to circulate for either 1 min or 30 min, the mice were euthanized, extensively perfused with PBS, and perfusion-fixed with carbodiimide/glutaraldehyde. Heart, brain, and glioma specimens were harvested and processed for NanoSIMS imaging. 12C14N or 1H images were used to visualize tissue morphology, and 2H/1H images were used to identify regions of 2H enrichment. The scale in the 2H/1H images of brain and glioma specimens ranges from 0.00018 to 0.0003 (i.e., from levels slightly above 2H natural abundance to levels twice as high as 2H natural abundance). The scale in the heart 2H/1H images ranges from 0.00018 to 0.0006. In mice euthanized 1 min after the [2H]TRLs injection, [2H]TRL margination was visualized along the luminal surface of glioma and heart capillaries, but not along the capillaries of normal brain (Figure 6A–B). After 1 min, deuterated lipids from the [2H]TRLs had already entered glioma cells and were even found in cytosolic neutral lipid droplets of those cells (Figure 6B). By contrast, 2H enrichment was virtually absent in normal brain. As expected (He et al., 2018a), we observed substantial amounts of [2H]TRL-derived lipids in cardiomyocytes, including in cytosolic lipid droplets. In gliomas harvested 30 min after the injection of [2H]TRLs, we observed similar findings: TRL margination along capillaries of gliomas and heart and the uptake of TRL-derived nutrients by glioma cells and cardiomyocytes (Figure 7). Again, [2H]TRL margination was absent in capillaries of the normal brain at the 30-min time point, and we did not find 2H enrichment in the parenchymal cells of the normal brain. We did, however, observe very low levels of 2H enrichment in capillary endothelial cells of normal brain. Given the absence of TRL margination in normal brain capillaries, we speculate that the very low amounts of 2H enrichment in brain capillary endothelial cells may relate to [2H]TRL processing in the periphery, followed by the uptake of unesterified [2H]fatty acids by endothelial cells of the brain.

NanoSIMS imaging reveals margination of [2H]TRLs along glioma capillaries and 2H enrichment in adjacent glioma cells.

Four-month-old C57BL/6 mice harboring CT-2A gliomas were fasted for 4 hr and then injected intravenously with 200 μl of [2H]TRLs. After 1 min, mice were euthanized and perfusion-fixed with carbodiimide/glutaraldehyde. Tissue sections were processed for NanoSIMS imaging. (A) NanoSIMS images showing margination of [2H]TRLs in glioma capillaries. 1H images were created to visualize tissue morphology (upper panels). Composite 2H/1H (red) and 1H (blue) images reveal [2H]TRLs (white arrows) in glioma and heart capillaries (middle and lower panels). The lower panels are close-up images of the regions outlined in the middle panels. 2H/1H ratio scales were set to show marginated TRLs. Scale bars, 4 μm. (B) NanoSIMS images showing 2H enrichment in glioma tissue. 12C14N images were generated to visualize tissue morphology. 2H/1H ratio images reveal margination of [2H]TRLs within the capillary lumen and 2H-enriched lipid droplets in gliomas and heart. There was no 2H enrichment in normal brain tissue. Scale bars, 4 μm. The bar graph shows the average fold change ± SD in the 2H/1H ratio above natural abundance. The experiment was performed in two mice with a minimum of seven images analyzed for each sample. Differences were assessed using a Student’s t-test with Welch’s correction.

https://doi.org/10.7554/eLife.47178.024
Figure 7 with 2 supplements see all
NanoSIMS imaging showing 2H enrichment in gliomas 30 min after an intravenous injection of [2H]TRLs.

Four-month-old C57BL/6 mice harboring CT-2A gliomas were fasted for 4 hr and then injected intravenously with 200 μl of [2H]TRLs. After 30 min, mice were euthanized and perfusion-fixed with carbodiimide/glutaraldehyde. Sections of glioma, brain, and heart were processed for NanoSIMS imaging. 12C14N images were created to visualize tissue morphology. 2H/1H ratio images reveal margination of [2H]TRLs along the capillary lumen (white arrows) and 2H enrichment in glioma and heart, including in cytosolic lipid droplets. Images of normal brain revealed slight 2H enrichment in capillary endothelial cells. Scale bars, 4 μm. The bar graph shows the average fold change ± SD in the 2H/1H ratio above natural abundance. The experiment was performed in two mice, with a minimum of seven images analyzed for each sample. Differences were assessed with a Student’s t-test with Welch’s correction.

https://doi.org/10.7554/eLife.47178.025

At both the 1-min and 30-min time points, we observed heterogeneity in 2H enrichment in glioma cells, with occasional perivascular cells exhibiting striking 2H enrichment. We do not know the identity of the highly enriched perivascular cells (i.e., whether they are tumor cells, pericytes, or macrophages), nor do we understand why some cells within the glioma took up more [2H]TRL-derived lipids than other cells.

As an experimental control, we injected a mouse with PBS alone rather than with [2H]TRLs. As expected, there was no 2H enrichment in the tissues of that mouse (Figure 7—figure supplement 1).

We performed an additional study in which [2H]TRLs were injected intravenously into a wild-type mouse and a Gpihbp1–/– mouse. After 15 min, the hearts and brains from these mice were harvested and processed for NanoSIMS imaging. The 2H/1H ratio images revealed 2H enrichment in the heart of the wild-type mouse but negligible 2H enrichment in the heart of the Gpihbp1–/– mouse (2H enrichment in cardiomyocyte lipid droplets was only ~10% greater than natural abundance) (Figure 7—figure supplement 2). In hindsight, the negligible amounts of 2H enrichment in the heart of the Gpihbp1–/– mouse was probably not surprising, given the very large pool of unlabeled triglycerides in the bloodstream of Gpihbp1–/– mice (~50–100-fold higher than that in wild-type mice). At the 15-min time point, we were unable to detect 2H enrichment in the brain of either the wild-type mouse or the Gpihbp1–/– mouse (Figure 7—figure supplement 2).

13C enrichment in gliomas following administration of 13C-labeled fatty acids or 13C-labeled glucose by gastric gavage

In addition to studies of gliomas after an intravenous injection of [2H]TRLs, we performed NanoSIMS imaging after administering 13C-labeled fatty acids or 13C-labeled glucose by gastric gavage (three doses over 36 hr) (Figure 8). In the case of the 13C-labeled fatty acid experiments, it is likely that most of the 13C-labeled lipids entered the bloodstream in chylomicrons. Once again, 12C14N images were useful for tissue morphology, and the 13C/12C ratio images were useful to identify regions of 13C enrichment. The scale for the 13C/12C images ranges from 0.0115 to 0.0150 (from slightly above 13C natural abundance to ~36% greater than natural abundance). After administering 13C-labeled fatty acids, 13C enrichment was observed in both glioma cells and in the capillary endothelial cells of gliomas (Figure 8A). In some images, 13C-enriched cytosolic lipid droplets were visible in glioma cells (Figure 8—figure supplement 1). 13C enrichment was virtually absent from normal brain (Figure 8A). However, after adjusting the scale of the NanoSIMS images, a small amount of 13C enrichment was observed in capillary endothelial cells within the brain parenchyma (Figure 8—figure supplement 2). As expected (He et al., 2018a), we observed substantial amounts of 13C enrichment in cardiomyocytes (Figure 8A).

Figure 8 with 4 supplements see all
Tissue uptake of fatty acid and glucose-derived nutrients by mice harboring CT-2A gliomas.

(A) NanoSIMS images showing 13C enrichment in mouse tissues (brain, glioma, and heart) after oral administration of 13C-labeled mixed fatty acids to mice (three 80-mg doses administered 12 h apart). 12C14N images were generated to visualize tissue morphology; 13C/12C ratio images were used to visualize 13C enrichment in tissues. Scale bars, 4 μm. (B) NanoSIMS images revealing 13C enrichment in tissues following oral administration of 13C-labeled glucose to mice (three 75-mg doses given 12-h apart). 12C14N images were generated to visualize tissue morphology; 13C/12C ratio images were generated to assess 13C enrichment in tissues. Scale bars, 4 μm. The bar graphs show the average 13C/12C ratio ± SD multiplied by 10,000 for fatty acids (left) and glucose (right). Each experiment was performed in two mice, with a minimum of seven images analyzed for each sample. Differences were assessed using a Student’s t-test with Welch’s correction.

https://doi.org/10.7554/eLife.47178.028

After administering [13C]glucose to mice, 13C enrichment was easily detectable in normal brain but was even ~20% higher in gliomas (Figure 8B). We also observed 13C enrichment in cardiomyocytes (Figure 8B). As expected, there was no 13C enrichment in the tissues of a mouse that was administered PBS alone (Figure 8—figure supplement 3).

To determine whether an absence of GPIHBP1 expression would influence the growth of glioma tumors, CT-2A glioma cells that had been stably transfected with a Gaussia luciferase reporter were injected into the brains of wild-type and Gpihbp1–/– mice (n = 11/group). Tumor burden was assessed in live animals by measuring luciferase activity in the blood (Mai et al., 2017; Tannous, 2009). We observed no statistically significant differences in tumor growth, tumor size, or survival between wild-type and Gpihbp1–/– mice (Figure 8—figure supplement 4). This result was not particularly surprising, given that gliomas have a robust capacity to utilize glucose-derived nutrients (Figure 8B).

Discussion

We sought to determine whether GPIHBP1, despite its complete absence from the capillaries of the brain, might nevertheless be expressed in the capillaries of gliomas. Using standard immunohistochemistry procedures, we documented GPIHBP1 expression in capillary endothelial cells of human gliomas and CT-2A-derived mouse gliomas. The expression of GPIHBP1 in glioma capillaries was intriguing, but the crucial issue is whether LPL would be bound to the GPIHBP1. Additional immunohistochemistry studies on mouse gliomas revealed that LPL colocalizes with GPIHBP1 on glioma capillaries, just as LPL colocalizes with GPIHBP1 in the capillaries of heart and brown adipose tissue (Young et al., 2011; Davies et al., 2010; Davies et al., 2012; Allan et al., 2017a; Fong et al., 2016; Allan et al., 2017b; Allan et al., 2016). The binding of LPL to GPIHBP1 was specific: the LPL-specific goat antibody did not detect LPL in the capillaries of gliomas in Gpihbp1–/– mice, nor did it detect any LPL in macrophages or hippocampal neurons of Lpl–/–MCK-hLPL mice. The colocalization of GPIHBP1 and LPL in the capillaries of gliomas implied that we might find evidence for TRL margination and processing in these tumors. Indeed, we observed both [2H]TRL margination along glioma capillaries and the entry of TRL-derived nutrients into glioma cells. Consistent with results of earlier studies (Goulbourne et al., 2014; He et al., 2018a), TRL margination was absent from the capillaries of normal brain, and we found no 2H enrichment in the brain parenchyma. We did, however, find very low levels of 2H enrichment in capillary endothelial cells of normal brain, perhaps as a result of the uptake of fatty acids that are derived from TRL processing in peripheral tissues. We observed consistent findings after administering [13C]fatty acids to mice by gastric gavage. In those experiments, we observed strong 13C enrichment in gliomas but no 13C enrichment in the normal brain (except for low levels of enrichment in capillary endothelial cells). After administering [13C]glucose by gavage, 13C enrichment was observed in both gliomas and normal brain. It is important to note that the [13C]fatty acids and the [13C]glucose were administered in three doses over 36 hr before harvesting tissues for NanoSIMS analyses, allowing ample time for the labeled nutrients to be utilized as fuel or to be converted into other nutrients (e.g., nonessential amino acids) (He et al., 2018a; Sidossis et al., 1995; Schneider and Potter, 1957). Thus, after administering 13C-labeled fatty acids or glucose to mice, the 13C in glioma cells was probably present in a variety of macromolecules (e.g., glucose, lipids, proteins, and nucleic acids).

Documenting GPIHBP1 and LPL in glioma capillaries, combined with the discovery that TRL-derived nutrients are taken up and utilized by glioma cells, opens a new chapter in glioma metabolism research (Figure 9). Laboratories that are interested in glioma metabolism have typically focused on the intrinsic metabolic properties of glioma cells and on how metabolic pathways in gliomas differ from those in normal brain (Lin et al., 2017; Guo et al., 2009a; Guo et al., 2009b; Guo et al., 2013; Gopal et al., 1963; Strickland and Stoll, 2017; Agnihotri and Zadeh, 2016). There have been suggestions, based on indirect observations of substrate utilization, that glioma tumors are capable of utilizing fatty acids for fuel and for anabolic processes (Lin et al., 2017; Guo et al., 2013; Mashimo et al., 2014; Ru et al., 2013; Zaidi et al., 2013). In those studies, however, the assumption was that the fatty acids probably originated from the tumor cells by de novo lipogenesis (Guo et al., 2011; Guo et al., 2009a; Guo et al., 2009b). No one, as far as we are aware, had ever considered the possibility that gliomas might be capable of taking up and utilizing nutrients from LPL-mediated intravascular processing of TRLs.

Intravascular lipolysis as a source of lipid nutrients for gliomas.

In normal brain (left panel), LPL is produced by astrocytes, neurons, oligodendrocytes, and fibroblasts. Because GPIHBP1 is not expressed in the capillaries of the brain parenchyma, we have proposed that LPL remains within the interstitial spaces of the brain (i.e., that it has an extravascular function) (Adeyo et al., 2012Young et al., 2011). In gliomas (right panel), GPIHBP1 is expressed in capillary endothelial cells, allowing GPIHBP1 to capture locally produced LPL and to shuttle it to the capillary lumen. Intravascular processing of triglyceride-rich lipoproteins in gliomas provides a source of lipid nutrients for glioma cells. HSPGs, heparan sulfate proteoglycans.

https://doi.org/10.7554/eLife.47178.033

In an ultrastructural study of human gliomas, Vaz et al. (1996) commented that the morphology of endothelial cells in gliomas resembles that of capillary endothelial cells in peripheral tissues, with euchromatin-rich nuclei, occasional fenestrations, and numerous pinocytotic vesicles within the cytoplasm. The expression of GPIHBP1 (a hallmark of capillary endothelial cells in peripheral tissues) in gliomas provides biochemical support for the notion that glioma capillaries resemble capillaries in peripheral tissues (Vaz et al., 1996). Our electron microscopy studies confirmed that the morphological features of glioma capillaries and normal brain capillaries differ substantially.

We have relatively few insights into the molecular basis for GPIHBP1 expression in glioma capillaries. One possibility is that the absence of a blood–brain barrier in glioma capillaries (Dubois et al., 2014; Wolburg et al., 2012; Liebner et al., 2000; Sage and Wilson, 1994) permits the exposure of endothelial cells to a paracrine factor that activates GPIHBP1 expression. Another possibility is that GPIHBP1 expression is stimulated by the expression of VEGF that is produced by glioma cells (Plate et al., 1994; Pietsch et al., 1997; Christov et al., 1998). In our studies, VEGF increased GPIHBP1 expression in the mouse brain endothelial cell line bEnd.3.

In the past, other laboratories have reported that glioma tumor cells can transdifferentiate into endothelial cells, thereby augmenting the vascular supply to tumors (Wang et al., 2010; Ricci-Vitiani et al., 2010; Soda et al., 2011). For example, endothelial cells in human glioblastomas were reported to harbor the same genetic alterations as the tumor cells, implying that at least some of the glioblastoma endothelial cells originate from stem cells within the tumor (Wang et al., 2010; Ricci-Vitiani et al., 2010). In another model (Soda et al., 2011), a Cre recombinase (Cre)-loxP–controlled lentiviral vector encoding activated forms of H-Ras and Akt was injected into the hippocampus of GFAP-Cre p53 mice, eliciting glioblastomas. In that model, the oncogenes were expressed in the GFAP+ cells, and the resulting tumors expressed GFP, H-Ras, and Akt and the loss of p53. Some GFP+ endothelial cells were observed in tumors, implying that these endothelial cells had originated from tumor cells. Furthermore, implanting a tumor cell line (generated from tumors induced with the same lentiviral vector) into the brain of immunocompromised mice was reported to yield tumors containing GFP+ endothelial cells. In our current studies, we observed no evidence of the differentiation of glioma cells into capillary endothelial cells. The glioma cell line that we used expressed blue fluorescent protein (BFP), but we did not find BFP expression in the capillary endothelial cells of gliomas.

Mass spectrometry–based analyses of homogenized tissue extracts from mouse gliomas and normal brain tissue, along with similar analyses of tumors from human patients, suggested differences in acetate oxidation in gliomas vs. normal brain (Mashimo et al., 2014). Although these studies of tissue extracts have been useful, they obviously cannot provide anatomical insights into metabolism. We have argued that NanoSIMS imaging studies are particularly useful when the goal is to understand metabolism at an anatomic level (cellular or subcellular) (He et al., 2018a). In the current studies, NanoSIMS imaging provided anatomic insights into glioma metabolism. For example, we observed TRL margination along the capillaries of gliomas but not in the capillaries of adjacent normal brain tissue. We also showed that the transport of TRL-derived nutrients across glioma capillaries and into glioma cells is rapid, occurring within 1 min, and that there is heterogeneity in nutrient uptake by different cells within the tumor. We found no uptake of TRL-derived nutrients by normal brain 1 or 15 min after the injection of [2H]TRLs and only very small amounts (confined to capillary endothelial cells) after 30 min. In addition, following the administration of [13C]glucose, we found more 13C enrichment in gliomas than in normal brain. As far as we are aware, our study is the first to use NanoSIMS analyses to investigate cancer metabolism in vivo. As we look to the future, we have little doubt that NanoSIMS imaging will be an important tool for understanding tumor metabolism, making it possible to investigate metabolic heterogeneity in tumor cells along with the metabolic properties of vascular cells, fibroblasts, and macrophages within the tumor. Nevertheless, it is important to point out that NanoSIMS imaging is not high-throughput, at least with the current instruments, and for that reason NanoSIMS imaging is best used (as in this study) to address discrete anatomic issues in metabolism. Examining large numbers of tumors or large numbers of mice would be difficult. Also, NanoSIMS imaging is very expensive.

Our studies have provided fresh insights into the uptake of lipid nutrients by gliomas, but many issues remain to be investigated. For example, in the current studies, we found numerous macrophages within gliomas, but we did not address differences in nutrient uptake by macrophages and glioma cells. In future studies, it should be possible to examine the uptake of TRL-derived nutrients into tumor cells, macrophages, and other immune cells within gliomas (by identifying specific cell types with 15N-labeled monoclonal antibodies or antibodies tagged with different lanthanide metals [Waentig et al., 2012; Kanje et al., 2016; Angelo et al., 2014; Keren et al., 2018]). It would also be desirable to determine whether the uptake of TRL-derived nutrients in gliomas correlates with the levels of GPIHBP1 and LPL in glioma capillaries (as quantified with LPL- and GPIHBP1-specific antibodies tagged with different lanthanide metals). Finally, it would be desirable to investigate whether the presence of GPIHBP1 and LPL in glioma capillaries could be exploited for patient care. For example, it is conceivable that fluorescently labeled GPIHBP1 antibodies or DiI-labeled TRLs could guide the surgical resection of tumors. In addition, a localized injection of GPIHBP1-specific monoclonal antibodies conjugated to chemotherapeutic agents into gliomas might be useful in targeting tumor vasculature (Schrama et al., 2006). A localized injection of gold-conjugated GPIHBP1-specific monoclonal antibodies could augment the efficacy of external beam radiotherapy (Haume et al., 2016; Hainfeld et al., 2004; Hainfeld et al., 2008).

Materials and methods

Key resources table
Reagent type
(species) or resource
DesignationSource or referenceIdentifiersAdditional
information
Genetic reagent (M. musculus)Gpihbp1–/–Beigneux et al., 2007RRID: MGI:3771172Dr. Stephen G Young (UCLA)
Genetic reagent (M. musculus)Lpl–/–MCK-hLPLLevak-Frank et al., 1995RRID: MGI:3624988Dr. Rudolph Zechner (Graz University)
Genetic reagent (M. musculus)ROSAmT/mGPdgfb-iCreT2Mathivet et al., 2017Dr. Holger Gerhardt (VIB KU-Leuven)
Cell line (M. musculus)CT-2ASeyfried et al., 1992Dr. Thomas Seyfried (Boston College)
Cell line (M. musculus)CT-2A–BFPPMID: 24658686Dr. Holger Gerhardt (VIB KU-Leuven)
Cell line (M. musculus)bEnd.3ATCCCatalog No. CRL-2299
RRID: CVCL_0170
Transfected construct (lentiviral plasmid)plenti-GLuc-IRES-EGFPTargeting SystemsCatalog No. GL-GFP
AntibodyRat monoclonal anti-mouse GPIHBP1 (11A12)Beigneux et al., 2009Dr. Stephen G Young (UCLA); IHC (10 μg/ml)
AntibodyMouse monoclonal anti-human
GPIHBP1 (RE3)
Hu et al., 2017Dr. Stephen G Young (UCLA); IHC (10 μg/ml)
AntibodyMouse monoclonal anti-human GPIHBP1 (RF4)Hu et al., 2017Dr. Stephen G Young (UCLA); IHC (10 μg/ml)
AntibodyMouse monoclonal anti-human GPIHBP1 (RG3)Hu et al., 2017Dr. Stephen G Young (UCLA); IHC (10 μg/ml)
AntibodyRabbit polyclonal anti-vWFDakoCatalog No. A0082
RRID: AB_2315602
IHC (1:200)
AntibodyGoat polyclonal anti-GFAPAbcamCatalog No. ab53554
RRID: AB_880202
IHC (1:200)
AntibodyRabbit polyclonal anti-GLUT1Millipore-SigmaCatalog No. 07–1401
RRID: AB_1587074
IHC (1:200)
AntibodyRabbit polyclonal anti-CD31AbcamCatalog No. ab28364
RRID: AB_726362
IHC (1:50)
AntibodyRat monoclonal anti-F4/80AbcamCatalog No. ab6640
RRID: AB_1140040
IHC (10 μg/ml)
AntibodyGoat polyclonal anti-mouse LPLPage et al., 2006Dr. André Bensadoun (Cornell); IHC (12 μg/ml)
AntibodyAlexa Fluor 488, 568, 647 secondariesThermoFisher ScientificIHC (1:500)
Commercial assay or kitImmPRESS Excel Staining KitVector LaboratoryCatalog No. MP-7602
Sequence-based reagentMouse Gpihbp1 primers5′-AGCAGGGACAGAGCACCTCT-3′ and 5′-AGACGAGCGTGATGCAGAAG-3′
Sequence-based reagentMouse Cd31 primers5′-AACCGTATCTCCAAAGCCAGT-3′ and 5′-CCAGACGACTGGAGGAGAACT-3′
Sequence-based reagentMouse Angpt2 primers5′-AACTCGCTCCTTCAGAAGCAGC-3′ and 5′-TTCCGCACAGTCTCTGAAGGTG-3′
Sequence-based reagentMouse Dusp5 primers5′-TCGCCTACAGACCAGCCTATGA-3′ and 5′-TGATGTGCAGGTTGGCGAGGAA-3′
Sequence-based reagentMouse Cxcr4 primers5′-GACTGGCATAGTCGGCAATGGA-3′ and 5′-CAAAGAGGAGGTCAGCCACTGA-3′
Sequence-based reagentMouse Lpl primers5′-AGGTGGACATCGGAGAACTG-3′ and 5′-TCCCTAGCACAGAAGATGACC-3′
Sequence-based reagentHuman LPL primers5′-TAGCTGGTCAGACTGGTGGA-3′ and 5′-TTCACAAATACCGCAGGTG-3′
Recombinant DNA reagentALO-D4 plasmidGay et al., 2015Dr. Arun Radhakrishnan (UT Southwestern)
Chemical compound, drugN-(3-Dimethylaminopropyl)-N′-ethylcarbodiimide hydrochloride (carbodiimide)Millipore-SigmaCatalog No. 03449
Chemical compound, drugGlutaraldehyde25% solutionElectron Microscopy SciencesCatalog No. 16220
Chemical compound, drugOsmium tetroxide 4% solutionElectron Microscopy SciencesCatalog No. 18459
Chemical compound, drugParaformaldehyde 16% solutionElectron Microscopy SciencesCatalog No. 15170
Chemical compound, drugEMbed 812Electron Microscopy SciencesCatalog No. 14120
Chemical compound, drugSodium cacodylate trihydrateElectron Microscopy SciencesCatalog No. 12300
Chemical compound, drugUranyl acetateSPI-ChemCatalog No. 02624AB
Chemical compound, drugDAPIThermoFisher ScientificCatalog No. 1306IHC (3 μg/ml)
Chemical compound, drugMouse VEGFMillipore-SigmaCatalog No. V4512
Software, algorithmLIMMARitchie et al., 2015RRID:
SCR_010943
OtherD-GLUCOSE (U-13C6, 99%)Cambridge Isotope LaboratoriesCatalog No. CLM-1396-PK
OtherMixed fatty acids (U-D, 96–98%)Cambridge Isotope LaboratoriesCatalog No. DLM-8572-PK
OtherMixed fatty acids (13C, 98%+)Cambridge Isotope LaboratoriesCatalog No. CLM-8455-PK

Immunohistochemical studies on human glioma specimens

Request a detailed protocol

Frozen surgical glioma specimens were obtained from the UCLA Department of Neurosurgery. Frozen autopsy control brain samples (frontal lobe, occipital lobe, and cerebellum) were obtained from the UCLA Section of Neuropathology. Samples were sectioned to 8 μm and placed on glass slides. All samples were fixed with 3% paraformaldehyde (PFA) in PBS/Ca/Mg and permeabilized in 0.2% Triton X-100 in PBS/Ca/Mg. Tissues were blocked with PBS/Ca/Mg containing 5% donkey serum and 0.2% bovine serum albumin (BSA) and incubated overnight at 4°C with one or more mouse monoclonal antibodies (mAbs) against human GPIHBP1 (RF4, RE3, RG3; 10 μg/ml) (Hu et al., 2017), a rabbit polyclonal antibody against von Willebrand factor (vWF) (Dako; 1:200), and a goat polyclonal antibody against human glial fibrillary acidic protein (GFAP) (Abcam; 1:500). In some experiments, recombinant soluble human GPIHBP1 (50 μg) was added to the primary antibody incubation. After washing the slides, 1-hr incubations were performed with an Alexa Fluor 647–conjugated donkey anti–mouse IgG (ThermoFisher Scientific; 1:500), an Alexa Fluor 488–conjugated donkey anti–rabbit IgG (ThermoFisher Scientific; 1:500), and an Alexa Fluor 568–conjugated donkey anti–goat IgG (ThermoFisher Scientific; 1:500). DNA was stained with 4′,6-diamidino-2-phenylindole (DAPI). Images were taken with an LSM700 confocal microscope with an Axiovert 200M stand and processed with Zen 2010 software (Zeiss).

Immunoperoxidase staining was performed with the ImmPRESS Excel Staining Kit (Vector Laboratories). Endogenous peroxidase activity was quenched with BLOXALL Blocking Solution (Vector Laboratories). After incubating sections in 10% normal horse serum, sections were incubated for 1 hr with mAb RF4 (5 μg/ml), followed by a 15-min incubation with a goat anti-mouse IgG (10 μg/ml, Vector Laboratories). Slides were then incubated for 30 min with a horseradish peroxidase–conjugated horse anti–goat IgG (ImmPRESS Excel Reagent, Vector Laboratories). After washing, the slides were incubated with ImmPACT DAB EqV (Vector Laboratories) until a color change was evident (~30 s). Finally, sections were counterstained with hematoxylin and mounted with Vectashield Mounting Media (Vector Laboratories). Images were recorded with a Nikon Eclipse E600 microscope (Plan Fluor 40×/0.50 NA or 100×/0.75 NA objectives) equipped with a DS-Fi2 camera (Nikon).

Genome dataset and gene-expression analyses

Request a detailed protocol

Cohorts for RNA-seq analysis were obtained from two databases: The Cancer Genome Atlas (TCGA) for tumor samples and Genotype-Tissue Expression (GTEx) for normal brain samples. Samples from TCGA (n = 157) and GTEx (n = 283) were processed with the TOIL pipeline as described (Vivian et al., 2017). A differential expression analysis of fatty acid metabolism genes was carried out with a linear model RNA-seq analysis software (LIMMA) (Ritchie et al., 2015). Genes were considered differentially expressed if the p-values were <0.05 and the log2 changes were >twofold. A heatmap was generated with the software R (Kolde, 2015).

Animal procedures and glioma implantation

Request a detailed protocol

Mice on a C57BL/6 background expressing both the ROSAmT/mG Cre-reporter (Muzumdar et al., 2007) and tamoxifen-inducible Pdgfb-iCreERT2 alleles (Claxton et al., 2008) were generated by breeding. In those mice, the administration of tamoxifen induces Cre recombinase expression in Pdgfb-positive cells. The recombination event results in the expression of EGFP in endothelial cells; all other cells express tdTomato. For the glioma implantation studies, mice (8–12-weeks-old) were injected intraperitoneally with tamoxifen (65 μg/g body weight, 4 injections in 2 weeks) before surgery. Mice were anesthetized with ketamine/xylazine, and a craniotomy was performed by drilling a 5‐mm hole between the lambdoid, sagittal, and coronal sutures. A blue fluorescent protein (BFP)-tagged CT‐2A glioblastoma spheroid (250-μm in diameter) (Seyfried et al., 1992; Oh et al., 2014) was injected into the cortex and sealed by cementing a glass coverslip on the skull. The CT-2A cell line was generated by Seyfried and coworkers through chemical induction with 20-methylcholanthrene in the brain of C57BL/6 mice and has been characterized extensively (Seyfried et al., 1992). In other experiments, CT-2A glioblastoma spheroids were implanted into the cortex of C57BL/6 wild-type mice and Gpihbp1–/– mice (Beigneux et al., 2007). Those procedures were performed as described previously (Stanchi et al., 2019).

Immunohistochemical studies on mouse gliomas

Request a detailed protocol

Mice harboring BFP-expressing CT-2A gliomas (Seyfried et al., 1992; Oh et al., 2014) were anesthetized with ketamine/xylazine and then injected intravenously (via the tail vein) with 100 μg of an Alexa Fluor 647–conjugated antibody against mouse GPIHBP1 (11A12) (Beigneux et al., 2009). After 1 min, the mice were perfused through the heart with 15 ml of PBS, followed by 10 ml of 2% PFA in PBS. Brain and glioma tissues were harvested and fixed overnight in 4% PFA. Tissue sections (200-μm-thick) were prepared with a vibratome. For immunofluorescence microscopy studies, the sections were fixed with 4% PFA in PBS and blocked and permeabilized in TNBT (0.1 M Tris, pH 7.4, 150 mM NaCl, 0.5% blocking reagent from Perkin Elmer, 0.5% Triton X‐100) for 4 hr at room temperature. Tissues were incubated with an antibody against GLUT1 (Millipore; 1:200) diluted in TNBT buffer overnight at 4°C, washed in TNT buffer (0.1 M Tris pH 7.4; 150 mM NaCl, 0.5% Triton X‐100) and incubated with an Alexa Fluor 488–conjugated donkey anti–rabbit IgG (ThermoFisher Scientific; 1:200). Tissues were washed and mounted in fluorescent mounting medium (Dako). Images were obtained with a Leica TCS SP8 confocal microscope.

For the analysis of tissues from mice that were not injected with anti-GPIHBP1 antibodies, tissues were embedded in OCT medium, and 10-µm sections were cut with a cryostat. Sections were fixed with 3% PFA in PBS/Ca/Mg, permeabilized with 0.2% Triton X-100 in PBS/Ca/Mg, and blocked with PBS/Ca/Mg containing 5% donkey serum and 0.2% BSA. Tissue sections were incubated overnight at 4°C with a rabbit antibody against CD31 (Abcam; 1:50), a goat antibody against mouse LPL (12 μg/ml) (Page et al., 2006), an Alexa Fluor 488–conjugated antibody against F4/80, or an Alexa Fluor 647–conjugated antibody against mouse GPIHBP1 (11A12, 10 μg/ml). After removing non-bound antibodies and washing the sections, unlabeled primary antibodies were detected with an Alexa Fluor 488–conjugated donkey anti–rabbit IgG (ThermoFisher Scientific; 1:500) or an Alexa Fluor 568–conjugated donkey ant–goat IgG (ThermoFisher Scientific; 1:500). DNA was stained with DAPI, and tissues were mounted with ProLong Gold mounting media (ThermoFisher Scientific). Images were recorded on an LSM 800 confocal microscope (Zeiss).

Immunocytochemistry studies on mouse peritoneal macrophages

Request a detailed protocol

Macrophages were collected by peritoneal lavage of C57BL/6 wild-type and Lpl–/–MCK-hLPL mice. Cells were centrifuged at 400 × g for 5 min at 4°C, washed with 5 ml of red blood cell lysing buffer (Sigma) for 5 min, washed twice with cold PBS, and then plated onto FBS-coated Petri dishes. Cells were cultured overnight in macrophage medium (Dulbecco Modified Eagle Medium with 10% FBS, 1% glutamine, and 1% sodium pyruvate). On the next day, macrophages were lifted with cold PBS containing 5 mM EDTA for 30 min at 4°C. Cells were then plated onto poly-D-lysine–coated glass coverslips (75,000 cells/coverslip) and incubated overnight in macrophage media. On the following day, the cells were washed three times for 10 min in PBS/Ca/Mg containing 0.2% BSA and then incubated with Alexa Fluor 568–labeled ALO-D4 (a modified cytolysin that binds to ‘accessible cholesterol’ in the plasma membrane) (Das et al., 2014; Das et al., 2013; Gay et al., 2015) for 2 hr at 4°C. Samples were washed three times for 1 min with PBS/Ca/Mg, fixed with 3% PFA in PBS/Ca/Mg, permeabilized with 0.2% Triton X-100 in PBS/Ca/Mg, and blocked with PBS/Ca/Mg containing 5% donkey serum and 0.2% BSA. Cells were then incubated with a goat antibody against mouse LPL (12 μg/ml) (Page et al., 2006) for 1 hr at room temperature followed by a 30-min incubation with an Alexa Fluor 647–labeled donkey anti–goat IgG (ThermoFisher Scientific; 1:500). DNA was stained with DAPI, and coverslips were mounted onto glass slides in ProLong Gold mounting media (ThermoFisher Scientific). Images were recorded with a Zeiss LSM700 confocal microscope.

Administration of [13C]fatty acids, [13C]glucose, and [2H]TRLs to mice

Request a detailed protocol

C57BL/6 mice with CT-2A gliomas (of three-week duration) were given 80 μl of [13C]fatty acids (~1 mg/μl; Cambridge Isotope Laboratories) or 80 μl of [13C]glucose (3 mg/kg body weight; Cambridge Isotope Laboratories) by oral gavage every 12 hr for 36 hr (three doses). To study TRL metabolism, mice were injected intravenously with a single bolus of [2H]TRLs (40 μg triglycerides in 100 µl) via the tail vein. The [2H]TRLs were isolated from the plasma of Gpihbp1–/– mice after administering deuterated fatty acids by gastric gavage (He et al., 2018a). After allowing the [2H]TRLs to circulate for 1 min or 30 min, the mice were perfused through the heart with 15 ml of ice-cold PBS/Ca/Mg at 3 ml/min (10 ml though the left ventricle and 5 ml through the right ventricle). Next, the mice were perfusion-fixed through the left ventricle with 10 ml of ice-cold 4% N-(3-dimethylaminopropyl)-N′-ethylcarbodiimide hydrochloride (‘carbodiimide;’ Sigma-Aldrich) (mass/vol) and 0.4% glutaraldehyde (Electron Microscopy Sciences) (vol/vol) in 0.1 M phosphate buffer. The heart, brain, and glioma tumors were collected and placed in 0.1 M phosphate buffer containing 4% carbodiimide and 2.5% glutaraldehyde for 2 hr at 4°C. Tissues were cut into 1-mm3 pieces and fixed overnight in 2.5% glutaraldehyde, 3.7% PFA, and 2.1% sucrose in 0.1 M sodium cacodylate (pH 7.4).

Preparation of tissue sections for NanoSIMS imaging and electron microscopy

Request a detailed protocol

After fixation, 1-mm3 pieces of tissue were rinsed three times (10 min each) in 0.1 M sodium cacodylate buffer (pH 7.4) and fixed with 2% OsO4 (Electron Microscopy Sciences) in 0.1 M sodium cacodylate on ice for 90 min. The samples were rinsed three times (10 min each) with distilled water and stained overnight with 2% uranyl acetate at 4°C. On the following day, the samples were rinsed three times for 10 min each with distilled water and then dehydrated with increasing amounts of ethanol (30%, 50%, 70%, 85%, 95%, and 100%; 3 × 10 min) before infiltration with Embed812 resin (Electron Microscopy Sciences) diluted in acetone (33% for 2 hr; 66% overnight; 100% for 3 hr). The samples were embedded in polyethylene molds (Electron Microscopy Sciences) with fresh resin and polymerized in a vacuum oven at 65°C for 48 hr. The polymerized blocks were then removed from the molds, trimmed, and sectioned.

For transmission electron microscopy, 65-nm sections were cut and collected on freshly glow-discharged copper grids (Ted Pella) that were coated with formvar and carbon. Sections were then stained with Reynold’s lead citrate solution for 10 min. Images were acquired with an FEI T12 transmission electron microscope set to 120 kV accelerating voltage and a Gatan 2K × 2K digital camera (Electron Imaging Center).

For NanoSIMS analyses, 500-nm sections were cut with a Leica UC6 ultramicrotome and collected on silicon wafers. Sections of tissue were coated with ~5 nm of platinum and analyzed with NanoSIMS 50L or NanoSIMS 50 instruments (CAMECA). Samples were scanned with a 16-KeV 133Cs+ beam, and secondary electrons (SEs) and secondary ions (1H, 2H, 12C, 13C, 12C14N) were collected. A 50 × 50-μm region of the section was pre-sputtered with a ∼1.2-nA beam current (primary aperture D1 = 1) to reach a dose of ~1 × 1017 ions/cm2 to remove the platinum coating and implant 133Cs+ in order to ensure a steady state of secondary ion release. A ∼40 × 40-μm region was imaged with an ∼3-pA beam current (primary aperture D1 = 2) and a dwell time of ~10 ms/pixel per frame for multiple frames. Both 256 × 256–pixel and 512 × 512–pixel images were obtained. Images were prepared using the OpenMIMS plugin in ImageJ. For image quantification, 2H/1H and 13C/12C ratios in the regions of interests were calculated with the OpenMIMS plugin and processed by GraphPad Prism 7.0.

Tumor studies in wild-type and Gpihbp1-deficient mice

Request a detailed protocol

Three-month-old C57BL/6 wild-type (five females, six males) and Gpihbp1–/– mice (six females, five males) were injected intracranially with CT-2A glioma cells stably expressing a Gaussia luciferase reporter gene (4 × 105 cells/mouse). Cells were injected 1 mm posterior and 2 mm lateral to the bregma at a depth of 2 mm. Tumor burden was monitored every three days by measuring Gaussia luciferase in the blood (Mai et al., 2017; Tannous, 2009). Mice were weighed at weekly intervals and were euthanized when they lost >20% of their body weight. After the mice were euthanized, their tumors and brains were weighed. All studies were approved by UCLA’s Animal Research Committee.

Gaussia luciferase measurements

Request a detailed protocol

To measure the levels of secreted Gaussia luciferase (sGluc), blood was obtained from the tail vein of mice and mixed with 50 mM EDTA to prevent coagulation. 5 μl of blood was transferred to a 96-well plate, and sGluc activity was measured by chemiluminescence after injecting 100 μl of 100 μM coelentarazine (Nanolight) (Mai et al., 2017; Tannous, 2009). Data were plotted as relative light units (RLU).

Quantifying mouse and human transcripts by qRT-PCR

Request a detailed protocol

C57BL/6 wild-type mice and Lpl–/–MCK-hLPL mice were anesthetized with isoflurane and perfused with PBS. Heart, brown adipose tissue (BAT), and quadricep were harvested and flash-frozen in liquid nitrogen. RNA was isolated with TRI reagent (Molecular Research), and quantitative (q)RT-PCR measurements were performed in triplicate with a 7900HT Fast real-time PCR system (Applied Biosystems). Gene expression was calculated with a comparative CT method and normalized to levels of cyclophilin A expression. Primers for mouse Gpihbp1, mouse Lpl, and human LPL are described in the 'Key Resources Table'.

VEGF treatment of brain endothelial cells

Request a detailed protocol

Mouse brain microvascular endothelial cells (bEnd.3; ATCC #CRL-2299) were plated into 6-well plates and grown in DMEM media containing 10% FBS, 1% glutamine, and 1% sodium pyruvate overnight. On the next day, cells were rinsed with PBS and incubated in medium containing recombinant mouse VEGF (100 ng/ml; Sigma) for another 24 hr. RNA was isolated with TRI reagent (Molecular Research), and qRT-PCR measurements were performed in triplicate with a 7900HT Fast real-time PCR system (Applied Biosystems). Gene expression was calculated with a comparative CT method and normalized to cyclophilin A expression. Primers for mouse Gpihbp1, Cd31, Angpt2, Cxcr4, and Dusp5 are described in the 'Key Resources Table'.

Cell lines

Request a detailed protocol

CT-2A cells were obtained originally from the Seyfried laboratory and have been extensively tested and characterized (Seyfried et al., 1992). These cells also robustly expressed GFAP. The bEnd.3 cells were obtained from ATCC with a proper ‘certificate of analysis’. All cell lines were negative for mycoplasma contamination.

Statistics

Statistical analyses of data were performed with GraphPad Prism 7.0 software. All data are shown as the means ± standard deviations. Differences were assessed using a Student’s t-test with Welch’s correction.

Study approval

Request a detailed protocol

All tissue samples from patients were obtained after informed consent and with approval from the UCLA Institutional Review Board (IRB; protocol 10–000655). Animal housing and experimental protocols were approved by UCLA’s Animal Research Committee (ARC; 2004-125-51, 2016–005) and the Institutional Animal Care and Research Advisory Committee of the KU Leuven (085/2016). The animals were housed in an AAALAC (Association for Assessment and Accreditation of Laboratory Animal Care International)-approved facility and cared for according to guidelines established by UCLA’s Animal Research Committee. The mice were fed a chow diet and housed in a barrier facility with a 12 hr light-dark cycle.

References

  1. 1
  2. 2
  3. 3
  4. 4
  5. 5
  6. 6
  7. 7
  8. 8
  9. 9
    Synthesis and regulation of lipoprotein lipase in the hippocampus
    1. O Ben-Zeev
    2. MH Doolittle
    3. N Singh
    4. CH Chang
    5. MC Schotz
    (1990)
    Journal of Lipid Research 31:1307–1313.
  10. 10
    Spinal cord of the rat contains more lipoprotein lipase than other brain regions
    1. DH Bessesen
    2. CL Richards
    3. J Etienne
    4. JW Goers
    5. RH Eckel
    (1993)
    Journal of Lipid Research 34:229–238.
  11. 11
  12. 12
  13. 13
  14. 14
  15. 15
  16. 16
  17. 17
  18. 18
  19. 19
  20. 20
  21. 21
  22. 22
  23. 23
  24. 24
  25. 25
    Localization of lipoprotein lipase mRNA in selected rat tissues
    1. IJ Goldberg
    2. DR Soprano
    3. ML Wyatt
    4. TM Vanni
    5. TG Kirchgessner
    6. MC Schotz
    (1989)
    Journal of Lipid Research 30:1569–1577.
  26. 26
  27. 27
  28. 28
  29. 29
  30. 30
  31. 31
  32. 32
  33. 33
  34. 34
  35. 35
  36. 36
  37. 37
  38. 38
  39. 39
  40. 40
  41. 41
  42. 42
  43. 43
  44. 44
  45. 45
  46. 46
  47. 47
    Pheatmap: Pretty Heatmaps, version 108
    1. R Kolde
    (2015)
    Pheatmap: Pretty Heatmaps.
  48. 48
  49. 49
  50. 50
  51. 51
  52. 52
  53. 53
  54. 54
  55. 55
  56. 56
  57. 57
  58. 58
    Augmentation of transvascular transport of macromolecules and nanoparticles in tumors using vascular endothelial growth factor
    1. WL Monsky
    2. D Fukumura
    3. T Gohongi
    4. M Ancukiewcz
    5. HA Weich
    6. VP Torchilin
    7. F Yuan
    8. RK Jain
    (1999)
    Cancer Research 59:4129–4135.
  59. 59
  60. 60
  61. 61
  62. 62
  63. 63
    Brain-type glucose transporter (GLUT-1) is selectively localized to the blood-brain barrier. Studies with quantitative western blotting and in situ hybridization
    1. WM Pardridge
    2. RJ Boado
    3. CR Farrell
    (1990)
    The Journal of Biological Chemistry 265:18035–18040.
  64. 64
  65. 65
  66. 66
  67. 67
  68. 68
  69. 69
    The blood-brain barrier: an important concept in neuroimaging
    1. MR Sage
    2. AJ Wilson
    (1994)
    AJNR. American Journal of Neuroradiology 15:601–622.
  70. 70
    Alternate pathways of glucose metabolism
    1. JH Schneider
    2. VR Potter
    (1957)
    Cancer Research 17:701–706.
  71. 71
  72. 72
  73. 73
  74. 74
  75. 75
  76. 76
  77. 77
  78. 78
  79. 79
  80. 80
  81. 81
  82. 82
  83. 83
  84. 84
  85. 85
  86. 86
    Lipoprotein lipase mRNA in neonatal and adult mouse tissues: comparison of normal and combined lipase deficiency (cld) mice assessed by in situ hybridization
    1. LK Yacoub
    2. TM Vanni
    3. IJ Goldberg
    (1990)
    Journal of Lipid Research 31:1845–1852.
  87. 87
  88. 88
    Vascular permeability and microcirculation of gliomas and mammary carcinomas transplanted in rat and mouse cranial windows
    1. F Yuan
    2. HA Salehi
    3. Y Boucher
    4. US Vasthare
    5. RF Tuma
    6. RK Jain
    (1994)
    Cancer Research 54:4564–4568.
  89. 89
  90. 90
    Differential permeability of the blood-brain barrier in experimental brain metastases produced by human neoplasms implanted into nude mice
    1. RD Zhang
    2. JE Price
    3. T Fujimaki
    4. CD Bucana
    5. IJ Fidler
    (1992)
    The American Journal of Pathology 141:1115–1124.
  91. 91

Decision letter

  1. Arun Radhakrishnan
    Reviewing Editor; University of Texas Southwestern Medical Center, United States
  2. Harry C Dietz
    Senior Editor; Howard Hughes Medical Institute and Institute of Genetic Medicine, Johns Hopkins University School of Medicine, United States
  3. Rodney Elwood Infante
    Reviewer; University of Texas Southwestern Medical Center, United States
  4. Rolf Brekken
    Reviewer; University of Texas Southwestern Medical Center, United States

In the interests of transparency, eLife includes the editorial decision letter and accompanying author responses. A lightly edited version of the letter sent to the authors after peer review is shown, indicating the most substantive concerns; minor comments are not usually included.

Thank you for submitting your article "GPIHBP1 expression in gliomas promotes utilization of lipoprotein-derived nutrients" for consideration by eLife. Your article has been reviewed by three peer reviewers, and the evaluation has been overseen by a Reviewing Editor and Harry Dietz as the Senior Editor. The following individuals involved in review of your submission have agreed to reveal their identity: Rodney Elwood Infante (Reviewer #1); Rolf Brekken (Reviewer #3).

The reviewers have discussed the reviews with one another and the Reviewing Editor has drafted this decision to help you prepare a revised submission.

Summary:

This well-presented study proposes an intriguing role for GPIHBP1-mediated LPL activity in the context of brain tumors. The authors show that GPIHBP1 is expressed in capillary endothelial cells of gliomas, but not in normal brains. Using NanoSIMS technology, at which the members of this group are experts, they show that GPIHBP1's interaction with LPL allows for gliomas to have increased triglyceride-rich lipoprotein uptake and utilization. They postulate that gliomas use this system as a source of nutrients for their growth. All reviewers agreed that the study was well-executed and that the findings were potentially significant, but all of them raised the same concern about whether GPIHBP1 was important for glioma development or progression. They welcomed a revised submission which addressed this concern, along with answers to some additional questions.

Essential revisions:

1) Please compare the tumor growth, number, and size in WT and GPIHBP1-deficient mice. If gliomas grow well in the absence of GPIHBP1, is this due to expression of Glut-1 on the vasculature in gliomas in GPIHBP1-deficient mice?

2) Provide some additional insight into the mechanism of increased GPIHBP1 expression in glioma associated vasculature. What is known about GPIHBP1 expression in other tissues and are any factors known to be associated with its expression?

3) Is it possible that the tail-vein injected Alexa Fluor 647-labeled GPIHBP1 antibody has access to gliomas and not to the normal brain tissue because of an impaired blood brain barrier in the case of the gliomas? Although there are figures where GPIHBP1 was stained after fixation, it is always done on slides with only normal tissue. An experiment similar to what is done for Glut1 in Figure 4 should be done for GPIHBP1 – slides containing both glioma and normal tissue where both GLUT1 and GPIHBP1 antibodies are used after fixation.

4) Does the expression of LPL affect the expression of GPIHBP1 in gliomas? The LPL knockout mouse model was used for IHC for normal brain tissue, but not for the glioma model (Figure 5—figure supplement 5).

[Editors' note: further revisions were requested prior to acceptance, as described below.]

Thank you for resubmitting your work entitled "GPIHBP1 expression in gliomas promotes utilization of lipoprotein-derived nutrients" for further consideration at eLife. Your revised article has been favorably evaluated by Harry Dietz (Senior Editor) and a Reviewing Editor.

The manuscript has been improved but there are some remaining issues that need to be addressed before acceptance, as outlined below:

The authors have addressed most of the points raised by the reviewers by providing new supporting data and reframing parts of the manuscript. The reviewers all agreed that this revised manuscript will be suitable for publication once one lingering concern is addressed. The authors clearly show that GPIHBP1 is not critical to glioma progression, at least in implant models (Figure 8—figure supplement 4), and provide a satisfactory explanation for why this may be so. Consequently, statements implying that GPIHBP1 expression in gliomas is important for tumor growth should be toned down or removed (e.g. in the impact statement and the last sentence of the Abstract). The authors may instead choose to highlight how the stark difference in GPIHBP1 expression in normal versus glioma capillaries could be utilized as a therapeutic tool (as they describe in the last paragraph of the Discussion).

https://doi.org/10.7554/eLife.47178.038

Author response

Summary:

This well-presented study proposes an intriguing role for GPIHBP1-mediated LPL activity in the context of brain tumors. The authors show that GPIHBP1 is expressed in capillary endothelial cells of gliomas, but not in normal brains. Using NanoSIMS technology, at which the members of this group are experts, they show that GPIHBP1's interaction with LPL allows for gliomas to have increased triglyceride-rich lipoprotein uptake and utilization. They postulate that gliomas use this system as a source of nutrients for their growth. All reviewers agreed that the study was well-executed and that the findings were potentially significant, but all of them raised the same concern about whether GPIHBP1 was important for glioma development or progression. They welcomed a revised submission which addressed this concern, along with answers to some additional questions.

We thank the reviewers and editors for their kind and generous comments about our manuscript. Over the past month, we have performed additional experiments in order to respond to the criticisms of the reviewers. In the revised manuscript, we believe that we have addressed all of the reviewers’ concerns. In addition, we added an additional experiment in which we performed NanoSIMS imaging of heart and brain of wild-type and Gpihbp1-deficient mice 15 min following the injection of [2H]TRLs (Figure 7—figure supplement 2).

Essential revisions:

1) Please compare the tumor growth, number, and size in WT and GPIHBP1-deficient mice. If gliomas grow well in the absence of GPIHBP1, is this due to expression of Glut-1 on the vasculature in gliomas in GPIHBP1-deficient mice?

We thank the reviewers for this suggestion. We injected CT-2A glioma cells that had been stably transfected with a Gaussia luciferase reporter (1, 2) into the brain of wild-type (n = 11) and Gpihbp1-deficient mice (n = 11). We monitored tumor mass by measuring the level of luciferase in the blood over a span of 3 weeks. In addition, we measured body weight at regular intervals, euthanizing mice when they lost more than 20% of their initial body weight. Finally, we weighed tumors after euthanizing the mice. We observed no statistically significant differences in tumor growth, tumor size, or survival between wild-type and Gpihbp1–/– mice. In the revised manuscript, we have included a supplemental figure showing these data (Figure 8—figure supplement 4).

The fact that we did not find significantly smaller tumors in Gpihbp1–/– mice was not surprising. Despite markedly elevated plasma triglyceride levels and inefficient processing of lipoproteins, Gpihbp1–/– mice appear healthy and have normal cardiac function. The heart normally uses large amounts of lipoprotein-derived triglycerides for fuel but manages to function normally in Gpihbp1–/– mice, likely because of the heart’s ability to compensate by utilizing glucose and/or albumin-bound fatty acids. Given those observations, we suspected that the gliomas in Gpihbp1–/– mice would be able to grow fairly normally—as a result of their ability to use glucose and/or glucose-derived metabolites as fuel. When [13C]glucose was administered to wild-type mice, NanoSIMS imaging revealed substantial 13C enrichment in tumors. Also, our paper demonstrated that glioma capillary endothelial cells express GLUT1. In the revised manuscript, we show additional studies that document GLUT1 expression in glioma capillaries of wild-type and Gpihbp1–/– mice (Figure 4—figure supplement 4).

Our studies provide strong support for the notion that GPIHBP1 expression in glioma capillaries facilitates uptake of lipoprotein-derived nutrients by tumor cells in wild-type mice. In the revised manuscript, we report that glioma tumors can grow in the absence of GPIHBP1, likely a result of the same sort of compensatory mechanisms that allow normal heart function in Gpihbp1-deficient mice.

2) Provide some additional insight into the mechanism of increased GPIHBP1 expression in glioma associated vasculature. What is known about GPIHBP1 expression in other tissues and are any factors known to be associated with its expression?

Thank you; this is an important issue. Thus far, mechanisms underlying GPIHBP1 expression in capillary endothelial cells are unclear. However, Chiu and coworkers recently reported that GPIHBP1 expression in cultured rat aortic endothelial cells can be induced by VEGF (3).

Over the past month, we have tested GPIHBP1 expression in the mouse brain endothelial cell line bEnd.3. We found that treating those cells with VEGF increased Gpihbp1 transcripts substantially. We have included those studies as a supplementary figure in the revised manuscript (Figure 3—figure supplement 3).

The induction of Gpihbp1 expression by VEGF could potentially explain why GPIHBP1 is expressed in glioma capillaries in vivo. Glioma cells are known to produce VEGF (4–6).

In terms of understanding Gpihbp1 expression, we would also draw the reviewers’ attention to our recent discovery of a Gpihbp1 enhancer element, located ∼3.6 kb upstream from exon 1 of mouse Gpihbp1 (7). Mice with a deletion of this enhancer had a ~90% decrease in Gpihbp1 expression in capillaries of the liver but only a ~50% decrease in Gpihbp1 transcripts in capillaries of heart and brown adipose tissue. The impact of the enhancer deletion on Gpihbp1 expression was insufficient to elicit hypertriglyceridemia. Our publication provided the first insight into DNA sequences regulating Gpihbp1 expression.

3) Is it possible that the tail-vein injected Alexa Fluor 647-labeled GPIHBP1 antibody has access to gliomas and not to the normal brain tissue because of an impaired blood brain barrier in the case of the gliomas? Although there are figures where GPIHBP1 was stained after fixation, it is always done on slides with only normal tissue. An experiment similar to what is done for Glut1 in Figure 4 should be done for GPIHBP1 – slides containing both glioma and normal tissue where both GLUT1 and GPIHBP1 antibodies are used after fixation.

We agree that the blood–brain barrier of gliomas can be leaky. However, there is no reason to believe that intravenously injected Alexa Fluor 647–labeled GPIHBP1 antibody would have greater access to glioma vasculature, simply because GPIHBP1 is located on the luminal surface of capillary endothelial cells. Thus, the injected antibodies would not need to cross the blood–brain barrier to bind to GPIHBP1. To further address the reviewer’s concerns, we performed staining of glioma and normal brain tissue with antibodies against GPIHBP1 and GLUT1 after tissue fixation (Figure 4—figure supplement 2). GPIHBP1 was present in endothelial cells of tumors but not in normal brain; GLUT1 was present in endothelial cells in both tumor and adjacent normal brain.

4) Does the expression of LPL affect the expression of GPIHBP1 in gliomas? The LPL knockout mouse model was used for IHC for normal brain tissue, but not for the glioma model (Figure 5—figure supplement 5).

We have no evidence that the expression of LPL influences the expression of GPIHBP1. GPIHBP1 and LPL in peripheral tissues are produced in different cell types—GPIHBP1 in endothelial cells and LPL in parenchymal cells. We no longer have access to adult LPL knockout mice. However, to address the reviewers’ concern, we examined transcript levels in archived tissue biopsies (heart, brown adipose tissue, skeletal muscle) from wild-type and Lpl–/–MCK-hLPL mice. The MCK-hLPL transgene is expressed at high levels in skeletal muscle and at much lower levels in the heart; the transgene is not expressed in brown adipose tissue. In these studies, our goal was to determine whether Gpihbp1 expression in skeletal muscle or heart would be altered by overexpression of human LPL. In heart and brown adipose tissue, Gpihbp1 expression levels were similar in wild-type mice and the Lpl–/–MCK-hLPL mice. In skeletal muscle, where hLPL levels are very high, Gpihbp1 expression was also similar in wild-type mice and the Lpl–/–MCK-hLPL mice. Thus, overexpression of LPL did not influence Gpihbp1 expression (Figure 5—figure supplement 8).

We would add that we recently documented normal plasma GPIHBP1 levels (a measurement that reflects GPIHBP1 expression in capillaries) in a human subject who was homozygous for a nonsense mutation in LPL (8), consistent with the notion that the amount of LPL expression does not affect GPIHBP1 expression in capillary endothelial cells.

In light of these findings, there is little reason to believe that the modest amounts of LPL expression in the brain would influence GPIHBP1 expression in gliomas.

References

1. Mai, W. X., L. Gosa, V. W. Daniels, L. Ta, J. E. Tsang, B. Higgins, W. B. Gilmore, N. A. Bayley, M. D. Harati, J. T. Lee, W. H. Yong, H. I. Kornblum, S. J. Bensinger, P. S. Mischel, P. N. Rao, P. M. Clark, T. F. Cloughesy, A. Letai, and D. A. Nathanson. 2017. Cytoplasmic p53 couples oncogene-driven glucose metabolism to apoptosis and is a therapeutic target in glioblastoma. Nat Med 23: 1342–1351.

2. Tannous, B. A. 2009. Gaussia luciferase reporter assay for monitoring biological processes in culture and in vivo. Nat Protoc 4: 582–591.

3. Chiu, A. P., A. Wan, N. Lal, D. Zhang, F. Wang, I. Vlodavsky, B. Hussein, and B. Rodrigues. 2016. Cardiomyocyte VEGF regulates endothelial cell GPIHBP1 to relocate lipoprotein lipase to the coronary lumen during diabetes mellitus. Arterioscler Thromb Vasc Biol 36: 145–155.

4. Plate, K. H., G. Breier, H. A. Weich, H. D. Mennel, and W. Risau. 1994. Vascular endothelial growth factor and glioma angiogenesis: coordinate induction of VEGF receptors, distribution of VEGF protein and possible in vivo regulatory mechanisms. Int J Cancer 59: 520–529.

5. Pietsch, T., M. M. Valter, H. K. Wolf, A. von Deimling, H. J. Huang, W. K. Cavenee, and O. D. Wiestler. 1997. Expression and distribution of vascular endothelial growth factor protein in human brain tumors. Acta Neuropathol 93: 109–117.

6. Christov, C., H. Adle-Biassette, C. Le Guerinel, S. Natchev, and R. K. Gherardi. 1998. Immunohistochemical detection of vascular endothelial growth factor (VEGF) in the vasculature of oligodendrogliomas. Neuropathol Appl Neurobiol 24: 29–35.

7. Allan, C. M., P. J. Heizer, Y. Tu, N. P. Sandoval, R. S. Jung, J. E. Morales, E. Sajti, T. D. Troutman, T. L. Saunders, D. A. Cusanovich, A. P. Beigneux, C. E. Romanoski, L. G. Fong, and S. G. Young. 2019. An upstream enhancer regulates Gpihbp1 expression in a tissue-specific manner. J Lipid Res 60: 869–879.

8. Beigneux, A. P., K. Miyashita, M. Ploug, D. J. Blom, M. Ai, M. F. Linton, W. Khovidhunkit, R. Dufour, A. Garg, M. A. McMahon, C. R. Pullinger, N. P. Sandoval, X. Hu, C. M. Allan, M. Larsson, T. Machida, M. Murakami, K. Reue, P. Tontonoz, I. J. Goldberg, P. Moulin, S. Charrière, L. G. Fong, K. Nakajima, and S. G. Young. 2017. Autoantibodies against GPIHBP1 as a cause of hypertriglyceridemia. N Engl J Med 376: 1647–1658.

[Editors' note: further revisions were requested prior to acceptance, as described below.]

The manuscript has been improved but there are some remaining issues that need to be addressed before acceptance, as outlined below:

The authors have addressed most of the points raised by the reviewers by providing new supporting data and reframing parts of the manuscript. The reviewers all agreed that this revised manuscript will be suitable for publication once one lingering concern is addressed. The authors clearly show that GPIHBP1 is not critical to glioma progression, at least in implant models (Figure 8—figure supplement 4), and provide a satisfactory explanation for why this may be so. Consequently, statements implying that GPIHBP1 expression in gliomas is important for tumor growth should be toned down or removed (e.g. in the impact statement and the last sentence of the Abstract). The authors may instead choose to highlight how the stark difference in GPIHBP1 expression in normal versus glioma capillaries could be utilized as a therapeutic tool (as they describe in the last paragraph of the Discussion).

In the current revision, we have removed all statements implying the possibility that GPIHBP1 could be relevant to the growth of gliomas, including in the “impact statement.” Instead, we simply state that GPIHBP1 expression facilitates TRL processing in gliomas, providing a source of lipid nutrients for glioma cells.

https://doi.org/10.7554/eLife.47178.039

Article and author information

Author details

  1. Xuchen Hu

    Department of Medicine, David Geffen School of Medicine, University of California, Los Angeles, Los Angeles, United States
    Contribution
    Conceptualization, Formal analysis, Validation, Investigation, Visualization, Writing—original draft, Writing—review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-0944-624X
  2. Ken Matsumoto

    VIB-KU Leuven Center for Cancer Biology (CCB), Leuven, Belgium
    Contribution
    Investigation, Visualization, Writing—review and editing
    Competing interests
    No competing interests declared
  3. Rachel S Jung

    Department of Medicine, David Geffen School of Medicine, University of California, Los Angeles, Los Angeles, United States
    Contribution
    Investigation
    Competing interests
    No competing interests declared
  4. Thomas A Weston

    Department of Medicine, David Geffen School of Medicine, University of California, Los Angeles, Los Angeles, United States
    Contribution
    Investigation, Writing—review and editing
    Competing interests
    No competing interests declared
  5. Patrick J Heizer

    Department of Medicine, David Geffen School of Medicine, University of California, Los Angeles, Los Angeles, United States
    Contribution
    Investigation
    Competing interests
    No competing interests declared
  6. Cuiwen He

    Department of Medicine, David Geffen School of Medicine, University of California, Los Angeles, Los Angeles, United States
    Contribution
    Investigation
    Competing interests
    No competing interests declared
  7. Norma P Sandoval

    Department of Medicine, David Geffen School of Medicine, University of California, Los Angeles, Los Angeles, United States
    Contribution
    Investigation
    Competing interests
    No competing interests declared
  8. Christopher M Allan

    Department of Medicine, David Geffen School of Medicine, University of California, Los Angeles, Los Angeles, United States
    Contribution
    Investigation, Writing—review and editing
    Competing interests
    No competing interests declared
  9. Yiping Tu

    Department of Medicine, David Geffen School of Medicine, University of California, Los Angeles, Los Angeles, United States
    Contribution
    Investigation
    Competing interests
    No competing interests declared
  10. Harry V Vinters

    Department of Pathology and Laboratory Medicine, David Geffen School of Medicine, University of California, Los Angeles, Los Angeles, United States
    Contribution
    Resources
    Competing interests
    No competing interests declared
  11. Linda M Liau

    1. Department of Neurosurgery, David Geffen School of Medicine, University of California, Los Angeles, Los Angeles, United States
    2. Jonsson Comprehensive Cancer Center (JCCC), David Geffen School of Medicine, University of California, Los Angeles, Los Angeles, United States
    Contribution
    Resources
    Competing interests
    No competing interests declared
  12. Rochelle M Ellison

    Department of Medicine, David Geffen School of Medicine, University of California, Los Angeles, Los Angeles, United States
    Contribution
    Investigation
    Competing interests
    No competing interests declared
  13. Jazmin E Morales

    Department of Medicine, David Geffen School of Medicine, University of California, Los Angeles, Los Angeles, United States
    Contribution
    Investigation
    Competing interests
    No competing interests declared
  14. Lynn J Baufeld

    1. Department of Molecular and Medical Pharmacology, David Geffen School of Medicine, University of California, Los Angeles, Los Angeles, United States
    2. Ahmanson Translational Imaging Division, David Geffen School of Medicine, University of California, Los Angeles, Los Angeles, United States
    Contribution
    Investigation
    Competing interests
    No competing interests declared
  15. Nicholas A Bayley

    1. Department of Molecular and Medical Pharmacology, David Geffen School of Medicine, University of California, Los Angeles, Los Angeles, United States
    2. Ahmanson Translational Imaging Division, David Geffen School of Medicine, University of California, Los Angeles, Los Angeles, United States
    Contribution
    Data curation, Formal analysis
    Competing interests
    No competing interests declared
  16. Liqun He

    Department of Immunology, Genetics and Pathology, Rudbeck Laboratory, Uppsala University, Uppsala, Sweden
    Contribution
    Formal analysis
    Competing interests
    No competing interests declared
  17. Christer Betsholtz

    1. Department of Immunology, Genetics and Pathology, Rudbeck Laboratory, Uppsala University, Uppsala, Sweden
    2. Integrated Cardio Metabolic Centre (ICMC), Karolinska Institutet, Huddinge, Sweden
    Contribution
    Supervision, Writing—review and editing
    Competing interests
    No competing interests declared
  18. Anne P Beigneux

    Department of Medicine, David Geffen School of Medicine, University of California, Los Angeles, Los Angeles, United States
    Contribution
    Supervision, Writing—review and editing
    Competing interests
    No competing interests declared
  19. David A Nathanson

    1. Department of Molecular and Medical Pharmacology, David Geffen School of Medicine, University of California, Los Angeles, Los Angeles, United States
    2. Ahmanson Translational Imaging Division, David Geffen School of Medicine, University of California, Los Angeles, Los Angeles, United States
    Contribution
    Resources, Supervision, Writing—review and editing
    Competing interests
    No competing interests declared
  20. Holger Gerhardt

    1. VIB-KU Leuven Center for Cancer Biology (CCB), Leuven, Belgium
    2. Max Delbrück Center for Molecular Medicine, Berlin, Germany
    Contribution
    Resources, Supervision, Funding acquisition, Writing—review and editing
    Competing interests
    Reviewing editor, eLife
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-3030-0384
  21. Stephen G Young

    1. Department of Medicine, David Geffen School of Medicine, University of California, Los Angeles, Los Angeles, United States
    2. Department of Human Genetics, David Geffen School of Medicine, University of California, Los Angeles, Los Angeles, United States
    Contribution
    Conceptualization, Resources, Formal analysis, Supervision, Funding acquisition, Methodology, Writing—original draft, Project administration, Writing—review and editing
    For correspondence
    sgyoung@mednet.ucla.edu
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0001-7270-3176
  22. Loren G Fong

    Department of Medicine, David Geffen School of Medicine, University of California, Los Angeles, Los Angeles, United States
    Contribution
    Conceptualization, Resources, Formal analysis, Supervision, Methodology, Writing—original draft, Project administration, Writing—review and editing
    For correspondence
    lfong@mednet.ucla.edu
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-4465-5290
  23. Haibo Jiang

    1. Department of Medicine, David Geffen School of Medicine, University of California, Los Angeles, Los Angeles, United States
    2. School of Molecular Sciences, University of Western Australia, Perth, Australia
    Contribution
    Resources, Formal analysis, Investigation, Writing—review and editing
    For correspondence
    haibo.jiang@uwa.edu.au
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-2384-4826

Funding

National Heart, Lung, and Blood Institute (HL090553)

  • Stephen G Young

National Heart, Lung, and Blood Institute (HL087228)

  • Stephen G Young

National Heart, Lung, and Blood Institute (HL125335)

  • Stephen G Young

Fondation Leducq (12CVD04)

  • Stephen G Young

National Institutes of Health (Ruth L Kirschstein National Research Service Award (T32HL69766))

  • Xuchen Hu

National Institute of General Medical Sciences (GM008042)

  • Xuchen Hu

National Cancer Institute (Brain Tumor SPORE grant (P50-CA211015))

  • Linda M Liau

Stichting Tegen Kanker (2012‐181)

  • Holger Gerhardt

Stichting Tegen Kanker (2018-074)

  • Holger Gerhardt

Japan Society for the Promotion of Science (Postdoctoral Fellowship)

  • Ken Matsumoto

University of California, Los Angeles (Medical Scientist Training Program)

  • Xuchen Hu

Australian Research Council (Discovery Early Career Researcher Award)

  • Haibo Jiang

Cancer Council Western Australia (Early Career Investigator Grant)

  • Haibo Jiang

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

This work was supported by grants from the NHLBI (HL090553, HL087228, HL125335), a Transatlantic Network Grant from the Leducq Foundation (12CVD04), and the Belgian Cancer Foundation (Stichting Tegen Kanker, grant 2012‐181, 2018–074). Xuchen Hu was supported by a Ruth L Kirschstein National Research Service Award (T32HL69766) and by the UCLA Medical Scientist Training Program. Ken Matsumoto was supported by a Japan Society for the Promotion of Science Postdoctoral Fellowship. Linda Liau was supported by a NCI Brain Tumor SPORE grant (P50-CA211015). Haibo Jiang was supported by an Australian Research Council Discovery Early Career Researcher Award and a Cancer Council Western Australia Early Career Investigator Grant. We thank Andre Bensadoun (Cornell University) for goat anti-mouse LPL antibodies; we also thank Anna Nowak (University of Western Australia) and the late Ben Barres (Stanford University) for helpful discussions.

Ethics

Human subjects: All tissue samples from patients were obtained after informed consent and with approval from the UCLA Institutional Review Board (IRB; protocol 10-000655).

Animal experimentation: Animal housing and experimental protocols were approved by UCLA's Animal Research Committee (ARC; 2004-125-51, 2016-005) and the Institutional Animal Care and Research Advisory Committee of the KU Leuven (085/2016). The animals were housed in an AAALAC (Association for Assessment and Accreditation of Laboratory Animal Care International)-approved facility and cared for according to guidelines established by UCLA's Animal Research Committee.

Senior Editor

  1. Harry C Dietz, Howard Hughes Medical Institute and Institute of Genetic Medicine, Johns Hopkins University School of Medicine, United States

Reviewing Editor

  1. Arun Radhakrishnan, University of Texas Southwestern Medical Center, United States

Reviewers

  1. Rodney Elwood Infante, University of Texas Southwestern Medical Center, United States
  2. Rolf Brekken, University of Texas Southwestern Medical Center, United States

Publication history

  1. Received: March 27, 2019
  2. Accepted: June 5, 2019
  3. Accepted Manuscript published: June 6, 2019 (version 1)
  4. Version of Record published: June 26, 2019 (version 2)

Copyright

© 2019, Hu et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 683
    Page views
  • 128
    Downloads
  • 0
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Download citations (links to download the citations from this article in formats compatible with various reference manager tools)

Open citations (links to open the citations from this article in various online reference manager services)

Further reading

    1. Biochemistry and Chemical Biology
    2. Cell Biology
    Timo Vögtle et al.
    Research Article
    1. Biochemistry and Chemical Biology
    2. Structural Biology and Molecular Biophysics
    Kenta Yamamoto et al.
    Research Article