Cribaro et al. acta neuropathol commun (2021) 9:24 https://doi.org/10.1186/s40478-020-01115-0 METHODOLOGY ARTICLE Open Access Three‑dimensional vascular microenvironment landscape in human glioblastoma George P. Cribaro1, Elena Saavedra‑López1, Leire Romarate1, Izaskun Mitxitorena1, Laura R. Díaz1, Paola V. Casanova1, Meritxell Roig‑Martínez1, José M. Gallego2, Ana Perez‑Vallés2 and Carlos Barcia1* Abstract The cellular complexity of glioblastoma microenvironments is still poorly understood. In-depth, cell-resolution tissue analyses of human material are rare but highly necessary to understand the biology of this deadly tumor. Here we present a unique 3D visualization revealing the cellular composition of human GBM in detail and considering its criti‑ cal association with the neo-vascular niche. Our images show a complex vascular map of human 3D biopsies with increased vascular heterogeneity and altered spatial relationship with astrocytes or glioma-cell counterparts. High- resolution analysis of the structural layers of the blood brain barrier showed a multilayered fenestration of endothe‑ lium and basement membrane. Careful examination of T cell position and migration relative to vascular walls revealed increased infiltration corresponding with tumor proliferation. In addition, the analysis of the myeloid landscape not only showed a volumetric increase in glioma-associated microglia and macrophages relative to GBM proliferation but also revealed distinct phenotypes in tumor nest and stroma. Images and data sets are available on demand as a resource for public access. Keywords: Glioblastoma, Blood vessels, Microglia, Macrophages, Astrocytes, T cells, Endothelium, Basement membrane, Tumor microenvironments Introduction Importantly, clinical histopathological examinations Despite the recent progress of successful immunothera- are generally based on thin, paraffin-embedded or cryo- peutic options for many types of cancers, glioblastoma preserved sections with simultaneous staining of one or (GBM) remains incurable [9, 16]. The aggressive nature two markers, only allowing a two-dimensional visualiza- of this brain tumor causes a rapid and severe spread of tion for rapid and operational analysis. In addition, in- malignant cells in brain tissue with irreparable loss. The depth studies of the GBM microenvironment performed extremely complex cellular structure of GBM that ena- in human tissue are limited and many conclusions are bles this invasiveness is still poorly understood and drawn from animal models or in vitro simulations [26, deserves a detailed microanatomical study to compre- 34]. Therefore, high-resolution microstructural studies of hend its biology. patient samples are critical to understand cellular com- position as well as how the tumor microenvironment is established in its real scenario. To do this, it is crucial to *Correspondence: carlos.barcia@uab.es consider the three-dimensional architectural preserva- 1 Department of Biochemistry and Molecular Biology, Institut de tion of the human tissue including, not only the paren- Neurociències, School of Medicine, Lab M2‑107, Office M2‑117, chyma but also the vascular niche so as to understand the Universitat Autònoma de Barcelona, 08193 Bellaterra, Barcelona, Spain Full list of author information is available at the end of the article full complexity of the tumor. © The Author(s) 2021. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creat​iveco​mmons​.org/licen​ses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creat​iveco​ mmons​.org/publi​cdoma​in/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data. Cribaro et al. acta neuropathol commun (2021) 9:24 Page 2 of 20 Thorough studies in GBM mouse models have shown tumors were excised and immediately fixed, cryopro- that the association of glioma cells and blood vessels is tected and sectioned in a cryostat (Leica Microsystems, essential for tumor-mass spread [10, 32], thus, special Wetzlar, Germany) thus preventing any significant tis- attention should be paid to analyzing the cellular prop- sue autolysis and preserving intact structure, an essential erties of vascular networks in GBM patient samples. requirement for the type of analysis described herein. Studying the elements of the blood brain barrier (BBB) Tumors were graded in-hospital and the Ki67 labeling or blood tumor barrier (BTB) in human samples is of index was assigned by the pathology department. All critical importance, as these may be relevant to design- tumors studied and imaged as part of this study were ing new therapeutic approaches [6]. The integrity of the grade IV glioma or GBM with Ki67 indices ranging from structural walls of the BBB could well have an impact on 10 to 50% (Additional file 1: Supplementary Table S1) treatment-strategy effectiveness [1], as the degree of BTB to compare with normal cortex (CTX). For nontumoral permeability could interfere with blood cell infiltration CTX, we used a normal brain from an individual who and contribute to the cellular conformation of the tumor. died from nonpathological causes. In this case, we were In particular, the GBM stroma, as opposed to the tumor able to select the frontal CTX, paralleling the area where nest, appears to be composed mainly of microglia and GBM are mostly located, and cut the tissue in thick sec- macrophages [7], but whether this occupies a determined tions following the same protocol as for tumor samples. spatial location within the tumor mass remains obscure Working with thick sections was meticulous. It and needs to be determined. required great care and significantly longer working Actually, both the extent to which the BTB is structur- times to preserve intact tissue structure, achieve full stain ally altered [30] and how this may hamper immune cell penetration and coverage and later acquire sufficient infiltration are still debated [19]. Therefore, further stud- image data. The method followed and described herein ies are necessary to understand the structural and cellular necessitated a continual balance between feasible and relationships of BTB features in human glioma [6] and optimal processing times to ensure quality results. how these properties may correlate with tumor aggres- siveness and patient prognosis. Multilabeling immunohistochemistry We present here a unique, multicolor, high-resolution The 60-μm tumor sections were cut serially through 3D view of human GBM microenvironments. We provide the entire sample, and multilabel immunofluorescence an atlas-like set of images and data using a highly detailed staining was performed by free-floating using various procedure that enables visualization of the cellular com- combinations of primary antibodies (Additional file 1: position of human tumor microenvironments in relation Supplementary Table S2) after testing and selecting opti- with neovasculature and the corresponding analysis of mal antibodies for each marker. numerical and volumetric parameters. In analyzing the Secondary fluorescent antibodies (Additional file 1: BBB or BTB structural walls, we display a vascular map of Supplementary Table S2) diluted in 1% horse serum GBM as well as its altered relationship with protoplasmic (HS) were used accordingly with the host of the primary astrocytes and analyze the fenestration of both endothe- antibodies and 4,6-diamidino-2-phenylindole (DAPI) lium and basement membrane. Furthermore, we examine (1:1000, Invitrogen, Carlsbad, CA, USA) was used to the permeability to lymphocytes and analyze the mye- stain nuclei. Immunohistochemical detection methods loid landscape, categorizing a distinct histocytometric were previously optimized to achieve full and homog- glioma-associated microglia and macrophages (GAMM) enous antibody penetration of the whole sections while profile for locating stroma and tumor nest. substantially sparing tissue damage and keeping struc- tural relationships intact, something absolutely critical to Materials and methods perform the analyses undertaken in this research. Human GBM samples The full immunohistochemistry (IHC) protocol In the present study we analyzed three-dimensional biop- spanned four days and comprised antigen retrieval in sies (3DBx) to spatially explore the tissue and overcome citrate buffer, blocking promiscuous/unspecific binding the limitations of regular 5–10 μm thin sections normally sites with HS, marking target structures with a combi- used for neuropathology and diagnosis. To do so, a spe- nation of primary and highly crossed absorbed second- cial protocol was set up to obtain optimally preserved ary antibodies and counterstaining nuclei with DAPI. volumetric sections. Human GBM samples were excised Depending on the structures or cell types to be marked surgically at the Valencia General Hospital. Tumors were and bearing in mind the need to avoid duplicating anti- removed and prepared following the protocol previously body species in an individual multistaining procedure, described by us [2] and approved by ethics committees the following antibody combinations were used: 1, COL- of the institutions involved. It should be emphasized that IV/GFAP/CD31/DAPI; 2, CD31/CD3/GFAP/DAPI; 3, Cribaro et al. acta neuropathol commun (2021) 9:24 Page 3 of 20 COL-IV/MHCII/GFAP/DAPI; and 4, CD31/Iba1/GFAP/ USA; IllucidaFX, Los Angeles, USA). Full descriptions DAPI. of methodology and parameters studied are given in the Incubation for 48 h with primary antibodies was fol- subsequent sections. lowed by 24 h of incubation with the appropriate second- ary antibodies. After washing, sections were incubated Parameter characterization and quantification with DAPI in solution for 30 min. The sections were The parameters measured included total cellularity; total (ProLong® washed again, mounted on non-gelatinized glass slides units (indivisible cell-size particles expressing a given and cover-slipped with antifading reagents ­ marker), volumetric surface area (external surface area Gold antifade, Invitrogen; Carlsbad, CA, USA and Fluo- of an isosurface) and volume (occupied by an isosurface) prep reagent, bioMérieux SA, F-69280 Marcy l’Etoile, in tissue blocks. Specifically measured were volumetric France) and once cured, examined by confocal micro- surface area and volume of GFAP; volumetric surface scope (Leica TCS SP5 or Zeiss LSM 700, at the Universi- area and volume of COL-IV, blood vessel calibers; ves- tat Autònoma de Barcelona Microscopy Service and the sel branching types and totals; vessel wall fluorescence Institut de Neurociències, respectively). variations; volumetric surface areas and volumes of CD31-positive endothelial cells; overlapping of CD31 and Confocal imaging COL-IV surfaces; Iba-1-positive units, volumetric sur- Images were acquired using a laser scanning confo- face areas and volumes; MHCII-positive units, volumet- cal microscope (Zeiss LSM 700) with four laser lines ric surface areas and volumes and overlapping of MHCII (405 nm, 488 nm, 555 nm and 639 nm), three objectives and GFAP structures; and total T cells (expressing CD3) (Plan-Apochromat 20x/0.8, Plan-Apochromat 40x/1.3 oil and their contact with or distance from blood vessel DIC and Plan-Apochromat 63x/1.4 oil DIC) and appro- endothelium. priate image acquisition software (ZEN2010, Carl Zeiss AG, Germany). Exploration and quantification methodologies Images for large photo mosaics were acquired with a In line with the various research objectives of this confocal microscope with motorized stage (Leica TCS study to closely examine the topographical and mor- SP5), five laser lines (458, 476, 488, 496 and 514 nm), six phological relationships of human GBM microenvi- objectives (Plan-Apochromat 10x/0.4 CS, Plan-Apochro- ronment components in 3D and compare them across mat 20x/0.7 CS, Plan-Apochromat 40x/1.25-0.75 oil CS, tumors with normal cortical tissue, we adopted the fol- Plan-Apochromat 63x/1.4-0.6 oil CS, Plan-Apochro- lowing strategy. First, we examined overall cellularity, mat 63x/1.4-0.6 glycerol Corr CS and HC PL Fluorar which is known to increase in GBM. We then centered 50x/0.80) and LAS AF software (Leica Microsystems on parameters described above for GFAP expression, CMS GmbH, Germany). as this is a canonical marker for GBM diagnosis. Next, given the fundamental role of vascular networks in GBM Image processing and close relationship with GFAP cells, we performed After acquisition, 3D stacks were processed for deconvo- a detailed analysis of BVs including separate COL-IV lution. This image processing technique was performed and CD31 parameters and surface overlap and blood to restore the imaged objects usually degraded by blur- vessel characteristics listed above. Microglia and mac- ring and noise, artifacts inherent to any image acquisition rophage populations were then examined for Iba-1 and system. The degree of blurring of a single, sub-resolution MHCII expression and contiguous MHCII and GFAP point-like object was considered as a measure of the opti- surfaces. Total T cells were counted, and T cell distribu- cal system’s quality and the blurred 3D image of this sin- tion was calculated along with distance from nearest BVs. gle point light source was defined by the Point Spread As a final component of our overall strategy, individual Function (PSF) [11]. Therefore, to counteract the effects parameters were analyzed statistically, and Pearson cor- of this blurring and to optimize image quality and sub- relation coefficients were calculated comparing individ- sequent quantifications, all images were processed using ual parameters among themselves. Details of each step in image processing software package (Huygens Profes- this strategy are presented in the following sections. sional, Scientific Volume Imaging b.v., Hilversum, The Netherlands) or image deconvolution software (Auto- Cellularity Quant X3, Bitplane AG, Zurich, Switzerland). Dense cellularity is characteristic of gliomas and measur- Once images were processed to maximize quality, the ing this parameter in the current context served a two- relevant parameters were measured using image visuali- fold purpose: to corroborate the Ki67% indices assigned zation and analysis software (Imaris 8.3.4, Bitplane AG, to the tumors when excised and, depending on the con- Zurich, Switzerland; Fiji ImageJ, Bethesda, Maryland, cordance of our cell counts with the hospital-designated Cribaro et al. acta neuropathol commun (2021) 9:24 Page 4 of 20 Ki67% index, to indirectly validate our analytic pro- single and multiple branches in each image were counted cedure. It also permitted us to examine how total cell using unbiased stereological criteria [21]. count varied with the Ki67% index and other param- eters. Spherical isosurfaces coinciding with individual Vessel wall integrity cells identified in each sample as DAPI-stained nuclei To quantify the vessel surface disruption, irregulari- were generated to estimate the total cell numbers (Addi- ties and fenestrations, we measured variations in fluo- tional file 2: Supplementary Figure S1) (Imaris, “Spots” rescence intensity along small segments of vessel walls module). assuming that vessel wall continuity and integrity would be reflected indirectly in this variation. Single 1-μm opti- GFAP area and density cal slices, were exported from confocal image stacks GFAP volumetric surface area and contained volume and small segments of five vessel walls in each sampling were measured in confocal images by importing them image were measured by tracing 40-μm lines along blood into 3D visualization software and creating 3D isosur- vessel walls using ImageJ software to generate a series of faces (Fig. 1) (Imaris, “Surfaces” module). values for fluorescence levels along the respective lines (ImageJ, “Plot Profile” function). Individual line lengths varied slightly but the profiles were harmonized after Collagen IV area and density exporting by only using the fluorescence intensity values We quantified the amount of COL-IV in confocal images corresponding to the first 27 microns measured for each by importing them into 3D visualization software and sample (Fig. 3). creating 3D isosurfaces based on detected fluorescence (Fig. 1) (Imaris, “Surfaces” module). Statistics for total CD31 area and density surface area, total volume were generated by the software To quantify any changes in CD31 expression relative to and exported as CSV files for calculations and statistical increasing tumorigenicity, analogous to the basement analysis. membrane analysis, we captured confocal images and imported them into image analysis software (Imaris, Vessel diameter “Surfaces” module) to create 3D isosurfaces and then To assess vessel diameters, 3D confocal images of tis- exported the isosurface areas and volumes for calcula- sue blocks were imported into image analysis software tions and statistical analyses. and the widths of all visible vessels in each image were measured (Imaris, “Measurement Points” tool) (Fig. 2). Surface overlap between CD31 and collagen IV Measurement data were exported to additional software In intact blood vessels, the vascular basement membrane to calculate statistical parameters (CSV file in MATLAB is lined with endothelial cells. To examine this, images [Natick, Massachusetts, USA]). were acquired from samples immunostained for both COL-IV and CD31 and imported into image analysis Branching software (Imaris or IllucidaFX). Three-dimensional iso- To quantify and categorize branching, vessel ramifica- surfaces were created from the COL-IV and CD31 chan- tions were observed three dimensionally and divided into nels and a mask was created separately for each channel primary (y- or t-shaped single bifurcations), secondary from the respective 3D isosurface to eliminate artifacts (w-shaped with two ramifications) and tertiary (irregu- from the original confocal channels. A new channel was lar multi-branching with three or more ramifications) then generated from the overlapping structures of the (Fig. 2). Confocal images were imported into the appro- masked COL-IV and CD31 channels in the confocal priate software (Fiji distribution of ImageJ) and the total (See figure on next page.) Fig. 1 Vascular map and GFAP’s altered spatial relationship in GBM. a GFAP distribution of characteristic protoplasmic astrocytes in normal CTX, and glioma cells in GBM-affected cases, displayed on original confocal images (top panel) and the corresponding rendered isosurfaces (bottom panel). b GFAP-expressing cells show a massive increase in GBM as reflected in volumetric surface area analysis relative to Ki67% and all tumor cases grouped. c Visualization of correlative vasculature associated with ­GFAP+ cells in normal brain and GBM. Merged views of GFAP from panel A, and blood vessels, evidenced by COL-IV (green) are shown. d 3D rendering of GFAP and COL-IV in normal CTX and GBM highlight the differences between normal and GBM blood vessels in relation with surrounding GFAP-expressing cells. e Detailed, higher magnification 3D renderings show altered spatial relationship of basement membrane and GFAP. f Hypermosaic of cortical sample of a normal brain CTX and GBM, and high-resolution mosaics from white inset. g Confocal alpha blending, 3D rendering of high-magnification images and skeletonization (including branching points color-coded by size) reveal dramatic changes in blood vessel distribution and architecture relative to increasing Ki67 index in GBM Cribaro et al. acta neuropathol commun (2021) 9:24 Page 5 of 20 Cribaro et al. acta neuropathol commun (2021) 9:24 Page 6 of 20 image stack and a 3D isosurface of this overlapping area Data analysis was created. From the latter, total overlapping areas were Statistical analyses extracted. Results for cellularity, vessel characteristics (diameters, branching and density) and GFAP and COL-IV volu- Glioma associated microglia and macrophages metric surface areas and volumes derive from samples Iba‑1 immunostained for COL-IV and GFAP and counter- Iba-1 volumetric surface area, units and volume were stained with DAPI with 50 data points (10 tissue blocks measured in confocal image stacks by importing them for normal CTX and 10 tissue blocks each for Ki67 into image analysis software and creating 3D isosurfaces indices 10, 20, 30 and 50%). Basement membrane con- (Imaris, “Surfaces” module). tinuity results are drawn from the same sample pool of tissue blocks with 50 data sets with each data set com- MHCII prising five separate sets of over 25 luminance meas- Human GBM biopsy samples were immunostained urement points. marking COL-IV, GFAP, MHCII and nuclei. MHCII vol- Results for CD31 surface area and multiple compari- umetric surface area, units and volume were then meas- sons between cellularity, vessel continuity, COL-IV ured in confocal image stacks by importing them into and CD31 surface areas, colocalization and GFAP sur- image analysis software (Bitplane’s Imaris or IllucidaFX) face area, derive from a sample set stained for COL-IV, and creating 3D isosurfaces (“Surface” module). A sepa- CD31 and GFAP and counterstained with DAPI with 24 rate colocalization channel was created and isosurfaced total data points for each variable, six for normal tis- to measure the total surface area of overlapping MHCII/ sue and 18 for tumor tissue blocks (six each for Ki67 GFAP areas. In addition, MHCII-rich and GFAP-rich indices of 10, 20 and 30%). Again, basement membrane areas were isolated by selecting regions of interest (ROI) continuity results are drawn from the same sample pool and the area and volume of COL-IV expressed in each with 24 data sets (six control tissue blocks and six each ROI was measured. Vessel diameters were measured sep- for Ki67 indices of 10, 20 and 30%) with each data set arately for each ROI. comprising five separate sets of 34 luminance measure- ment points. T cells: CD3 Total T cell counts, and T cell extravasation and migra- CD3′s antigen specificity and its universal expression at tion data were drawn from a sample pool immunostained all T cell developmental stages make it an ideal immu- for CD31, CD3 and GFAP and counterstained with DAPI. nological marker to detect T cells in tissue sections. To Thirty data sets were generated and analyzed (six sepa- explore T-cell populations in GBM, tumor tissue sam- rate tissue block areas for Ki67 10% and 12 areas each for ples were immunostained for CD3 and acquired confocal 20 and 30%). images were imported into image analysis software (Ima- Data for Iba-1 and MHCII derive from two sam- ris). The objective was to measure the extent of extrava- ple groups (each with six normal and 24 tissue blocks; sation (i.e., movement of T cells away from vessels) and tumors had Ki67% indices of 10, 20 and 30%). Iba-1 sam- see if that bore any relation to membrane degradation in ples were immunostained for CD31, Iba-1 and GFAP and severe tumors. Total T cells in each image were counted counterstained with DAPI. MHCII samples were immu- using image analysis software (Imaris, “Spots” module) nostained for COL-IV, MHCII and GFAP and counter- and the distances from or contacts with the nearest blood stained with DAPI. vessels were measured in three-dimensional visualization For all statistical comparisons, we used an unpaired, (Imaris, “Measurement Points” tool). two-tailed student t-test with Welch’s correction, or (See figure on next page.) Fig. 2 Volumetric increase of GBM vasculature and heterogeneity is due to increased calibers and primary branching. a Creation of COL-IV isosurfaces in normal CTX and GBM. b Volumetric surface increases in GBM. c Measurements of diameter of COL-IV+ vessels in normal CTX (N. CTX) and GBM. d Quantifications of diameters in GBM compared to normal CTX. e, f and g Alpha blending of primary, secondary and tertiary branching points. h Percentage of branching types (single and multiple) in normal CTX and GBM. i Representative image of isosurface rendering and quantification of branching points. j Quantification of primary, secondary and tertiary branching points. k Representative images of branching points in GBM from primary to tertiary (top panel), and looping, intussusception and silent collateral vessels (bottom panel, from left to right respectively). l Quantification of silent collateral BV in normal and GBM Cribaro et al. acta neuropathol commun (2021) 9:24 Page 7 of 20 Cribaro et al. acta neuropathol commun (2021) 9:24 Page 8 of 20 one-way ANOVA or Kruskal–Wallis test (Dunnett’s, correlative increase in cellularity with respect to increas- Tukey’s, Dunn’s or Sidak’s multiple comparisons test) ing Ki67 index (Additional file 2: Supplementary Figure with *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001, S1). unless otherwise stated in the text. Data are repre- sented as mean ± SEM. F-tests were used to exam- GFAP map and expression ine the similarity of the variance of the compared Using GFAP expression, one of the most accepted GBM groups, and the appropriate tests were performed markers in histopathology, GBM-affected cortical areas accordingly. were mapped and tumor location was made evident given the distinct morphology of resident astrocytes in Results different cortical layers and the tumor mass itself sur- To analyze the cellular complexity of brain tumor micro- rounded by layers of reactive astrocytes (Additional file 4: environments, GBM samples obtained from surgical Supplementary Figure S3). Tumor mass could be clearly resection were rapidly processed to preserve the tissue detected as a heterogeneous and disorganized bulk of structural integrity. Then, after cryoprotection, samples cells with shapes ranging from fibrous to swollen-eccen- were cut in 60-μm thick sections (for 3D visualization) tric gemistocytic formations. Sampling of GBM 3DBx in and processed for high penetration multi-labeling immu- cases with increasing Ki67 indices showed higher GFAP nohistofluorescence. High-resolution 3D scans were density and an increase in the volumetric surface area of captured in x–y-z planes covering the section thick- the ­GFAP+ elements (Fig. 1a and b). Close proximity to ness. Systematic 3D sampling of biopsies was also done interlaminar and varicose-projecting astrocytes served to build hypermosaics. Three-dimensional stacks were as a hint to histologically locate the tumor, frequently deconvoluted and rendered in appropriate software to hosted in deeper layers of the CTX and consistent with generate 3D visualizations to analyze numerical and vol- the potential origin described in previous studies [14]. umetric parameters as well as cellular topographical rela- tionships (Additional file 3: Supplementary Figure S2). Basement membrane vascular network We performed an extensive 3D mapping of GBM sam- To visualize the blood vessels, staining collagen IV (COL- ples at cellular resolution using multilabeling staining for IV)-rich basement membrane provided a clear and reli- vascular structures and tumor-cell types covering both able outline of the vasculature both in control CTX and peritumoral areas and tumor cores. Only a selection of GBM (Fig. 1c–g). In some GBM cases, there was massive images is displayed in this resource paper, and original COL-IV deposition, in some rare areas with barely recog- image files in .lsm, .lif or.ims format can be examined on nizable vessels (Additional file 5: Supplementary Figure request. S4). Prior clinical neuropathological examination of the Visualizing the vascular 3D footprint revealed not only samples had determined the classical GBM features the complexity of interacting structures in normal CTX including pluricellularity, formation of hypoxic pseu- but also the altered spatial relationship between base- dopalisades, aberrant mitoses, and glomeruloid ves- ment membrane and G ­ FAP+ cells in GBM (Fig. 1d). + sels. Importantly, expression of classical GBM markers ­GFAP cells in GBM appeared as a dense mass with vari- GFAP and Vimentin was clearly detected in the 3DBx. able morphology with irregular fibrillary processes and Moreover, canonical gemistocytic formations were also gemistocytes (Additional file 4: Supplementary Figure identified on examination at high-resolution (Additional S3 and Fig. 1d and e) in contrast with regular protoplas- file 2: Supplementary Figure S1). Before proceeding with mic astrocytes in normal CTX. G ­ FAP+ glioma cells were further in-depth imaging and to validate the homology often in apposition and surrounding BVs, compatible and consistency of the clinical neuropathological assess- with the phenomenon of tumor cell motility along BV ment, full sets of confocal images were evaluated histo- via the outer side of the basement membrane, events that pathologically with our automated tools to corroborate a are compatible with previous observations seen in glioma (See figure on next page.) Fig. 3 GBM BVs show uneven COL-IV distribution reflecting basement membrane disruption and fenestration. a Comparison of high-resolution confocal mosaic in normal CTX and GBM. Original confocal overview and volumetric isosurfaces are shown (large increments in frame are 50 μm). b Luminance surface plot measured along the basement membrane of normal CTX vessel and GBM-associated vessels according to the proliferation index Ki67. c Contrasted transparency of basement membrane in normal CTX (N. CTX) and GBM. d Aspect of altered COL-IV distribution in GBM according to increasing Ki67. e Measurements of relative basement membrane fluorescence in normal CTX (N. CTX) and GBM. f Standard deviation of relative fluorescence of normal CTX and GBM basement membrane Cribaro et al. acta neuropathol commun (2021) 9:24 Page 9 of 20 Cribaro et al. acta neuropathol commun (2021) 9:24 Page 10 of 20 mouse models [32, 33]. This blood vessel-GFAP cell inter- as looping, splitting (intussusception) or activation action was also seen in pseudopalisades, being fibrous in of silent collateral vessels [12], we acquired images of the hypercellular areas (Additional file 6: Supplementary these structures from GBM 3DBx for analysis. We Figure S5), consistent with the escaping behavior of GBM imaged arrangements that were consistent with looping cells in these structures [3, 29]. and also with the formation of COL-IV pillars compat- To map and understand the vascular network, we ible with intussusception (Fig. 2k). We also observed built three-dimensional hypermosaics that exhibit the numerous silent collateral vessels (Small vessels ≤ 5 μm heterogeneity of GBM vasculature compared to unal- diameter insufficient to accommodate blood cell pas- tered normal CTX. Non-neoplastic CTX appeared sage) although quantification of the latter did not consistently irrigated with a regular vessel network. In resolve whether or not this strategy is used, as the pro- contrast, GBM showed uneven vessel distribution with portion appears to be similar in GBM and normal CTX irregular vessel shapes (Fig. 3f ). 3D rendering and shad- (Fig. 2l). owing allowed correct visualization of this heterogene- High resolution, in-depth imaging and 3D render- ity, which increased in accordance with the Ki67 index ing, and shadowing of COL-IV revealed a patent struc- (Fig. 1f and g), which demonstrates in tissue that the tural difference between normal and GBM-affected impaired structural adaptation and transformation of CTX (Fig. 3a). Healthy CTX displayed a regular vascular tumor-induced angiogenesis, predicted by computa- network with uniform appearance of COL-IV, whereas tional simulations [25], correlate with malignancy. GBM-affected CTX showed a heterogeneous distribu- By generating 3D isosurfaces, we were able to meas- tion and disrupted basement membrane (Fig. 3a) with a ure COL-IV volumetric surface area, demonstrating fenestrated appearance. Analysis of fluorescent-labeled an increase in volume in GBM, especially in highly COL-IV luminance showed uniform levels in normal proliferative cases (50% Ki67) (Fig. 2a and b). As this CTX while GBM vessels displayed drastic peaks and val- increased heterogeneity may be due to an increase in leys (Fig. 3b). Alpha blending of COL-IV images shows blood vessel caliber, we analyzed the occurrences of the disparity between the crêpe-paper-like appearance of blood vessels in three diameter ranges, smll, medium normal vessels and the disrupted aspect of GBM vessels and large, and found vessel diameters were increasingly (Fig. 3c). Interestingly, this COL-IV disruption appears to larger in more proliferative tumors (Fig. 2c and d). be more dramatic in more proliferative tumors (Fig. 3d). Because this heterogeneity could be due to the gen- Analysis of relative fluorescence provided clear evidence eration of new ramifications, we also analyzed and of this feature (Fig. 3e), showing a significant increase in quantified vessel branching. To do so, we arbitrar- relative fluorescence variation possibly indicating micro- ily categorized branching type in 3D images accord- scopic disruption of the basement membrane in human ing to the number of branches emanating from each GBM (Fig. 3f ). branching point sampled [primary (y- or t-shaped), with two branches; secondary (w-shaped), with three Endothelial vascular network branches and tertiary (irregular, > 3), with four or more Because basement membrane was seen as highly fenes- branches]. Interestingly, branches generated were trated, we analyzed whether this phenomenon occurs mostly of the primary type whereas secondary and ter- equally in the adjoining endothelium. To do so, we tiary were both minimal (Fig. 2e–j). These results sug- labeled the endothelium with CD31 (Fig. 4a), a reli- gest that the major strategy for generating new vessels able marker, paralleling previous vascular measurements in human GBM is single branching, in contrast with made in glioma mouse model [32], and quantified the other potential strategies [4, 8]. volumetric area of the vascular network (Fig. 4b). We saw Because angiogenesis, besides sprouting, can be that endothelium increases in highly proliferative tumors, accomplished by several alternative strategies, such but the global increase was of a lesser extent than that of (See figure on next page.) Fig. 4 GBM endothelium appears disrupted from the basement membrane resulting in BTB multilayer fenestration. a Confocal images and isosurface renderings of ­CD31+ endothelium for normal CTX and two GBM cases with 20 and 30% Ki67 indices. b Increased volumetric surface in high Ki67% cases. c Alpha blending of COL-IV and CD31 in normal CTX and GBM. Overview is shown in top panel and higher magnification in bottom panel. d Detail of different vascular structures in GBM labeled with COL-IV and CD31. E. COL-IV and CD31 colocalization plots in normal CTX (N. CTX) and GBM with 10 and 30% Ki67. F. Diagram of the basement membrane and endothelium in normal CTX (CTRL) and GBM. g Confocal images of COL-IV and CD31, 3D isosurface rendering and surface overlap in normal CTX and two cases of GBM with 20 and 30% Ki67. h Venn diagrams showing the 3D surface overlap of COL-IV and CD31 in N. CTX and GBM cases. i Percentage of overlapping COL-IV and CD31 3D surfaces in normal CTX and GBM-affected CTX Cribaro et al. acta neuropathol commun (2021) 9:24 Page 11 of 20 Cribaro et al. acta neuropathol commun (2021) 9:24 Page 12 of 20 the basement membrane. Double labeling of basement by labeling CD3, and combined with endothelium marker membrane and endothelium in GBM-affected cortices CD31, GFAP to detect the tumor mass, and DAPI coun- showed that the two did not seem to match completely, terstaining to identify individual cells (Fig. 6a and b). T in contrast with the regular overlap seen in normal CTX cells and vascular endothelium were segmented and spe- (Fig. 4c and Additional file 7: Supplementary video 1). cific 3D isosurfaces were generated to perform quantifi- Particularly, two different scenarios could be observed cations. Total number of T cells was increased in all GBM in GBM cases, COL-IV visualized without overlapping cases, compared to normal CTX, but no differences were endothelium, and endothelium seen without overlay- seen between groups according to Ki67 index (Fig. 6c). ing basement membrane (Fig. 4c, d and Additional file 8: High-resolution analysis allowed the precise detection Supplementary Video 2 and Additional file 9: Supple- of T cells attached to the endothelium (extravasating) or mentary Video 3). In some cases, endothelial cells were infiltrated into the tumor mass (migrated) (Fig. 6d). This highly proliferative, resembling the first stages of glo- allowed us to analyze three dimensionally the distance T merulus formation, and in other cases occupying canoni- cells travelled from the endothelium into the tumor. The cal glomeruloid vascular tufts in large numbers with no number of extravasating T cells was similar in all tumor corresponding neoplastic cells (Additional file 10: Sup- cases suggesting an inflammatory-mediated lympho- plementary Figure S6 and Additional file 11: Supplemen- cyte attraction and initiation of diapedesis. However, the tary Video 4). Further analysis of the overlapping voxels number of migrated cells appears to increase in more between COL-VI and CD31 isosurfaces revealed a homo- proliferative tumors (Fig. 6e). Interestingly, the distance geneous interaction in normal CTX, with almost 100% traveled appears to increase in tumors with higher prolif- isosurface overlap compared with tumor-bearing CTX eration index (Fig. 6f ). (Fig. 4e–g). Detailed quantification of the interacting 3D isosurfaces demonstrated an intimate interaction in con- Mapping the myeloid landscape trol CTX in contrast with an increasingly disrupted one Along with T cells, monocytes also permeate the tumor in GBM cases (Fig. 4h and i). parenchyma, becoming macrophages associated with the As we had observed dramatic fenestration of endothe- tumor and joining the resident macrophage-like popu- lium and basement membrane in GBM compared to lation, microglia [7]. The two cell-types, are hardly dis- normal CTX, we were interested in examining its rela- tinguishable when in the tumor core and are normally tionship with the external wall of the BBB or BTB by labe- pooled as GAMMs [24]. Thus, to identify GAMMs, we ling astrocytes or glioma cells with GFAP. This becomes first used the pan marker Iba-1. Cortical microglia in especially relevant considering that previous studies in normal CTX showed the classical morphology, display- glioma mice models showed that astrocytic endfeet are ing a homogeneous network of thin processes (Fig. 7a displaced by glioma cells during the co-option process and b). In GBM samples, a gradual morphological spec- through initial tumor growth [32]. trum was identified. The peripheral rim of the tumor Combined labeling of three primary BBB components, showed layers of activated GAMMs, with morphologies endothelium, basement membrane and astrocytes or compatible with a pre-activated phenotype with larger glioma cells (Fig. 5a and b) and detailed high-resolution cell bodies and thick branches. By contrast, in the tumor 3D reconstruction of normal CTX and GBM (Fig. 5c and core GAMMs showed fully activated shapes without d), depicted complex, multilayered fenestration in GBM branches and with a rounded cell-body outline (Fig. 7a compared to the regular and continuous structure in nor- and b). Intermediate zone GAMMs displayed transitional mal CTX with clear variations in GFAP layers (Fig. 5c–e). shapes. Throughout the tumor, Iba-1+ GAMMs appear In normal CTX, protoplasmic astrocytes were seen in distributed in all niches, with no distinctive or particu- close apposition to homogeneously built vessels (Fig. 5a, lar association, i.e. located either near or distant to blood c and e) whereas GBM cases showed gemistocytic gli- vessels (Fig. 7c–d). Quantification of Iba-1+ GAMMs, oma cells coopting the external walls of disrupted vessels using 3D isosurfaces revealed an increase corresponding (Fig. 5b, d and e) where anomalous, not easily discern- with the Ki67 proliferation index (Fig. 7e and f ), which is ible GFAP-expressing astrocytic-like structures were concordant with the idea that GAMMs may contribute to recognized. tumor survival. To further characterize pro-inflammatory GAMMs, T cell infiltration and migration we used MHCII (HLA-DR) immunofluorescence, a As the multilayered fenestration described above may marker classically used for activated states of microglia affect immune cell permeation of tumor parenchyma, a and brain macrophages [17]. MHCII was detected in recent case in point [19], we analyzed T cell homing or peripheral areas and tumor core, showing particularly infiltration in GBM 3DBx. T cell presence was detected strong intensity in perivascular areas (Fig. 7g and h) in Cribaro et al. acta neuropathol commun (2021) 9:24 Page 13 of 20 Fig. 5 Disrupted BTB structural elements in GBM. a Characteristic structure of brain BV elements in normal CTX. Overview of the architectural integrity of brain BV formed by protoplasmic astrocytic endfeet, COL-IV basement membrane and CD31 endothelium (1). 3D detail of a BV-related protoplasmic astrocyte (2). Higher magnification of the brain vessel integrity (3). Diagram of the normal CTX BBB elements (4). b Characteristic disrupted structure of brain BV elements in GBM. Overview of the altered vascular structure in GBM gemistocytic tumor nest (1). Detail of classical GBM gemistocyte on disorganized basement membrane (2). Detail of the intermingled elements of the GBM associated vessels (3). Diagram of the GBM BTB elements (4). c 3D reconstruction of a normal cortical BV (Large increments in frame are 5 μm). d 3D reconstruction of a GBM-associated BV (Small increments in frame are 5 μm). e Diagram illustrating the alterations in GBM-associated BV versus non-tumor BV Cribaro et al. acta neuropathol commun (2021) 9:24 Page 14 of 20 Fig. 6 Increased T cell infiltration and migration in GBM. a Representative T cell infiltration in G ­ FAP+ areas. Higher magnification of a ­CD3+ T cell is seen in the bottom panel. b Representative images of ­CD3+ T cell infiltration (yellow) and GFAP expression (red) in GBM patients with 10, 20 or 30% Ki67 indices as determined by neuropathological examination. Bottom panel shows the rendering of T cells (magenta) and C ­ D31+ endothelium (green). c T cell quantification in GBM patients. d Three-dimensional analysis of T cell position with respect to endothelium distinguishes extravasating T cells (in diapedesis) or migrated into the tumorigenic parenchyma. e Quantification of extravasating and migrated T cells. f Quantification of the distance travelled from the closest endothelium (See figure on next page.) Fig. 7 Myeloid landscape in GBM. a Morphological spectrum of GAMMs in GBM compared with cortical microglia in normal CTX. Tumor periphery, intermediate areas and core are shown. b Computerized camera lucida of characteristic morphological spectrum of GAMMs. c Iba-1+ GAMMs (magenta) in GBM tumor core (GFAP) in proximity to BV (CD31). d Higher magnification of Iba-1+ GAMMs (magenta) in GFAP-rich (red) area. e Rendering of Iba-1 (magenta), CD31 vessels (green) and GFAP (red) isosurfaces, together with cell-size DAPI spots (blue). f Quantification of Iba-1+ units, volumetric surface area, and volume in normal CTX and GBM cases. g Visualization of D ­ API+ nuclei (blue), G ­ FAP+ cells (red), COL-IV (green) vessels, and ­MHCII+ cells (magenta) in peripheral areas and tumor core. h ­MHCII+ cells in close proximity to blood vessels. i Representative image of mutually exclusive GFAP (red) and MHCII (magenta) isosurfaces in GBM. j Quantification of MHCII units, volumetric surface and volume in normal brain and GBM cases Cribaro et al. acta neuropathol commun (2021) 9:24 Page 15 of 20 Cribaro et al. acta neuropathol commun (2021) 9:24 Page 16 of 20 close apposition to the basement membrane, compat- which should be considered for further studies or when ible with perivascular macrophages. Certain areas dis- taking therapeutic action. played ­MHCII+ cells homing through permeable blood vessels (Fig. 7h). A similar phenomenon could also be Discussion seen in vessels surrounded by necrotic pseudopalisades We present here a reliable, multicolor labeling method to (Additional file 6: Supplementary Figure S5e). Quantifi- visualize the cellular microenvironments of GBM three cation of MHCII expression revealed increasing num- dimensionally. We ruled out other standard multiplex bers in more proliferative tumors and interestingly the procedures especially those requiring aggressive micro- areas occupied by MHCII were particularly elevated in waving of samples in some steps between labeling and tumors with the highest Ki67 Index (Fig. 7j). staining, which would have damaged the tissue structure Most importantly, the expression of MHCII was low or in floating, thick 3DBx. Our images and novel analysis undetectable in GFAP-expressing areas, and conversely, shed light on the architectural heterogeneity and cellular ­GFAP+ cells were not detected in MHCII-rich areas, complexity of human GBM microenvironments and are showing mutually exclusive regions with no clear overlap made available as a resource for further studies. and differentiating the tumor nest from the stroma (Fig. 8a Classical GFAP labeling allowed us to identify and map and b and Additional file 12: Supplementary Video 5). an array of morphological changes in GBM and rela- Both areas were equally perfused by blood vessels as tionships with blood vessels. Importantly, these changes COL-IV isosurfaces revealed similar volumetric parame- correlate with vascular heterogeneity, which is character- ters in each, although, the caliber of vessels within MHCII- ized by the increase in branching points and variation of rich regions was higher (Fig. 8c). To validate the differential calibers. expression of MHCII in GAMMs, we combined Iba-1 and The detailed analysis of endothelium, basement mem- MHCII labeling. Cells were contoured in confocal stacks brane and astrocytic endfeet show that the structural and the level of MHCII fluorescence was sufficient to dis- layers of the BBB are altered in a particular manner, dis- criminate two subpopulations of Iba-1+ cells according to playing a multilayered fenestration. These results suggest the fluorescence intensity of MHCII channel (Fig. 8d). At that vascular integrity is compromised in GBM, although higher resolution, both GAMM subpopulations could be this may still be a matter of debate [19, 30]. Most inter- easily identified (Fig. 8e). Then, by creating cell-size isosur- estingly, this basement membrane and endothelium faces, histocytometry plots could be generated, allowing us fenestration in GBM, each with a different disruption to sort Iba-1+ ­MHCIIHigh stromal GAMMs (Fig. 8f and g). pattern, may have important consequences in immune Stroma in GBM has scarcely been investigated, cell permeation. and although it is thought to be mostly composed of Our analyses demonstrate that T cells home into the GAMMs [7], its topographical differentiation from the tumor parenchyma and remarkably, tumor proliferation tumor nest is still unclear. This method allows the use seems to positively correlate with the level of migration of a marker that may separate the populations of the of lymphocytes within the tumor. This actually challenges two environments (Fig. 8h). the conception of malfunctioning vessel wall and poor Overall, this resource article provides an atlas-like entry of immune cells in tumors [19], as T-cell adhesion image collection accompanied by precise measure- and transmigration appear to be functional in GBM. ments of numerical and volumetric parameters. We This is actually consistent with previous results from our hope this may help to further understand human GBM group, in which we showed that T cells enter the GBM microenvironment complexity and to shed light on parenchyma as nonproductive lymphocytes in a patrol- GBM’s architectural and cellular composition, both of ling mode, displaying a kinaptic state, which entails increased motility and random walking within the tumor (See figure on next page.) Fig. 8 Distinct neuro-inflammatory profile of GAMMs in stroma and tumor nest. a GFAP (magenta) and MHCII (green) label tumor nest and stroma, respectively, as mutually exclusive niches in GBM. COL-IV blood vessels (blue) and cell nuclei (gray) are also marked. b Vascular volume of stroma (MHCII-rich) and tumor nest (GFAP-rich) suggested equal perfusion. c Vessels in stroma (MHCII-rich) have larger calibers than in tumor nest (GFAP-rich). d Segmentation by contour of MHCII (red) and Iba-1 (green) cells from a GBM confocal stack. DAPI was used as counterstain (cyan). Fluorescence intensity of the green and red channels plotted in ­MHCIIHigh and ­MHCIILow profiles. e Top panels show confocal alpha-blended images of Iba-1 (green) and MHCII (red) labeling combination in GBM samples to visualize colocalizing cells (yellow: colocalization). Bottom panels show a high-resolution optical slide of an Iba-1/MHCII-expressing GAMM. DAPI (blue) was used for counterstaining. f Categorization of segmented M ­ HCII+ cells according to MHCII intensity. Selection of segmented Iba-1+ with high intensity of MHCII. g Left. Cytohistometry graph of Iba-1+ (green) and ­MHCII+ (red) cells showing the selection of Iba-1+/MHCIIHigh (white) as stromal GAMMs. Right. Selected cells highlighted in the image (yellow). h Diagram of the GAMM neuroinflammatory profile distribution in tumor nest and stroma Cribaro et al. acta neuropathol commun (2021) 9:24 Page 17 of 20 Cribaro et al. acta neuropathol commun (2021) 9:24 Page 18 of 20 [5]. At least for GBM, the BTB does not seem to hamper more proliferative tumors (calculation of GFAP/MHCII immune cell infiltration. C­ D31+ endothelium in GBM measured volume ratios shows: TSR for Ki67 30%, 3; appears to function properly as an adhesion layer [23] to Ki67 20%, 8; Ki67 10%, 33). facilitate leukocyte diapedesis (including lymphocytes Increased numbers of ­ GFAP+ cells, T cells, and and monocytes) and entry. But most importantly, the GAMMs, correlated with increasing vascular net- endothelial fenestrations that expose the basement mem- work complexity, which implies an increase in neu- brane to circulating immune cells may also be functional. roinflammation and tissue deterioration due to BTB Lymphocytes can also use these open areas for extrava- disruption. Importantly, our 3D quantifications show sation since VLA-1 integrins in lymphocytes can directly accurate measurements expressed in precise volumetric bind COL-IV to facilitate their infiltration [27, 28]. units in contrast with previous estimations made solely in Our results also show a high density of GAMMs deriv- two dimensions (i.e., 6 µm-thick sections) [20]. ing from resident and blood infiltrate monocytes. Quan- Our data provide an in-depth analysis of GBM in thick tifications consistently indicate that the increase in sections, giving a more comprehensive understanding of GAMMs correlates with tumor proliferation. However, it the cellular distribution within human GBM. The tech- remains unclear why a higher density of GAMMs reflects nique enables 3D tissue imaging providing a novel visu- a detrimental cellular composition. GAMMs (both alization of the cellular heterogeneity of this aggressive peripheral macrophages and brain-intrinsic microglia) and deadly tumor. have been proposed as regulators of the transition from Herein we display only a small representative selection proneural to mesenchymal GBM subtype, mesenchymal from a large repository of images that are available on being the signature with poor prognosis in this transi- request. tion [31]. In our GBM cases, increased Iba-1 expression is seen in more proliferative tumors, which is consistent Supplementary information with the increased transcriptional levels of Iba-1 found in Supplementary information accompanies this paper at https​://doi. the mesenchymal GBM subtype [31]. org/10.1186/s4047​8-020-01115​-0. As GAMMs are the major source of non-neoplastic cells, they are considered the major cellular component Additional file 1: Supplementary Table S1, Table S2. Neuropathologi‑ of the tumor stroma [7]. However, given that the loca- cal data of the GBM patients and antibodies used in the methodology. tion of the stroma versus tumor nest in GBM has not Additional file 2: Supplementary Figure S1. Human samples selected been clearly delineated, it is unclear how this may con- for the study comply with bona fide GBM neuropathology. a Cellularity in normal CTX (N. CTX) compared with GBM, evidenced by thionine stain‑ tribute to a negative outcome. It is generally assumed for ing. b Presence of canonical hypoxic pseudopalisades in GBM biopsies. c other tumor types that the location of the microenviron- Examples of aberrant mitoses. d Glomeruloid vessels. e Overexpression ment regulates tumor-associated macrophages function of GFAP and vimentin (VIM) evidenced by immunohistofluorescence and nuclear counterstaining (DAPI). f Gemistocytic formations in two repre‑ [15, 35], however, the role of the stromal phenotype of sentative GBM samples evidenced by GFAP and DAPI. g Quantification of GAMMs in human GBM requires further analysis to be cellularity using computer-generated “spots”. h Comparison of cellularity better defined. in normal CTX (N. CTX) and GBM (confocal image and 3D segmentation). i Representative images of increasing cellularity with Ki67 percentage Stroma and tumor nest are well mapped in many provided by the clinical neuropathology. j Quantification of average cel‑ types of tumors [18], and fibroblasts appear to build lularity by 3D stereological dissector in analyzed samples. k Correlation of stroma in breast cancer, gastric cancer, epithelial can- the average cellularity per 3D dissector with the Ki67-group index. cer and many others. Nevertheless, the cellular com- Additional file 3: Supplementary Figure S2. Methodological Diagram (workflow). a Patients diagnosed with GBM underwent surgical resec‑ position of stroma in brain tumors has scarcely been tion and biopsies were prepared and preserved for thick sectioning. b explored. Our data shows that GAMMs populate Multi-color free-floating immunohistofluorescence was performed on both microenvironments with increasing density in thick sections. c High-resolution confocal images were captured on the samples considering x, y and z axes to generate 3D tissue blocks. d 3D sys‑ more proliferative tumors, however, according to our tematic sampling of the sections was conducted covering GBM biopsies. results, the difference in phenotype between stroma e Images were processed by deconvolution defined by PSF. f Blend views and tumor nest seems crucial. Our results indicate that and isosurfaces were conducted with specialized rendering software, to generate maps of measurable structures. g 3D rendering and visualization the nature of the stromal areas can be distinguished were performed considering the BV footprint. h Mosaics were built by by the expression of MHCII in human GBM. Interest- computerized stitching of the tissue blocks systematic sampling. i Differ‑ ingly, levels of different MHCII parameters increase ent vascular parameters were measured as well as the spatial relationships with other cellular structures or populations. with the Ki67 index, which conforms with the idea that the tumor-stroma ratio (TSR) is associated with poor Additional file 4: Supplementary Figure S3. Identification of cortical location of GBM cells by GFAP marker. a Primate-specific ­GFAP+ interlami‑ prognosis, a low TSR being indicative of aggressive- nar astrocytes extend long thin processes from cortical layer I to layer III ness for some tumors [13, 22]. Our data here suggest or IV. b Varicose-projecting and interlaminar astrocytes mingle in layer III. that this also happens in GBM, showing low ratios in c Protoplasmic astrocytes appear from layer II surrounding blood vessels. Cribaro et al. acta neuropathol commun (2021) 9:24 Page 19 of 20 the technical support provided by Naroa Mendizuri with the processing and d Detail of a protoplasmic astrocyte. e Varicose-projecting astrocytes. f multilabeling of GBM biopsies. Detail of varicose projection. g Peri-tumoral reactive astrocytes in deeper layers. h Protoplasmic astrocyte in deeper layer. i and j Detail of peritu‑ Authors’ contributions moral reactive astrocytes. k Polarized varicose astrocyte. l Tumoral massive GPC, ES, LR, IM, LR, PC and MR performed the main research of this manu‑ GFAP expression. m Gemistocytic formations in tumor nest. n Diagram of script, including processing the samples for multilabeling, microscopy distribution of GFAP-expressing cells in GBM-affected CTX. imaging, analyzing, and generating the 3-D renderings. JMG conducted the Additional file 5: Supplementary Figure S4. Aberrant COL-IV deposition brain surgeries and prepared the biopsies. APV assessed the neuropathologi‑ in GBM. a Concentric multilayered deposition (1), concentric layers and cal diagnosis of the samples. CB designed and coordinated the research. The granulated dispersed COL-IV deposition (2), aberrant deposition in large manuscript was written by CB and GP and revised and edited by all coauthors. vascular structures (3). b Multilayered and perforated glomeruloid vessel All authors read and approved the final manuscript. within a ­GFAP+ tumor nest region. c Mosaic of hypervascularized node in GBM, ­GFAP+ (red) neoplastic areas with high deposition of COL-IV (green); Funding ­CD31+ endothelium (magenta) showing hypertrophic limits of a vascular This project was supported by grants from the Spanish Ministry of Economy barrier. and Competitiveness, and the European Regional Development Fund (Fondo Europeo de Desarrollo Regional, FEDER) (Reference Grants: RYC-2010-06729, Additional file 6: Supplementary Figure S5. Vascular structure and loca‑ SAF2013-45178-P, PEJ-2014-P00015, and SAF2015-64123-P), Generalitat de tion in pseudopalisades. a Diagram of the pseudopalisades (PP) structures Catalunya (Reference Grant: 2014 SGR-984), Spanish Ministry of Science, surrounding necrotic focus (NF) and central blood vessel (BV). b Detail Innovation and Universities (Reference Grant: PGC2018-096003-B-I00) and by of the central BV and the hyper-cellular area forming the PP. c Rendering the Asociación Española Contra el Cancer (AECC). of two pseudopalisades with central NF. d Hypermosaic of a dense area of hypoxic pseudopalisades showing structural relationship of BVs with Availability of data and material PPs stained by DAPI, GFAP, COL-IV and CD31. e MHCII macrophages are The datasets during and/or analysed during the current study are available relegated to the vascular areas in the PPs. from the corresponding author on reasonable request. Additional file 7: Supplementary Video 1. Protoplasmic astrocyte and blood vessels in normal CTX. 3D rotation and clipping plane of a GFAP+ Ethics approval and consent to participate astrocyte (red) in close proximity to blood vessels evidenced by COL-IV Ethics approval was obtained from the Research and Ethics Committee of the (green) and CD31 (magenta). Nuclei are stained with DAPI (blue). two Institutions involved. Additional file 8: Supplementary Video 2. Blood vessel and glioma Consent for publication ­ FAP+ glioma cells (red) cells in GBM. 3D rotation and clipping plane of G Not applicable. in close proximity to a large vessel evidenced by COL-IV (green) and CD31 (magenta). Nuclei are counterstained with DAPI (blue). 3D rendering from Competing interests Patient 11. The authors declare that they have no competing interests. Additional file 9: Supplementary Video 3. Blood vessel and glioma cells ­ FAP+ glioma cells surround‑ in GBM. 3D rotation and clipping plane of G Author details 1 ing a branched blood vessel immunostained for COL-IV (green) and CD31 Department of Biochemistry and Molecular Biology, Institut de Neurocièn‑ (magenta). Nuclei are stained with DAPI (blue). 3D rendering from patient cies, School of Medicine, Lab M2‑107, Office M2‑117, Universitat Autònoma de 3.1. Barcelona, 08193 Bellaterra, Barcelona, Spain. 2 Department of Neurosurgery & Department of Pathology, Valencia General Hospital, Valencia, Spain. Additional file 10: Supplementary Figure S6. Glomeruloid vessels. a Three-dimensional reconstruction of a glomeruloid vessel in GBM patient Received: 25 November 2020 Accepted: 28 December 2020 to visualize ­GFAP+ cells (red), basement membrane (green), endothelium (magenta) and nuclei (blue). b Visualization of maximum projection and sections of the glomeruloid structure. c Distinct distribution of basement membrane and endothelium in glomeruloid vessels. d Relative fluores‑ cence intensity profile of 4 channels from yellow line in c. References Additional file 11: Supplementary Video 4. Confocal planes through 1. Arvanitis CD, Ferraro GB, Jain RK (2020) The blood–brain barrier and a 3D stack of a glomeruloid vessel from GBM. Endothelial cells are immu‑ blood–tumour barrier in brain tumours and metastases. Nat Rev Cancer nostained for CD31 (magenta), and basement membrane is labeled with 20:26–41. https​://doi.org/10.1038/s4156​8-019-0205-x COL-IV antibodies (green). Glioma cells are immunostained for GFAP and 2. Barcia C Jr, Gomez A, Gallego-Sanchez JM, Perez-Valles A, Castro MG, nuclei are evidenced by DAPI (blue). Both zenithal view (left) and rotated Lowenstein PR, Barcia C Sr, Herrero MT (2009) Infiltrating CTLs in human view (right) are displayed. glioblastoma establish immunological synapses with tumorigenic cells. Additional file 12: Supplementary Video 5. 3D rotation of a GBM sam‑ Am J Pathol 175:786–798. https​://doi.org/10.2353/ajpat​h.2009.08103​4 ple containing tumor nest and stroma. Isosurfaces of GFAP-rich (red) and 3. Brat DJ, Castellano-Sanchez AA, Hunter SB, Pecot M, Cohen C, Hammond MHCII-rich (magenta) areas are shown as well as COL-IV+ blood vessels. EH, Devi SN, Kaur B, Van Meir EG (2004) Pseudopalisades in glioblastoma are hypoxic, express extracellular matrix proteases, and are formed by an actively migrating cell population. Can Res 64:920–927. https​://doi. Abbreviations org/10.1158/0008-5472.can-03-2073 GFAP: Glial fibrillary acidic proteinCOL-IV; COL-IV: Collagen IV or Type-IV colla‑ 4. Das S, Marsden PA (2013) Angiogenesis in glioblastoma. N Engl J Med gen; CD31: Cluster of differentiation 31; Iba-1: Ionized calcium binding adaptor 369:1561–1563. https​://doi.org/10.1056/NEJMc​ibr13​09402​ molecule 1; MHCII: Major histocompatibility complex class II; CD3: Cluster of 5. Diaz LR, Saavedra-Lopez E, Romarate L, Mitxitorena I, Casanova PV, differentiation 3. Cribaro GP, Gallego JM, Perez-Valles A, Forteza-Vila J, Alfaro-Cervello C, Garcia-Verdugo JM, Barcia C Sr, Barcia C Jr (2018) Imbalance of immu‑ nological synapse-kinapse states reflects tumor escape to immunity in Acknowledgements glioblastoma. JCI Insight. https​://doi.org/10.1172/jci.insig​ht.12075​7 We would like to thank all the personnel from the Administration and 6. Griveau A, Seano G, Shelton SJ, Kupp R, Jahangiri A, Obernier K, Krishnan Technical Laboratories of the Institut de Neurociències, for the help provided S, Lindberg OR, Yuen TJ, Tien AC, Sabo JK, Wang N, Chen I, Kloepper J, at the Universitat Autònoma de Barcelona, very especially to the excellent Larrouquere L, Ghosh M, Tirosh I, Huillard E, Alvarez-Buylla A, Oldham technicians at the Microscopy Core, Núria Barba and Saioa Mendizuri and also MC, Persson AI, Weiss WA, Batchelor TT, Stemmer-Rachamimov A, Suva Cribaro et al. acta neuropathol commun (2021) 9:24 Page 20 of 20 25. Pries AR, Cornelissen AJ, Sloot AA, Hinkeldey M, Dreher MR, Hopfner M, ML, Phillips JJ, Aghi MK, Mehta S, Jain RK, Rowitch DH (2018) A glial Dewhirst MW, Secomb TW (2009) Structural adaptation and heteroge‑ signature and Wnt7 signaling regulate glioma-vascular interactions and neity of normal and tumor microvascular networks. PLoS Comput Biol tumor microenvironment. Cancer Cell 33(874–889):e877. https​://doi. 5:e1000394. https​://doi.org/10.1371/journ​al.pcbi.10003​94 org/10.1016/j.ccell​.2018.03.020 26. Quail DF, Joyce JA (2017) The microenvironmental landscape of 7. Hambardzumyan D, Gutmann DH, Kettenmann H (2016) The role of brain tumors. Cancer Cell 31:326–341. https​://doi.org/10.1016/j.ccell​ microglia and macrophages in glioma maintenance and progression. Nat .2017.02.009 Neurosci 19:20–27. https​://doi.org/10.1038/nn.4185 27. Richter M, Ray SJ, Chapman TJ, Austin SJ, Rebhahn J, Mosmann TR, 8. Hardee ME, Zagzag D (2012) Mechanisms of glioma-associated neovas‑ Gardner H, Kotelianski V, deFougerolles AR, Topham DJ (2007) Collagen cularization. Am J Pathol 181:1126–1141. https​://doi.org/10.1016/j.ajpat​ distribution and expression of collagen-binding alpha1beta1 (VLA-1) and h.2012.06.030 alpha2beta1 (VLA-2) integrins on CD4 and CD8 T cells during influenza 9. Jackson CM, Choi J, Lim M (2019) Mechanisms of immunotherapy resist‑ infection. J Immunol 178:4506–4516. https​://doi.org/10.4049/jimmu​ ance: lessons from glioblastoma. Nat Immunol 20:1100–1109. https​://doi. nol.178.7.4506 org/10.1038/s4159​0-019-0433-y 28. Roberts AI, Brolin RE, Ebert EC (1999) Integrin alpha1beta1 (VLA-1) medi‑ 10. Jain RK, di Tomaso E, Duda DG, Loeffler JS, Sorensen AG, Batchelor TT ates adhesion of activated intraepithelial lymphocytes to collagen. Immu‑ (2007) Angiogenesis in brain tumours. Nat Rev Neurosci 8:610–622. https​ nology 97:679–685. https​://doi.org/10.1046/j.1365-2567.1999.00812​.x ://doi.org/10.1038/nrn21​75 29. Saavedra-López E, Roig-Martínez M, Cribaro GP, Casanova PV, Gallego 11. Jonkman J, Brown CM, Wright GD, Anderson KI, North AJ (2020) Tutorial: JM, Pérez-Vallés A, Barcia C (2019) Phagocytic glioblastoma-associated guidance for quantitative confocal microscopy. Nat Protoc 15:1585–1611. microglia and macrophages populate invading pseudopalisades. Brain https​://doi.org/10.1038/s4159​6-020-0313-9 Commun. https​://doi.org/10.1093/brain​comms​/fcz04​3 12. Kreuger J, Phillipson M (2016) Targeting vascular and leukocyte commu‑ 30. Sindhwani S, Syed AM, Ngai J, Kingston BR, Maiorino L, Rothschild J, nication in angiogenesis, inflammation and fibrosis. Nat Rev Drug Discov MacMillan P, Zhang Y, Rajesh NU, Hoang T, Wu JLY, Wilhelm S, Zilman A, 15:125–142. https​://doi.org/10.1038/nrd.2015.2 Gadde S, Sulaiman A, Ouyang B, Lin Z, Wang L, Egeblad M, Chan WCW 13. Lee D, Ham IH, Son SY, Han SU, Kim YB, Hur H (2017) Intratumor stromal (2020) The entry of nanoparticles into solid tumours. Nat Mater. https​:// proportion predicts aggressive phenotype of gastric signet ring cell car‑ doi.org/10.1038/s4156​3-019-0566-2 cinomas. Gastric Cancer: Off J Int Gastric Cancer Assoc Jpn Gastric Cancer 31. Wang Q, Hu B, Hu X, Kim H, Squatrito M, Scarpace L, deCarvalho AC, Lyu Assoc 20:591–601. https​://doi.org/10.1007/s1012​0-016-0669-2 S, Li P, Li Y, Barthel F, Cho HJ, Lin YH, Satani N, Martinez-Ledesma E, Zheng 14. Lee JH, Lee JE, Kahng JY, Kim SH, Park JS, Yoon SJ, Um JY, Kim WK, Lee JK, S, Chang E, Sauve CG, Olar A, Lan ZD, Finocchiaro G, Phillips JJ, Berger Park J, Kim EH, Lee JH, Lee JH, Chung WS, Ju YS, Park SH, Chang JH, Kang MS, Gabrusiewicz KR, Wang G, Eskilsson E, Hu J, Mikkelsen T, DePinho SG, Lee JH (2018) Human glioblastoma arises from subventricular zone RA, Muller F, Heimberger AB, Sulman EP, Nam DH, Verhaak RGW (2017) cells with low-level driver mutations. Nature 560:243–247. https​://doi. Tumor evolution of glioma-intrinsic gene expression subtypes associates org/10.1038/s4158​6-018-0389-3 with immunological changes in the microenvironment. Cancer Cell 15. Lewis CE, Pollard JW (2006) Distinct role of macrophages in differ‑ 32(42–56):e46. https​://doi.org/10.1016/j.ccell​.2017.06.003 ent tumor microenvironments. Can Res 66:605–612. https​://doi. 32. Watkins S, Robel S, Kimbrough IF, Robert SM, Ellis-Davies G, Sontheimer org/10.1158/0008-5472.CAN-05-4005 H (2014) Disruption of astrocyte-vascular coupling and the blood-brain 16. Lim M, Xia Y, Bettegowda C, Weller M (2018) Current state of immuno‑ barrier by invading glioma cells. Nat Commun 5:4196. https​://doi. therapy for glioblastoma. Nat Rev Clin Oncol 15:422–442. https​://doi. org/10.1038/ncomm​s5196​ org/10.1038/s4157​1-018-0003-5 33. Winkler F, Kienast Y, Fuhrmann M, Von Baumgarten L, Burgold S, Mittereg‑ 17. McGeer PL, Itagaki S, McGeer EG (1988) Expression of the histocompat‑ ger G, Kretzschmar H, Herms J (2009) Imaging glioma cell invasion in vivo ibility glycoprotein HLA-DR in neurological disease. Acta Neuropathol reveals mechanisms of dissemination and peritumoral angiogenesis. Glia 76:550–557. https​://doi.org/10.1007/bf006​89592​ 57:1306–1315. https​://doi.org/10.1002/glia.20850​ 18. Mueller MM, Fusenig NE (2004) Friends or foes—bipolar effects of 34. Wolf KJ, Chen J, Coombes JD, Aghi MK, Kumar S (2019) Dissecting and the tumour stroma in cancer. Nat Rev Cancer 4:839–849. https​://doi. rebuilding the glioblastoma microenvironment with engineered materi‑ org/10.1038/nrc14​77 als. Nat Rev Mater 4:651–668. https​://doi.org/10.1038/s4157​8-019-0135-y 19. Munn LL, Jain RK (2019) Vascular regulation of antitumor immunity. Sci‑ 35. Yang M, McKay D, Pollard JW, Lewis CE (2018) Diverse functions of mac‑ ence 365:544–545. https​://doi.org/10.1126/scien​ce.aaw78​75 rophages in different tumor microenvironments. Can Res 78:5492–5503. 20. Nishie A, Ono M, Shono T, Fukushi J, Otsubo M, Onoue H, Ito Y, Inamura https​://doi.org/10.1158/0008-5472.CAN-18-1367 T, Ikezaki K, Fukui M, Iwaki T, Kuwano M (1999) Macrophage infiltration and heme oxygenase-1 expression correlate with angiogenesis in human gliomas. Clin Cancer Res: Off J Am Assoc Cancer Res 5:1107–1113 Publisher’s Note 21. Nyengaard JR, Bendtsen TF, Gundersen HJ (1988) Stereological estimation Springer Nature remains neutral with regard to jurisdictional claims in pub‑ of the number of capillaries, exemplified by renal glomeruli. APMIS Suppl lished maps and institutional affiliations. 4:92–99 22. Peng C, Liu J, Yang G, Li Y (2018) The tumor-stromal ratio as a strong prog‑ nosticator for advanced gastric cancer patients: proposal of a new TSNM staging system. J Gastroenterol 53:606–617. https​://doi.org/10.1007/ s0053​5-017-1379-1 23. Piali L, Hammel P, Uherek C, Bachmann F, Gisler RH, Dunon D, Imhof BA Ready to submit your research ? Choose BMC and benefit from: (1995) CD31/PECAM-1 is a ligand for alpha v beta 3 integrin involved in adhesion of leukocytes to endothelium. J Cell Biol 130:451–460. https​:// • fast, convenient online submission doi.org/10.1083/jcb.130.2.451 • thorough peer review by experienced researchers in your field 24. Poon CC, Sarkar S, Yong VW, Kelly JJP (2017) Glioblastoma-associated • rapid publication on acceptance microglia and macrophages: targets for therapies to improve prognosis. Brain: J Neurol 140:1548–1560. https​://doi.org/10.1093/brain​/aww35​5 • support for research data, including large and complex data types • gold Open Access which fosters wider collaboration and increased citations • maximum visibility for your research: over 100M website views per year At BMC, research is always in progress. Learn more biomedcentral.com/submissions