https://orcid.org/0000-0002-9464-6544
Abstract
Glioblastomas are characterized by aggressive and infiltrative growth, and by striking heterogeneity. The aim of this study was to investigate whether tumor cell proliferation and invasion are interrelated, or rather distinct features of different cell populations.
Tumor cell invasion and proliferation were longitudinally determined in real-time using 3D in vivo 2-photon laser scanning microscopy over weeks. Glioblastoma cells expressed fluorescent markers that permitted the identification of their mitotic history or their cycling versus non-cycling cell state.
Live reporter systems were established that allowed us to dynamically determine the invasive behavior, and previous or actual proliferation of distinct glioblastoma cells, in different tumor regions and disease stages over time. Particularly invasive tumor cells that migrated far away from the main tumor mass, when followed over weeks, had a history of marked proliferation and maintained their proliferative capacity during brain colonization. Infiltrating cells showed fewer connections to the multicellular tumor cell network, a typical feature of gliomas. Once tumor cells colonized a new brain region, their phenotype progressively transitioned into tumor microtube-rich, interconnected, slower-cycling glioblastoma cells. Analysis of resected human glioblastomas confirmed a higher proliferative potential of tumor cells from the invasion zone.
The detection of glioblastoma cells that harbor both particularly high proliferative and invasive capabilities during brain tumor progression provides valuable insights into the interrelatedness of proliferation and migration—2 central traits of malignancy in glioma. This contributes to our understanding of how the brain is efficiently colonized in this disease.
(1) Particularly invasive glioblastoma cells are also particularly proliferative.
(2) Different patterns of tumor cell proliferation can be tracked live and in real time.
(3) Tumor network-integrated glioblastoma cells are more quiescent.
We have recently shown that brain invasion is achieved by a subpopulation of tumor cells with neuronal features. After colonization of new areas, these tumor cells became more stationary and interconnected with their long membrane protrusions, tumor microtubes, to an interconnected multicellular tumor network. It remained unclear how the second key hallmark of malignancy, growth by tumor cell proliferation, related to these findings. Here we demonstrate that brain invasion and proliferation are achieved by the same subpopulation of tumor cells, while the tumor network-connected glioblastoma cells are more slow-cycling. These findings provide a relevant addition to our understanding of how brain tumor cells colonize the brain. They also increase our understanding of the biological role of different tumor cell subpopulations during glioblastoma progression and support the rationale to therapeutically target the forefront of brain colonization in glioma.
The diffuse infiltration of the entire brain, which is typical for glioblastoma,2–4 the most common and aggressive form of primary brain tumor, leads to inevitable recurrence. Tumor progression is determined by both cancer cell proliferation and migration. The infiltrative nature of glioblastomas and other incurable gliomas limits the completeness of surgical resection and makes adjuvant treatments such as radiotherapy and chemotherapy indispensable although they can also not prevent disease recurrence. Gliomas consist of multiple cell types that are highly heterogeneous on a molecular, functional, and structural level.1,5–11 Furthermore, they extend long membrane protrusions, tumor microtubes (TMs), that are important for tumor cell invasion, proliferation, dissemination in the brain, and connection to a multicellular communicating network that resists radio- and chemotherapy, and can even repair itself.1,7,12–19 We recently discovered that a distinct network-unconnected subpopulation of glioblastoma cells (GBCs) transcriptionally reside in a neural progenitor-like state and do receive neuronal synaptic input that promotes cell invasion through reutilization of mechanisms established during neuronal development.1
In general, migration and cell division are considered mutually exclusive cellular programs that cannot coexist at one particular moment of time,2,20 because cytoskeletal machineries and existing finite free energy resources cannot be used simultaneously for proliferation and migration.21–23 This has been termed “go-or-grow.” The go-or-grow concept has frequently fueled the assumption that highly migratory GBCs are, in general, not relevantly contributing to overall proliferation.
Here we investigate whether migrating GBCs are indeed less proliferative, or whether both central traits of the disease are rather positively correlated during the time course of brain colonization. Taking advantage of refined in vivo 2-photon microscopy (IV2PM) technologies, we followed tumor cells within tumor microregions over many weeks to answer this question.
Methods
Cell Culture and Lentiviral Transductions
Primary human glioblastoma cells (S24, P3 GBCs) were kept under stem-like neurosphere conditions (P3 was a kind gift from Hrvoje Miletic, K.G.).
Tumor cells were transduced with lentiviral vectors for multicolor imaging. For cytosolic tdTomato expression, the LeGo-T2 vector (Addgene 27342, RRID: Addgene_27342 https://www.addgene.org/27342/) was used. LeGo-T2 was a gift from B. Fehse.24 Nuclear GFP expression was achieved by transduction with Tet-Off-H2B-GFP vector (kind gift from B. Falkowska-Hansen and P. Boukamp).25 EF1-mVenus-p27K (kind gift from M. Bentires-Alj) created a mVenus nuclear signal in non-cycling cells.26 GBCs expressing cytosolic tagGFP in rpwLenti-PGK-R4-GW-R3-SV40-Puro were kindly provided by the Cellular Tools Core Facility, DKFZ Heidelberg, Germany.
All cells used in this study were regularly tested for mycoplasma infections. Glioblastoma origin was verified and subtypes were determined as described before.1,7,15
Sampling of Patient Glioblastoma Tumor Tissue
Patient-derived tumor sections and primary cell cultures from tumor bulk (TB) (T1 contrast-enhancing (CE) solid tumor) and the infiltration zone (IZ) (FLAIR-hyperintensive area outside the CE tumor) were precisely collected during surgery by navigation sampling (Brainlab AG, Munich, Germany).
Methods for derivation and expansion of clinical cell samples were previously described.27 The study protocols conformed to the ethical guidelines of the Declaration of Helsinki and were approved by the institutional ethical review board of Heidelberg University (2018-843R-MA) and the University of Bonn, (#182/08), respectively. Patients or their guardians provided informed consent.
Fluorescence Staining and Imaging
For in vivo EdU incorporation, animals were injected i.p. with 50 mg/kg EdU (#BCK488-IV-IM-S, Baseclick GmbH, Neuried, Germany) and euthanized 4 hours after application.
Samples (patient sections and xenografts) were subjected to fluorescence staining according to standard protocols. Human nestin (RRID:AB_308832 http://antibodyregistry.org/AB_308832), Ki67 (RRID:AB_443209 http://antibodyregistry.org/AB_443209), turboGFP (RRID:AB_2540616 http://antibodyregistry.org/AB_2540616), LipiLight-488 (Idylle, Paris, France), EdU fluorescence chemistry (#BCK488-IV-IM-S; Baseclick GmbH), and Hoechst 33342 were applied for staining. The human nestin antibody was used to identify patient-derived tumor cells in our mouse model and was utilized to stain GBCs in patient tumor samples. We confirmed the suitability of the human nestin antibody, and the expression of nestin in the different glioblastoma cell states (Supplementary Figure S1; Supplementary Material and Methods).
Tumor cells were incubated in spheroid culture or high-glucose (5,85 mg/mL), stem-like monolayer culture. To induce the cessation of nuclear GFP production, cells were incubated with 1 ng/mL doxycycline (Dox, Sigma, St. Louis, Missouri, USA)
Images were acquired using a Zeiss LSM 710 confocal microscope.
Cranial Window Preparation and In Vivo Tumor initiation
All animal procedures were performed in accordance with the institutional laboratory animal research guidelines after approval by the Regierungspräsidium Karlsruhe, Germany (G188-12, G132-16). See Supplementary Material and Methods for information on surgical procedures.
Dox was added to the drinking water at a final concentration of 200 µg/mL in deionized water containing 5% sucrose.
In Vivo 2-Photon Laser Scanning Microscopy
We acquired in vivo 2-photon laser scanning microscopy (IV2PM) images using a LSM 7MP microscope (Zeiss, Jena, Germany) as described previously,7 utilizing a Chameleon UltraII laser (Coherent, Santa Clara, California, USA) for fluorophore excitation. We set the laser wavelength to 850 nm for detection of GFP, mVenus, and tetramethylrhodamine-isocyanate-Dextran and measured tdTomato at 950 nm respectively.
Image Processing and Quantifications
IV2PM images were acquired with Zeiss ZEN Black software and processed with Fiji 2.0.0 (RRID:SCR_002285 http://fiji.sc) or Imaris (Bitplane, Zurich, Switzerland) to clear unspecific background and to crop identical areas for comparison of longitudinally acquired data. GFP intensities and cell numbers were measured manually in Fiji 2.0.0.
Quantification of Patient-Derived Tumor Cell Expansion Kinetics
Tissue samples from the TB and the IZ were triturated to a single-cell suspension. For expansion, cells were suspended in controlled neurobasal media containing B27, Laminin, and N2 supplements, and were provided with EGF and FGF under adherent conditions. Growth kinetics were recorded between passages 5 and 10. These culture conditions select for GBCs. During the initial 5 passages, the tumor cell content increases particularly in the culture initiated from the sample retrieved from the resection cavity. We previously determined empirically that it requires 5 passages for the tumor cell content to become similar in the cultures derived from both the resection wall and the tumor core.27 Cell vitality was confirmed by trypan blue exclusion. For controlled expansion, 13 400 cells/cm2 were plated on 6 cm dishes. Cells were expanded to 95%–100% confluence for up to 6 days, then harvested, counted, and reseeded at identical conditions. Expansion time, cell density and the expansion stability index (R2) and population doublings were documented as previously described.27 Expansion stability was assessed using linear regression analysis to demonstrate continuous proliferation over time. An index of R2 close to 1 indicates a consistent proliferative activity over time.
Statistics
SigmaPlot Software (Systat Software, Inc., Erkrath, Germany) and Prism (GraphPad Software, San Diego, California) and R statistical programming language (v.4.2.2, R Core Team, Vienna, Austria, RStudio IDE v.2022.12.0 + 353) were used for statistical analysis. Data were tested for normality using the Shapiro–Wilk test. Statistical significance was assessed by the 2-sided Student’s t-test, Mann–Whitney test, and the Fisher’s exact test. For analysis of variance between more than 2 groups a one–way ANOVA or an ANOVA on the ranks was performed. The association of proliferation and GBC density as well as the association of nuclear GFP intensity and the distance to the TB were statistically evaluated using non-parametric Spearman’s rank correlation coefficients (ρ) in a group-wise manner at a 2-sided < 0.05 significance level and visualized as scatter plots. Additionally, corresponding exploratory linear regression lines with 95% CI ranges were plotted within the figure28 (Figures 1, 3, and 5, Supplementary Figure S2). All plots were generated using the ggplot2 grammar of graphics and ggpubr packages.
Tumor cell proliferation in patient-derived tumor samples can be found both in the tumor bulk and the infiltration zone. (A) Axial T1 contrast-enhanced (CE) MRI and T2 fluid-attenuated inversion recovery (FLAIR) MRI scans of MA01. Insets represent CE and FLAIR-hyperintensive but non-contrast-enhancing tumor (non-CE). (B) Representative spatially annotated immunofluorescence staining corresponding to the outlined regions (insets) within the MRIs in A. Nestin (green) marks tumor cells (see Supplementary Figure S1 for sensitivity and specificity of nestin as a marker for all cell states of GBCs). Ki67 (red) is expressed throughout the active cell cycle (G1, S, G2, and M phase). Hoechst dye (blue) stains nuclei. Arrowheads indicate Ki67 positive tumor cell nuclei. (C) Proliferation index relative to tumor cell density within representative single-plane scans of 200 µm2 in human tumor samples from 4 patients (MA1-4; Proliferation indices derived from CE tumor regions are encircled; Spearman rank correlation ρ and P values are given, the colored area along each regression line marks the 95% confidence interval; n = 10–15 images per patient tumor sample). Abbreviations: IZ, infiltration zone; TB, tumor bulk.
See Supplementary Material for further details regarding materials and methods.
Results
Proliferation Versus Tumor Cell Density in Sections of Mouse and Human Glioblastoma
First, we sought to estimate the proliferative capacity of different tumor regions from patient-derived glioblastoma cell xenografts and in glioblastoma tumor samples from 4 patients. Ki67 and EdU were used as proliferation markers, which mark cycling cells and S-phase cells, respectively (Figure 1, Supplementary Figure S2). In comparison to tissue within the IZ in the xenograft tumors (Supplementary Figure S2) and more diffusely growing non-CE (non-CE) tumor regions in humans (Figure 1), the cell number-corrected proliferation index did not consistently differ in the TB of xenotransplants and spatially annotated human CE glioblastoma bulk samples (Figure 1C, Supplementary Figure S2E-F). Together these data imply that GBCs can in principle proliferate in all tumor regions, including the invasive parts. However, these analyses lack sufficient temporal and spatial resolution and therefore did not clarify how invasion and proliferation might be interrelated.
A Model to Trace the Proliferative History of GBCs
Therefore, to unequivocally determine the interrelation between tumor cell proliferation and tumor cell infiltration in glioblastoma, we established a nuclear-localized label-retaining GFP system (Tet-Off-H2B-GFP; Figure 2A-B). Flow cytometry confirmed stable nuclear GFP expression (Figure 2C). After a chase period during exposure to Dox in vitro and in vivo, actively proliferating cells progressively lost the nuclear GFP intensity, which should approximately halve with every cell division (Figure 2B), both in spheroid cultures in vitro (Figure 2D), and also during tumor growth in the live mouse brain in the same tumor microregion, with re-formation of nuclear GFP signal 56 days later when Dox treatment had been stopped (Figure 2E). Live imaging confirmed about 50% reduction of the GFP signal after cell division, both in serum-free GBC monolayer (Figure 2F) and in vivo (Figure 2G). The Ki67 positive (Ki67+) growth fraction of GBCs stably transduced with Tet-Off-H2B-GFP did not differ significantly from the respective parental wild-type cell line, neither in immunohistochemistry nor in flow cytometry (Figure 2H-I). This model, therefore, seemed very well suitable to track GBCs’ proliferative history in the live mouse brain.
The Tet-Off-H2B-GFP System is a suitable methodology to track GBCs’ proliferative history. (A) Lentiviral constructs used for constitutive cytoplasmic tdTomato and doxycycline (Dox) dependent nuclear GFP expression. When present, Dox suspends GFP expression. (B) The mitotic history of each cell can be tracked by its nuclear GFP intensity over time because GFP intensity per cell is roughly halved with every cell division. (C) Exemplary flow cytometry analysis of S24 and P3 tdTomato GBCs transduced with Tet-Off-H2B-GFP show stable nuclear GFP expression (n = 5–6 replicates per GBC). (D-G) GFP intensity was reduced over time in the presence of Dox when cultivated in vitro, both in 2 and 3 dimensions, and when grown in vivo. Left panels show representative micrographs at different time points from the respective experiments; associated box plots summarize measurements of relative GFP intensity. (D) Reduction of GFP intensity in S24 spheroids in the presence of Dox (n = 10 spheroids, 185–200 GBCs quantified per time point). (E) Representative in vivo 2-photon laser scanning microscopy (IV2PM) of S24 GBCs (z = 40 µm). No Dox was administered during the first 38 days after tumor cell implantation. After 4 weeks of Dox exposure, GFP was significantly reduced at day 70. GFP expression resumed after Dox was withheld (each time point n = 4 regions in n = 3 mice, 347–433 GBCs quantified). (F) Relative change of GFP intensity in 23 P3 GBCs grown in the serum-free monolayer that underwent cell division and 30 P3 GBCs that did not divide within a time frame of 7 hours. In vitro time series. Mean GFP intensity of the 2 daughter GBCs, normalized to their cytoplasmic tdTomato, was measured as a fraction of the corresponding parental GBC. (G) Data obtained by IV2PM time-series (z = 40 µm) of S24 GBCs in vivo. Relative GFP reduction was quantified in 7 dividing S24 and 9 dividing P3 GBCs compared to 29 and 30 non-dividing GBCs, respectively. GFP values were normalized to cytoplasmatic tdTomato of the respective cell. (H) Analysis of the Ki67+ growth indices of the parental S24 and P3 wild-type GBC orthotopically growing in vivo show no significant difference compared to the GBCs stably transduced with the lentiviral constructs used in this study (n = 3 mice for the wild-type GBCs and one mouse for each stably transduced GBC model, S24 GBCs P = .132; P3 GBCs P = .172) (I) Exemplary flow cytometry analysis of S24 and P3 Tet-Off-H2B-GFP tdTomato GBCs revealed no significant differences in the Ki67+ growth fraction compared to the respective parental wild-type cell line (n = 3–4 replicates per GBC; S24 GBCs P = .499; P3 GBCs P = .663). ANOVA on ranks with Dunn’s multiple comparison method was used for the data presented in D-E and H. Two-tailed unpaired Student’s t-test was used for the data presented in F-G and I. Values in C and I are presented as mean ± SD. Arrowheads indicate dividing cells.
Invasive GBCs Have a History of High Proliferation
To thoroughly analyze the proliferative history of migrating tumor cells that are responsible for the notoriously diffuse brain infiltration,4 we implanted GBCs harboring the Tet-Off-H2B-GFP system into the brain of a mouse with a chronic cranial window. We performed weekly time-lapse IV2PM of identical microregions over 4 weeks starting 7 days after tumor implantation (Figure 3, Supplementary Figure S3A-B). Dox treatment was initiated at the day of tumor implantation. Tumor cells at the leading edge of the IZ were predominantly deprived of nuclear GFP while GBCs within the TB retained more GFP, indicating that fewer cell divisions occurred among tumor cells that remained more stationary (Figure 3A). When followed over 4 weeks, the difference in relative GFP intensities in constitutively tdTomato-positive GBCs between bulk and invasion zone was gradually increasing, supporting a cumulative loss of GFP signal in the particularly invasive tumor cells by an increasing number of cell divisions (Figure 3B-D, Supplementary Figure S3A-B). Infiltration started within days after tumor implantation (Supplementary Figure S3C). Of note, the combination of little bulk tumor growth in the first 4 weeks after implantation (Figure 2B) with clear proliferative capacity in this area (Supplementary Figure S2) speaks for a situation in which cell division in the TB contributes to brain colonization during this phase of tumor establishment: By generating daughter cells that rapidly colonize the surrounding brain. Together this data provides clear evidence that, unexpectedly, the most invasive glioblastoma cells have a particularly proliferative history, too, and that this biological characteristic stays true over tumor progression.
Invasive GBCs in the mouse brain have a particularly active proliferative history. (A) Representative in vivo images of S24 and P3 GBC tumor bulk and invasion zones. GBCs are transduced with Tet-Off-H2B-GFP. Dox was administered throughout tumor development. Its high tumor cell density characterizes the tumor bulk, on the left, with most cells interconnected within the tumor cell network. The invasion zone, on the right, is sparsely populated with GBCs. The S24 image is derived from the larger D28 image in (B), outlined with a dashed rectangle. The infiltrating GBCs show a neural progenitor-like morphology with tumor network-unconnected 1 or 2 tumor microtubes (B) Sequential weekly images of identical microregions covering a section of the outer tumor bulk border and the infiltration zone of S24 GBCs during 4 weeks of tumor progression (z = 90 µm). See Supplementary Figure S3A for images of the individual channels. (C,D) Quantification of GFP intensity in individual S24 (C) and P3 (D) GBCs in vivo. Right panels: Scatter blots illustrate the correlation between distance of tumor cell infiltration and GFP intensity of individual GBCs. GFP intensity was normalized to the 10% GBCs with the highest intensity values at each time-point (n = 3 mice per cell line, S24 196-473 GBCs quantified for each time-point, P3 194-421 GBCs quantified for each time-point). Two-sided Student’s t-tests were used to compare GFP intensity within the tumor bulk and the infiltration zone (box blots). Spearman’s rank correlation coefficient, linear regression lines, and 95% CI ranges are given to analyze the relation of nuclear GFP intensity and distance to tumor bulk of individual GBCs at indicated time points (dot blots). Abbreviations: IZ, infiltration zone; TB, tumor bulk.
Proliferative History and TM Network Integration
Previous studies on tumor cell subtypes in human glioma tissue and on GBC lines in vivo have identified striking morphological tumor cell heterogeneity: (1) round sessile GBCs extending no TMs, (2) uni- or bipolar GBCs extending 1 or 2 TMs morphologically resembling neural progenitor-like cells,1 and (3) highly interconnected tumor cells with multiple TMs (>4 TMs) (Supplementary Figure S1D).1,7,12–15 Within the sparsely colonized IZ, we found a compelling dominance of the neural progenitor-like tumor cell with 1 or 2 TMs (Figure 4A,C-D and Supplementary Figure S3E-F), confirming our previous results.1 Progressive colonization of the IZ led to both an increase in the number of tumor cells per identical 3D tumor microregion and a concurrent reduction in mean nuclear GFP intensity (Figure 4B and Supplementary Figure S3D), suggesting ongoing proliferation during brain colonization. With increasing tumor cell densities, we observed a morphological evolution from unconnected GBCs to highly network-connected tumor cells with multiple TMs (>4 TM) (Figure 4E, and Supplementary Figure S3G).7,12 GBCs with >4 TMs are relatively stationary (Figure 4C and Supplementary Figure S3E). Twenty-eight days after implantation, the stationary, network-connected GBCs with multiple TMs had a higher nuclear GFP intensity than the particularly invasive, largely unconnected GBCs, which extend 1 or 2 TMs (Figure 4F-G and Supplementary Figure S3H-I). When tracked over the course of 4 weeks, the loss in average nuclear GFP intensity of invasive GBCs with 1 or 2 TMs steadily increased, reaching a 5-fold lower GFP intensity at day 28 compared to the increasing number of strongly connected GBCs with >4 TMs (Figure 4F and Supplementary Figure S3H). In line, nuclear GFP intensity in the TB remained relatively stable or dropped only moderately during the 4 weeks of tumor progression (Supplementary Figure S3J), which is in contrast to its stark drop in the tumor IZ (Figure 3C-D). We observed a similar situation in the IZ of more progressed tumors when followed from day 43 to day 64 after tumor implantation (Figure 4H-J and Supplementary Figure S4). Together this demonstrates that glioblastoma cells with 1 or 2 TMs are particularly invasive and proliferative, while highly TM-connected glioblastoma cells with >4 TMs are more stationary and slow-cycling.
TM-dependent GBC subpopulations differ in their proliferative potential. (A) Left, a representative IV2PM of a S24 GBC tumor growing in vivo (z = 90 µm). The tumor bulk is comprised of densely arranged glioma tumor cells that form multiple tumor microtubes (TMs) to build a highly interconnected tumor cell network. The interconnectivity decreases towards the transition zone, the surface of the main tumor (inset 1). Infiltrating GBCs (inset 2) are rarely interconnected. Inset 1: Connected GBCs with multiple TMs coexist with neural progenitor-like tumor cells (white arrowhead). The connected GBCs maintained their nuclear GFP (outlined with dashed white circle). The neural progenitor-like tumor cells displayed a reduced nuclear GFP intensity. Inset 2: The neural progenitor-like tumor cells dominate the infiltration zone and have mostly lost their nuclear GFP. (B) Quantification of S24 nuclear GFP (box blots) and S24 tumor cell density (black dots show mean GBC density ± SEM) in identical 3D infiltration zone microregions measured weekly for 28 days (S24 nuclear GFP intensity decreases significantly over time, Kruskal–Wallice one way ANOVA on ranks; the mean tumor cell density increased but did not reach statistical significance over the course of 4 weeks P = .178; n = 3 mice, one way ANOVA). (C) Left, box plot, quantification of relative GFP intensity of S24 GBCs in the tumor infiltration zone on day 28 after implantation, among tumor cells grouped according to the number of TMs they possess; right, dot blots, disaggregation of tumor cell GFP intensity and infiltration distance within the groups defined by their number of TMs. Compared to S24 GBCs with >4 TMs (mean infiltration distance of 65 µm), S24 GBCs without TMs at the time of image acquisition (mean infiltration distance of 134 µm; P < .001) and the neural progenitor-like S24 GBC with 1 or 2 TMs (mean infiltration distance of 197 µm, P < .001) covered a longer infiltration distance. GFP intensity was normalized to the 10% GBCs with the highest intensity values (n = 3 mice, 289 GBCs analyzed, t-test, Mann–Whitney rank sum test). (D) Changes in prevalence of the various morphological S24 GBC subtypes within the infiltration zone. Over time the fraction of the neural progenitor-like tumor cells decreases to the benefit of the glioma cell subtype with more TMs. (E) Changes in tumor cell interconnectivity over time. S24 GBCs form more interconnections over time to build a multicellular network of tumor cells (connection, C). (F) Measurement of nuclear GFP intensity in the S24 neural progenitor-like GBCs as a fraction of the corresponding GBCs with >4 TMs (horizontal line). For the comparison, mean GFP values of GBCs with >4 TMs were set to 1 at each time point. As no GBCs with >4 TMs were found at day 7, their mean GFP intensity value at day 14 was used instead (n = 3 mice, 38–197 S24 neural progenitor-like GBCs measured; t-test, Mann–Whitney rank sum test, data are presented as mean ± SEM). (G) S24 GBCs of the infiltration zone were divided into low and high GFP intensity groups relative to the median nuclear GFP intensity and were further stratified according to the number of TMs they possessed. The low GFP group included 180 GBCs and the high GFP group (equal to or higher than median GFP intensity) included 183 GBCs (n = 3 mice, 363 S24 GBCs quantified, Student’s t-test, columns are mean ± SEM). (H) Representative IV2PM illustrating the dynamic changes of nuclear GFP expression within the infiltration zone of a mature S24 GBC tumor at day 43 and day 64 (z = 40 µm). Dietary Dox administration was started at day 43. On day 43 the neural progenitor-like tumor cell dominated the infiltration zone. On day 64, after 4 weeks of brain colonization in the presence of Dox, GBCs have morphologically evolved into tumor cells now frequently attributed with more interconnecting TMs forming a multicellular tumor network. (I) Measurement of nuclear GFP intensity in mature tumor in vivo in the S24 neural progenitor-like glioma cell as fraction of the corresponding mean nuclear GFP intensity in GBCs with >4 TMs (horizontal line). Analysis was performed as described above for F (n = 3 mice, 109–150 neural progenitor-like GBCs measured; t-test, Mann–Whitney rank sum test, data are presented as mean ± SEM). (J) Analysis of nuclear GFP intensity in a mature S24 GBC tumor on day 64 after 4 weeks of dietary Dox. Quantification as described above for G. The low GFP and high GFP groups each included 173 GBCs (n = 3 mice, 346 S24 GBCs quantified, Student’s t-test). Abbreviations: C, connection; IZ, infiltration zone; TB, tumor bulk.
The perivascular niche can harbor long-term quiescent and particularly resistant glioma cells.17 Proposed mechanisms of GBC brain invasion include perivascular routes along blood vessels, routes along basement membranes and along white matter tracts, and diffuse infiltration through the extracellular matrix.29,30 Therefore, we compared the proliferation and migration histories of tumor cells within the perivascular and parenchymal compartments of the IZ and implanted tumors. Twenty-eight days after implantation, we found no significant difference in either proliferation or migration histories (Supplementary Figure S5), with GFP retention patterns comparable between both compartments. This speaks for general applicability of proliferation and migration being predominantly found in the same glioma cell subpopulation, independent of microanatomical niches.
Single-Cell Analyses
To further characterize the GBCs that morphologically resemble the neural precursor-like cells with 1 or 2 TMs and the predominantly network-connected GBCs with 3 or more TMs, we developed an in vitro monolayer model that reflects phenotypic characteristics observed within the in vivo IZ and tumor core, respectively.31,32 We found a vast majority of GBCs grown under sparse, IZ-like conditions extended 1 or 2 TMs, whereas GBCs with 3 or more TMs were enriched in the dense, tumor-core-like condition (Figure 5A-B). In this monolayer model, S24 GBCs form a particularly highly connected GBC network that closely mimics the characteristics of glioblastoma networks of in vivo model systems and patients. We subsequently assessed the transcriptional profiles of a total of 2691 S24 GBCs with single-cell resolution (RNA sequencing) and obtained 1347 cells. We applied previously described GBC signatures astrocytic-like (AC), mesenchymal-like (MES), oligodendrocyte progenitor-like (OPC), neuronal progenitor-like (NPC) and G1S and G2M cycling states to every cell.6 The frequency of cycling cells (G1S and G2M) was higher for GBCs grown under sparse conditions (Figure 5C). Importantly, the NPC/OPC cell state, the most frequent 1 for invasive cells that typically extend 1 or 2 TMs,1 was also enriched in the sparse condition (Figure 5D). Together this data supports both the OPC/NPC identity and the higher proliferative potential of invasive GBCs.
Glioblastoma cell states are connected to invasivity, and second reporter system. (A) Representative confocal microscopy images of a dense and sparse serum-free S24 GBC monolayer labeled with LipiLight-488 membrane dye. (B) Pie chart illustrating the frequency of neural precursor-like cells based on their ≤2 TM morphology in sparse and dense serum-free S24 GBC monolayer. The rate of S24 GBCs with ≤2 TM is significantly higher under sparse conditions (P < .00001, Fisher’s exact test). (C) All cells were assigned to either G1S or G2M cycling or non-cycling cell states. Frequency of cycling cells (combined G1S and G2M cells) is enhanced in the sparse condition (Fisher´s exact test). (D) Fold change of pooled NPC/OPC population versus AC/MES in sparse and dense conditions after reassigning G1S and G2M GBCs to the best matching non-cycling cell state shows a significant enrichment of the NPC/OPC population in sparse culture conditions (Fisher´s exact test). (E) Scheme of EF1-mVenus-p27K expressing nuclear mVenus in non-cycling cells.26 Proliferating cells are tdTomato+mVenus-; non-proliferating cells are tdTomato+mVenus+. (F) Flow cytometry analyses of S24 serum-free neurosphere cultures stably expressing EF1-mVenus-p27K. The Ki67+ growth fraction within the cycling mVenus- GBC subpopulation and within the mVenus+ GBC subpopulation (a total of 34 742 GBCs quantified). (G) Scatter blot showing the proliferation index (mVenus−) relative to tumor cell density within representative 3D microregions of 400 µm2 and z = 30 µm in GBC xenografts (Spearman rank correlation ρ and P values are given, the colored area along the regression line marks the 95% confidence interval; n = 31 images from 4 mice). (H) Representative IV2PM images of a S24 EF1-mVenus-p27K GBC tumor on day 51 (z = 60 µm). Various tumor cell densities represent areas of the infiltration zone (IZ; top row) and the tumor bulk (TB; bottom row). The infiltration zone was almost exclusively composed of unconnected, GBCs with 1 or 2 TMs. Arrowheads indicate proliferating cells. Abbreviations: IZ, infiltration zone; TB, tumor bulk.
Real-Time Tracking of Proliferation During Brain Infiltration
To confirm our key findings, we used an additional live reporter system (EF1-mVenus-p27K) that harnesses the fluorescent protein mVenus whose expression was rendered strictly dependent on the nuclear cell cycle repressor p27 by co-expression as a fusion protein. Thus, this construct enables us to distinguish actively cycling mVenus negative (mVenus−) from non-cycling mVenus positive (mVenus+) GBCs (Figure 5E). In S24 EF1-mVenus-p27K stably transduced GBC neurospheres, proliferation based on Ki67+ and mVenus− in flow cytometry (Figure 5F) were similar to their wild-type parental (Figure 2I) cell line. Confirming the suitability of this reporter system, mVenus− GBCs were strongly enriched for Ki67 positivity, while mVenus+ GBCs were enriched for low Ki67 expression (Figure 5F). Importantly, actively cycling tumor cells were significantly enriched in the tumor IZ (Figure 5G-H), demonstrating active proliferation during the process of ongoing brain invasion. Particularly the neural progenitor-like GBCs with 1 or 2 TMs1 were frequently found cycling (mVenus−) within the IZ (Figure 5H).
Higher Proliferative Capacity of GBCs from the Human IZ
Lastly, to test whether indications for a positive interrelation of proliferation and migration can be found in the human situation, too, we investigated whether a principal difference in proliferative potential between patient-derived GBCs of the TB compared to GBCs from the IZ can be found in resected tumor material. For this, T1 CE TB areas were separated from T2 FLAIR-positive IZs (Figure 6A). Intraoperative mapping facilitated the isolation and subsequent cultivation of glioblastoma cells from both areas (Figure 6B; n = 54 samples from tumor cores; n = 158 samples from IZs; from 54 patients). Remarkably, tumor cells isolated from the IZ revealed a significantly higher proliferative potential, as determined by increased growth kinetics (Figure 6C) and a reduced expansion time (Figure 6D). This feature remained stable over 5 passages, both in GBCs from the tumor core and the IZ (Figure 6E) and was not related to relevant differences in cell density (Figure 6F). This functional data supports a transferability of the mouse experimental data to the human disease.
Proliferative capabilities of GBCs derived from the resected human infiltration zone are higher than those from the tumor bulk. (A) Representative pre-operative MRI of a glioblastoma patient (left: Axial contrast-enhanced T1-weighted image, right: Axial FLAIR image); insets indicate sampling sites. (B) Scheme illustrates derivation of clinical cell samples. Samples were expanded separately for 5 passages before analysis. (C-F) Proliferation kinetics between passages 5 and 10. (C) Cell growth kinetics comparing patient-derived tumor cells from the core of the tumor-population doublings (PDs) per day (mean 0.19 PD/day) to PD of tumor cells derived from the periphery of the tumor (mean 0.29 PDs/day; tumor bulk (TB) n = 54 samples; infiltration zone (IZ) n = 158 samples from the 54 patients). (D) Quantification of GBC expansion time necessary for 5 passages (GBCs from tumor bulk, mean 32 days; GBCs from infiltration zone, mean 19 days; tumor bulk n = 54; infiltration zone n = 158). (E,F) Differences in proliferation and expansion kinetics were not due to differences in growth kinetic stability through continuous in vitro passaging as determined by linear regression analysis (P = .0556; tumor bulk n = 14; infiltration zone n = 54) or differences in cell density in culture dishes (P = .05531; tumor bulk n = 54; infiltration zone n = 158). Note, a regression index of R2 close to 1 indicates a consistent proliferative activity over time. Abbreviations: IZ, infiltration zone; PD, population doubling of propagated cultures; TB, tumor bulk.
Discussion
The data presented in this manuscript favors both invasive and proliferative traits as a key features of the brain-colonizing, network-unconnected tumor cell subpopulation. We have characterized this subpopulation before as being enriched with neural progenitor-like (OPC- and NPC-like) tumor cell states.1 These insights have implications for our current understanding of how efficient brain colonization by glioblastoma cells is achieved, and how to better target both central traits of malignancy in the future. Our observations are in line with previous transcriptional profiling of orthotopically growing xenografts that revealed an increase in migration/invasion signature in actively proliferating cells, rather than in non-proliferating cells33 and they are also in line with single-cell RNA sequencing data that demonstrated increased proliferation signatures in the NPC- and OPC-like subpopulation of tumor cell states.6 All in all, the live tracking of proliferative features of glioblastoma cells extends the current go versus grow concept: while it is very plausible that go-OR-grow is effective at one particular moment of time, the data presented here demonstrates a go-AND-grow situation during brain colonization.
We demonstrate here that the increasing network integration of isolated invasive tumor cells over time, as shown before1 and confirmed by the findings reported here, is associated with a decrease in tumor cell proliferation. This is interesting because the network-connected subpopulation was also characterized by an increase in resistance to standard treatments such as radiotherapy, chemotherapy, and surgery.7,12–14 In other words, the GBC network provides a niche for slower-cycling tumor cells. Slow-cycling cellular behavior has been associated with cellular resistance to cytotoxic therapy due to decreased genotoxic vulnerability.14,34–41
While the highly invasive and proliferative neural progenitor-like tumor cell is likely an easier target for anti-proliferative chemotherapy,7 it can easily escape surgical resection and likely also focal radiotherapy. The findings of this study add to our understanding of the malignant potential of the neural progenitor-like tumor cell subpopulation, which is insofar an important addition to the resilient, treatment-resistance off more established tumor areas where tumor cell networks prevail.7,12–14,17,18 Thus, we clearly need to devote more attention to the invasive and proliferative tumor cell subpopulation when designing new treatment regimens for patients suffering from glioblastoma in the future. Surgery will continue to be a major player in treating glioblastoma as it can remove the particularly network-connected GBCs; however, all current data indicates that it cannot remove all tumor cell networks which are present in T2-negative brain areas, too, at least in mice. Therefore, even supramarginal resection inevitably leaves behind some tumor cell networks – and even more so the diffusely infiltrated tumor cells.2–4
In summary, this study shows that the network-unconnected and neural progenitor-like GBCs are not only particularly invasive1 but also harbor a particularly high proliferative capacity, making them to the plausible major driver of overall brain colonization. These network-unconnected GBCs are however more susceptible to chemo- and radiotherapy.7,14 It is tempting to speculate that the invasive neural precursor-like cells should ideally be targeted with brain-systemic treatments as early as possible before they transition into an interconnected GBC network.
Funding
German Research Foundation [Deutsche Forschungsgemeinschaft, SFB1389 to M.R., E.J., M.O., V.V., W.W., F.W. and VE1373/2-1 to V.V.], grant from the Bundesministerium für Bildung und Forschung [FKZ03V0785 to B.S., M.G.], Olympia-Morata-Program/University of Heidelberg (to M.R.), Else-Kröner-Fresenius Stiftung [2020-EKEA.135 to V.V.], Heidelberg University (Physician Scientist Program to V.V.).
Acknowledgments
We thank E. Maier and S. Weil for technical assistance and advice and T. Ratliff for reading and commenting on the manuscript. We acknowledge the support of the DKFZ Light Microscopy Facility, the DKFZ animal core facilities. We thank Rainer Will and the Cellular Tools Core Facility, DKFZ Heidelberg, Germany.
Conflict of interest statement
F.W., E.J., M.O. and W.W. report the patent (WO2017020982A1) “‘Agents for use in the treatment of glioma.’” F.W. is co-founder of DC Europa Ltd (a company trading under the name Divide and Conquer) that is developing new medicines for the treatment of glioma. Divide and Conquer also provides research funding to F.W.’s lab under a research collaboration agreement.
Authorship
M.R. and F.W. have full access to all the data and take responsibility for the integrity and accuracy of the data analysis; Concept and design: M.R., K.K.J, B.S., F.W.; Acquisition, analysis, or interpretation of data: M.R., K.K.J., D.C.H., L.R., M.S., L.H., H.M., D.D.A., M.C.S., S.U., E.J., M.O., G.S., M.E.M., V.V., M.G., F.W.; Drafting of the manuscript: M.R., F.W.; Critical revision of the manuscript for important intellectual content: M.R., D.C.H., L.R., M.S., L.H., M.C.S., E.J., G.S., M.E.M., V.V., M.G., N.E., B.S., W.W., F.W.; Statistical analysis: M.R., K.K.J., D.C.H., L.H., M.E.M.; Obtained funding: M.R., E.J., M.O., M.G., N.E., B.S., W.W., F.W.; Administrative, technical, or material: M.R., K.K.J., D.C.H., L.R., M.S., L.H., H.M., T.K., D.D.A., M.C.S., S.U., E.J., M.O., G.S., M.E.M., V.V., M.G., N.E., B.S., W.W., F.W.; Supervision: M.R., E.J., M.O., N.E., W.W., F.W.
Data Availability
The data that support the findings of this study are available from the corresponding author, [F.W.], upon reasonable request.
References
Author notes
Kianush Karimian-Jazi and Dirk C. Hoffmann Contributed equally.
