Introduction
Low-grade gliomas (pLGG) are the most common pediatric brain tumors and account for about 30% of all pediatric brain tumors [
1]. pLGG, commonly displaying a hallmark MAPK activation, comprise a set of different WHO grade I and II entities [
2]. As in other fields of cancer research, pLGG cell lines suitable for
in-vitro preclinical experiments are needed for drug development and in-depth mechanistic understanding. In sharp contrast to the high incidence of pLGG, in particular of pilocytic astrocytomas (PA; WHO grade I), the most common subgroup of pLGG, the number of available reliable patient-derived preclinical models is low [
2,
3]. Only a handful of patient-derived pLGG
in-vitro models have been published. The BT-40 cell line [
4] harbors a
BRAF V600E-mutation and a
CDKN2A-deletion and molecularly reflects a pleomorphic xanthoastrocytoma (PXA) WHO grade II. The PA patient-derived cell line Res186 [
5] was shown to have a
PTEN-deletion but no MAPK-alteration, findings untypical for PA. The recently published cell line JHH_NF1_PA1 [
6], derived from a 14-year-old NF1-patient´s PA, was established using a conditional reprogramming cell culture approach to sustain growth of senescence-prone cells. Loss of
NF1, absence of atypical alterations (such as
ATRX-loss or
CDKN2A-loss) and genetic stability over time support JHH_NF1_PA1 to be a true PA model. Other approaches to culture patient-derived pLGG cells beyond classical culture methods include mouse brain-slice overlay cultures [
7] or synthetic extracellular matrices with astrocytes [
8]. These latter techniques amend the portfolio of preclinical pLGG models but are not ideally suitable for large-scale comprehensive preclinical testing.
A main reason for the failure of PA-cultures in conventional cell culture approaches adapted from higher-grade tumors [
6] is oncogene-induced senescence (OIS), described in primary PA and PA short-term cultures [
9]. To surmount the obstacle of pLGG paucity we used Simian Virus 40 Large T antigen (SV40-TAg) to reversibly circumvent OIS in primary PA cultures [
10]. The generated patient-derived PA cell line, DKFZ-BT66, turned out to be a valuable tool for preclinical drug testing and studies on OIS [
11‐
13]. DKFZ-BT66 was the proof of concept that pLGG-model generation from primary material using SV40-TAg is feasible. However, the success of generating PA cell lines using SV40-TAg was not easily repeated. To improve the model establishment efficiency, we sought to identify factors that influence the success of generating PA cell lines using SV40-TAg, and investigated sample intrinsic (quality of the sample; e.g., sample size, origin, tumor cell content) and extrinsic (quality of the workflow, e.g., culture conditions, enrichment for tumor cells, efficient viral transduction) factors. To this aim, we standardized the modeling workflow and implemented ddPCR, a technique that only requires minimum amounts of genomic DNA input [
14], as a tool to closely monitor each step of the culturing process. Here, we report on the performance and outcome of this workflow as well as factors influencing modeling success.
Materials and methods
Patient samples
Primary pLGG tumor material was collected during therapeutic surgical intervention in Heidelberg or Tübingen, Germany. Samples were transferred into shiping vials containing unsupplemented NeurobsaI™ medium (21,103,049, Gibco) and stored at 4 °C before processing. Samples from Tübingen were shipped overnight at room temperature. Informed consent for sample collection, use of material and clinical data was obtained within the study S-304/2014 (V2/V3), approved by the institutional review board of the University of Heidelberg. The primary patient tumor material was molecularly analyzed (DNA-methylation and gene panel sequencing) within the PTT2.0 study [
15] or the LOGGIC Core co-clinical biobank [
16]. Methylation scores were obtained via the brain tumor classifiers (V11b4, 12.3 or 12.5) (
www.molecularneuropathology.org).
Primary culture and cell lines
Processing of primary samples and HEK293T cells was described before [
17]. Fully supplemented Astrocyte Basal Medium (ABM) [
17] or Neural Stem Cell Medium (NSM) (500 ml DMEM/F12 (1:1) + GlutaMAX (31331-028, Gibco) supplemented with 10ml B27-Supplement w/o vitamine A (12587-010 Gibco), 20ng/ml Human bFGF (100 µg/ml in PBS; AF-100-18B Peprotech) and 20ng/ml Human EGF (100 µg/ml in PBS; AF-100-15 Peprotech) was used for primary culture.
A2B5 selection
A2B5 positive selection was performed following manufacturer´s instructions with anti-A2B5 MicroBeads (human, mouse, rat; #130-093-392) on a MACS® MultiStand (#130-042-303) using MS Columns (#130-042-201), all from Miltenyi Biotec, Germany. Prior to magnetic separation, tumor cell suspensions were treated with 20 µl of FcR Blocking Reagent (#130-059-901) for 10 min and subsequently with 20 µl of Anti-A2B5 MicroBeads for 15 min. After magnetic separation, cells (from both fractions, positive-selected and flow through, respectively) were centrifuged at 300×g for 10 min, immediately resuspended in culture medium (ABM or NSM) and transferred to a 37.0 °C incubator with humidified atmosphere and 5% CO2.
pCW57.1 dsGFP-TAg plasmid, virus production and lentiviral transduction
The generation of the doxycycline-inducible lentiviral expression vector for expression of SV40 large T (SV40-TAg), pCW57.1 dsGFP-TAg, and generation of viral supernatant was described in detail before [
17]. Virus titers were determined after infection of HEK293T cells with serial dilutions of the viral supernatants by counting GFP or RFP positive colonies. Primary pLGG cultures were incubated with pure viral supernatant for six hours before top-up with target cell medium (ABM or NSM). Medium change to full target cell medium was performed after 24 h. The maximum number of sequential infections per sample and condition was three.
gDNA extraction and ddPCR
Genomic DNA of pLGG tumor cell suspensions was extracted with the QIAamp DNA Mini Kit (Qiagen; #51304). DNA was eluted with 20 µl nuclease-free water.
ddPCR for the detection of the
BRAF-duplication and the
BRAF V600E-mutation was performed as described before [
14,
17]. Determination of tumor cell fraction (TCF) based on ddPCR results was done as follows 1)
BRAF-duplicated tumors: BRAF-duplicated tumor cells have a BRAF exon 14 copy number of 3. Microenvironmental non-tumor cells have a BRAF exon 14 copy number of 2. The mean copy number of BRAF exon 14 in a given mixture of tumor cells and non-tumor cells the is, depending on the TCF, between 2 and 3. Accordingly, the TCF could be calculated using the formula:
$$2 - BRAF\,exon\,14\,copy\,number$$
The copy number of BRAF exon 14 was calculated using the ratio of BRAF exon 14 copies (duplicated in case of BRAF-duplication) to BRAF exon 3 (reference; not duplicated in case of BRAF-duplication) using the formula:
$$\frac{{Concentration\,BRAF\,exon\,14\,\left[ {copies/\mu l} \right]}}{{Concentration\,BRAF\,exon\,3\left[ {copies/\mu l} \right]}}\, \cdot 2$$
2)
BRAF V600E-mutation: The calculation of TCF in a
BRAF V600E-positive tumor sample assumed heterozygosity of the BRAF-V600E mutation. It was calculated using the ratio of the concentration of
BRAF V600E copies to the concentration of BRAF wildtype (WT) copies. applying the following formula:
$$\frac{{\frac{{Concentration\,BRAF\,V600E\,\left[ {\frac{{copies}}{{\mu l}}} \right]}}{{Concentration\,BRAF\,WT\,\left[ {\frac{{copies}}{{\mu l}}} \right]}}\, \cdot \,2}}{{\left( {1 + \,\frac{{Concentration\,BRAF\,V600E\,\left[ {\frac{{copies}}{{\mu l}}} \right]}}{{Cconcentration\,BRAF\,WT\,\left[ {\frac{{copies}}{{\mu l}}} \right]}}} \right)}}$$
To calculate the TCF in % the TCF determined using the formulas above was multiplied by 100.
Graphs, flow-charts and statistics
Flow-diagrams were generated with Biorender (
https://www.biorender.com) using the Hopp Children´s Cancer Center´s institutional account. Differences between two groups were compared in R Studio (R Version 4.1.0) using an unpaired
t test in the “stats” package. Comparisons between multiple groups were done using one-way ANOVA followed by Tukey’s ‘Honest Significant Difference’ test in the “stats” package. Graphs were generated using R package “ggplot2” (v 3.3.5). Boxplots visualize the minimum value, the first quartile (25th percentile), the median, the third quartile (75th percentile) and the maximum value.
Discussion
Expression of SV40-TAg in primary PA cells allows to circumvent OIS and generate proliferating PA cell lines. The resulting PA
in-vitro models are of high translational significance: As shown before [
10,
17], the generated cell lines clearly reflect PA biology, even though they are genetically modified. While expression of SV40-TAg may result in different outcomes (e.g. cell death, apoptosis, transformation or no effect) in different tissues [
20], for pLGGs we have previously shown that SV40-TAg overcomes OIS but does not transform or immortalize PA cells. SV40-TAg transduced PA cells have a stable genome, re-enter OIS upon SV40-TAg withdrawal and enter replicative senescence at the end of their lifespan. Unlike any other PA model to date, inducible SV40-TAg models allow to also study the senescent state of the cells after re-induction of OIS. Long-term expandability overcomes the limitation of low proliferation rates and short lifespans observed in other primary modeling approaches [
21]. SV40-TAg PA models are less cost effective compared to e.g. cultures in synthetic extracellular matrices [
8]. Taken together, they are excellent and unique tools for standardized studies of tumor cell intrinsic mechanisms and preclinical drug testing to uncover new therapeutic approaches, such as rational combination treatments [
12] or senolytic BH3-mimetics [
17].
Only a fraction of primary PA samples were successfully converted into proliferating tumor cell lines using the SV40-TAg approach in our hands. In the present cohort, monitoring by ddPCR was applied, aiming at a more profound understanding of modeling processes and identification of determinants of success and failure. Of note, the ddPCR assays used detected BRAF-duplications and the BRAF V600E-mutation. Two of the 18 primary samples (11%) (pLGG8 and pLGG10) had different MAPK-alterations and therefore could not be monitored by ddPCR. Furthermore, due to small sample sizes and resulting low amounts of material, gDNA-samples for ddPCR could not be collected at all intended timepoints for all samples.
ddPCR identified the underlying MAPK-alteration in all 13 tested primary PA suspensions (p0). TCF in the primary suspensions was median 55% and in line with published tumor/microenvironment ratios in PA [
18,
22]. This confirmed a high quality of the primary surgical samples and excluded quality of the primary samples as a major factor of modeling success and failure.
As expected, and in line with previous observations [
6] none of the 18 primary samples gave rise to a proliferating cell lines without SV40-TAg transduction. Successful transduction was therefore a prerequisite for model generation. The plasmid pCW57.1 dsGFP-TAg yielded titers that were consistent throughout different preparations and therefore unlikely to be a major variable of modeling success. The comparably low titers obtained were most likely due to the large SV40-TAg insert (> 8kbp) which reduces the lentiviral packaging efficiency [
23]. The supernatant allowed for successful transduction of 59% of the infected primary cultures in ABM. This rate is lower compared to published studies in glioblastoma for example, where infection was shown to be successful in 100% of primary samples [
24]. This difference might be explained by the different biology of pLGG and the aforementioned low viral titers. Only primary cultures in ABM but not under stem cell conditions (NSM) were successfully transduced. Transduction protocols especially adapted to stem cell conditions [
25] might therefore be an option to successfully transduce primary pLGG cells cultured in NSM. NSM was not shown to be inferior in its ability to maintain tumor cell survival in short-term culture and successful transduction of these cultures might increase the number of successful models.
Sample intrinsic factors like sex of the patient, tumor localization, genetic alteration or initial size of the sample likely did not influence the modeling success. Initial TCF of the sample was not a factor indicative of later success. A priori enrichment of TCF by e.g. A2B5-selection is therefore not necessary or advisable and we did not focus on improvement of this selection process. The age of the patients at sampling was associated with modeling outcome. Only samples from younger patients (≤ 5 years) generated pLGG-models after infection. In line with this observation, our first reported PA cell line (DKFZ-BT66) was also derived from a young patient (two years) [
10]. Of note, a young age alone was not a guarantee for modeling success because some samples from patients of young age also failed. An earlier study had shown that pLGGs from young patients (< 4 years) had significantly longer telomeres compared to older patients (> 10 years) [
26] and older patients had lower chance to relapse. Our own previous cell line study proved dependence on replicative senescence in later passages of DKFZ-BT66 despite SV40-TAg expression [
10], indicating that SV40-TAg can override OIS but not replicative senescence in PA. This might explain, why model generation based on SV40-TAg expression is more likely to be successful in samples from younger aged patients, because their tumor cells still inherit a higher replicative capacity due to longer telomeres. Moreover, our limited available data indicated a generally stronger ability of tumor cells from younger patients to survive/proliferate short-term in culture without SV40-TAg. This points at a comparably higher intrinsic proliferative capacity in the very first days of culture in younger PA cells. Although lentiviral gene-transfer is known to be successful in non-dividing cells [
27] the higher proliferative capacity of younger PA cells in turn might create a more susceptible state for successful transduction and outgrowth of PA tumor cells over co-transduced microenvironmental cells in the primary culture.
Taken together, based on the data presented in this study patient age is the only major sample intrinsic factor influencing success rate of PA cell line generation using SV40-Tag. Several extrinsic factors contributing to a higher likelihood of successful model establishment, like type of medium, efficacy of transduction and maintenance of TCF, should be considered to optimize the outcome. Monitoring of PA cultures by ddPCR is strongly recommended to control for quality.
Publisher’s Note
Springer nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.