A standardized and reproducible protocol for serum-free monolayer culturing of primary paediatric brain tumours to be utilized for therapeutic assays

In vitro cultured brain tumour cells are indispensable tools for drug screening and therapeutic development. Serum-free culture conditions tentatively preserve the features of the original tumour, but commonly comprise neurosphere propagation, which is a technically challenging procedure. Here, we define a simple, non-expensive and reproducible serum-free cell culture protocol for establishment and propagation of primary paediatric brain tumour cultures as adherent monolayers. The success rates for establishment of primary cultures (including medulloblastomas, atypical rhabdoid tumour, ependymomas and astrocytomas) were 65% (11/17) and 78% (14/18) for sphere cultures and monolayers respectively. Monolayer culturing was particularly feasible for less aggressive tumour subsets, where neurosphere cultures could not be generated. We show by immunofluorescent labelling that monolayers display phenotypic similarities with corresponding sphere cultures and primary tumours, and secrete clinically relevant inflammatory factors, including PGE2, VEGF, IL-6, IL-8 and IL-15. Moreover, secretion of PGE2 was considerably reduced by treatment with the COX-2 inhibitor Valdecoxib, demonstrating the functional utility of our newly established monolayer for preclinical therapeutic assays. Our findings suggest that this culture method could increase the availability and comparability of clinically representative in vitro models of paediatric brain tumours, and encourages further molecular evaluation of serum-free monolayer cultures.


Establishment of tumour cell monolayers and 3-D tumour spheres from surgical material.
We obtained fresh primary brain tumour tissues from 18 paediatric patients, including 10 medulloblastomas (MB), 1 atypical rhabdoid tumour (ATRT), 2 anaplastic astrocytomas (AA), 2 anaplastic ependymomas (AEP), 1 ependymoma (EP) and 2 pilocytic astrocytomas (PA), and attempted to establish monolayer cultures and 3-D sphere cultures in parallel from these samples. For a summary of patient-related information, see Table 1. In brief, single-cell dissociated cells were seeded in serum-free cell culture medium supplemented with growth factors (referred to as UC, see Material and Methods) in cell culture flasks or 6-well plates without adherence (Ultra-Low ™ ) to enable spheroid formation, or in culture flasks or 6-well plates with adherence (Cell-Bind ™ ) to force cells to grow as adherent monolayers.
Due to the varying volumes of tissues obtained from surgery, the yields of tumour cells from surgical material were highly variable and not comparable between individual tumours. However, even from small tissue samples, >5 × 10 6 viable cells could generally be obtained from each tumour (see tumour cell counts in Table 1). The loose tissue texture and high tumour cell density of medulloblastomas made it possible to obtain a large number of viable cells from small tissue samples, while other tumour types contained more debris and stromal cells, which made it difficult to estimate the true number of pure tumour cells. However, approximately 3 × 10 6 cells were seeded per T75 cm 2 flask or 400 000 cells per 6-well. Fresh medium was added every 3-4 day and cells were passaged when reaching high confluence (adherent cells), or when dark spots (a sign of cell-death due to lack of nutrition) appeared in the spheroids. When cells were transferred to new cell culture vials, at least 1/3 of conditioned medium was added.
All tumour types have sphere-forming capacity on low-adherence surface, but only subsets of high-grade tumours can be propagated as spheres. Traditionally, the neurosphere assay has been utilized as a culture method to preserve the TIC population of the original tumour. When single-cell dissociated tumour tissue was seeded in UC on Ultra-Low ™ adherent surface, all tumours (18/18) initially formed spheres after 1-5 days in culture, despite grade of malignancy (Fig. 1a), if seeded at high confluence (200 000 cells/ml). However, only 65% (11/17) of tumours could be further propagated (>passage 3) as sphere cultures (Table 1). Primary spheres obtained from grade I-II tumours did not proliferate, had a loose consistency and eventually dissociated. Upon passaging, no secondary spheres were formed. In contrast, spheres derived from subsets of medulloblastomas, anaplastic astrocytomas and ependymomas were highly proliferative (Fig. 1b) and could be passaged for >3 generations (see Table 1 and thawed for further propagation and analyses. Over multiple passages, spheres were however difficult to dissociate and consequently difficult to count and seed for assays. High confluence seeding in UC medium on adherent surface enables cell culture establishment from most primary brain tumour subsets. By seeding primary tumour cells on Cell-Bind ™ surface, monolayer cultures were successfully established (>passage 5) from 14/18 (78%) primary tumours, including medulloblastomas, grade II-III ependymomas, high-grade atrocytomas and atypical rhabdoid tumour, see Table 1 and Fig. 2a. High-grade tumours initially formed spheres even in Cell-Bind ™ plates, but spheres attached after some time in culture. For medulloblastomas, it took up to several weeks for spheres to attach and generate monolayer cultures. 2 medulloblastomas (molecularly characterized as group 3 and group 4 respectively 32 ) did not attach, and could only be propagated as sphere cultures. 4 tumours (betacatenin-activated medulloblastoma, anaplastic astrocytomas/ependymomas) could be propagated as monolayers, but not as spheres. 2/18 tumours, both diagnosed as grade I pilocytic astrocytomas, did initially attach to the plate surface, but did not proliferate for >3 passages. These cell cultures could only be further propagated following addition of FBS to the cell culture medium (data not shown).
Proliferation of adherent brain tumour cells (>passage 5) was demonstrated by Ki67 labelling, and assessed by cell counting between passages. Although confluent cultures contained a substantial amount of proliferating cells (Fig. 2b), tumour cells did not generate colonies when seeded at low densities (1-100 cells/96 well plates). Proliferation rate varied between tumour samples, but was generally slow and fluctuated over time. Doubling time in exponential growth phase was estimated to 3 and 13 days respectively for a fast-growing and a slow-growing cell culture respectively (MB-LU 140, AEP-LU 158, see Fig. 2b).
Although cells were generally cultured for 1-2 months before being cryopreserved, several monolayer cell lines (AA-LU 74, MB-LU 140, AEP-LU 149, AEP-LU 158, MB-LU 159) have been passaged for >20 generations at our lab, and frozen and thawed at multiple occasions without apparent effects on cell viability. Ideally, cell cultures should be established from primary material and passaged for 2-3 passages before being cryopreserved. If cells are single-cell dissociated at surgery and frozen without preculturing, they are more difficult to thaw and propagate at a later occasion.
Sphere cultures and adherent tumour cell cultures maintain the phenotypic characteristics of the original tumours. Serum-cultured adherent glioma cell lines phenotypically diverge from their matched sphere counterparts 6 . To determine if monolayer paediatric brain tumour cells maintained their original phenotype in vitro, we used immunofluorecent labelling to investigate the expression of a panel of neural (nestin, nf-200, β -III-tubulin and GFAP) and putative tumour progenitor/stem cell markers (CD15, CD29, CD44 and CD133) in subsets of tumour monolayer cultures. In comparison we labelled the corresponding primary tumours (cryosections) and tumour-derived sphere cultures, and qualitatively scored expression of markers as homogenously or heterogeneously positive (see Table 2). Except for CD15, immature and mature markers were generally not enriched by either culturing method, but were rather expressed in a tumour specific fashion in both spheres and adherent cultures ( Table 2 and representative images in Figs 3 and 4). The neural stem cell marker nestin was highly expressed in all in vitro cultures (Fig. 3a), suggesting an immature phenotype of cultured cells. Neuronal maturation markers (nf-200 and β -III-tubulin) were however detected in subpopulations of cells within spheroids, as well as on occasional cells in adherent cultures (Fig. 3b), implying that some degree of in vitro-induced cell differentiation will occur regardless of culturing method. Primary tumours expressing the astrocyte marker GFAP maintained the expression in corresponding monolayer and sphere cultures (Fig. 3c), whereas GFAP negative primary tumours remained negative in cell cultures (Fig. 3d). CD29 and CD133 were expressed by subpopulations of cells, both in vivo and in vitro (Fig. 4a,b). In contrast, CD44 was expressed by a majority of astrocytoma cells (in vivo) and ependymoma cells (in vitro and in vivo). In the medulloblastoma cryosection, CD44 was less prominent, but both sphere and adherent cell cultures were enriched for CD44 (Fig. 4c). Finally, 2/18 tumours in our cohort (MB-LU 159 and MB-LU 72) displayed nuclear β -catenin staining in cryosections. Both tumours maintained nuclear β -catenin on adherent cells (data not shown).
In contrast to all other markers investigated, CD15 was strictly associated with sphere cultures and no or only occasional cells expressed CD15 in adherent cell cultures (Fig. 4d). 5/7 sphere cultures were highly enriched for CD15, even when CD15 was not expressed in the primary tumour (MB-LU 159). Spheres derived from MB-LU 70 and AA-LU 74 were CD15 negative, as were the primary tumours. The pilocytic astrocytomas (PA-LU 226 and PA-LU 228), which could not be propagated in vitro in serum-free medium, displayed high expression of CD15 in cryosections ( Supplementary Fig. S1). Monolayer tumour cells cultured in UC medium express PGE 2 and its rate-limiting enzyme COX-2 and are suitable for functional assays. COX-2 is the rate-limiting enzyme for production of the key immunosuppressive factor prostaglandin E 2 (PGE 2 ), and we have previously developed immunotherapeutic treatment strategies including COX-2 inhibition in experimental brain tumour models 27,28 . PGE 2 is detected in plasma isolated from paediatric brain tumour patients at the time of surgery (MB-LU 128, MB-LU 140, MB-LU 159, AEP-LU 158; Fig. 5a), suggesting that paediatric patients could benefit from such a treatment approach.
To investigate the functional utility of our newly established monolayer cultures for preclinical assays, we first demonstrated COX-2 expression by immunolabelling of primary tumours and corresponding in vitro cultures from medulloblastomas, anaplastic astrocytoma and anaplastic ependymoma. As shown in Fig. 5b and Table 2, COX-2 expression was detected in all tumour specimens and in the corresponding cell cultures. Next, supernatants from monolayer cell cultures established from three paediatric patients was analysed with ELISA. PGE 2 was secreted at substantial levels in vitro by one ependymoma and one medulloblastoma culture, and secretion was considerably reduced upon treatment with the selective COX-2 inhibitor Valdecoxib (  Wilcoxon matched pairs test). Notably, the experiment was performed at three independent timepoints where cells were cryopreserved and thawed in between, with only small variance in PGE 2 levels between experiments.

Cytokine profiling of tumour tissues and monolayer tumour cells.
To further assess the inflammatory fidelity of monolayer cells to primary tumours, we characterized the secretory profile of cultured medulloblastomas (n = 3), astrocytoma (n = 1) and ependymomas (n = 2), using a high-sensitivity cytokine multiplex assay (Meso Scale Discovery). 100,000 monolayer cells were cultured in 1 ml UC medium between 24 hours and 7 days, and supernatants were collected and analysed for the presence of IL-6, IL-8, IL-15, IL-16, TNF-α and VEGF, which are known mediators of angiogenesis, proliferation, inflammation and immune suppression. In parallel, we characterized the secretome of lysed primary tumour tissue (encompassing both stromal and tumour cells) obtained from two tumour patients (AEP-LU 158 and MB-LU 140). Measurements were performed three times between passage 5 and 15 for each tumour culture.
All in vitro cell cultures and tumour tissues highly expressed vascular endothelial growth factor (VEGF) (Fig. 6a). The medulloblastoma and ependymoma cultures and tissues also secreted IL-6, IL-8 and IL-15, while the astrocytoma culture secreted no or very low levels of these interleukins (Fig. 6b-d). IL-16 and TNF-α were undetectable in all cell culture supernatants (data not shown), but was secreted in both tumour tissues (IL-16: 5337 ± 33 and 3121 ± 94 pg/ml respectively, TNF-α: 0.30 ± 0.11 and 1.07 ± 0.17 pg/ml respectively). Although the exact quantities of cytokines secreted by tissues and cells were not comparable in this experimental setting, it should be noted that the levels of IL-8 was up to 50x higher in cultured cell samples than in a single tissue samples while levels of other cytokines remained comparable (Fig. 6c). IL-6 secretion in MB-LU 140 tissue could be detected, but was below the lower quantitative limit of this assay (Fig. 6b).

Discussion
There is currently a lack of clinically relevant in vitro models of paediatric brain tumours, as well as comprehensive culturing protocols to generate such models. A recent review of the literature estimated the number of established paediatric brain tumour cell lines to approximately 60, of which only a handful are used in a majority of publications 33 . Here, we describe a standardized protocol for establishing and propagating primary paediatric brain tumour cell cultures, by utilizing a simple serum-free cell culture medium and cell culture plastics with high adherence. This approach enabled cell culture establishment from most primary brain tumour subsets, including ones that failed to generate sphere cultures. Monolayer cells could be robustly passaged for an extended time period, displayed phenotypic similarities with the original tumours, and expressed clinically relevant inflammatory factors in vitro.
The traditional approach to establish primary brain tumour cultures as neurospheres is not always successful. Certain brain tumour types, including low-grade tumours and distinct subgroups of high-grade tumours 11,12 have proved to be particularly unsuitable for sphere growth. It has been suggested that sphere-forming ability reflects high tumour stem cell content and poor prognosis in brain tumour patients 12,34 . Similarly, we observed that although all tumours initially formed spheres in high-density cell cultures, 'spheres' derived from low-grade astrocytomas, ependymomas and a beta-catenin activated medulloblastoma (associated with good prognosis 35 ) could not be serially passaged, and initial sphere formation likely reflected fusion rather than sphere-forming ability. In contrast, monolayer cell cultures were successfully established from 78% of primary paediatric brain tumours, including the ones that would not generate neurosphere cultures. This culturing approach could thus be especially useful in establishment of cell cultures from less aggressive tumours, which may otherwise be difficult to propagate in serum-free medium. While monolayer tumour cells grew slowly in serum-free medium compared to traditional serum-cultured cell lines, they could be expanded into numbers sufficient for in vitro assays, xenografting in immunodeficient mice or therapeutic immunizations, in approximately 2-3 months. Notably, most medulloblastomas in our cohort also generated robust sphere cultures, and two medulloblastomas (group 3 and 4 respectively 32 ) in our cohort could only be propagated as spheres and did not attach to poly-L-lysine-coated plastic (data not shown) or Cell-Bind ™ . The proneness of primary glioblastomas to generate monolayer and sphere cultures respectively has been reported to reflect molecular subgroup 11 and the same may be true for medulloblastomas. While the neurosphere assay has long been considered the golden standard for preserving the TIC population of the primary tumour, recent data has shown that adherent brain tumour cultures maintain tumorigenic capacity when cultured in serum-free medium 16,36 . Attempts have been made to culture adherent stem and tumour cells on different attachment substrates, such as laminin, collagen and poly-lysine. The choice of substrate dictates cell fate, and the optimal substrate for maintaining monolayer tumour stem and progenitor cell cultures is yet to be determined. While coating with laminin maintains NSCs in vitro 15 , other substrates induce neuronal differentiation of stem cells 37 . Here, we have utilized cell culture flasks or 6-well plates with high adherence plastic (Cell-Bind ™ ), which is a cheap and standardized alternative to biological coating. Although the current study does not provide conclusive evidence for preservation of TICs, we find that our monolayer cultures display extensive self-renewal capacity and are immunophenotypically similar to sphere cultures over serial passages. Additional studies should be conducted to assess if cells also maintain tumorigenic capacity in vivo.
CD15 has previously been suggested to be a marker for TICs in adult glioblastomas, ependymomas and medulloblastomas [38][39][40] . Interestingly, we find that CD15 is highly enriched on most cultured tumour spheres even when it is not expressed in the primary tumour, suggesting upregulation following sphere formation. In contrast, low malignant pilocytic astrocytoma cryosections displayed large areas of intense CD15 labelling, but corresponding cultures did not form spheres in culture. In consistency with our  findings, previous studies have also demonstrated an intense CD15 labelling in low malignant human brain tumours, and a lack of CD15 positivity in subsets of high malignant brain tumours 38,40,41 . In addition, we found that monolayer cells were generally CD15 negative, but did express nestin and did not display considerable expression of maturation markers compared with corresponding spheres. The most frequently used CD15 antibody clones in the literature (Hi98, MC-480 and MMA) react with the cell surface carbohydrate epitope 3-fucosyl-N-acetyllactosamine, which is expressed not only on NSCs and potential TICs, but also on mature human astrocytes and neurons 42,43 . Altogether, these findings question the role of CD15 alone as a robust TIC marker in human brain tumours, and also indicate that alterations in CD15 expression could be related to adherence in addition to cellular maturation state.
Immunotherapeutic approaches for malignant brain tumours have shown limited clinical efficacy, despite promising experimental results, partly due to the poor resemblance between preclinical models and primary tumours. In vitro culturing in serum-containing medium has been shown to change the immunological phenotype of brain tumour cells, including upregulation of MHC class I and ICAM-1, and altered cytokine secretion 44 . In the current study, we have utilized Lonzas Ultra-Culture serum-free medium, which has previously been shown to preserve tumour antigens of primary human glioma cells to be utilized for clinical immunotherapy 7 . We screened a subset of monolayer cultures for secretion of a panel of inflammatory factors, and compared to the secretory profile of fresh frozen unsorted tumour tissue comprising both tumour and stromal cells. All brain tumour cultures investigated in this study displayed high levels of VEGF and PGE 2 , which are well known therapeutic targets and mediators of angiogenesis and immune suppression in malignant paediatric brain tumours 45,46 . We also report IL-15 secretion by brain tumour cells in vivo and in vitro. IL-15 is ubiquitously expressed in the brain and promotes neuron survival and neuroinflammation in multiple sclerosis and experimental autoimmune encephalomyelitis 47,48 , although its functions in CNS malignancies are not known. Moreover, we detected substantial secretion of IL-8 in medulloblastoma and ependymoma tissues and cultures, confirming previous reports of IL-8 expression in these tumour types in vivo 49 . The quantitatively high overall levels of IL-8 in vitro however indicate that IL-8 secretion may be partly enhanced by cell culturing. The same could be true for IL-6, which was detected in one medulloblastoma cell culture where IL-6 was not detected at significant levels in the corresponding primary tumour. Finally, IL-16 and TNF-α were secreted at high levels in tumour tissues, but were undetectable in all tumour cultures, in consistency with reports that IL-16 and TNF-α expression in human and mouse gliomas predominantly originates in tumour-associated macrophages/microglia 50,51 .
Regardless of culture method, tumour cells indisputably change when transferred to in vitro conditions. Concerns have been raised whether adherent brain tumour cell lines in serum-free conditions are equivalent to corresponding sphere cultures 52 , but experimental evidence is scarce at this point. Nevertheless, monolayer cell culturing is a standard method at most laboratories and the practical benefits of using adherent cells rather than spheres are easily appreciated. Our results so far show that this is a feasible method for establishment, propagation and assessment of paediatric brain tumour cells, and encourages further molecular studies of serum-free monolayer cultures.

Methods
Patients and tumour sampling. All experiments were performed in accordance with national regulations and were approved by the Local Ethical Review Board of Lund, Sweden (ETIK2008/642) and the Ethical Research Board of the Medical Faculty at Lund University, Lund, Sweden (serial number LU1028-03). Brain tumour and blood samples were obtained from 18 paediatric patients (<18 years) at the time of surgical resection, and tumours were histologically classified according to WHO criteria. For a summary of patient information, see Table 1. All patients and/or their parents gave their informed consent prior to inclusion in the study.
Tumour dissociation and cell culturing. Parts of the fresh tumour tissue was prepared for mesoscale analysis or cryosectioning, see below. The remaining part of the tumour material was single-cell dissociated for cryopreservation and cell culturing. Tumours were cut in small pieces using a sterile scalpel-blade in a petridish containing Accutase ® (Sigma-Aldrich, Stockholm, Sweden) or TrypLE ™ Express (Gibco ® , Life Technologies, Stockholm, Sweden) and incubated for 15 minutes at 37 °C. Tumours were then further dissociated in 5 ml cell culture medium by pipetting using a 5 ml serological pipet, and mechanically minced through a BD Falcon ™ cell strainer (75 μ m) (BD Biosciences Pharmingen, Stockholm, Sweden) using a 2 ml syringe plunger. Cell culture medium was used to rinse the strainer to approximately 8 ml in total of cell suspension. Cells were centrifuged for 8 min at 1200 rpm. The supernatant was removed and cells were resuspended in medium and counted.
For thawing, cell vials were held in running warm water until the frozen pellet disappeared. The cell suspension was transferred to 2 ml cold UC and kept on ice for 5 min. 2 ml cold UC was added to the cell suspension, and cells were left on ice for another 2 min., followed by a final addition of 5 ml cold UC. Cells were centrifuged for 8 min. at 1200 rpm, resuspended in UC and seeded in cell culture vials as described above.
Proliferation and clonogenicity assays. Adherent tumour cells >p5 were seeded in either T75 75 cm 2 Cell-Bind ™ cell culture flasks at a concentration of 3-4 × 10 6 cells/10 ml UC, or in 6-well Cell-Bind ™ cell culture wells at a concentration of 150 000-400 000 cells/2 ml UC. Cells were passaged as described above, counted every 10 days for 40 to 60 days and seeded in new cell culture vials. Doubling-time was calculated according to the following formula; DT= Tln2/ln(number of cells end/ number of cells start) where DT is doubling-time, and T is total number of days.

Immunofluorescent labelling of tumour cryosections and cultured cells. Fresh tumour tissue
was snap-frozen in isopentane, sectioned and stained as previously described 53 . For immunofluorescent labelling of cultured tumour cells, 8-well multichamber slides (BD Biosciences) were coated with poly-L lysine (GIBCO ® , Life Technologies) for 4 h, and spheres (>p3) or adherent cells (>p5) attached overnight in a cell culture incubator. Cells were fixed in 4% paraformaldehyde for 30 min, incubated with 0.3% Triton-X100 for 5 min when staining for intracellular markers, blocked with 5% goat serum (DAKO, Stockholm, Sweden) and incubated with primary antibodies for 2,5 h at 37 °C. Finally, cells were labelled with a secondary antibody for 30 min and mounted with ProLong Gold anti fade containing DAPI (Molecular Probes, Life Technologies). Alternatively, tumour spheres were pelleted, embedded in Cryomount (Histolab Products AB, Gothenburg, Sweden), sectioned and stained as described for cryosections.