The lncRNA TP73-AS1 is linked to aggressiveness in glioblastoma and promotes temozolomide resistance in glioblastoma cancer stem cells

Importantly, we demonstrate that high expression levels of TP73-AS1 are associated with poor prognosis in three nonoverlapping, independent GBM cohorts (Fig. 1c). By analyzing TCGA survival and expression data, we found that high TP73-AS1 expression is correlated with poor patient outcome in the “mesenchymal” and “proneural” subtypes but did not correlate with survival in the “classical” or “neural” subtypes (Fig. 1d). In glioma, high expression of TP73-AS1 is significantly correlated with poor patient outcome in all tested cases, namely, IDH wt and mutant tumors, astrocytoma, and oligodendroglioma (Fig. 1e).

Collectively, these data suggest that TP73-AS1 comprises a clinically relevant lncRNA, potentially contributing to the aggressive tumor biology of GBM.

We interrogated published data established with patient-derived gCSC growing under adherent conditions^4,25–27. We found that TP73-AS1 is highly expressed in gCSC and neuronal stem cells, as compared with the normal cortex (Fig. 2b). Due to the high expression of TP73-AS1 in gCSC, we subsequently used a gene silencing approach in these cells to study the functions of TP73-AS1.

In order to examine the role of TP73-AS1 in gCSC, we used two established gCSCs, G26 and G7^4,25. As the biological function of some lncRNAs is tightly linked to their transcription^28–30, we performed CRISPR inhibition (CRISPRi) as a gene knockdown (kd) approach^31,32. In CRISPRi, a guide RNA (gRNA) targeting the region close to the transcriptional start site of a specific gene is expressed in a cell. The gRNA binds to a DNAse dead mutant CAS9 (dCAS9) fused to the transcriptional inhibitor domain KREB and the complex silences transcription of the targeted gene. We used a lentivirus approach to generate a stable gCSC, G26, and G7, expressing a doxycycline-inducible dCAS9-KREB³³ in combination with either one of the two gRNAs targeting TP73-AS1 (g#1 and g#2) or scramble controls (gSCR). Indeed, following doxycycline induction, qRT-PCR analysis showed significant reduction in TP73-AS1 expression in the gCSC models expressing either one of the two gRNAs as compared with controls (Fig. 2c).

The lncRNA TP73-AS1 has an ~200-bp 3′ overlap with the 3′ untranslated region of an adjacent gene, TP73 (p73) (Fig. 2a). As a member of the p53 family of transcription factors³⁵, p73 has been implicated in several biological categories, in particular in cancer³⁶, brain development³⁷, and importantly, in the stemness of neuronal progenitor cells^38,39 and invasiveness of GBM tumor cells⁴⁰. Because many lncRNAs are known to directly regulate the expression of adjacent genes^28–30, we investigated whether TP73-AS1 regulates the expression of TP73. First, we tested whether silencing of TP73-AS1 affected the expression of TP73 in our gCSC models. However, we were not able to detect consistent changes in TP73 expression. While in G7 cells, TP73-AS1 kd led to a reduction of TP73 expression, but the levels of TP73 in G26 cells decreased upon silencing of TP73-AS1 using g#1, while it increased in g#2 cells (SUP Fig. 1a).

Next, we asked whether the expression of TP73-AS1 is correlated with that of TP73 in expression data obtained from various (627) human cell lines and two primary GBM tumor datasets. We found that the expression of these two genes is not correlated (SUP 1b). We therefore conclude that it is unlikely that TP73-AS1 directly regulates the expression of its neighboring geneTP73.

Treatment with TMZ is the most common therapeutic approach used for GBM patients; therefore, it is critical to better understand the mechanisms underlying TMZ resistance in gCSC⁴¹. Because the expression of TP73-AS1 is negatively correlated with patient survival, we asked whether TP73-AS1 is involved in TMZ resistance in gCSC.

To confirm these findings, we measured the viability of G7 and G26 gCSC treated with an escalating dose of TMZ for 5 days and found that the kd of TP73-AS1 in both gCSCs led to an increased sensitivity to TMZ (Fig. 4c).

We investigated the impact of TP73-AS1 on gCSC proliferation using a BrdU assay. We found that there was no significant difference between TP73-AS1 kd and control cells under resting conditions, and that proliferation was inhibited in all cases 2 days following TMZ treatment (SUP Fig. 3a). In addition, we measured reactive oxygen species (ROS) in TP73-AS1 depleted and control gCSC cells, using DCFDA (chloromethyl-2′,7′-dichlorofluorescein diacetate) and FACS, and found that in G7 cells, there were no differences in ROS levels in all tested conditions. In contrast, TP73-AS1 depletion resulted in increased ROS levels following 2 days of TMZ treatment and in non-treated G26 cells (SUP Fig. 3b).

These data indicate that TP73-AS1 contributes to TMZ resistance in gCSC and that the protective functions of TP73-AS1 are not linked to regulation of proliferation. Furthermore, TP73-AS1 protective functions in gCSC may be due to curbing ROS levels in a cell-line-specific manner.

O-6-Methylguanine-DNA methyltransferase (MGMT) is known to repair DNA damage caused by alkylating agents, including TMZ, and it is regulated in GBM at the level of promotor methylation⁴², and MGMT methylation status is linked to the response to TMZ in patients. We asked what is the status of the promotor of MGMT in our gCSC using published data^43–45 and found that the promoter is methylated in both cell lines (SUP Fig. 3c), suggesting that the increased sensitivity of the TP73-AS1 kd cells is not likely to be related to MGMT methylation.

To learn more about the cellular pathways interacting with TP73-AS1, we compared the transcriptome of TP73-AS1 kd vs. control G7 and G26 gCSC upon TMZ treatment and control conditions using RNAseq. According to our principal component analysis (PCA), we found that all three biological replicates from each experimental condition grouped together indicate that the major differences between samples are due to the biological perturbations caused by the knockdown of TP73-AS1 or TMZ treatment (SUP Fig. 4a).

Next, we asked which of the transcripts are affected by TP73-AS1 kd in each cell line upon TMZ treatment or in control conditions (SUP Table 1). To this end, we clustered the genes, which are significantly (DeSeq2 with a cutoff of the adjusted p value < 0.05) up- and downregulated upon TP73-AS1 kd. We found that there were profound transcriptional changes affecting over 2000 genes that were significantly changed in each experimental condition. For validation, we chose four transcripts, measured their expression using qRT-PCR, and found that their expression pattern matched the pattern found in the RNAseq data (Sup Fig. 5 and Fig. 6a).

Under TMZ treatment, genes belonging to neuronal differentiation and function were also upregulated in TP73-AS1 kd cells in both gCSC models (Fig. 5c and SUP Table 2). Interestingly, transcripts related to metabolism, mitochondria, and nucleotide metabolism in particular, were downregulated upon TP73-AS1 kd in both cell lines (Fig. 5c and Table 2). These findings may explain how TP73-AS1 kd leads to increased sensitivity to the alkylating agent TMZ in gCSC, as chemotherapy resistance is tightly linked to cellular metabolism⁴⁶. Together, our transcriptomic analysis reveals that TP73-AS1 is a major regulator in gCSC by orchestrating multiple cellular pathways, including neuronal and metabolic pathways, providing protection against TMZ.

To test if ALDH1A1 downregulation explains, at least in part, the protective functions of TP73-AS1 in gCSC, we treated gCSC with the TMZ (300 μM) in the presence or absence of the ALDH1 inhibitor, DEAB (200 μM), for 5 days, after which we measured cell death using PI and FACS (Fig. 6b). We found that DEAB increased the sensitivity of G7 cells to TMZ, confirming the protective functions of ALDH1 in gCSC and providing support to the model where TP73-AS1 enhances the resistance of gCSC to TMZ by promoting the expression of ALDH1A1.

G26 and G7 were a kind gift from Steven Pollard. Cells were grown on plates coated with Laminin (Sigma-Aldrich, L2020), as detailed in ref. ⁴.

Cells were cultured and differentiated in a 96-well plate covered with poly-L-ornithine hydrobromide (Sigma-Aldrich; P3655) and Laminin (10% in PBS). Cells were grown overnight, after which they were fixed with 4% paraformaldehyde (paraformaldehyde 16%, Alfa Aesar) in PBS for 30 min and permeabilized with permeabilization buffer (0.5% Triton X-100 (Sigma-Aldrich) in PBS). The cells were washed with washing solution (25% BSA (bovine serum albumin), Sigma-Aldrich) and 0.1% Tween-20 (Sigma-Aldrich, P9416; PBS). The cells were blocked for 1 h with 10% goat serum in wash buffer and subsequently in washing solution. The blocking buffer was removed and the cells were labeled with primary antibodies (anti-nestin 1:100; Abcam (10C2); anti-TUBIII 1:75 (Merck (MAB1637); anti-GFAP 1:100 (Santa Cruz (2E1)) diluted in blocking buffer and incubated overnight at 4 °C. The cells were incubated at room temperature (RT) for 30 min and then washed using washing solution. A secondary antibody (ENCO, Alexa 549) was added (1:200 in blocking solution) for 1 h at RT. The cells were washed with the washing solution and stained with DAPI for 1 min. The cells were washed three times with DDW. Images were taken using a fluorescent microscope.

The BrdU proliferation assay was performed using BrdU Cell Proliferation Assay Kit (Biovision Incorporated) following the manufacturer’s instructions. Cells were cultured in 96-well plates covered with poly-L-ornithine and laminin (10% in PBS) overnight, after which the cells were labeled with 5-bromo-2-deoxyuridine (BrdU) for 3 h and stained. Cells were visualized under a fluorescent microscope and the number of BrdU-positive cells and the total cell number were determined in each field.

The RNA from cultured cells was purified using the PureLink RNa Mini Kit (Thermo Fisher Scientific). From a starting amount of 100–200 ng of RNA, cDNA was generated using the cDNA synthesis kit (E6300S, BioLabs). Briefly, 20 µl of the cDNA synthesis reaction was subjected to the following conditions: 5 min at 25 °C, 60 min at 72 °C, and 5 min at 96 °C. cDNA is synthesized in vitro, from an mRNA template using the enzyme reverse transcriptase, resulting in single-stranded cDNA production. qPCR was carried out on the produced cDNA to allow for detection of mRNA expression levels in the cells of interest. The PCR conditions were as follows: initialization step at 95 °C for 10 min, 46 cycles of the denaturation step at 60 °C for 30 s, annealing step at 60 °C for 30 s, and elongation step at 72 °C for 1 s, followed by the final elongation at 40 °C for 10 min to ensure that any remaining single-stranded DNA is fully extended. Most of the forward and the reverse primers were designed to span different exons whenever possible. All primers were purchased from Integrated DNA Technologies (IDT). qRT-PCR was performed using an iCycler (Bio-Rad Laboratories) with a threshold cycle number determined with the use of iCycler software version. Reactions were performed in triplicate and threshold cycle numbers were averaged. The results were normalized to L32 and POL2.

TP73-AS1: probe #29, primer1 ctccggacactgtgttttctc, primer2 tcttttaaggcggccatatc; P73: probe #60, primer1 cacgtttgagcacctctgg, primer2 cgcccaccacctcattatt; L32: probe #33, primer1 gcacactgactacagccttga, primer2 tacccaggtttggaggtgtg; ALDH1A1 #14: primer1 tttggtggattcaagatgtctg, primer2 cactgtgactgttttgacctctg; FNDC5 #8: primer1 taccaaaacaccccttctgg, primer2 tcttcctgtccgtggtgaat; ZNF536 #22: primer1 gagtcccagtcggtgagc, primer2 gctctcctcggtgacgttag; WNT5A #11: primer1 ttctggctccacttgttgct, primer2 gccaaagccactaggaagaac.

psPAX2 was a gift from Didier Trono (Addgene plasmid # 12260). pMD2.G was a gift from Didier Trono (Addgene plasmid # 12259). pHAGE TRE dCas9-KRAB was a gift from Rene Maehr and Scot Wolfe (Addgene plasmid # 50917). pLKO.1-puro U6 sgRNA BfuAI large stuffer was a gift from Scot Wolfe (Addgene plasmid # 52628). We used pLKO.1-puro U6 sgRNA BfuAI large stuffer to clone our gRNA sequences according to the instructions available on the addgene page. The sequences used were as follows:

scramble/control: ACCGCGCCAAACGTGCCCTGACGG; TP73-AS1 g#1: GCAGTCGGGGCTGACGGCGG; TP73-AS1 g#2: CCTAGATGGGAGCCGGGGAT.

Lentiviruses were generated using our standard protocol in 293T cells, as detailed in HACE1 reduces oxidative stress and mutant Huntingtin toxicity by promoting the NRF2 response⁶⁴. Cells were grown in DMEM, 10% FBS, and transfected using CalFectin (Signagen) and plasmids at a ratio of 1:2:3 (PAX2; pMD2.G; transfer vector). Media was changed after 24 h and the viruses were collected after 48 h. The collected viruses were stored at −80 °C. G7 and G26 cells were grown to 80% confluence on 6-cm dishes, then infected with a vector ratio of 1:10 (cas9, gRNA), and left for 24 h in a 37 °C incubator. Following infection, the lentivirus-containing medium was removed and replaced by fresh medium and incubated at 37 °C with the selection antibiotic using puromycin (1 μg/ml) and G418 (1500 μg/ml).

The depletion of the TP73-AS1 was done using the CRISPRi method combined with viral infection. gRNA were designed to target the promoter region of the TP73-AS1 gene. Knockdown induction was done using doxycycline at a concentration of 2 µg/ml (Sigma-Aldrich, D3072). Gene knockdown was measured using qRT-PCR.

Cells are plated into a 96-well plate over a range of densities from high to low (number of cells/well = 1, 5, 10, and 20). The cells were allowed to grow for 21 days and form spheres. To determine the probability of sphere formation, the wells were scored for the presence or absence of sphere growth. The natural log fraction of the sphere-negative wells was plotted on a linear scale versus the density (cells/well). The frequency of sphere-forming cells in a particular cell type was determined using the ELDA web tool at http://bioinf.wehi.edu.au/software/elda↗.

The kd of TP73-AS1 was induced by addition of doxycycline for 10 days. The cells were plated in 12-well plates and TMZ (Sigma-Aldrich, T2577) or DMSO was added. Seven to ten days post treatment, the cells were washed and trichloroacetic acid (TCA) was added for an hour at 4 °C. The plates were washed and air dried. Crystal Violet (C0075, Sigma-Aldrich) was added for half an hour and additional rinsing was conducted. After dehydration, 10% acetic acid was added and the optical density was measured at 570 nm. The results were normalized to scramble gRNA.

Cell apoptosis was determined using flow cytometry. Cells were harvested 10 days post doxycycline treatment using Accutase (Biological Industries, Israel) and the cells were treated with TMZ for 7 days and harvested later. The cells were stained using an apoptosis detection kit (Miltenyi Biotec). The cells were resuspended in binding buffer, and were then stained with Annexin V FITC and propidium iodide (PI). The cells were analyzed using a flow cytometer (Sysmex) equipped with FCS Express software (De Novo Software) according to the manufacturer’s instructions. For the DEAB treatments, G7 cells were treated with 200 μM of DEAB (D86256↗, Sigma-Aldrich), an ALDH1A1 inhibitor, and with 300 μM of TMZ for 5 days, after which cell death was measured using PI staining and FACS.

Cell quantitation of ROS was determined using flow cytometry. Cells were harvested 10 days post doxycycline treatment using Accutase (Biological Industries, Israel) and the cells were treated with TMZ for 2 days and harvested later. The cells were stained using a ROS detection kit (D6883, Sigma-Aldrich). The cells were treated with DCFDA for 1 h, and then harvested and washed with PBS. The cells were analyzed using a flow cytometer (Sysmex) equipped with FCS Express software (De Novo Software) according to the manufacturer’s instructions.

Libraries were prepared using QuantSeq 3′ mRNA-Seq Library Prep Kit FWD for Illumina (Lexogene) according to the manufacturer’s protocol. Samples were pooled and sequencing was performed using Nextseq5000 using all four lanes.

Thirty-three samples were sequenced in the present study. These are composed of three biological replicates of G7 cells expressing gRNA SCR (scrambled) and gRNA#1 in TMZ treated and untreated conditions and three biological replicates of G26 cells expressing gRNA SCR, gRNA#1, and gRNA#2 under TMZ treatment and non-treated conditions. Sequencing resulted in an average of 9,655,276 reads per sample with an average sequence length of 86 bp. Raw reads were then trimmed for sequencing adaptors, poly-A, as well as poor-quality bases using Trim galore (v0.4.2; https://www.bioinformatics.babraham.ac.uk/projects/trim_galore/↗) and Cutadapt (v1.12.1; github.com/easybuilders/easybuild-easyconfigs/tree/master/easybuild/easyconfigs/c/cutadapt). After QC, an average of 9,561,457 reads per sample remained for downstream analysis with an average sequence length of 78 bp. Then mapping and gene-level read count estimation were performed using STAR (v2.5.3a) and RSEM (v1.2.31), respectively, against the human reference genome (GRCh38). The DeSeq2 R package was used to normalize gene counts and perform differential gene expression, followed by clustering and visualizations using the DeSeq2 VSD function (Variance Stabilizing Transformation). Clustering analysis was performed by employing a hierarchical clustering method (using Pearson correlation clustering and “ward.D2” agglomeration method), while the number of clusters were assessed using the “eclust” function from the “factoextra” R package. The “clusterprofiler” R package was used for Gene Ontology enrichment analysis, while gene annotation was retrieved from the “org.Hs.eg.db” R package. PCA was generated for the normalized count data using the plotPCA function from the DeSeq2 R package using the top 500 variable genes.

Cells were grown in a medium containing antibiotic selection. KD was induced with doxycycline for 10 days, after which the cells were replated into 24-well plates in the presence or absence of TMZ (Sigma-Aldrich, T2577). One week post treatment, the cells were washed with trichloroacetic acid (TCA) for an hour at 4 °C. The plates were washed and air dried. Crystal Violet (C0075, Sigma-Aldrich) was added for half an hour and additional rinsing took place. Following dehydration, 10% acetic acid was added and the OD was measured at 590 nm. The results were normalized to scramble gRNA.

Supplemental material SUP table 1 SUP table 2 SUP Figure 1 SUP Figure 2 SUP Figure 3 SUP Figure 4 SUP Figure 5