ATRX loss induces multiple hallmarks of the alternative lengthening of telomeres (ALT) phenotype in human glioma cell lines in a cell line-specific manner

Cancers must maintain their telomeres at lengths sufficient for cell survival. In several cancer subtypes, a recombination-like mechanism termed alternative lengthening of telomeres (ALT), is frequently used for telomere length maintenance. Cancers utilizing ALT often have lost functional ATRX, a chromatin remodeling protein, through mutation or deletion, thereby strongly implicating ATRX as an ALT suppressor. Herein, we have generated functional ATRX knockouts in four telomerase-positive, ALT-negative human glioma cell lines: MOG-G-UVW, SF188, U-251 and UW479. After loss of ATRX, two of the four cell lines (U-251 and UW479) show multiple characteristics of ALT-positive cells, including ultrabright telomeric DNA foci, ALT-associated PML bodies, and c-circles. However, telomerase activity and overall telomere length heterogeneity are unaffected after ATRX loss, regardless of cellular context. The two cell lines that showed ALT hallmarks after complete ATRX loss also did so upon ATRX depletion via shRNA-mediated knockdown. These results suggest that other genomic or epigenetic events, in addition to ATRX loss, are necessary for the induction of ALT in human cancer.


Introduction
Telomeres consist of multiple kilobases of repeated TTAGGG sequence at the ends of chromosomes and are protected by a sequence-specific protein cap [1]. Due to the limitations of cellular replication machinery, in the absence of a telomere length maintenance mechanism, telomeres will shorten with each cell division. In proliferating normal somatic cells, one or more telomeres will ultimately shorten to a critical length, causing cells to undergo senescence PLOS  or apoptosis [2][3][4]. Cancer cells, given their limitless proliferation potential, evolve mechanisms to abrogate these responses and maintain their telomeres above these critical lengths. While most cancer subtypes are known to activate telomerase, which directly adds telomere DNA repeats onto the ends of chromosomes, for telomere maintenance, 5-10% of all cancers utilize a telomerase-independent mechanism, termed alternative lengthening of telomeres (ALT) [5]. While rare in cancers overall, ALT is prevalent in certain cancer subtypes, including sarcomas, pancreatic neuroendocrine tumors, and gliomas [6]. It is widely thought that ALT arises through a recombination-based mechanism, rather than direct enzymatic telomere elongation [5]. Recent evidence has implicated the break-induced repair machinery in the ALT mechanism, further suggesting a DNA-damage response component to the ALT pathway [7,8]. In addition to the lack of telomerase activity, several other key hallmarks of ALT-positive cell populations have been identified, including the presence of ultrabright telomeric DNA foci and ALT-associated promyelocytic leukemia (PML) bodies (APBs) in a subset of cells [9], generation of circular, extrachromosomal telomeric DNA species (c-circles) [10], as well as extreme overall telomere length heterogeneity [11]. Furthermore, cancers that utilize the ALT telomere maintenance mechanism typically display loss of ATRX or, more rarely, DAXX [12][13][14][15][16]. ATRX and DAXX function together as a chromatin remodeling complex that load the histone variant H3.3 into telomeric and other repetitive heterochromatic regions [17][18][19][20]. In addition, ATRX itself is recruited to telomeres after the detection of replication stress, where it is thought to assist in stress relief via its intrinsic helicase activity [19,21,22]. Therefore, the effects of ATRX loss on telomeres may be due to either improper maintenance of telomeric heterochromatin, improper resolution of replication stress at telomeres, or both [23]. Cancers with ATRX loss display large, ultrabright telomeric DNA foci that are strongly correlated with the presence of ALT [12]. Furthermore, in adult high-grade gliomas, inactivating mutations in ATRX are mutually exclusive with activating mutations in the TERT promoter, providing genetic evidence that ATRX loss contributes to telomere maintenance via the ALT phenotype in this cancer subtype [15].
The strong link between ATRX loss and ALT in clinical specimens implies that ATRX is a suppressor of the ALT mechanism. However, existing in vitro data do not conclusively support this notion. Most cell lines that utilize the ALT mechanism have lost ATRX [24], while forced expression of ATRX in ALT-positive, ATRX-negative cell lines diminished the presence of ALT-associated features [25,26]; these data suggest that ATRX loss is necessary for ALT to occur. However, several prior studies have either knocked out or knocked down ATRX in telomerase-positive cell contexts, but no evidence of the ALT phenotype has resulted from these efforts [24,25,[27][28][29]. These data strongly suggest that ATRX loss alone is not sufficient for ALT to occur and that there are unidentified cooperating alterations in cancer cells that allow for the development of ALT hallmarks.
In order to better understand the apparent role of ATRX as an ALT suppressor in cancer, we have generated clonal, ATRX-knockout human high-grade glioma-derived cell lines. Highgrade gliomas, encompassing grade III anaplastic astrocytomas and grade IV glioblastomas, display high frequencies of ALT, with rates being higher in pediatric cases [6, 16, 30-34]therefore, high-grade gliomas represent an ALT-competent cancer subtype. Herein, we have identified, for the first time, two cell lines, U-251 and UW479, which generate multiple ALTassociated hallmarks after ATRX loss. These new in vitro tools will further our understanding of the molecular mechanisms of ALT, facilitating identification of new anti-cancer therapies for patients with ALT-positive cancers, most of whom currently face dismal prognoses and no effective treatment options.

Results
In order to assess the telomere-associated effects of ATRX loss, we chose a glioma in vitro model system, due to the relatively high frequency of ATRX loss and ALT in glioma [6,12,13,16,[30][31][32][33][34]. To this end, we obtained a panel of high-grade glioma cell lines, derived from adult (MOG-G-UVW, SF295, and U-251) or pediatric (CHLA-200, KNS42, SF188, UW479) patients. Relevant mutations in these cell lines are listed in Table 1: activating TERT promoter mutations were identified in five of the seven cell lines, and previously identified mutations in H3F3A, encoding histone H3.3, and TP53 were validated [35][36][37][38][39][40][41][42]. All of the glioma lines assessed retain expression of both ATRX and DAXX, as assessed by immunoblotting ( Fig 1A) and immunohistochemistry (S1 Fig), while U2-OS, a well-characterized ALT-positive osteosarcoma cell line harboring an ATRX deletion [24,43], lacks ATRX expression (Fig 1A). It is of note, however, that KNS42 has been reported to harbor a single nucleotide polymorphism (SNP) within its ATRX locus (Q891E, rs3088074) of unknown significance [44]. Furthermore, telomerase activity, measured by telomere repeat amplification protocol (TRAP), is detectable in all seven glioma cell lines and, as expected, is absent in U2-OS ( Fig 1B). Conversely, all seven glioma cell lines lack characteristics of ALT activation, such as c-circles, assessed by dot blot, (Fig 1C) and ultrabright telomeric DNA foci, as detected by telomere-specific FISH ( Fig  1D), both of which are apparent in U2-OS (Fig 1C and 1D). Therefore, we conclude that all seven high-grade glioma cell lines used in this study are bona fide ALT-negative, telomerasepositive cell line models.
In order to assess the effects of ATRX loss on ALT in glioma cell lines, we sought to generate clonal cell lines with functional knockout of ATRX in these seven cell lines. We used a dual-nickase CRISPR approach targeting exon 9 of ATRX followed by the generation of ATRX KO subclones, as well as clones from empty vector (EV) controls. Although knockout of ATRX was unsuccessful in SF295, CHLA-200, and KNS42, we successfully isolated at least two ATRX KO clones each from MOG-G-UVW, U-251, SF188, and UW479. Immunohistochemistry of ATRX KO clones reveals complete loss of nuclear immunostaining of ATRX compared to EV control clones (Fig 2A), while immunoblotting revealed a dramatic loss of ATRX, as well ( Fig 2B). It is worth noting that antibodies recognizing distinct epitopes were used for the immunohistochemistry and immunoblotting experiments, indicating that the results obtained were not due to an antibody-specific effect. Both antibodies that were used recognize antigens downstream of exon 9 -the targeted region for our gene-editing approach. Therefore, we  [42] cannot exclude the possibility that a severely truncated ATRX is still present. However, all functional domains of ATRX (the ADD domain, ATP-binding helicase, and C-terminal helicase) are located downstream of the region targeted in this study. Therefore, we are confident that a severely truncated protein would be rendered completely nonfunctional. ATRX KO clones were further confirmed by Sanger sequencing of individual ATRX amplicons after colony PCR (Table 2). Nine out of eleven ATRX KO clones displayed only out-offrame sequences within the targeted region of ATRX, while the remaining two (SF188 ATRX KO 1 and U-251 ATRX KO 5) displayed heterozygous in-frame events that resulted in protein loss. Importantly, no wild-type ATRX sequences were observed in these clones. Taken together, these data confirm the generation of multiple, independent functional ATRX knockouts in MOG-G-UVW, SF188, U-251, and UW479.  Having produced functional knockouts of ATRX, we sought to determine the effect of ATRX loss on telomere maintenance in multiple clones of each of these cell lines. Specifically, we examined well-established hallmarks of the ALT telomere maintenance mechanism. Consistent with previous observations in other cell line models [24,25,[27][28][29], MOG-G-UVW and SF188 ATRX KO clones did not display ultrabright telomeric foci ( Fig 3A). In contrast, these characteristic markers of ALT were detectable in ATRX KO clones derived from U-251 and UW479 (Fig 3B), but not in empty vector control clones when assessed one month after clonal expansion (S2 Fig). In addition, each U-251 and UW479 ATRX KO clone harbored ALTassociated PML bodies ( Fig 3C). Furthermore, c-circles were detected in each U-251 ATRX KO clone and in one out of two UW479 ATRX KO clones, but not in any ATRX KO clones derived Cell-line specific effects of ATRX loss in glioma from MOG-G-UVW or SF188 (Fig 3D). While the c-circle levels are quite low relative to U2-OS, samples with c-circle levels of over 4.5% of U2-OS have previously been designated ccircle positive, confirming that samples with comparatively low c-circle levels can still be considered positive [45]. Furthermore, the ALT hallmarks that we have observed were apparent even after several months in culture after clonal expansion. Therefore, while ATRX loss in MOG-G-UVW and SF188 did not lead to overt changes in telomere maintenance, loss of ATRX in U-251 and UW479 resulted in the rapid and sustained appearance of ALT-associated hallmarks.
The frequency of ultrabright telomere foci was quantified in each ATRX KO clone, revealing a range in prevalence of 0.1-2.4% of cells containing these foci in the U-251 clones and 0.02-0.12% in the UW479 clones ( Fig 4A). While these levels are quite low, no cells containing ultrabright telomeric foci were identified in empty vector clones-ultrabright telomeric DNA foci were significantly increased in ATRX KO cells compared to empty vector clones for both U-251 (p = 0.0004) and UW479 (p = 0.01, Wilcoxon rank-sum analysis). The degree to which these foci colocalized with PML to form APBs was also quantified. A subset of ultrabright telomeric DNA foci colocalized with PML to form APBs in each ATRX KO clone ( Fig 4B), with 29-84 percent of focus-positive cells also containing APBs ( Fig 4C). Heterogeneity in PML and telomere focus colocalization was observed both between cells and within single cells, as observed previously [12]. Furthermore, the levels of c-circles in these clones were quantified relative to a standard curve derived from U2-OS using densitometry analysis. Here, again, the levels of c-circles were low, ranging from 1-5% of the levels found in U2-OS ( Fig 4D). Despite the low levels of these hallmarks that were observed in our ATRX KO clones, neither c-circles nor ultrabright telomeric foci were observed in the parental cell lines (Fig 1C and 1D) or empty vector control clones (S1 Fig, Fig 3D), strongly suggesting that the low levels of ALTassociated hallmarks that are observed in these lines are, indeed, due to ATRX loss.
Another hallmark of ALT is overall telomere length heterogeneity. Therefore, we assessed the effect of ATRX loss on this ALT characteristic in ATRX KO clones derived from U-251 and UW479, as well. Despite the emergence of other hallmarks of ALT in these ATRX KO  As shown in Fig 1B, both U-251 and UW479 are telomerase-positive glioma cell lines. However, hallmarks of ALT are induced after loss of ATRX in these lines (Fig 3B and 3D). Therefore, we sought to determine the potential effect of ATRX loss on telomerase activity in these clones. As shown in Fig 5, while there are slight changes in telomerase activity across the clones (presumably due to clonal variability), all ATRX KO clones retain telomerase activity. Therefore, the emergence of ALT characteristics in U-251 and UW479 ATRX loss is not due to a global loss of telomerase activity; rather, ATRX KO clones derived from these cell lines appear to be utilizing the telomerase enzyme to maintain telomere length, while also generating features associated with the ALT telomere maintenance mechanism.
In adult glioma, tumors with mutation of ATRX also tend to have mutation of TP53 [15]. Both U-251 and UW479 harbor mutations in TP53 (Table 1). While SF188 does as well, it is intriguing that MOG-G-UVW, derived from an adult glioma, is TP53-wildtype (Table 1). One interpretation of these data is that mutation of TP53 cooperates with ATRX loss in the development of ALT in this cancer subtype. To assess whether p53 alterations cooperate with ATRX loss in promoting ALT in adult glioma, we overexpressed a dominant-negative p53 (TP53 R273H ) in MOG-G-UVW ATRX KO  Therefore, p53 status cannot explain the differences that we have observed regarding whether an adult glioma cell line will display ALT hallmarks after ATRX loss. In addition, recent work has suggested that downregulation of RAP1 and XRCC1 licenses IDH1-mutant glioma cells to engage the ALT mechanism after ATRX loss [47]. As such, we assessed the levels of RAP1 and XRCC1 in our EV and ATRX KO clones by immunoblotting. Differences that were observed in RAP1 or XRCC1 levels did not correspond to the presence or absence of ALT hallmarks (S5 Fig), suggesting that these proteins are not downregulated to lead to ALT characteristics after ATRX loss in U-251 and UW479, which are IDH1-wildtype. However, the possibility remains that focal losses of these proteins contribute to the development of ALT characteristics in our ATRX KO clones. Telomere FISH plus DAPI nuclear staining was performed on EV and ATRX KO clones, and 36-100 images (magnification = 400X) per experiment were obtained via scanning microscopy. Total cell number was determined by nuclear segmentation in TissueQuest software, and a minimum of 1000 cells were included in each analysis. (A) Cells containing ultrabright telomeric foci were identified by image analysis using pixel intensity and particle size thresholds. Analysis was limited to segmented nuclei. For both U-251 and UW479, ATRX KO clones have significantly more cells containing ultrabright telomeric DNA foci than EV clones (p = 0.0004 and 0.01, respectively). Significance was calculated using Wilcoxon rank-sum analysis. Data are from three independent measurements, and error bars represent standard error of the mean. (B) Telomere FISH plus PML immunofluorescence and DAPI nuclear staining was performed on EV and ATRX KO clones, and 36-100 images (magnification = 400X) per experiment were obtained via scanning microscopy. Total cell number was determined by nuclear segmentation in TissueQuest software, and a minimum of 1000 cells were analyzed for each experiment. Cells containing ultrabright telomeric foci were identified by image analysis using pixel intensity and particle size, and colocalization events were identified using the Image J Colocalization plugin [46]. Analysis was limited to segmented nuclei. The percent of all cells containing ultrabright foci and APBs were calculated (B), as was the percent of total focus-positive cells containing an APB (C). Data are from two independent experiments, and error bars represent standard error of the mean. (D) C-circle levels were quantified in ATRX KO cells by densitometry and compared to U2-OS. Data were generated from three independent measurements. Error bars represent standard error of the mean. Furthermore, we sought to evaluate the emergence, if any, of telomere-specific DNA damage after ATRX loss. As such, we assessed the levels of telomere-dysfunction induced foci (TIFs) in our EV and ATRX KO clones by assessing colocalization between telomeres and phospho-H2A.X (S6 Fig). When compared to EV clones, significant increases in TIF frequency Cell-line specific effects of ATRX loss in glioma were observed in MOG-G-UVW ATRX KO 1 (p < 0.0001 compared to both EV 1 and EV 2), SF188 ATRX KO 1 (p = 0.04 compared to EV 1), U-251 ATRX KO 1, 2, and 3 (p < 0.0001 compared to both EV 1 and EV 2), and UW479 ATRX KO 1 (p < 0.0001 compared to EV 1). Interestingly, significant decreases in TIF frequency were also observed in some clones: SF188 ATRX KO 2 (p < 0.0001 compared to both EV 1 and EV 2) and UW479 ATRX KO 2 (p < 0.0001 compared to EV 2). These results, as well as the significant difference observed between UW479 EV 1 and EV 2 (p < 0.0001), strongly suggest that the degree of telomere-specific damage observed in these clones is a result of clonal variability and not ATRX loss.
In addition, previous work has identified break-induced repair as a potential mechanism for ALT [7,8]. POLD3, a component of DNA polymerase delta, was found to be necessary for break-induced telomere synthesis in ALT-positive cell lines and was found to localize to sites of induced telomeric damage [7]. In order to determine if ATRX loss affects the degree of POLD3 colocalization with telomeres, we performed simultaneous telomere FISH and immunofluorescence against POLD3 (S6 Fig). We observed a speckled, pan-nuclear localization of POLD3 rather than discrete nuclear foci in both our EV and ATRX KO populations (Part A of S7 Fig). This staining pattern makes it difficult to evaluate colocalization between POLD3 signals and individual telomeres; however, the consistent staining pattern observed across our EV and ATRX KO clones indicates that POLD3 does not undergo a dramatic re-localization to telomeres after ATRX loss. In addition, we did not observe a consistent colocalization pattern with ALT-associated telomeric DNA foci in ATRX KO clones (Part B of S7 Fig).
In light of our observation that U-251 and UW479 develop ALT hallmarks after ATRX knockout, we sought to determine the effect of ATRX dosage on the behavior of these cell lines: in contrast to ATRX elimination, is reduction in ATRX levels sufficient to induce ALT characteristics in these cell lines? To answer this question, we stably knocked down ATRX expression in these two cell lines, as well as MOG-G-UVW and SF188, using 3 unique shRNAs against ATRX. Reduction in ATRX protein levels was confirmed by immunohistochemistry ( Fig 6A) and immunoblotting (Fig 6B) compared to empty vector control (pLKO.1). ALTassociated hallmarks were assessed after selection and expansion of stable shATRX cell lines. Knockdown of ATRX yielded similar results to completely knocking out ATRX: immediately after ATRX knockdown and selection, U-251 and UW479 cells displayed ultrabright telomeric foci ( Fig 7A) with PML colocalization (Fig 7B), while MOG-G-UVW and SF188 did not display this feature. Emergence of c-circles in U-251 and UW479 after ATRX knockdown was also observed but was more variable compared to ATRX KO . ATRX knockdown did not yield detectable c-circles in U-251, and induced positivity in one out of three shATRX lines from UW479 ( Fig 7C). As expected, MOG-G-UVW and SF188 did not display ultrabright telomeric foci (Fig 7A) or c-circles (Fig 7C). Interestingly, for U-251, while the ALT-associated ultrabright foci were immediately apparent, this feature was not stable over time, in contrast to the ATRX KO cells: U-251 shATRX-11 had lost ultrabright telomeric foci within ten passages (S8 Fig), while retaining reduction in ATRX levels. Still, the immediate presence of ultrabright telomere DNA foci and APBs in all three shRNA populations suggests that simply reducing the levels of ATRX is sufficient to transiently yield a subset of ALT characteristics under the appropriate conditions. Furthermore, these results indicate that the ALT features observed in U-251 and UW479 after ATRX knockout were not the result of off-target effects of the CRISPR system and can be attributed to ATRX loss.
In an attempt to identify additional glioma cell lines that show ALT hallmarks after ATRX loss, we stably knocked down ATRX in SF295, CHLA-200, and KNS42, cell lines in which we were unable to recover CRISPR-mediated ATRX knockouts. Knockdown was confirmed by immunohistochemistry (Part A of S9 Therefore, through our modulation of ATRX expression in glioma cell lines, we have confirmed that most cell lines do not develop ALT characteristics after ATRX loss. However, herein, we have identified for the first time two cell lines, U-251 and UW479, in which ALT hallmarks arise after partial or complete ATRX loss. These two cell lines are intriguing models for the future study of ATRX loss and ALT in glioma.

Discussion
In this study, we have examined the effects of ATRX loss on the ALT telomere maintenance mechanism in telomerase-positive human glioma cell lines. While five of the seven cell lines examined herein do not display characteristics of ALT after ATRX knockout and/or knockdown, two cell lines-U-251 and UW479 -display multiple hallmarks of ALT immediately after functional ATRX loss. Specifically, ATRX KO clones derived from U-251 and UW479 acquire ultrabright telomeric DNA foci, APBs, and c-circles (Fig 3, Fig 7). These two cell lines will be valuable tools for the future study of ALT and ATRX loss in cancer.
Ours is the first study to induce hallmarks of ALT via knockout of ATRX in established human cell lines already utilizing telomerase as their telomere length maintenance system, without altering telomerase itself. Previous induction of an ALT phenotype using in vitro Cell-line specific effects of ATRX loss in glioma models have been in the context of telomerase loss. Napier et al. induced ALT after shRNAmediated ATRX knockdown in telomerase-negative primary fibroblasts, pre-immortalization [25], while Min et al. induced ALT after knocking out the telomerase subunit TERC in telomerase-positive cell lines [48]. In addition, O'Sullivan et al. stimulated the emergence of ALT characteristics after siRNA-mediated depletion of both ASF1a and ASF1b, accompanied by an intrinsic loss of TERT expression and telomerase activity [23]. Because the ATRX KO clones derived from U-251 and UW479 retain enzymatically active telomerase (Fig 5), we cannot claim that the ATRX KO clones are ALT-positive, despite the emergence of multiple ALT-associated hallmarks (Fig 3). The original, and most stringent, definition of ALT is a functional one, relying on demonstrating telomere maintenance over significant cell division in the absence of telomerase [11]. We cannot exclude the possibility that a sub-population of cells in ATRX KO clones derived from U-251 and UW479 have lost telomerase activity and that these cells are utilizing the ALT mechanism and are the source of the ALT-associated characteristics that we detect. Without a robust method to reliably detect the individual components of the telomerase holoenzyme in situ, along with c-circles and ultrabright telomeric DNA foci in the same cells, we cannot yet determine whether there is intracellular heterogeneity in telomerase activity associated with ALT-activity among these cells that is subtle enough to evade detection in a bulk, in vitro assessment. Still, our data do support the notion that a telomerase-positive Cell-line specific effects of ATRX loss in glioma cellular population can generate ALT hallmarks after ATRX loss under the appropriate cellular context. This observation is in agreement with the prior finding that ALT and telomerase activity can coexist [49].
We have observed multiple hallmarks of the ALT mechanism in ATRX KO clones derived from U-251 and UW479 -ultrabright telomeric DNA foci, APBs, and c-circles (Fig 3). It should be noted that these features have been found in contexts other than ALT. For example, APBs have been observed in mouse embryonic stem cells [50]. However, evidence of direct telomeric recombination has been detected in mouse somatic cells, suggesting that mice may simultaneously utilize ALT and telomerase to lengthen their telomeres [51]; in contrast, this phenomenon has not been observed in human cells. An additional potential source of ALT hallmarks is telomere trimming-a homeostatic mechanism to prevent telomeres from overlengthening [52]. Previous work has shown the presence of APBs [53] and c-circles [54] under instances of excessive telomere elongation leading to telomere trimming. However, telomere trimming is associated with an overall lengthening of telomeres, a feature that we do not observe in our ATRX KO cells (S3 Fig). Therefore, we believe that the simultaneous presence of three ALT associated features (ultrabright telomeric DNA foci, APBs, and c-circles) after the loss of a known ALT suppressor, ATRX, is more in line with an ALT-like phenotype (while not yet fully ALT due to the retention of telomerase) than these alternative explanations.
While it is clear that ATRX loss induces the formation of ALT-associated telomere DNA foci, APBs, and c-circles in U-251 and UW479, other hallmarks of the ALT phenotypenamely, telomere length heterogeneity-were not apparent (S3 Fig). This discrepancy may be due to several reasons. First, the frequency of cells containing increased telomere content associated with telomere DNA foci are quite rare in these clones (Fig 4), and the effects of these foci on overall telomere heterogeneity may be too small to measure in a bulk analysis. Second, while ultrabright telomeric foci, APBs, and c-circles arose nearly immediately after ATRX knockout (Fig 3), it is possible that the dramatic heterogeneity in telomere length typically observed in ALT-positive specimens requires a longer culture time to develop; however, knockdown of ASF1a and ASF1b in HeLa cells resulted in dramatic increases in telomere length heterogeneity within 14 days [23], so this possibility is unlikely. Lastly, the retention of telomerase activity in these cell lines (Fig 5) may be preventing an expansion in telomere length distribution. If telomerase is still directly acting to maintain the telomeres in these cells at constant lengths, there may not be the dramatic shift in telomere lengths as is typically observed in true ALT-positive cells, where telomerase activity is absent. Ectopic expression of telomerase in ALT-positive cells causes preferential extension of the shortest telomeres in the population, reducing the number of telomere ends that are potentially recognized as DNA damage [55]. Therefore, loss of telomerase activity may be necessary for a higher degree of telomere shortening to yield both critically short telomeres and excessively long telomeres that result from recombination-based lengthening of telomeres. Possibly, this process is initiated by DNA damage signaling from chromosome ends that have lost their protective telomere caps, ultimately yielding ALT-associated extreme telomere length heterogeneity.
Herein, we have also assessed whether complete ATRX loss is necessary for the induction of ALT-associated features, or if partial suppression is sufficient. While full ATRX knockout leads to ALT features in U-251 and UW479 (Fig 3), shRNA-mediated reduction in ATRX levels also induces these characteristics in these two cell lines (Fig 7). These findings confirm that the phenotypes observed in U-251 and UW479 ATRX KO clones are due to ATRX loss and not an off-target effect of CRISPR; in addition, these results have direct relevance to the current use of ATRX loss as assessed by immunohistochemistry as a diagnostic biomarker in glioma [56]. Further study will help clarify the link between partial ATRX loss in gliomas with the presence of ALT in order to more effectively use ATRX as a surrogate marker for the ALT pathway for diagnostic [56], prognostic [34,57], and, potentially, therapeutic purposes [29,58].
Regardless of how telomerase activity contributes to the observed ALT features in U-251 and UW479 after ATRX loss, the features that arose-ultrabright telomeric DNA foci, APBs, and c-circles-were present nearly immediately after loss of ATRX, either through CRISPRmediated knockout (Fig 3) or shRNA-mediated knockdown (Fig 7). Therefore, U-251 and UW479 are the first identified human cancer cell lines that show ALT features after in vitro knockout of ATRX, a well-known and clinically relevant ALT suppressor [12,25,26]. The rapid emergence and long-term persistence of ALT-associated hallmarks in these cell lines after ATRX loss suggests that the telomeres in these cell lines are primed to undergo an ALTlike process, and that ATRX loss immediately allowed this mechanism to activate. In other words, there must be additional genomic and/or epigenetic events that occur in ALT-positive cancers, cooperating with ATRX loss to allow the ALT mechanism to occur. For example, either additional ALT suppressors, like ATRX, must be lost, or heretofore unidentified ALTfacilitating factors must be present. U-251 and UW479 are novel tools that will enable this investigation-further study of the underlying genomic and/or epigenetic events present in these cell lines (in comparison to cell lines that do not show ALT features after ATRX loss) will allow for a greater understanding of relevant ALT activators and suppressors in cancer. The identification of U-251 and UW479, due to their unique telomere phenotype after ATRX loss, provides a key step in the elucidation of the molecular mechanism(s) of ALT and the identification of promising therapeutic targets for ALT-positive cancers [29,58].
In cancers, ATRX loss and ALT are tightly linked, implicating ATRX as a suppressor of the ALT pathway. Despite the strong association between ATRX loss and ALT in cancer, prior studies have not observed ALT characteristics after ATRX knockout or knockdown, much like we have observed with the MOG-G-UVW and SF188 cell lines. In contrast, here, we have identified two ALT-negative glioma cell lines-U-251 and UW479 -in which ATRX loss is sufficient for the generation of ALT hallmarks de novo. Because these two cell lines displayed ALT characteristics after both knockdown and knockout of ATRX, it is likely that their telomeres are already primed to undergo an ALT-like process. The identification of these cell lines will allow for the comprehensive analysis of additional genomic and/or epigenetic events-in addition to ATRX loss-that are required for suppression or licensing of the ALT mechanism.

Cell culture
U2-OS cells were purchased from the American Type Culture Collection (ATCC; Manassas, VA). U-251 and SF295 were kindly provided by Dr. Angelo DeMarzo (Johns Hopkins University School of Medicine, Baltimore, MD), who obtained the lines as part of the NCI-60 cancer cell line panel from the National Cancer Institute. MOG-G-UVW was obtained from Sigma. CHLA-200 was obtained from the Children's Oncology Group. KNS42 was obtained from the Japan Cancer Research Resources cell bank. SF188 and UW479 were kindly provided by Dr. Chris Jones (Institute of Cancer Research, Sutton, UK). Cells were maintained in a humidified incubator kept at 37˚C with 5% CO 2 . MOG-G-UVW was grown in DMEM/F-12 (Thermo-Fisher) supplemented with 10% FBS and 2 mM L-Glutamine (ThermoFisher), SF295 and U-251 were grown in RPMI (ThermoFisher) supplemented with 10% FBS, CHLA-200 was grown in IMDM (ThermoFisher) supplemented with 20% FBS, 4 mM L-Glutamine, and 1X Insulin-Transferrin-Selenium, and KNS42, SF188, and UW479 were grown in DMEM/F-12 (ThermoFisher) supplemented with 10% FBS. All media contained 1% Penicillin/Streptomycin, 1% Amphotericin B, 10 μg/mL Gentamicin, and 5 μg/mL Plasmocin (Invivogen).
Identities of all cell lines used in this study were confirmed by short tandem repeat (STR) profiling using the GenePrint 10 and/or the PowerPlex 18D kits (Promega). Cells were confirmed to be mycoplasma-negative using the VenorGeM assay (Millipore-Sigma).

Telomere repeat amplification protocol (TRAP)
TRAP was performed based on the previously described protocol [59]. Briefly, 10 6 cells were lysed in NP-40 lysis buffer. Each assay was performed using an equivalent of 2500 cells, with RNase treated cells included as a negative control. The reactions were run on 20% TBE precast gel (ThermoFisher). Gels were imaged using a ChemiDoc Touch (BioRad). Signals were quantified by densitometry analysis using ImageJ and Fiji software [60,61]. Background intensity quantified from the no template control was subtracted from the band intensity measured for telomerase activity from cell lysates. Relative telomerase activity was determined by dividing telomerase products (background adjusted) by the band intensity of the 36 bp internal standard control.

C-circle assay
C-circles were detected based on the previously described assay [62]. Briefly, genomic DNA was extracted from asynchronously growing cells using the DNEasy Blood and Tissue Kit (Qiagen) and purified using the QiaQuick PCR Purification Kit (Qiagen). DNA from each sample was added to a rolling circle amplification reaction in the presence or absence of phi29 polymerase (New England Biolabs). Products were blotted onto a positively-charged nylon membrane (Roche) and detected using a DIG-labeled telomere probe ((CCCTAA) 4 ; Integrated DNA Technologies). Signal was detected using anti-DIG Fab Fragments (Roche, Cat#1109327 4910) and CDP-Star (Roche). Blots were either exposed to film or developed using a Chemi-Doc Touch (BioRad). For quantification of c-circle levels, a standard curve of U2-OS, an ALTpositive, c-circle positive cell line was established by creating a dilution series of input DNA (ranging from 150 ng to 0.78 ng) was amplified and then blotted on the same membrane as the samples of interest. Signals were quantified by densitometry using ImageJ and Fiji software [60,61].

Telomere FISH and immunofluorescence
To generate formalin-fixed, paraffin-embedded (FFPE) blocks of cell lines, cell plugs were generated by centrifuging cells onto a 2% agarose barrier in a 0.2 mL PCR tube and immersed in 10% phosphate-buffered formalin for 48 hours.

Sequencing
Genomic DNA was isolated from asynchronously growing cells using the DNEasy Blood and Tissue Kit (Qiagen). Twenty nanograms of genomic DNA was amplified using GoTaq Green polymerase (Promega) using cycling conditions optimal for each primer pair (S1 Table). PCR products were purified using QIAQuick PCR Purification Kit (Qiagen) and sequenced by Eurofins Genomics.

Quantification of telomere FISH and PML immunofluorescence
Combined telomere FISH and PML immunofluorescence was performed on cytospins as described above. The TissueFAXS Plus (Tissue Gnostics) automated microscopy workstation was utilized for slide-based imaging. The platform contains an 8-slide ultra-precise motorized stage for high-throughput scanning and utilizes a Zeiss Z2 Axioimager microscope with high quality optics. A separate high-performance workstation with TissueQuest software (Tissue Gnostics) was used to analyze the fluorescent images. For each cell line, automated, precise nuclear segmentation was performed to generate total cell number. Images containing the segmentation masks, along with telomere and PML signals, were exported for analysis using using ImageJ software [60]. Background subtraction was performed for PML images using a rolling ball algorithm, and a threshold was applied to the remaining spots. Rolling ball and threshold values were empirically determined for each set of images. Colocalization between telomere foci and PML bodies was identified using the Colocalization plugin for ImageJ [46]. All analyses were limited to segmented nuclei.

TRF Southern blotting
Cells were resuspended in proteinase K lysis buffer (10 mM Tris-HCl, pH 8; 100 mM EDTA; 0.5% SDS; 20 μg/mL RNase A; 100 μg/mL proteinase K) and incubated at 37˚C with agitation for five hours. Two equal volume phenol:chloroform:isoamyl alcohol (25:24:1, pH 8) extractions were performed, followed by two ethanol precipitations and final resuspension in H 2 O. DNA was digested overnight with AluI and MboI (New England Biolabs). Two micrograms of digested DNA were run on an 0.8% agarose gel for 4 hours at 100 V. The gel was depurinated, denatured, and neutralized prior to transfer onto a nylon membrane using the Turboblotter transfer system (GE). Detection was performed using a DIG-labeled telomere probe ((CCCTAA) 4 ; Integrated DNA Technologies). Signal was detected using CDP-Star (Roche). Blots were either exposed to film or developed using a ChemiDoc Touch (BioRad).

TIF analysis
Telomere-specific FISH and immunofluorescence against phospho-H2A.X was performed on formalin-fixed, paraffin-embedded cell blocks as described above. Slides were imaged using a TissueFAXS Plus (Tissue Gnostics) automated microscopy workstation, and images were obtained using TissueQuest software (Tissue Gnostics) on a separate workstation, as described above. Images containing telomere and phospho-H2A.X signals were exported for analysis using using ImageJ software [60]. For both telomeres and phospho-H2A.X signals, background subtraction was performed using a rolling ball algorithm, and an intensity threshold was applied to the remaining spots. Threshold values were empirically determined for each set of images. Finally, in order to exclude ultrabright telomeric foci and cells expressing abnormally high levels of phospho-H2A.X, only spots smaller than 20 pixels were included in the final analysis. Colocalization between telomeres and phospho-H2A.X puncta were identified using the Colocalization plugin for ImageJ [46].

ATRX knockdown
shRNA constructs targeting ATRX (TCN000034811, TCN000034790 and TRCN0000013592), as well as pLKO.1 empty vector (SHC001), were obtained from Millipore-Sigma. Lentiviral particles for each shATRX construct and the pLKO.1 empty vector were produced by co-transfecting these plasmids with a VSVG packaging plasmid into 293T cells using Lipofectamine 2000 (ThermoFisher). Supernatants were collected at 48 and 72 hours post-transfection and concentrated with PEG8000 precipitation. Glioma cells were transduced with shATRX or empty vector viral particles for 48 hours prior to selection with puromycin.

Dominant-negative mutant TP53 overexpression
Cells were transfected with a R273H dominant-negative mutant TP53 expression construct, which was a gift from Bert Vogelstein (Addgene #16439) [66], using Lipofectamine 3000. Stably transfected cells were selected by treatment with G418-Sulfate. RNA was extracted from cells using PerfectPure (5-Prime) and converted to cDNA using RealMasterScript Supermix (5-Prime), and p53 transcripts containing the dominant-negative mutation were detected by Sanger sequencing.

Microscopy
Fluorescent images were obtained using a Nikon 50i epifluorescence microscope equipped with X-Cite series 120 illuminator (EXFO Photonics Solutions Inc.) and appropriate fluorescence excitation/emission filters. Grayscale images were captured using Nikon NIS-Elements software and an attached Photometrics CoolsnapEZ digital camera, pseudo-colored and merged. Alternatively, fluorescent images were obtained using a TissueFAXS Plus (Tissue Gnostics) automated microscopy workstation with a Pixelfly USB monochrome camera (PCO). Fluorescent background was subtracted using Adobe Photoshop software. Brightfield images were obtained using a Nikon 50i epifluorescence microscope, Nikon NIS-Elements software, and an attached Nikon Digital Sight DS-Fi1 camera.

Statistical analysis
Significance was determined using Wilcoxon rank-sum analysis using the R environment [67]. P values less than 0.05 were considered significant. All graphs were generated using Graphpad Prism software (versions 7 and 8).
Supporting information S1 Table. Sequencing specifications. PCR primer sequences and associated PCR amplification conditions. (DOCX) Combined telomere-specific FISH and immunofluorescence against phospho-H2A.X was performed in EV and ATRX KO clones, and 36 images (magnification = 400X) per experiment were obtained via scanning microscopy. A minimum of 2000 cells were analyzed for each clone. Telomeres and phospho-H2A.X puncta were identified by setting pixel intensity thresholds after background subtraction. Ultrabright telomeric foci and cells overexpressing phospho-H2A.X were excluded from analysis by eliminating signals larger than 20 pixels. Colocalization events were identified using the Image J Colocalization plugin [46], and percent colocalization was calculated as a fraction of total telomeres. Significance was calculated using a one-way ANOVA incorporating a Tukey's multiple comparisons test. Asterisks ( Ã ) indicate significant difference from the EV1 clone, while pound signs (#) indicate significant difference from the EV2 clone. Error bars represent standard deviation. (TIF)

S7 Fig. ATRX loss does not induce POLD3 focus formation.
Combined telomere-specific FISH and immunofluorescence against POLD3 was performed in EV and ATRX KO . A) In both EV and ATRX KO clones, a pan-nuclear, speckled pattern was observed for POLD3. Representative images (magnification = 400X) for EV and ATRX KO clones from MOG-G-UVW, U-251, and UW479 are shown. B) No consistent pattern of colocalization between POLD3 and ALT-associated telomeric DNA foci was observed. Representative images (magnification = 400X) of cells from U-251 ATRX KO