Gonadal sex patterns p21-induced cellular senescence in mouse and human glioblastoma

Males exhibit higher incidence and worse prognosis for the majority of cancers, including glioblastoma (GBM). Disparate survival may be related to sex-biased responses to treatment, including radiation. Using a mouse model of GBM, we show that female cells are more sensitive to radiation, and that senescence represents a major component of the radiation therapeutic response in both sexes. Correlation analyses revealed that the CDK inhibitor p21 and irradiation induced senescence were differentially regulated between male and female cells. Indeed, female cellular senescence was more sensitive to changes in p21 levels, a finding that was observed in both wildtype and transformed murine astrocytes and patient-derived GBM cell lines. Using a novel Four Core Genotypes model of GBM, we further show that sex differences in p21-induced senescence are patterned by gonadal sex. These data suggest that sex differences in p21 induced senescence contribute to the female survival advantage in GBM.

lines. Using a novel Four Core Genotypes model of GBM, we further show that sex differences in p21-induced senescence are patterned by gonadal sex. These data suggest that sex differences in p21 induced senescence contribute to the female survival advantage in GBM.

Introduction
Sex differences are observed in the majority of diseases, including cancer. Across a wide range of ages and cultures, male sex is associated with higher incidence and worse prognosis for most tumor types [1][2][3] . Glioblastoma (GBM), the most common primary malignant brain tumor, follows this same pattern -women are both less likely to develop GBM and have a significant survival advantage compared to men [4][5][6] . Sex differences in survival may result from differences in the response to standard of care therapy, which for GBM includes surgical resection, followed by treatment with radiation and chemotherapy. Previous research from our lab found that female GBM patients had a greater decline in tumor growth velocity after treatment with radiation and chemotherapy than male patients, and that when these measures were used to stratify patients, there was a significant association with survival in female but not male patients 7 . Furthermore, we identified unique molecular pathways associated with improved survival in male and female patients, and expression of genes in these pathways correlated with the sensitivity to a range of chemotherapeutic agents 7 . While this study advanced our understanding of the relationship between chemotherapy and cellular responses in male and female GBM, it did not focus on radiation, the backbone of GBM treatment, which is currently applied to male and female patients equally.
In cancer treatment, the goal is to stop tumor cell proliferation. While one mechanism to achieve this is through triggering cell death/apoptosis, another possibility is to activate cellular senescence -a cell fate decision that results in irreversible cell cycle arrest 8 . Senescence is a known outcome of radiation treatment, and at least one study has reported that this is the dominant response in glioblastoma 9 . Senescence is often described as a double-edged sword, since senescent cells secrete a broad variety of factors with complex pro-and anti-tumorigenic effects [10][11][12] . However, the cell intrinsic aspect of senescence, specifically the terminal cell cycle exit of cells harboring mutated DNA, serves a beneficial function in tumor prevention and treatment, and enhancing this response could be a strategy to improve treatment efficacy.
Senescence is primarily regulated by two central pathways: p53/p21 WAF1/Cip1 and p16 INK4A /Rb 8,10,11 . Importantly, we have previously identified sex differences in the regulation of p21, p16, and RB in a GBM model, with female cells being more likely to upregulate these pathways in response to cellular stress 13,14 . Whether these differences influence the cellular response to radiation in males and females is currently unknown. In this study, we show that primary male and female human GBM lines have unique molecular pathways that contribute to radiation sensitivity. Using an in vitro mouse model of GBM, we find that females are more sensitive to radiation, and that senescence is a major component of the radiation response in both males and females. Using a correlation-based approach, we identify an association between p21 and irradiation induced cellular senescence that differs in males and females. The same levels of p21 correlate with higher percentages of senescent cells in females than in males -a finding that was observed in both wildtype and transformed cells, as well as in patient-derived GBM cell lines. Finally, using a novel Four Core Genotypes (FCG) model of GBM, we investigate the biological mechanisms underlying this sex difference, and show that sex differences in p21-induced senescence are patterned by gonadal sex.

Unique molecular pathways contribute to radiation response in male and female primary human GBM cell lines
Building upon our previously published work, which identified unique gene signatures associated with sensitivity to a range of chemotherapeutic agents in male and female primary human GBM cell lines 7 , we sought to determine whether these pathways similarly influenced sensitivity to radiation. We performed irradiation dose response curves with four male and five female primary human GBM lines (Supplementary Table 1; Supplementary Fig. 1), and calculated an individual IC 50 value for each line. As previously observed for other DNA damaging agents, there was no significant difference in the median IC 50 values for male and female lines (Fig. 1a).
Our previous work identified a set of differentially regulated genes associated with better  Table 2). In both cases, downregulation of the majority of genes in the pathway was associated with improved survival 7 .
We previously observed that low expression of the MC5 gene set was associated with low IC 50 for multiple chemotherapies in male cell lines, while low expression of the FC3 gene set was associated with low IC 50 for multiple chemotherapies in female cell lines -a finding consistent with the observation that downregulation of the MC5 and FC3 genes was associated with improved survival in male or female GBM patients respectively 7 . To determine whether these molecular pathways may also be contributing to the male and female irradiation response, we calculated Spearman rank correlation coefficients between IC 50 values and the expression levels for MC5 and FC3 genes, measured using the Illumina HumanHT-12 v4 expression microarray.
For the female cell lines, we saw a similar pattern to that observed with other DNA damaging agents (see Fig. 7B, C in reference 7 ). There was a mild positive correlation between FC3 gene expression and irradiation IC 50 , indicating that low expression of these genes was associated with low IC 50 , or better response to irradiation. There was also a negative correlation between MC5 gene expression and IC 50 , indicating that high expression of the MC5 genes was associated with low IC 50 (i.e. better response to irradiation) in the female lines (Fig. 1b). In contrast, the male cell lines showed no significant correlation between MC5 gene expression and IC 50 in the male lines, while there was a significant negative correlation between FC3 gene expression and IC 50 , indicating that high expression of the FC3 genes was associated with better response to irradiation in the male lines (Fig 1b). Neither male nor female lines showed any significant correlation between IC 50 and the expression levels of randomly selected gene sets, serving as a negative control. These results suggest that the response to radiation therapy may be driven by different intracellular pathways in males and females, and that efforts to improve response to therapy may be advanced by identifying the mechanism(s) underlying this sex difference.

Female mouse GBM model astrocytes are more sensitive to radiation treatment
In order to better understand the mechanisms underlying radiation response in male and female GBM, we utilized an in vitro mouse model. This model consists of murine astrocytes with loss of function of the tumor suppressors Nf1 and p53 (Nf1-/-DNp53) 13 . When implanted intracranially, both male and female cells form tumors that histologically resemble high grade gliomas, although the frequency of tumor formation differs by cell sex 13 . Additionally, this model displays sex differences in gene expression that are concordant with sex differences in human GBM patient gene expression 14 . We performed irradiation dose response curves with male and female Nf1-/-DNp53 astrocytes, and found a significant sex difference (Fig. 2a). Female cells were more sensitive to irradiation, a finding consistent with results from human GBM that suggest female GBM patients are more responsive to standard of care therapy (radiation and chemotherapy) 7 .
To further assess sex differences in radiation sensitivity, we irradiated male and female Nf1-/-DNp53 astrocytes with 0, 3, or 9 Gy, then used live cell imaging to track cell growth for 3 days (Fig. 2b). As previously reported 13 , in the absence of radiation treatment male cells grew faster than female cells. While both male and female cells showed impairment in growth after irradiation with 9 Gy, only female cells showed a decline in growth after treatment at the lower dose of 3 Gy (Fig. 2b). This suggests that female GBM model astrocytes are more sensitive to low dose irradiation. We next used the colony formation assay to assess clonogenic survival following radiation treatment. Female Nf1-/-DNp53 astrocytes had a greater decrease in colony counts with radiation treatment than male Nf1-/-DNp53 astrocytes (Fig. 2c) 15 . We focused on the two mechanisms that result in durable therapeutic responses to cancer therapy -apoptosis and senescence. To assess apoptosis, we measured levels of cleaved caspase-3 five days after irradiation (Fig. 3a, b). Overall, we observed a small, but significant treatment-induced increase in cleaved caspase-3 by Two-Way ANOVA. There was no effect of sex on response (Fig. 3b). Measurement of cleaved PARP levels did not confirm the treatment effect ( Supplementary Fig. 2a, b), nor did we see any significant change in either apoptosis marker at 24 hours after irradiation ( Supplementary Fig. 2a, c, d). Thus, apoptosis does not appear to be the primary mechanism responsible for the decrease in cell number following irradiation.
To measure senescence, we stained cells for senescence-associated β-galactosidase (SA-βgal), the most widely used biomarker of senescence 11,16,17 , five days after irradiation, then quantified the percentage of SA-β-gal positive cells (Fig. 3c, d). In response to irradiation, there was a clear, dose dependent increase in the percentage of senescent cells in both male and female Nf1-/-DNp53 astrocytes (Fig. 3d). Consistent with an increase in senescent cells, we observed a change in cell morphology after irradiation 11,16 , with a greater percentage of cells appearing enlarged and irregular in shape ( Supplementary Fig. 3). In addition, when we stained for the cell proliferation marker Ki67, we saw a decrease in the percent of positive cells after irradiation in both sexes, although this decrease was more variable in the male cell lines (Fig.   3e, f). Together, these findings are consistent with an increase in senescence. To further confirm the senescence response, we assessed expression levels of two genes previously identified as part of an ionizing radiation induced senescence (IRIS) signature that was shared across multiple cell types and time points 18 . In agreement with this previous report, we observed an increase in expression of Ccnd1 (Cyclin D1) (Fig. 3g) and a decrease in expression of Cdkn1b (p27) (Fig. 3h) 5 days after irradiation in both male and female Nf1-/-DNp53 astrocytes, providing further support that radiation induced senescence is occurring in these cells. These findings suggest that senescence is a central component of the response to irradiation in Nf1-/-DNp53 astrocytes, and that this, rather than apoptosis, primarily drives the decrease in cell growth after radiation.

Expression of p21 24 hours after irradiation correlates with the senescence response observed at 5 days
Two critical pathways for the induction of cellular senescence are the p16 INK4A /Rb and p53/p21 WAF1/Cip1 pathways 10,11 . While the Nf1-/-DNp53 astrocytes express a dominant negative p53, and thus lack p53 function, they retain the ability to upregulate p21 in response to DNA damage 14 , presumably through p53-independent mechanisms. In order to better understand which of these two pathways may underlie induction of senescence following irradiation, we used qPCR to measure expression levels of Cdkn2a (p16) and Cdkn1a (p21) 24 hours after treatment of cells with 0, 6, or 9 Gy (for nomenclature clarification, we will use the gene name (Cdkn2a, Cdkn1a) when referring to mRNA expression levels, and the protein name (p16, p21) when referring to protein expression levels). We then correlated these values with the percentage of cells that were SA-β-gal positive at 5 days. Because our lab has previously identified sex differences in the regulation of both p16 and p21 in Nf1-/-DNp53 astrocytes 14 , we looked at the relationship between these two measures and senescence in males and females separately. Cdkn2a levels 24 hours after irradiation showed no significant correlation with the senescence response at 5 days in either males or females (Fig. 4a, Supplementary Table 3). In contrast, levels of Cdkn1a significantly correlated with the percent of SA-β-gal positive cells in both males (r=0.59) and females (r=0.79) (Fig. 4b, Supplementary Table 3). This suggests that early upregulation of p21, but not p16, contributes to the senescence response following irradiation in mouse GBM astrocytes. This is consistent with previous research from our lab, which found that in response to treatment with the DNA damaging agent etoposide, levels of p21, but not p16, increased in Nf1-/-DNp53 cells 14 .
Downstream of p21 is the cyclin dependent kinase Cdk2, and it is through inhibition of Cdk2 that p21 primarily acts to maintain Rb in a hypophosphorylated state 19,20 , a critical step in senescence induction 11 . Additionally, it has been reported that the p21/Cdk2 ratio is the primary determinant of the senescent fate decision in human fibroblasts following irradiation 21 Table 3). We found that the Cdkn1a/Cdk2 ratio significantly correlated with SAβ-gal positivity in both males (r=0.53) and females (r=0.80).

Expression of p21 differentially correlates with SA-β-gal positivity in male and female mouse GBM model astrocytes
We next sought to determine whether the correlation between p21/Cdk2 ratio and senescence was also observed at the protein level, and whether this relationship was maintained at later timepoints, once senescence was established. To this end, we measured p21 and Cdk2 levels by western blot, 24 hours and 5 days after irradiation with 0, 6, and 8 Gy (Fig. 4d-f).  expression at 5 days corresponded to higher levels of senescence in females than in males, as measured by SA-β-gal. This suggests that not only does the p21/Cdk2 ratio play a role in the maintenance of senescence following irradiation -but that there may be a sex difference in sensitivity to p21/Cdk2 levels, and/or that p21 may be playing a greater role in irradiation induced senescence in female GBM cells, than in male GBM cells.

SA-β-gal
We wondered whether the sex difference in the relationship between p21 and SA-β-gal that we observed in our mouse GBM model represents a fundamental sex difference, present in normal astrocytes, or is unique to transformed cells and the loss of p53 function. To address this question, we isolated astrocytes from the cortex of male and female postnatal day 1 C57Bl6 mouse pups. Astrocytes from each pup were cultured independently and split to allow for corresponding measures of SA-β-gal and collection of RNA or protein. We irradiated these wildtype (WT) astrocytes with 0 or 10 Gy, then collected RNA or protein 24 hours later, and stained for SA-β-gal at 7 days (Fig. 5a). The irradiation dose and time until SA-β-gal measurement were increased to adjust for the much slower division rate of wildtype astrocytes compared to Nf1-/-DNp53 astrocytes. Quantification of the percentage of SA-β-gal positive cells confirmed a significant increase following irradiation in both male and female wildtype astrocytes ( Fig. 5b).
We used qPCR to measure expression levels of Cdkn2a (p16) and Cdkn1a (p21) mRNA at 24 hours after irradiation, then correlated these measures with the corresponding SA-β-gal percentage. As expected, Cdkn2a levels did not significantly correlate with SA-β-gal positivity in either male or female WT astrocytes (Fig. 5c To determine whether the sex difference in the relationship between p21 expression and SA-βgal was maintained at the protein level, we measured p21 and Cdk2 protein levels by western blot (Fig. 5f), then correlated these with the percent of SA-β-gal positive cells at 7 days. Both p21 alone, and the p21/Cdk2 ratio, significantly correlated with senescence in female WT astrocytes (r=0.67 and r=0.53 respectively), but not male WT astrocytes (Fig. 5g, h, Supplementary Table 3). These results suggest that p21 may be playing a greater role in senescence induction in female astrocytes than in male astrocytes following irradiation, and that this is occurring regardless of whether the cells are transformed or not.

Sex differences in senescence are observed in wildtype astrocytes with repeated in vitro passaging
In initial studies to assess the effects of treatment induced senescence in wildtype astrocytes, we observed that as the cells reached higher passages, a baseline sex difference in senescence began to emerge -with females having higher percentages of SA-β-gal positive cells than males in the untreated condition. To assess this more directly, we cultured male and female wildtype astrocytes and stained for SA-β-gal when the cells were at either low (p2 or p3) or high (p5) passage ( Supplementary Fig. 4a). At greater than five passages, the wildtype astrocytes showed a dramatic decrease in growth, with the majority of cells adopting an enlarged, flattened, irregular morphology, and were unable to be passaged further.
At p2/3, the ratio of SA-β-gal positive cells (female/male) was approximately 1, indicating equivalent levels of senescent cells in male and female astrocyte cultures (Supplementary Fig.   4b). At p5, this ratio was significantly increased, indicating higher percentages of senescent cells in female cultures ( Supplementary Fig. 4b). Thus, female mouse astrocytes undergo senescence more frequently than male mouse astrocytes with repeated passaging. Cell cycle analysis of male and female wildtype astrocytes by flow cytometry found no difference in cell cycle distribution or cell size based on sex (data not shown), suggesting the difference in senescence is not due to baseline differences in cell proliferation rates. Whether it reflects sex differences in the propensity to undergo replicative senescence or in the sensitivity to oxidative damage, resulting from culture at 20% oxygen 22 , remains to be determined. However, this suggests that sex differences may extend to other senescence paradigms and phenotypes, and highlights the need to study these in both sexes separately.

SA-β-gal
We next assessed whether sex differences in the role of p21 in irradiation induced senescence extended to human GBM. For this purpose, we utilized the same primary human GBM lines used to correlate irradiation sensitivity with expression of the MC5 and FC3 gene sets. We irradiated each male and female line with 0 or 6 Gy and stained for SA-β-gal 5 days later. The 6 Gy dose was chosen based on the human GBM irradiation dose response curves; it was above the mean IC 50 value for both sexes, but lower than the dose at which most lines showed a plateau in response. There was a significant increase in the percent of SA-β-gal positive cells following irradiation in both male and female human GBM lines (Fig. 6a).
Using protein samples collected 24 hours and 5 days after irradiation, we measured levels of p21 and Cdk2 by western blot (Fig. 6b). At 24 hours, both p21 alone (  Table 3). This same pattern was observed at 5 days, with p21 and p21/Cdk2 significantly correlating with SA-β-gal for female (r=0.83 and r=0.84, respectively) but not male lines (r=-0.09 and r=-0.27, respectively, p21 male vs. female correlation difference p=0.0461, p21/Cdk2 male vs. female correlation difference p=0.0192) (Fig. 6e, f, Supplementary Table 3). Thus, in both mouse and human, transformed and wildtype astrocytes, the relationship between p21 and senescence, as measured by SA-βgal, differs between males and females.

Knockdown of p21 decreases senescence in irradiated female mouse GBM model astrocytes
To further test the role of p21 in senescence in irradiated male and female cells, we utilized an Nf1-/-DNp53 p21 knockdown line previously developed in our lab using CRISPR/Cas9 14 . We confirmed knockdown by irradiating Cas9 and p21 KD cells with 0 or 8 Gy and measuring levels of p21 protein 24 hours later. Both male and female p21 KD lines had a strong reduction in p21 levels -less than 30% of control levels ( Supplementary Fig. 5a, b). We then irradiated male and female Cas9 and p21 KD cells with 0 or 8 Gy and stained for SA-β-gal 5 days later (Fig. 7a).
The female p21 KD line had a significant reduction in the percent of SA-β-gal positive cells compared to female Cas9 after irradiation, decreasing to levels similar to those of male cells ( Fig. 7b). This supports the idea that p21 expression contributes significantly to the levels of senescence after irradiation in females, but not in males. Surprisingly, despite the decrease in SA-β-gal in female p21 KD cells, they still retain a significant senescence response, and in fact decrease only to the levels of senescence in male cells. This suggests that additional pathways, beyond p21, are contributing to senescence after irradiation.

The Four Core Genotypes mouse model can be used to interrogate the developmental mechanisms underlying sex differences
Multiple mechanisms can potentially underlie an observed sex difference: 1) acute differences in circulating gonadal hormone levels, 2) organizational/epigenetic effects of gonadal hormone exposure in utero, and 3) differences in the expression levels of genes on the X and Y chromosomes 23 . The sex difference in the relationship between p21 and SA-β-gal that we observe is present in vitro, with male and female cells grown in identical media. Thus, it is unlikely that the differences in p21 are driven by effects of circulating estrogen or testosterone.
To investigate whether organizational effects or sex chromosomes are responsible, we utilized a transgenic mouse model known as the Four Core Genotypes (FCG) model 24 . In the FCG model, the Sry gene, which encodes the testis-determining factor, is deleted from the Y chromosome and inserted onto an autosome, allowing for the separation of gonadal and chromosomal sex.
Crossing XY -/Sry+ males with normal XX females results in mice of four genotypes: XY -/Sry+ (XY with testes -normal male), XY -/Sry-(XY with ovaries), XX/Sry+ (XX with testes), and XX/Sry-(XX with ovaries -normal female) (Fig. 8a). Mice that inherit the Sry gene, regardless of chromosomal sex, develop testes and are exposed to masculinizing levels of gonadal hormones during in utero development. Mice that lack the Sry gene develop ovaries and display phenotypes associated with a feminized brain 24 . We crossed FCG XY -/Sry+ mice with Cas9 expressing female (XX) mice and isolated astrocytes from the postnatal day 1 pups. We then used CRISPR/Cas9 to delete Nf1 and p53 from these astrocytes, mimicking our Nf1-/-DNp53 GBM model ( Supplementary Fig. 6).

Gonadal sex patterns the relationship between p21 and SA-β-gal positivity
Using our newly developed FCG GBM model, we irradiated the cells with 0 or 8 Gy and then stained for SA-β-gal 5 days later (Fig. 8b). The percentage of SA-β-gal positive cells increased following irradiation in all four genotypes (Fig. 8c), although due to high variability this did not reach significance in the XY+ genotype. Using qPCR, we measured levels of Cdkn1a (p21) mRNA expression 24 hours after irradiation and correlated this with the percent of SA-β-gal positive cells at 5 days (Fig. 8d, left, Supplementary Table 3). To better evaluate whether sex chromosomes or gonadal sex determines the relationship between p21 and SA-β-gal, we grouped the genotypes based on these factors and then compared XY vs XX (Fig. 8d, center, Supplementary astrocytes isolated from mice that were gonadally female had a significant relationship between p21 and senescence, as measured by SA-β-gal. This suggests that gonadal sex, and the epigenetic effects of in utero gonadal hormones, patterns the role of p21 in senescence induction in response to irradiation.

Discussion
Glioblastoma remains an incurable disease, with limited treatment options. Surgical resection, followed by radiation and chemotherapy remains the most effective treatment strategy. Better understanding of the mechanisms underlying the therapeutic response to radiation may help enhance treatment efficacy. Here we show that radiation response correlates with different molecular signatures in male and female human GBM lines. Using a mouse model of GBM, we further identify increased sensitivity to radiation in female GBM astrocytes, and determine that senescence is a major component of the radiation response in both sexes. With a correlationbased approach, we identified p21 as a likely mediator of irradiation induced senescence.
Strikingly, the relationship between p21 and the senescence marker SA-β-gal significantly differed in male and female cells -a finding that was observed in mouse GBM model astrocytes, mouse wildtype astrocytes, and primary human GBM lines. Female cells had higher levels of senescence, as measured by SA-β-gal, in response to the same levels of p21suggesting that p21 plays a more critical role in irradiation induced cellular senescence in female than in male cells. This finding was confirmed by p21 knockdown, which decreased the percentage of senescent cells in female GBM model astrocytes to that of males. Finally, using a novel FCG model of GBM, which enables mechanistic dissection of the biological underpinnings of a sex difference, we determined that the relationship between p21 and irradiation induced cellular senescence is patterned by gonadal sex. This finding will help guide future studies that aim to uncover the pathways regulating the p21-senescence axis and eventually modulate this interaction to improve radiation response in both sexes.
Our results highlight some intriguing directions for future studies. Interestingly, when p21 was knocked down in female GBM model astrocytes, it did not eliminate the senescence response to irradiation, but did abrogate the sex difference in response. This suggests that there are additional pathway(s), beyond p21, contributing to senescence induction after radiation, and that these are shared by males and females. Which signaling pathways, and whether they are always active or are upregulated with p21 loss, remains to be determined. Another avenue of future direction is uncovering how gonadal sex patterns the relationship between p21 and senescence. Our findings suggest that the testosterone surge in utero (or lack thereof in females) is responsible for the sex differences in the p21-senescence axis. The hormone surge has been shown to exert long term organizational effects in the brain through epigenetic mechanisms 1,25 . However, the differences we observe are not explained through simple regulation of accessibility at the p21 locus, since equivalent levels of p21 in males and females correspond to different percentages of SA-β-gal positive cells. Instead, differences in hormone exposure in utero may regulate proteins that influence p21 localization or activity through posttranslational modifications [26][27][28][29] . Future studies will investigate this possibility, with the goal of enhancing the senescence response after irradiation.
We also observed a sex difference in the levels of SA-β-gal cells in male and female wildtype astrocytes with continued passaging in vitro. This suggests female cells may be more susceptible to replicative senescence and/or to oxidative stress, since cells were grown at supraphysiological oxygen levels 22 . This has important implications for the fields of aging and neurodegenerative disease. Senescent cells contribute to a wide range of normal and pathological changes with aging 30,31 , and are increasingly thought to play a role in neurodegenerative disease 32,33 . Sex differences in p16 and p21 expression in aged mice in vivo have been reported, although this study actually found higher levels of these in male mice than female mice 34 . In our study, the same levels of p21 were associated with increased SA-β-gal cells in females, raising the possibility that this difference may not actually translate into increased senescent cells in males. Alternatively, it is possible that the senescent cell phenotype may also differ between male and female cells. The senescence associated secretory phenotype varies based on cell type and senescence induction mechanism 18,35 , and so could reasonably be influenced by cell sex. This could potentially result in different rates of clearance for male and female senescent cells in vivo, or mean that the same levels of senescence have different effects on the surrounding tissue depending on sex.
One of the potential limitations of our study is that we rely on SA-β-gal for our primary measure of senescence. While SA-β-gal is the most widely used marker of senescence, there is debate about its specificity, since it is not required for senescence induction and has been observed in non-senescent cells 16 . In support of the idea that SA-β-gal does correspond to senescence in our study: 1) it increased with exposure to radiation in a dose dependent manner, 2) along with increases in SA-β-gal, we saw more cells with an enlarged, irregular shape and a decrease in cells positive for the proliferation marker Ki67, 3) expression changes consistent with ionizing radiation induced senescence were observed in our cells, and 4) levels of the cell cycle inhibitor p21 also increased. Regardless of whether SA-β-gal is a measure of true senescence, our findings still have important implications for the field. When interpreting studies that use p21 as a marker of senescence, it may be necessary to consider the sex of the subjects, and whether the same pathway could have differences in downstream outcomes between male and female cells.
Finally, our results highlight the utility of correlation analyses when studying sex differences.
Most phenotypes are not truly sexually dimorphic, but represent a spectrum of values in males and females. While the mean values differ, and the two ends of the spectrum are populated primarily by individuals of one sex or another, there is considerable overlap. In addition, the variability between individuals of a single sex may obscure the differences between the two sexes, requiring a large sample size to detect the difference. With our approach, we discovered a difference in the relationship between p21 and SA-β-gal in males and females that was not apparent when we simply compared mean p21 levels alone. This could offer a powerful investigation strategy that takes advantage of individual variability rather than viewing it as a liability.
In summary, we have uncovered a novel sex difference in the relationship between p21 and irradiation induced cellular senescence that has potential implications for the fields of cancer research, neuroscience, and aging. Better understanding of this mechanism could identify novel approaches to improve response to cancer therapy.

Primary human GBM lines
Primary human GBM lines were kindly provided by Dr. Albert H. Kim. Specimens for culture were obtained prospectively at the time of surgery and cell lines were established as described 36  detached with Accutase (Sigma) and split 1:2 to 1:4 depending on cell line. One female human line (B51) was excluded from the SA-β-gal studies, since these cells did not attach to glass coverslips, even when coverslips were coated with laminin.

Mouse wildtype astrocytes
All animals were used in accordance with an Animal Studies Protocol approved by the Animal Studies Committee of the Washington University School of Medicine, per the recommendations of the Guide for the Care and Use of Laboratory Animals (NIH). Mouse wildtype astrocytes were isolated from the cortex of postnatal day 1 C57BL/6J pups as described 13 . Briefly, the cortices were dissected from the rest of the brain in cold HBSS and the meninges were removed.

Mouse GBM model astrocytes
Nf1-/-DNp53 astrocytes were generated as described 13 . Briefly, astrocytes were isolated from the cortex of postnatal day 1 Nf1 flox/flox ;GFAP-Cre mouse pups. Sex was determined by genotyping pup tail DNA for the X and Y chromosome paralogs Kdm5c (Jarid1c) and Kdm5d (Jarid1d) 37 . Astrocytes from at least three male and three female pups were then pooled by sex. astrocytes that received the pX330-Nf1-p53 plasmid were immortalized and had a significant growth advantage. Since wildtype astrocytes stop dividing after passage 5-6, growth advantage was used to select for astrocytes with successful CRISPR mutation/deletion of Nf1 and p53.
FCG-Cas9 Nf1/p53 CRISPR astrocytes were passaged at least 3 times, then protein was collected for confirmation of Nf1 and p53 knockdown by western blot.

For irradiation experiments, cells were irradiated using an RS 2000 X-ray irradiator (Rad Source
Technologies) unless otherwise specified in the specific methods section. Radiation was delivered at a dose rate of ~1.8 Gy/min with 160 kVp X-rays. Control cells were transported to the irradiator and sat on the bench for the same length of time as cells were in the irradiator.

Human GBM irradiation dose response curves
Human GBM cells were lifted with Accutase, counted, and 20,000 cells per well were plated in

Mouse GBM irradiation dose response curves
Irradiation dose response curves for mouse GBM model astrocytes were performed using the sulforhodamine B assay as described, with minor modifications 42

Cell growth assays via live cell imaging
Cells were plated at 1000 cells per well into a 96-well plate with 5 technical replicates plated per condition. Cells were irradiated using a GammaCell 40 Irradiator (Best Theratronics). Nonirradiated plates were transported to the radiation facility, but not exposed. The 96-well plates were then immediately placed into the IncuCyte ZOOM live cell imaging system (Satroius).
Phase-contrast images were taken every 4 hours for a total of 72 hours with 4 scan areas taken per well. Cell confluence was analyzed using the IncuCyte ZOOM analysis software. Percent confluence over time was used as a measure of longitudinal cell growth.

Clonogenic Assay
Cells were plated at 500 cells per well into 6-well plates.

Immunocytochemistry for Ki67
Nf1-/-DNp53 astrocytes were plated in 6-well plates and irradiated the following day.

RNA isolation and cDNA preparation -Mouse GBM model astrocytes
RNA was isolated from Nf1-/-DNp53 astrocytes and FCG GBM model astrocytes using TRIzol Reagent (Invitrogen), following the manufacturer's protocol. Briefly, cells were grown in 10 cm dishes to ~70% confluence, then irradiated. At desired timepoint after irradiation, media was aspirated, and cells were washed once with cold PBS. 1 ml of Trizol was added to the dish; cells were scraped into Trizol, allowed to sit 5 minutes at room temperature, then transferred to a 1.7 ml tube. 200 μl of chloroform was added, and samples were shaken vigorously for 15 seconds, then sat at room temperature for 3 minutes. Samples were spun at 12,000xg for 15 minutes at 4°C. The aqueous phase was transferred to a new tube, and 500 μl of isopropyl alcohol was added to precipitate RNA. Samples were incubated for 10 minutes at room temperature, then spun at 12,000xg for 10 minutes at 4°C. Supernatant was aspirated, and pellet was washed with aspirated. After drying, pellet was resuspended in 35 μl molecular biology grade water (DNase/RNase free). RNA concentration was measured with a NanoDrop 1000 spectrophotometer (Thermo Scientific). RNA was treated with Amplification Grade DNAse I (Invitrogen) to eliminate genomic DNA, and cDNA was generated using the SuperScript III First-Strand Synthesis System (Invitrogen), according to the manufacturer's instructions.

RNA isolation and cDNA preparation -Mouse wildtype astrocytes
RNA was isolated from mouse wildtype astrocytes using the QIAGEN RNeasy Mini Kit, according to the manufacturer's instructions. Briefly, astrocytes were grown in 10 cm Primaria dishes to ~70% confluence, then irradiated. At desired timepoint after irradiation, media was aspirated, and cells were washed once with cold PBS. 350 μl of RLT buffer + βmercaptoethanol was added and cells were scraped into RLT buffer, then transferred to a 1. The slope difference between the male and female-specific slopes was derived with 95% confidence interval (CI) and tested against 0 by the 2-sided Wald test. Significance was claimed at p<0.05. Detailed slope and correlation estimation results can be found in Supplementary   Table 3. For the Nf1-/-DNp53 mRNA correlations only, mRNA expression levels from cells irradiated with 0, 6 or 9 Gy were correlated with SA-β-gal results from cells irradiated with 0, 6, or 8 Gy -thus resulting in a correlation consisting of 0, 6, and 8/9 Gy doses. For all other correlations, irradiation levels were exactly the same for protein/mRNA measures and SA-β-gal.

Data Availability
The data that support the findings of this study are available from the corresponding author upon reasonable request.