Mechanisms and consequences of ATMIN repression in hypoxic conditions: roles for p53 and HIF-1

Hypoxia-induced replication stress is one of the most physiologically relevant signals known to activate ATM in tumors. Recently, the ATM interactor (ATMIN) was identified as critical for replication stress-induced activation of ATM in response to aphidicolin and hydroxyurea. This suggests an essential role for ATMIN in ATM regulation during hypoxia, which induces replication stress. However, ATMIN also has a role in base excision repair, a process that has been demonstrated to be repressed and less efficient in hypoxic conditions. Here, we demonstrate that ATMIN is dispensable for ATM activation in hypoxia and in contrast to ATM, does not affect cell survival and radiosensitivity in hypoxia. Instead, we show that in hypoxic conditions ATMIN expression is repressed. Repression of ATMIN in hypoxia is mediated by both p53 and HIF-1α in an oxygen dependent manner. The biological consequence of ATMIN repression in hypoxia is decreased expression of the target gene, DYNLL1. An expression signature associated with p53 activity was negatively correlated with DYNLL1 expression in patient samples further supporting the p53 dependent repression of DYNLL1. Together, these data demonstrate multiple mechanisms of ATMIN repression in hypoxia with consequences including impaired BER and down regulation of the ATMIN transcriptional target, DYNLL1.

treatment, and for recruitment of 53BP1 to foci after both MMS and APH treatment 22,[26][27][28][29] . A role for ATMIN in maintaining genomic stability was also demonstrated to be important in the suppression of B cell lymphoma 30 . Finally, ATMIN, which includes a Zn 2+ finger domain, has been shown to act as a transcription factor driving expression of a very limited number of target genes including DYNLL1, which plays an important role in lung morphogenesis 28,29 .
Here, we demonstrate that ATMIN was not required for ATM activation in hypoxia and was instead repressed in hypoxic conditions therefore contributing to the hypoxia-mediated repression of DNA repair. The mechanism of repression is complex and is in part dependent on both HIF-1 and p53, demonstrating that repression of ATMIN is critical to the biological response to hypoxia. While this may be due to the role of ATMIN in DNA repair, we also demonstrate for the first time that expression of the ATMIN target gene DYNLL1 is altered in response to hypoxia.

ATMIN is dispensable for ATM activation in hypoxia.
To test the hypothesis that ATMIN might be required for hypoxia-induced ATM activation, we used ATMIN siRNA to deplete ATMIN levels in RKO cells. The cells were then exposed to hypoxia (< 0.1% O 2 ) for 3 h, as we have shown previously that ATM is activated during this period 13 . As expected, a robust induction of phosphorylated ATM (ATM-S1981) and the ATM target Kap-1 (Kap1-S824) was observed in response to hypoxia in the cells treated with the control siRNA. Surprisingly, a similar induction of ATM activity was also observed in the cells with depleted levels of ATMIN (Fig. 1A). To ensure that ATMIN did not have role in sustaining hypoxia-induced ATM activation, experiments over a longer time period were also carried out and again no dependence on ATMIN was observed ( Figure S1). This finding  24 h post transfection, cells were exposed for up to 3 h of hypoxia (< 0.1% O 2 ). (B) Colony survival assay of RKO cells treated with 10 μ M KU-55933, exposed to 6 h hypoxia (< 0.1% O 2 ), and irradiated with 0-6 Gy. Post treatment, cells were allowed to form colonies for 8 days under normal tissue culture conditions. Mean and standard deviation (SD) from 3 independent experiments is shown. (C) RKO cells transfected with either scrambled (Scr) siRNA or ATMIN siRNA were exposed to 6 h of hypoxia treatment (< 0.1% O 2 ) or left at normoxia (21% O 2 ), followed by irradiation (4 Gy). Post radiation, cells were reoxygenated and kept under normal tissue culture condition to allow colony formation. (D) Colony survival assay of RKO cells transfected with ATMIN siRNA, exposed for up to 36 h of hypoxia (< 0.1% O 2 ). For all colony formation data mean surviving fractions are shown ± SD from three independent experiments. was also verified with two single siRNAs to ATMIN ( Figure S2). As a control, we verified that loss of ATMIN decreased ATM activity in response to APH-induced replication stress ( Figure S3) 22,31 . Loss of ATM has a dramatic and well-characterized effect on radiosensitivity, although this has not been as extensively demonstrated under hypoxic conditions 32 . Here, we used an ATM inhibitor (KU-55933) in hypoxic conditions and exposed RKO cells to radiation (0-6 Gy). As expected, the cells treated with the ATM inhibitor were significantly more sensitive to radiation even in hypoxic conditions (SER 37 = 2.1) (Figs 1B and S4). We then asked if ATMIN might contribute to the role of ATM in radiation response under hypoxic conditions by carrying out the same experiment in cells treated with a siRNA against ATMIN. Loss of ATMIN did not affect the radiation sensitivity of cells irradiated under hypoxic conditions suggesting that ATM function could not have been significantly impacted by ATMIN depletion (Fig. 1C). Previous studies have shown that loss of ATM activity leads to increased sensitivity to hypoxia/reoxygenation 15,33 . Therefore, if ATMIN contributes to ATM activity in hypoxia, loss of ATMIN would also be expected to increase sensitivity to hypoxia/reoxygenation. Again, ATMIN was depleted using siRNA in RKO cells and a colony survival assay in response to hypoxia/reoxygenation was carried out. Depletion of ATMIN had no significant effect on cell viability in response to hypoxia compared to wild type RKO cells, which again suggests that ATMIN does not contribute to ATM activity during hypoxia (Fig. 1D). To further support this conclusion, we investigated the induction of apoptosis 6 hours after hypoxic exposure and found that loss of ATMIN did not significantly affect the fraction of apoptotic cells ( Figure S5). Altogether, these data demonstrate that ATMIN is not required for the activation of ATM in response to hypoxia-induced replication stress.
ATMIN is repressed in an oxygen dependent manner through p53 and HIF-1α. While investigating the potential role of ATMIN on ATM activation in hypoxia we noticed that ATMIN levels decreased over time in hypoxia ( Figure S1). Both cancer (RKO, HCT116) and non-cancerous (RPE1-hTERT) cell lines were exposed to hypoxia (< 0.1% O 2 ) for up to 24 h and analyzed for ATMIN expression by western blotting. In each case the levels of ATMIN protein were significantly decreased in hypoxia ( Fig. 2A-C). Next, we tested the oxygen dependency of ATMIN decrease by comparing exposure to severe (< 0.1% O 2 ) and mild hypoxia (2% O 2 ). The levels of ATMIN decreased at both oxygen tensions, although more profoundly at < 0.1% O 2 (27% and 55% decrease at 2% O 2 and < 0.1% O 2 , respectively, when compared to ATMIN levels at 21% O 2 ) (Fig. 2D). In addition, we investigated the effect of commonly used hypoxia mimetics, desferoxamine (DFO) and CoCl 2 , and found that DFO treatment partially repressed ATMIN (25% decrease), while CoCl 2 had no effect on ATMIN protein (Fig. 2D). As the levels of ATMIN have not been reported to change in response to stress previously, we asked if this also occurred upon exposure to alternative sources of replication stress. RKO cells were exposed to Hu or APH for a period of 3-6 h and the levels of ATMIN protein determined. No apparent decrease in ATMIN was detected in response to either treatment suggesting that hypoxia has a unique effect on ATMIN levels (Fig. 2E). To rule out repression of ATMIN with slower kinetics compared to hypoxia, no decrease in ATMIN expression was observed when cells were treated with Hu and APH for up to 18 h ( Figure S6). Together, this data demonstrate that ATMIN is repressed in response to a range of hypoxic conditions (< 0.1-2% O 2 ) but this effect is more profound at the more severe level of hypoxia.
As HIF-1 has been demonstrated to play a role in the repression of other DNA repair proteins in hypoxia we investigated a potential role for HIF-1 in ATMIN repression 11 . However, as the repression of ATMIN was more significant in severe hypoxia compared to milder levels we also included p53 in our analysis, as p53 activity is also restricted to severely hypoxic conditions 34 . We used isogenic cell lines HCT116 p53 +/+ and p53 −/− or RKO HIF-1α +/+ and RKO HIF-1α −/− to initially investigate this hypothesis. We observed that ATMIN repression in hypoxia was partially reversed when either p53 or HIF-1α were deleted ( Figure S7A,B). To investigate further the contribution of p53 and HIF-1 to ATMIN repression, we used siRNA to knock down HIF-1α in a p53 null background and compared ATMIN protein levels after hypoxia. We observed that the combined depletion of p53 and HIF-1α had the largest effect on rescuing ATMIN protein levels when compared to knock-down or deletion of either transcription factor alone (Fig. 2F). These results suggest that both p53 and HIF-1α independently contribute to repression of ATMIN in hypoxia.

Hypoxia represses ATMIN expression at the protein level without increasing degradation rates. As both transcription factors HIF-1 and p53 play a role in ATMIN repression we investigated ATMIN
mRNA expression in hypoxia. ATMIN mRNA was measured by qPCR in response to hypoxia in 5 cells lines (RKO p53wt , HCT116 p53wt , U87 p53wt , MCF-7 p53wt and U2OS p53wt ) all of which have a normal HIF-1 response. ATMIN mRNA levels were not significantly affected by hypoxia in the majority of the cell lines tested with the exception of HCT116 cells where mRNA decreased initially (after 8 h) and then basal levels were restored. These results demonstrate that in the cell lines tested, neither p53 nor HIF-1 directly or indirectly repress ATMIN transcription (Figs 3A-C and S8A,B). This was further confirmed by analyzing ATMIN mRNA expression in cells lacking either p53 (HCT116 p53 −/− ) or HIF-1α (RKO HIF-1α −/− ), again no hypoxia-dependent increase in expression was observed ( Figure S8C,D).
ATMIN contains a C-terminal PEST domain, which is associated with proteasomal degradation 21 . Given the stability of ATMIN mRNA in hypoxia we hypothesized that ATMIN could be repressed through increased protein degradation in hypoxia. We measured ATMIN in cells exposed to a proteasomal inhibitor (MG-132) either in normoxia or in hypoxia. The hypoxia-mediated repression of ATMIN was abrogated by the addition of proteasomal inhibitor MG-132 confirming that ATMIN levels are regulated by proteasomal degradation (Fig. 3D). However, ATMIN also accumulated in normoxic conditions (21% O 2 ) in the presence of MG-132, suggesting that ATMIN is not specifically targeted to the proteasome in hypoxic conditions. To verify this further we measured the half-life of ATMIN in both normoxia and hypoxia by blocking protein translation with cycloheximide. We determined that the half-life of ATMIN (approximately 4 hours) was not altered by exposure to hypoxia (Figs 3E and S9).
Scientific RepoRts | 6:21698 | DOI: 10.1038/srep21698 Together, these findings suggest that in response to hypoxia ATMIN translation is repressed as opposed to altered transcription or degradation rates.
Biological consequences of hypoxia-mediated repression of ATMIN include the down regulation of ATMIN target DYNLL1. ATMIN has been described as having a role in BER although this role is somewhat unclear 26,27 . BER is known to be functionally repressed in hypoxic conditions and consequently hypoxic cells are less able to repair damage induced by agents such as MMS 5 . To determine if the loss of ATMIN expression contributes to the increased sensitivity to MMS observed in hypoxic cells we exposed cells treated with ATMIN siRNA to range of hypoxic conditions and MMS and carried out a colony survival assay. In agreement with previous reports we observed an increased sensitivity of hypoxic cells to MMS 5 . We also found a significant decrease in cell survival of cells treated with MMS when ATMIN levels were depleted (Fig. 4A) 22,26,27 . This effect was decreased in cells exposed to 2% O 2 and completely abolished in cells treated with < 0.1% O 2 (Fig. 4A), where ATMIN was most repressed as a result of the hypoxia exposure.
ATMIN directly transactivates DYNLL1, therefore suggesting the novel hypothesis that DYNLL1 mRNA levels might be regulated in an oxygen dependent manner 29 . In agreement with previous reports, we confirmed that DYNLL1 expression is dependent on ATMIN levels in RKO cells (Fig. 4B). Next, we exposed cells to hypoxia, and observed an oxygen-dependent down-regulation of DYNLL1 expression, again with a more profound decrease at < 0.1% O 2 (Fig. 4C), which correlates with ATMIN levels (Fig. 2D). The hypoxia-mediated decrease in DYNLL1 expression was rescued by over-expression of ATMIN in these conditions (Figs 4D and S10). As the decrease of  HCT116 (B) and RPE1-hTERT (C) cells were exposed to hypoxia (< 0.1% O 2 ) for the times indicated and western blot analysis was carried out using the antibodies indicated (β -actin was used as a loading control). (D) RKO cells were treated as indicated for a period of 16 h. The hypoxia mimetic drugs were used as follows; 100 μ M DFO and 150 μ M CoCl 2 . Western blots using the antibodies indicated are shown. (E) RKO cells were exposed to DMSO, 1 mM Hu or 5 μ g/ml APH for the indicated amount of time and western blot analysis was carried out for ATMIN, RPA 32 and β -actin (loading control). (F) HCT116 p53 −/− cells were transfected with either HIF-1α siRNA or Scr siRNA and HCT116 p53 +/+ cells were transfected with Scr siRNA. Cells were exposed to hypoxia (< 0.1% O 2 ) for the times indicated and protein lysates analyzed by western blotting with the antibodies shown.
Scientific RepoRts | 6:21698 | DOI: 10.1038/srep21698 ATMIN in severe hypoxia was p53 dependent (Figs 2F and S7A), we asked if DYNLL1 was also regulated in a p53-dependent manner in hypoxia. Therefore we tested the levels of DYNLL1 mRNA in RKO cells treated with p53 siRNA. As predicted, hypoxia-mediated repression of DYNLL1 was partially but significantly rescued in the absence of p53 (Fig. 4E). This study highlights an entirely novel link between ATMIN regulation of DYNLL1 and tumor hypoxia.
To date no validated hypoxia signatures, which are specific to the levels of hypoxia which induce replication stress i.e. < 0.1% O 2 exist. However, we recently identified a group of six genes, which are induced by p53 in response to hypoxia (< 0.1% O 2 ) and can predict p53 status in vivo in multiple cancer types 19 . Therefore, we used this group of genes to investigate a potential correlation between DYNLL1 expression and hypoxic p53 signaling in patient samples. Through analysis of the TCGA breast and lung cancer patient cohorts we found that DYNLL1 expression significantly and inversely correlated with the hypoxia-inducible p53-dependent group of genes (Pearson r = − 0.09197, p = 0.0013 for breast and Pearson r = − 0.2618, p < 0001 for lung cancers, respectively) (Fig. 4E), suggesting that hypoxia-and p53-dependent repression of the ATMIN target DYNLL1 occurs in human cancers. Altogether, our data shows that repression of ATMIN in hypoxia is likely to have multiple consequences, including contributing to the repression of DNA repair pathways and down regulation of the ATMIN target, DYNLL1.

Discussion
Despite having a role in ATM activation in response to replication stress, ATMIN does not play a role in hypoxia-induced ATM activation. Instead, we show for the first time that ATMIN is repressed in response to hypoxia and most significantly under severe hypoxia (< 0.1% O 2 ). Our data demonstrates that ATMIN repression is in part dependent on both p53 and HIF-1α . The involvement of both transcription factors suggests that ATMIN would be repressed in a broad range of hypoxic conditions and therefore in large areas of tumor. Repression of ATMIN in hypoxia is suggested to contribute to down regulation of BER and decreased expression of DYNLL1, as summarized in the proposed model (Fig. 5). Importantly, we show that repression of DYNLL1 can be inversely (D) RKO cells were transfected with pMSCV-ATMIN vector or an empty vector control. 24 h later cells were transferred to hypoxic conditions (< 0.1% O2) for 8 h. qPCR was then carried out for DYNLL1. (E) DYNLL1 mRNA was analyzed by qPCR in RKO cells transfected with either Scr or p53 siRNA and exposed to 16 h of hypoxia (< 0.1% O 2 ). All qPCR bar graphs show mean mRNA expression ± SD from three independent experiments. (F,G) Expression of DYNLL1 (Log 10 conversion) in the breast (E) and lung (F) TCGA datasets is shown against hypoxia-inducible p53 signature (Log 10 conversion). One-tailed p value is shown on each graph for each Pearson r (correlation coefficient). correlated with p53 signaling in human cancers suggesting that the consequences of ATMIN repression are likely to play a role in tumors in vivo.
The mechanism of repression of ATMIN protein in hypoxia remains unclear, although dependence on p53 and HIF-1α in this process may suggest involvement of micro-RNAs (miRs). Both p53 and HIF-1α are known to induce expression of multiple miRs, which are likely to inhibit protein translation 35,36 . Analysis of the ATMIN mRNA sequence using available online resources (e.g. www.microrna.org) showed that multiple p53 or HIF-dependent miRs are predicted to bind to the ATMIN transcript, including miR-373 (a HIF-1α target) and miR-34 (a p53 target) 12,37,38 . In addition, a recent study, which screened for the miR-124 targets involved in the DNA damage response, found ATMIN to be regulated by miR-124 39 . Importantly, miR-124 is a known direct target of p53 suggesting that it is likely to be involved in down-regulation of ATMIN under hypoxia 40 . It is noteworthy that the dependence on p53 and HIF-1 for repression of ATMIN in hypoxia was less significant after longer exposure times to hypoxia (> 16 h). This finding suggests that in chronic hypoxic conditions ATMIN might be repressed through an alternative mechanism. A recent study characterized the repression of the BER pathway in chronic hypoxia (72 h 0.2% O 2 ) as a consequence of inhibited protein translation 5 . Together, these findings suggest that the repression of ATMIN occurs in hypoxia through mechanisms dependent on HIF-1, p53 and eventual decrease in translation.
Finally, we show that ATMIN repression in hypoxia is linked with down regulation of DYNLL1 expression. Numerous functions have been attributed to DYNLL1, including regulation of mitosis, spindle orientation, macropinocytosis, nuclear localization/translocation of some proteins e.g. 53BP1, PAK1 or ciliogenesis [41][42][43][44] . Our findings suggest that some or all of these processes will be altered in response to hypoxia through the decreased expression of DYNLL1. In support of this conclusion we have shown that 53BP1 does not form foci in response to hypoxia (< 0.1% O 2 ) and so there would be no requirement for this function of DYNNL1 13 . Most importantly our data demonstrate that in the absence of functional p53 the hypoxia-mediated repression DYNNL1 described here is alleviated. The specific roles of DYNNL1 in hypoxia driven tumor progression therefore warrant further study.

Methods
Cell lines, transfections and drug treatments. RKO Colony survival. RKO cells were transfected with 50 nM ATMIN siRNA and the next day re-plated at a density of 250-2000 cells/well in a 6-well plate. After 2 h, cells were exposed to < 0.1% O 2 for 0-36 h and/or drug treatment and then left for 7-9 days to form colonies, which were visualized by crystal violet staining.
Hypoxia and radiation. Hypoxic treatments at < 0.1% O 2 were carried out in a Bactron chamber (Shel Lab) and at 2% O 2 in a Don Whitley H35 Hypoxystation. Radiation doses were delivered in a GSM D1 137-Cesium gamma irradiator with at 0-6 Gy doses as previously described 47 . Western blotting. Protein lysates were extracted with SDS lysis buffer (10 mM Tris-Cl, pH 7.5, 0.1 mM EDTA, 0.1 mM EGTA, 0.5% SDS, 0.1 mM β -mercaptoethanol, protease/phosphatase inhibitors) and western blotting carried out as described previously 34 . Western blots were imaged using the Odyssey infrared system (LI-COR). The following antibodies were used: ATMIN (Millipore), HIF-1α (BD Biosciences), p53 (Santa Cruz), Figure 5. ATMIN expression is repressed in response to hypoxia through mechanisms which include both HIF-1a and p53. The involvement of both HIF-1α and p53 in the repression of ATMIN suggests that this would occur in a significant fraction of the tumour, although p53-mediated repression was more robust. The consequences of loss of ATMIN include increased sensitivity to DNA damaging agents requiring effective BER for repair and decreased expression of DYNLL1. Some of the potential consequences of loss of DYNNL1 in hypoxia are indicated and warrant further study.
Expression analysis in cancer datasets. Raw RNA-sequencing (RNA-seq) data for 1105 Breast invasive carcinoma and 522 Lung adenocarcinoma tumors were downloaded from the TCGA project (accessed through cBioportal: http://www.cbioportal.org/) on 10 th September 2015 and 12 th September 2015 respectively. To examine tumor-associated p53-activity (referred to as p53 signature in the figure), raw data for each sequenced gene were rescaled to set the median equal to 1, and p53-activity was quantified by averaging the normalized expression of 6 p53 target genes, associated with hypoxia-induced p53 activity (encoding BTG2, CYFIP2, INPP5D, KANK3, PHLDA3 and SULF2) 19 . Log 10 conversion of the p53 signature was plotted against Log 10 conversion of raw data for DYNLL1 (also rescaled to set the median equal to 1). One-tailed p value shown on each graph for each Pearson r (correlation coefficient).
Statistical analysis. Statistical analysis of clonogenic assays, mRNA expression and expression analysis in cancer datasets was carried out using Graphpad Prism v6.03 software. For colony survival assays the plating efficiency per condition was determined by dividing the total number of counted colonies by the total number of plated cells. Surviving fractions (SF) were calculated by dividing the plating efficiency of treated cells by the plating efficiency of the control cells. Comparison of groups was done using an unpaired t-test with Welch's correction. Data was fitted according to the linear quadratic model using the formula SF = e^ − (α D + β D 2 ). Sensitizer enhancement ratio (SER) was calculated as SER 37 = D 0 (without sensitizer)/D 0 (with sensitizer) for the same biological effect at SF = 37%. qPCR data was analyzed on 7500 Software V.2.0.5 (Applied Biosystems) using the Δ Δ Ct method. An unpaired t-test with Welch's correction was used. Statistical significance is marked as follows: *p < 0.05, **p < 0.01, ***p < 0.001.