Targeting Homologous Recombination in Notch-Driven C. elegans Stem Cell and Human Tumors

Mammalian NOTCH1-4 receptors are all associated with human malignancy, although exact roles remain enigmatic. Here we employ glp-1(ar202), a temperature-sensitive gain-of-function C. elegans NOTCH mutant, to delineate NOTCH-driven tumor responses to radiotherapy. At ≤20°C, glp-1(ar202) is wild-type, whereas at 25°C it forms a germline stem cell⁄progenitor cell tumor reminiscent of human cancer. We identify a NOTCH tumor phenotype in which all tumor cells traffic rapidly to G2⁄M post-irradiation, attempt to repair DNA strand breaks exclusively via homology-driven repair, and when this fails die by mitotic death. Homology-driven repair inactivation is dramatically radiosensitizing. We show that these concepts translate directly to human cancer models.


Worm RNAi by feeding
RNAi was performed essentially as per [29]. Single colonies of HT115 bacteria containing L4440 plasmids with cloned fragments corresponding to target genes were from Vidal and Ahringer RNAi feeding libraries. Each RNAi reagent was verified by DNA sequencing. Young adult hermaphrodites were placed onto NGM plates seeded with dsRNA-expressing or empty vector control bacteria (RNAi feeding plate). After overnight incubation, worms were transferred to an identical fresh RNAi feeding plate and allowed to lay eggs for 2h. RNAi phenotypes of synchronized F1 progeny were examined at the indicated times post radiation.

Quantitative PCR
Worms were collected in Trizol reagent (Invitrogen) and subjected to three rounds of freeze cracking by alternating between liquid nitrogen and room temperature. Crude RNA extracts were collected and purified with RNeasy Mini Kit (Qiagen) according to manufacturer's instructions. 1 μg of total RNA was reverse-transcribed in 20 μl using the Thermoscript RT-PCR system (Invitrogen) at 50°C for 1h. Quantitative PCR was performed on the Applied Biosystems 7500 FAST Real Time PCR instrument with Taqman Gene Expression assay system. The IDs of C. elegans gene expression assay are: mre-11-Ce02480998_g1; rad-51-Ce02458920_g1; atl-1-Ce02479867_g1; mus-101-Ce02413322_g1; cku-80-Ce02445546_g1; lig-4-Ce02449042_g1; hsr-9-Ce02412427_g1; rad-50-Ce02482582_g1; mec-7-Ce02497588_g1. Expression level of each sample was standardized to C. elegans mec-7 endogenous control standard. Knockdown was calculated as percentage remaining gene expression normalized to relevant non-silenced control.

Germ cell quantification
Worms were fixed in ethanol and stained with DAPI using Vectashield (Vector Laboratories Inc.). Z-stack images were acquired with a 20x water objective at 2-μm intervals using a Leica Confocal Microscope. To quantify C. elegans germ cell nuclear numbers each entire z-stack was loaded into Volocity (version 5.3.1) as a single lei file. Then the entire area of visible DAPIstained germ cells in one gonad arm was selected for analysis. If the two gonads were uneven size, germ cells from both gonads were measured and averaged. In the selected gonadal area threshold intensity was set high enough such that the program identified individual cells and excluded spaces between cells. The Volocity Program requires an approximate size guide to find objects. We determined the approximate nuclear volume experimentally by measuring volume from high magnification images of DAPI stained nuclei (63x, zoom 5). At least 100 nuclei from 4-5 worms per condition were measured. Volocity quantification was verified by hand counting of~20 gonads from glp-1(ar202). Generally, Volocity numbers were lower than hand counts, but differed by <5%.

glp-1(ar202) tumor cell cycle arrest
Adult worms, raised at 15°C, were transferred to 25°C and allowed to lay eggs for 1.5h. After hatching, mid-L4 progeny were transferred to fresh plates and either irradiated at 480Gy, requiring 4h, and allowed to recover for 8h, or not. Relative germline nuclear DNA quantity was determined as per [30] with the following modifications: worms were stained with DAPI (Vector Laboratories Inc.) and all nuclei were quantified from position of cell diameter (CD) 6 through CD 15 from the distal tip, or CD -1 to -10 from the proximal end of the oviduct, which produced statistically-indistinguishable DNA content distributions. 2N DNA content was established from non-mitotic somatic cells of the vulva and uterus in the same animal and from sets of daughter chromosomes of anaphase germ nuclei, and was verified using 4N nuclei (metaphase figures and pachytene nuclei). To obtain the haploid equivalent, the total fluorescence from each germ cell nucleus was divided by one half of the 2N value obtained from the somatic cells. Every nucleus was measured from the distal tip to the first cell diameter within four cell diameters of the transition zone (to avoid meiotic S) as described previously [31].

Germ cell and somatic cell radiosensitivity assays
Radiation-induced germ cell apoptosis was analyzed as per [28]. Worms were synchronized at 25°C and irradiated at the L4 stage. Germ cell corpses were scored in the distal pachytene region of one gonad arm of wild-type worms, and in both distal and proximal regions of one gonad arm of glp-1(ar202). Radiation-induced somatic phenotypes were assessed by vulval morphology in adults derived from 120Gy-irradiated late-stage embryos (at 4h after egg laying). Vulval phenotypes are scored as wild-type or abnormal including protruding vulva (Pvl), vulvaless (Vul), ruptured vulva (Rup) and uncoordinated (Unc) using Nomarksi microscopy. Phenotype percentages were derived from animals surviving until adulthood. To examine meiotic chromosomes, L4 hermaphrodites were subjected to 120Gy, and after 18h DAPI-stained oocytes at diakinesis were evaluated under a Zeiss fluorescence equipped with epifluorescence filters.

Worm longevity studies
Assays were performed at 25°C. Synchronized L4-stage worms, timed to egg laying, were placed on seeded plates on day one. Adults were transferred from progeny onto fresh plates every other day until reproduction ceased. Data, derived from animals scored daily as dead or alive, is plotted as Kaplan-Meier survival curves using Graphpad Prism.

RAD51 shRNA
CUTLL-1 cells were infected with GIPZ Lentiviral particles expressing human RAD51 shRNA or non-silencing shRNA (Open Biosystems Inc. RAD51 clone ID V2LHS_171184). Stable cell lines were selected by addition of 1 μg⁄ml puromycin and GFP expression. Efficiency of RAD51 knockdown was measured by quantitative PCR as above. Human RAD51 expression level was normalized to human TATA-binding protein (TBP) expression (Open Biosystems, Inc. RAD51 assay ID is Hs-00153418 and TBP assay ID is Hs-433769-0711011).

XRCC4 shRNA
shRNA sequences were predicted by the Designer of Small Interfering RNAs (DSIR) software (http://biodev.extra.cea.fr/DSIR/DSIR.html). Multiple shRNA sequences were tested in order to achieve high knockdown efficiency. The shRNA constructs were cloned into the pHAGEpuro vector and transfected into 293T cells with delta 8.9 and pMDG vectors to produce lentivirus. CUTLL-1 cells were infected with unconcentrated virus overnight at 37°C and puromycin selected the next day. Efficiency of XRCC4 knockdown measured by quantitative PCR was 65% compared to empty vector-treated CUTLL-1 cells. Level of human XRCC4 expression was normalized to human TATA-binding protein (TBP) expression (Open Biosystems Inc. XRCC4 assay ID is Hs-01104868).

Clonogenic survival assay
Cells (0.5x10 6 ⁄ml complete media) were subjected to escalating radiation doses. At 1h post irradiation, cells were added into Methylcellulose Medium (Stemcell Technologies) working solution containing 20% fetal bovine serum according to manufacturer's instructions. The cell suspension was seeded onto 35 mm dishes in triplicate and after 11-14 days, surviving colonies, defined as a minimum of 50 cells, were counted using a stereoscopic microscope (Nikon TMS). Surviving fraction (SF) was calculated as number of colonies formed⁄number of cells seeded x plating efficiency. Radiation dose survival curves were fitted to the LQ standard model [34] using GraphPad Prism 6. D 0 (the dose required to reduce the fraction of surviving cells to 37% of its previous value) and D q (a threshold dose below which there is no effect) were calculated as Nomiya T described [34]. To test radiation-drug combination effect, cells were treated with Mirin (provided by the Organic Synthesis Core Facility, MSK) for 1h preceding irradiation, followed by a 12-day drug-free clonogenic assay.

Notch-driven tumor irradiation studies
6-8 week old non-obese diabetic⁄severe combined immunodeficient (NOD-SCID) female mice were purchased from Taconic Farms Inc. Mice were housed at the MSK animal core facility. Xenografted tumors were generated in murine right flanks using 5x10 6 CUTLL-1 cells infected with GIPZ shRNA non-silencing lentiviral particles or cells infected with GIPZ human RAD-51 shRNA lentiviral particles, selected as described above. At 100-150 mm 3 , tumors were irradiated using a Philips MG-324 X-ray unit at 117.5 cGy⁄min (50 cm source to skin distance). Tumor volumes were measured 2x per week for at least 15 weeks. Euthanasia is performed by exposing mice to 100% carbon dioxide in a cage or euthanasia chamber as recommended in The American Veterinary Medical Association (AVMA) Guidelines for the Euthanasia of Animals (2013, pp. 26, M1.6).
This study was carried out as recommended in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health. The protocol was approved by the Institutional Animal Care and Use Committee of Memorial Sloan-Kettering Cancer Center (IACUC protocol 92-10-038). All procedures performed comply with provisions of the Animal Welfare Act. Memorial Sloan-Kettering Cancer Center's animal care and use program is administered by the Research Animal Resource Center (RARC). The program has been fully accredited by the Association of Assessment and Accreditation of Animal Care, International (AAALAC) since 1967, is registered with the USDA, and has an approved assurance on file with the Office of Laboratory Animal Welfare, NIH (OLAW).

Statistical Analysis
Statistical significance was determined by a two-tailed Student t-test using GraphPad Prism software (GraphPad, San Diego, CA, USA). Results are presented as mean ± standard error. The P value in clonogenic survival of CUTLL-1 cells was calculated from the confidence interval as defined by Altman and Bland [35].
We and others reported that wild-type worms show dose-dependent germline apoptosis after irradiation, confined to cells in meiotic prophase just distal to the gonad arm bend [28,39,41]. To determine if radiation-induced apoptotic cell death contributes to germ cell loss  in ar202, we irradiated worms and examined germ cell apoptosis at 24h and 48h post radiation. However, there was little ar202 germ cell apoptosis after 240Gy (Fig 2D) or 480Gy (not shown). Since caspase gene ced-3 is required for radiation-induced germline apoptosis [42], ced-3 was inactivated by 2 approaches in glp-1(ar202), either by generating a ced-3(n717);glp-1 (ar202) double mutant or by RNAi, and germ cell number was scored after irradiation. Inactivation of caspase-mediated cell death by either approach did not alter ar202 radiation response (Fig 2E, S2 Fig), indicating radiation-induced germline loss in ar202 is non-apoptotic.

glp-1(ar202) germline tumor cells engage homology-directed repair (HDR) for radioprotection
An alternative death pathway might entail reproductive (mitotic) cell death, an outcome of failure of cycling cells to adequately repair DNA DSBs, usually by coordinate activation of NHEJ and HDR [43]. To explore mechanisms of DSB repair in glp-1(ar202), we employed RNAi knockdown of the conserved DDR repair machinery. Quantitative PCR confirmed RNAi knockdown efficiency (S3 Fig). Table 1 and Fig 3A summarize impact of DDR gene silencing. RNAi depletion of 5⁄6 HDR genes (mre-11, rad-51, rad-54, mus-101, atl-1, but not rad-50), and the npp-15 ortholog of human NUP133, a mammalian nuclear pore component [44], conferred radiosensitivity. Unlike other HDR genes, rad-50 knockdown in mutant glp-1(ar202) does not enhance radiosensitivity in mitotic germline tumors, although rad-50 gene expression was reduced after RNAi by 81±8% in ar202 (S3 Fig), indicating that C. elegans RAD-50 may not play a role in radiation-induced DSB repair in mitotic germ cells. This result is consistent with findings from Villeneuve and co-workers that showed RAD-50 is required for loading RAD-51 onto radiation-induced DSBs in meiotic but not mitotic germ cells [45].
To address whether ar202 germline tumors express NHEJ genes, we employed the temperature-sensitive germ cell-deficient mutant glp-4(bn2) [46]. S1 Table shows that when glp-4(bn2) animals are grown at the permissive temperature, and therefore contain a germ line, they express key NHEJ genes lig-4 and cku-80, as well as HDR genes mus-101, rad-51 and atl-1, at mutant worms were synchronized at 25°C and irradiated with 240Gy at the L4 stage. Germline apoptosis was scored in one gonad loop per worm. Incidence of germ cell death was quantified by dividing number of apoptotic germ cells by total germ cells. Data (mean±s.e.m) are from 10-12 worms⁄group. (E) Inactivation of apoptosis does not alter ar202 response to radiation. glp-1(ar202) and glp-1(ar202);ced-3(n717) double mutant worms were irradiated at the L4 stage. Data (mean±s.e.m) are from 9-12 worms⁄group. Note the line of "glp-1 480Gy" is hidden behind the line of "glp-1;ced-3 480Gy".
doi:10.1371/journal.pone.0127862.g002  Fig 1A. Mitotic germ cells reside between the distal end of the gonad (indicated by bold asterisk in bottom panel) and the transition zone [10], which characteristically contains crescent-shaped nuclei (arrow). *p<0.05 and **p<0.01 vs. empty vector control. (F) Knockdown of NHEJ genes results in vulval much higher levels than animals grown at the restrictive temperature, which lack a germ line. Gene expression levels in somatic tissue and germ line could also be affected by culturing animals at the different temperatures, although this is unlikely in our study. We conclude, therefore that NHEJ genes are, in fact, enriched in the germ line, while post-mitotic somatic cells in adult worms express minimal amounts. Consistent with these data, we recently reported mitotically-active cells of murine small intestinal crypts aggressively repair radiation DNA damage, while post-mitotic villus cells do not [23].
To obtain functional evidence that RNAi feeding adequately inactivated respective NHEJ DSB repair genes, we examined consequence of inactivating NHEJ genes on somatic development in irradiated wild-type worms. For these studies, N2 embryos grown in lig-4 RNAi plates were collected at 4h post egg laying, a time preceding vulval development, and irradiated with 120Gy. At 96h after 120Gy, minimal overall damage was detected in N2 worms even with rad-51 silencing, while lig-4 or cku-80 knockdown-worms displayed abnormal vulval development (Fig 3F, upper panel, p<0.01 for lig-4; p<0.05 for cku-80), with increased penetrance of somatic defects (lower panel) [47,48]. Taken together, our results suggest that failure of germline tumors to use NHEJ to repair radiation-induced DSBs results from lack of engagement of NHEJ repair machinery, rather than lack of availability of NHEJ repair genes in the germline. abnormalities post irradiation. Phenotypes were evaluated 120h post 120Gy using 75-85 worms⁄group. Somatic developmental phenotypes were quantified as wild-type vulva (WT), protruding vulva (Pvl), vulvaless (Vul), ruptured vulva (Rup) and uncoordinated (Unc). *p<0.05. (G) Knockdown of HDR genes in wild-type worms results in highly-abnormal oocyte chromosome morphology post irradiation. Chromosome morphology was quantified in the two oocytes (circled) closest to the spermatheca (arrow in right upper panel) at 18h post irradiation. Quantification of these data is included in Table 2 Investigation of germline chromosomal aberrations produced results consistent with this finding as only HDR gene inactivation yielded post-radiation germline chromosomal aberrations ( Fig 3G). Diakinesis oocytes in control worms usually display the normal number of six bivalents (visualized by DAPI, corresponding to six paired homologs attached by chiasmata) at 18h after 120Gy. While neither cku-80 nor lig-4 RNAi impacted this post-radiation pattern ( Table 2), rad-51 RNAi yielded high frequency of clustered chromosomes. Loss-of-function mre-11 displayed, in addition to clumping, twelve univalents within oocytes [49] (Fig 3G and  Table 2). Altogether these studies suggest an exclusive role for HDR in the reparative response of Notch-responsive proliferating germ cells to ionizing radiation. Furthermore, the NHEJ apparatus appears available in the germline but apparently not engaged for DSB repair, suggesting NHEJ is actively suppressed in germ cells, consistent with prior reports [47,50].

Inactivation of HDR radiosensitizes human Notch-driven cancer
Aberrant Notch activation occurs in diverse human cancers, such as in breast cancer and T-ALL [2,5], although the role of Notch in human cancer remains enigmatic and therapeutic gain has not yet been realized by targeting a Notch phenotype [51]. To test whether inhibiting HDR radiosensitizes Notch-driven human malignancy, we employed the T-cell lymphoblastic lymphoma cell line CUTLL-1 [26], which harbors a t(7;9) translocation producing hyperactive NOTCH1, similar to glp-1(ar202). Irradiated CUTLL-1 cells display fewer cells in G1⁄S relative to G2 with G2 phase cells increasing from 9.2% at baseline to >55% at 24h after 4Gy, which persists for 48h (Fig 4A). To silence RAD51, CUTLL-1 cells, infected with human RAD51GIPZ lentiviral shRNA, were puromycin selected, leading to 33% stable RAD51 reduction (S4 Fig). RAD51 shRNA-expressing CUTLL-1 cells displayed significantly-reduced colony formation with D 0 of the radiation dose-response curve shifting from 0.59 to 0.40 (p<0.001), and minimal impact on D q (Fig 4B, left). A similar result was obtained by administering the small molecule MRE11⁄HDR inhibitor Mirin [52]. Irradiated-CUTLL-1 cells, pre-treated for 1h with 50 μM Mirin, a dose that does not affect cell survival (S5 Fig), followed by a 12-day drug-free clonogenic assay, exhibited radiosensitization comparable to genetic RAD51 knockdown (D 0 decreasing from 0.77 to 0.47 with Mirin; Fig 4B right). In contrast, knockdown of the critical NHEJ repair gene XRCC4 was not radiosensitizing (S6 Fig). To test whether targeting HDR would enhance in vivo-radiosensitivity in Notch-driven cancer, RAD51 shRNA-expressing CUTLL-1 cells, grown as chloromas in the flanks of immunodeficient (NOD-SCID) mice, were irradiated at 100-150 mm 3 . Initial studies established the Targeting HDR in Notch-Driven Tumors 50% tumor control dose (TCD 50 ), a standard readout of radiotherapy effectiveness [53], as 13.8Gy for CUTLL-1 tumors (Fig 4C). A 12Gy-dose was selected to evaluate impact of RAD51 inactivation. RAD51-shRNA-expressing CUTLL-1 xenografts responded to 12Gy more robustly than non-silenced control CUTLL-1 tumors (p<0.001), with all RAD51 shRNAexpressing CUTLL-1 tumors showing complete responses by 10 days. Further, over 15 weeks, 83% of RAD51 shRNA-expressing CUTLL-1 tumors achieved autopsy-confirmed cure, while only 33% of CUTLL-1 tumors expressing non-silencing shRNA achieved cure (Fig 4D), equivalent to a 1.5-fold dose-modifying factor for radiosensitization based on HDR inactivation. Tumor radiosensitization is of fundamental importance to radiation oncologic research, although successes have been modest, as tumor-specific DDR phenotypes tractable for Targeting HDR in Notch-Driven Tumors pharmacologic intervention remain poorly defined. Here, we characterize a radiation phenotype in a NOTCH-driven C. elegans stem cell tumor that predicts pharmacologic and genetic outcome of human NOTCH-driven tumor radiosensitization. These studies provide a basis for clinical strategies for improved NOTCH-directed cancer therapy using agents currently under development that target HDR.