A PHF8 Homolog in C. elegans Promotes DNA Repair via Homologous Recombination

PHF8 is a JmjC domain-containing histone demethylase, defects in which are associated with X-linked mental retardation. In this study, we examined the roles of two PHF8 homologs, JMJD-1.1 and JMJD-1.2, in the model organism C. elegans in response to DNA damage. A deletion mutation in either of the genes led to hypersensitivity to interstrand DNA crosslinks (ICLs), while only mutation of jmjd-1.1 resulted in hypersensitivity to double-strand DNA breaks (DSBs). In response to ICLs, JMJD-1.1 did not affect the focus formation of FCD-2, a homolog of FANCD2, a key protein in the Fanconi anemia pathway. However, the dynamic behavior of RPA-1 and RAD-51 was affected by the mutation: the accumulations of both proteins at ICLs appeared normal, but their subsequent disappearance was retarded, suggesting that later steps of homologous recombination were defective. Similar changes in the dynamic behavior of RPA-1 and RAD-51 were seen in response to DSBs, supporting a role of JMJD-1.1 in homologous recombination. Such a role was also supported by our finding that the hypersensitivity of jmjd-1.1 worms to ICLs was rescued by knockdown of lig-4, a homolog of Ligase 4 active in nonhomologous end-joining. The hypersensitivity of jmjd-1.1 worms to ICLs was increased by rad-54 knockdown, suggesting that JMJD-1.1 acts in parallel with RAD-54 in modulating chromatin structure. Indeed, the level of histone H3 Lys9 tri-methylation, a marker of heterochromatin, was higher in jmjd-1.1 cells than in wild-type cells. We conclude that the histone demethylase JMJD-1.1 influences homologous recombination either by relaxing heterochromatin structure or by indirectly regulating the expression of multiple genes affecting DNA repair.


Introduction
Histones H3 and H4 are methylated on several amino acid residues, and this methylation can proceed up to the triple level at individual amino acids, usually lysines or less frequently arginines. Histone methylation patterns differ greatly between active and inactive genes, suggesting that histone methylation has a great impact on chromatin structure [1]. A well-known example is the histone H3 Lys9 tri-methylation and Lys4 hypomethylation in heterochromatin [2].
DNA damage signaling as well as gene expression, is influenced by histone methylation, as in the case of 53BP1 binding to chromatin containing methylated H3K79 or H4K20 [3,4]. The DNA damage checkpoint protein MDC1 is demethylated in response to double-strand DNA breaks, and this allows its ubiquitination [5]. The methylation of the tumor suppressor p53 promotes its association with 53BP1 leading to transcriptional activation and apoptosis [6].
Histones methylated on lysines are demethylated by two classes of enzymes, LSD and JmjC demethylases, with very different reaction mechanisms [1,7,8]. The LSD (lysine-specific demethylase) family has only two members in mammals and removes methyl groups by oxidizing amines using FAD and oxygen. The other family, JmjC (with Jumonji C domains), has more than twenty members, and requires Fe 2+ and α-ketoglutarate for catalysis. A JmjC demethylase, PHF8, targets histone H3 mono-and di-methyl Lys9 (H3K9me1 and H3K9me2), and its mutation is associated with X-linked mental retardation (XLMR) [9]. In the model organism C. elegans, there are two close homologs of PHF8, JMJD-1.1 and JMJD-1.2, encoded by F43G6.6 and F29B9.2 open reading frames, respectively. Interestingly, JMJD-1.2 is expressed in neurons and is involved in the movement of worms, which agrees with the fact that its human homolog PHF8 is associated with XLMR. JMJD-1.2 targets H3K9me2 like PHF8, as well as H3 di-methyl Lys27 (H3K27me2) [9], and induces the expression of target genes by binding to H3K4me3 via a PHD domain [10].
In this work, we investigated whether and how the two PHF8 homologs in C. elegans, JMJD-1.1 and JMJD-1.2, affect worm survival in response to DNA damage. JMJD-1.2 demethylates H3K9me2 and H3K27me2, and its closest homolog JMJD-1.1 is likely to have a similar specificity for particular histone residues. It is known that the di-and tri-methylation of H3K9 and H3K27 have been associated with inactive genes and heterochromatin [1]. Therefore, it seemed very likely that the two PHF8 homologs influence responses to DNA damage by modulating chromatin structure. Indeed, we show here that the two proteins, especially JMJD-1.1, influence DNA repair in response to interstrand DNA crosslinks and double-strand DNA breaks.
Results JMJD-1.1 is required for resistance to interstrand DNA crosslinks and double-strand DNA breaks in C. elegans JMJD-1.1 and JMJD-1.2 proteins are very similar in amino acid sequences to human PHF8, which is a histone demethylase associated with X-linked mental retardation. There is more than 35% identity between PHF8 and C. elegans homologs over a 380 amino acid stretch and 55% in the JmjC domain (Fig 1). JMJD-1.1, in particular, has 60% similarity to JMJD-1.2 over the entire length of the polypeptide, suggesting that the two C. elegans proteins are probably paralogs. We obtained jmjd-1.1(tm3980) and jmjd-1.2(tm3713) deletion mutant worms generated by the National Bioresource Project (Japan) and outcrossed them several times with wildtype N2 strain to remove any background mutations. The two mutants have deletions of more than 500 nucleotides containing exons, and are very likely to be null mutants. In the case of jmjd-1.1(tm3980), two successive exons are deleted, and the most likely transcript is the second exon joined to the fifth exon. Actually, the transcript was found to encode a prematurely terminated polypeptide of 135 amino acids due to a frameshift, as verified by reverse transcription followed by a gene-specific polymerase chain reaction (S1 Fig). In addition, the level of the transcript in mutant worms amounted to only 20% of that in wild-type N2, as measured by quantitative polymerase chain reaction after reverse transcription (S1 Fig). Therefore, the greatly reduced mRNA level and the premature termination eliminating the JmjC domain suggest that the jmjd-1.1(tm3980) allele is completely null.
In order to probe the roles of the histone demethylases in DNA damage responses, the jmjd-1.1(tm3980) and jmjd-1.2(tm3713) mutants were subjected to various types of DNA damage at the L4 larval stage (Fig 2). In C. elegans hermaphrodite gonads, proliferating germ cells undergo promeiotic development to produce oocytes and sperms, and after self-fertilization, early-embryos are laid by the worms. Embryos that developed from treated germ cells in the proliferating region of the gonad were scored for hatching to L1 larvae. Besides the two jmjd-1 mutants, two other mutants were also used as positive controls for different types of DNA damage, along with wild-type N2 as a negative control. FCD-2 (a C. elegans FANCD2 homolog) is a key component of the Fanconi anemia (FA) pathway, which participates in the repair of interstrand crosslinks (ICLs) [11]. Therefore, fcd-2(tm1298) mutant worms were used as a positive control for hypersensitivity to ICLs. BRC-1 (a C. elegans BRCA1 homolog) has an indispensable role in promoting end-resection in the initial stage of homologous recombination (HR) [12], and the mutant brc-1(tm1145) was used as a positive control for hypersensitivity to DSBs.
When jmjd-1.1(tm3980) or jmjd-1.2(tm3713) worms were exposed to ICLs induced by photo-activated psoralen, both mutants were hypersensitive compared to the wild-type (Fig  2A). The jmjd-1.1(tm3980) mutant showed a higher sensitivity to ICLs than jmjd-1.2(tm3713) (p = 0.0268, two-way ANOVA test), implying a more important role of JMJD-1.1 in repair of ICLs. Since the two mutants were both hypersensitive to ICLs, we tested whether the two corresponding proteins had overlapping functions. For this purpose, we reduced jmjd-1.1 expression by RNA interference (RNAi) in wild-type, jmjd-1.1, and jmjd-1.2 strains. Knockdown of   for 40 min and exposed to UVA light (150 J/m 2 ). (B) L4 stage worms were exposed to γ-rays at 60 Gy. In every experiment, eggs were collected between 24 and 40 h post treatment, and their survival was scored 24 h later. Mutations for fcd-2 and brc-1 (C. elegans FANCD2 and BRCA1 homologs, respectively) were used as positive controls for the corresponding type of DNA damage. Error bars indicate SEM. p values were obtained by calculating the difference (Δ) in embryonic survival between 0 and 100 TMP for each strain, and comparing the Δ values in the strains by two-way ANOVA.

JMJD-1.1 is a downstream effector of the Fanconi anemia pathway during ICL repair
In mammalian cells, the FA pathway participates in the initial stage of ICL repair, which involves the recognition of ICLs and their conversion to DSBs. FANCD2 is a key player in the FA pathway, shunting the DSB intermediate to the HR pathway [13][14][15]. The accumulation of nuclear foci of FCD-2, a FANCD2 homolog in C. elegans, at ICL sites is a good indicator of successful activation of the FA pathway [16]. We found that the extent of FCD-2 focus formation after ICL treatment in the proliferating mitotic germ cells of jmjd-1.1(tm3980) and wild-type worms was not significantly different (Fig 3, p>0.09 at all the time points, Student's t-test). In addition, the immuno-signal declined gradually to similar extents in the two strains until 24 h. These results indicate that JMJD-1.1 does not affect FCD-2 activation and has a role downstream of the FA pathway in dealing with ICLs.

Prominent role of JMJD-1.1 in homologous recombination during ICL and DSB repairs
ICLs are converted to DSBs in the FA pathway as a result of DNA cleavages by endonucleases such as XPF, MUS81, SLX1, and FAN1 [17,18]. In order for DSBs to be repaired by HR, extensive end resection of DSBs needs to take place; then RPA binds to the exposed single-strand DNA, and RAD51 replaces RPA prior to DNA strand invasion [14]. Mammalian cells defective in HR exhibit prolonged accumulation (or slow dissociation) of RPA and RAD51 proteins at DNA lesions [19,20]. Therefore, we investigated the accumulation and dissociation kinetics of RPA-1 and RAD-51 (C. elegans RPA70 and RAD51 homologs, respectively) in mitotic germ cells of wild-type and jmjd-1.1(tm3980) worms. Although these proteins accumulated normally at 6 h after ICL treatment (also at 9 h for RAD-51), striking defects in HR were revealed by the subsequent sluggish dissociation of these proteins in the mutant worms (Fig 4): accumulation of RPA-1 and RAD-51 were still observed in the mutant worms at 18 h post treatment, but they disappeared almost completely in wild-type worms. Similar results were observed for the repair of DSBs induced by γ-rays (S3 Fig). The formation of RPA-1 and RAD-51 foci in mitotic germ cells was normal in the mutant, but their dissociation was slowed down (S3 Fig). In summary, JMJD-1.1 is needed for repair of both ICLs and DSBs via HR, specifically in the step following DSB end resection and subsequent RPA-1/RAD-51 loading.
The putative histone demethylase JMJD-1.1 modulates heterochromatin structure Spatiotemporal regulation of chromatin structure is critical for the DNA repair machinery to perform its work efficiently [21,22]. We therefore examined the level of histone H3 tri-methyl Lys9 (H3K9me3), a well-established heterochromatin marker, in mitotic germ cells after ICL or DSB formation. The immuno-signal for H3K9me3 became weaker after DNA damage induction, but it recovered gradually with the progression of DNA repair in both wild-type and mutant cells (Fig 5A and S4 Fig). However, the amount of heterochromatin in the mitotic germ cells of mutant worms was higher than in wild-type worms both before and after the induction of ICLs and DSBs. The differences in H3K9me3 level were also evident in a western analysis of worm extracts (Fig 5B), where the signal for H3K9me3 in the jmjd-1.1 mutant was about 1.5 fold that in the wild type both before and after ICL treatment. The higher level of heterochromatin in the mutant is consistent with the kinetics of RPA-1 and RAD-51 dissociation from DNA lesions in Fig 4 and S3 Fig. Wild-type worms nearly completed repair of DNA damage within 12 h, as can be seen from the fact that most of RPA-1/RAD-51 had dissociated from the DNA lesions by 12 h post treatment. In contrast, DNA damage in the mutant worms was still under repair for up to 18 h post treatment, as shown by the presence of by non-detached RPA-1/RAD-51. The higher level of heterochromatin, both before and after DNA damage, is likely to be one of the main factors inhibiting DNA repair in the mutant worms. Based on a previous study of the dual demethylase activity of JMJD-1.2 on H3K9me2 and H3K27me2 [9], we carried out immuno-staining to assess any changes in the levels of these two histone signatures in mitotic germ cells. Contrary to our expectation, neither differed between wild-type and jmjd-1.1 worms (S5 Fig; data not shown for H3K27me2), suggesting that JMJD-1.1 targets other signatures such as H3K9me3.
JMJD-1.1 affects the same step in the HR pathway as RAD-54 We further addressed the epistatic relationship between JMJD-1.1 and genes of the two main DSB repair pathways, HR and NHEJ (nonhomologous end-joining). For this analysis, we depleted C. elegans rad-54 or lig-4 (C. elegans RAD54 or LIG4 homologs, respectively). RAD54 and LIG4 participate in HR and NHEJ, respectively, in mammalian cells, and their conservation in C. elegans provides excellent ways of turning on and off each of these DSB repair pathways. In several reports, hypersensitivities to DNA damage caused by the absence of key HR factors were reversed by depletion or mutation of crucial components of NHEJ [13,23]. Thus in mammalian systems, phenotypes due to BRCA1 mutation are rescued by mutation of 53BP1 which inhibits the end-resection of HR and promotes NHEJ [24,25]. Likewise in C. elegans, the hypersensitivity of BRCA1 worms to DSBs was rescued by a LIG4 mutation, and that of RAD54 knockdown worms was reversed by 53BP1 or LIG4 mutations [26]. Similarly, the hypersensitivity of FANCD2 worms and Fanconi anemia patient cells to ICLs was rescued by inhibiting NHEJ [13]. Therefore, we examined the effect of depleting either lig-4 or rad-54 in jmjd-1.1 mutants. Since jmjd-1.1(tm3980) worms are impaired in HR, we expected that the embryonic  lethality caused by ICLs might be reversed by depleting lig-4. Moreover, if the main role of JMJD-1.1 were in HR, rad-54 depletion in the mutant background might not have any additional effects on overall embryonic viability. As expected, embryonic lethality due to the jmjd-1.1 mutation was fully rescued by lig-4 knockdown, supporting the view that JMJD-1.1 is involved in HR during ICL repair (Fig 6A, p = 0.0038, jmjd-1.1(tm3980);lig-4(RNAi) vs. jmjd-1.1 (tm3980)). However, rad-54 depletion in the jmjd-1.1 mutant background led to massive lethality: death induced by ICLs was 27±2% in the wild type, 56±5% in jmjd-1.1(tm3980), 38±4% in rad-54(RNAi), and 76±4% in jmjd-1.1(tm3980);rad-54(RNAi). This suggested that JMJD-1.1 plays a role in parallel to RAD-54 in HR (Fig 6B, p = 0.0177, jmjd-1.1(tm3980);rad-54(RNAi) vs. jmjd-1.1(tm3980); p = 0.0019, jmjd-1.1(tm3980);rad-54(RNAi) vs. rad-54(RNAi)). One possible explanation may be that JMJD-1.1 and RAD-54 act at the same step of HR, but have different effects on chromatin structure. In agreement with this view, we have found that knockdown of rad-54 or use of a hypomorphic mutation of rad-54 slows down the disappearance of RAD-51 foci after γ-irradiation (data not shown), as does the jmjd-1.1 mutation.

Discussion
A recent report describing the similarities between human histone demethylase PHF8 and its C. elegans homolog JMJD-1.2, and the physiological importance of these enzymes led us to examine the roles of the JmjC domain-containing histone demethylases in response to DNA damage [9]. C. elegans has two homologs of human PHF8, the JmjC domain-containing proteins JMJD-1.1 and JMJD-1.2. PHF8 and the two C. elegans homologs share two conserved domains: the PHD-finger recognizes and selects target substrate specific for histone demethylase, while JmjC is the catalytic domain (Fig 1) [9]. In this study, we have linked the putative histone demethylase JMJD-1.1 to DNA damage responses, and this in turn led to understanding its activity in terms of chromatin structure. jmjd-1.1(tm3980) worms exhibit hypersensitivity to ICLs and DSBs (Fig 2), but not to pyrimidine dimers induced by UVC radiation (data not shown). The exceptionally severe effect on embryonic survival after ICL treatment demonstrated that JMJD-1.1 is dedicated more to ICL repair than to DSB repair. However, interestingly, jmjd-1.2(tm3713) worms showed a milder degree of hypersensitivity to ICLs than jmjd-1.1 (tm3980) worms and no hypersensitivity to DSBs (Fig 2).
The jmjd-1.1 mutation did not affect FCD-2 (FANCD2 homolog) focus formation in response to ICLs, indicating that the initial FA pathway converting ICLs to DSBs was not affected by the mutation. Therefore, the HR pathway acting subsequently to FA pathway was suspected to be defective in the mutant. Indeed, the disappearance of RAD-51 foci not their accumulation was retarded (Fig 4B). The slow disappearance of RAD-51 foci was also observed in jmjd-1.1 after γ-ray treatment, reflecting defects in HR progression (S3 Fig). In addition to RAD-51 foci, RPA-1 foci, which usually form at earlier times than those of RAD-51, also disappeared slowly in the mutant. Thus, JMJD-1.1 appears to function in HR during ICL and DSB repair, at a step after RAD-51 loading onto single-strand DNA. In addition to JMJD-1.1 and JMJD-1.2, a few C. elegans histone demethylases have been reported to be important for genome stability previously. A H3K4 demethylase SPR-5 is required for normal meiotic homologous recombination and for resistance to double-strand DNA breaks [27,28]. JMJD-2 demethylates H3K9 and H3K36, and its depletion leads to increased apoptosis and RAD-51 focus formation in pachytene-stage germ cells [29].
Immunostaining of C. elegans germ cells against the heterochromatin signature, histone H3K9me3, strongly supported the notion that the difference in the dissociation kinetics of RPA-1 and RAD-51 between wild-type and jmjd-1.1 worms derived from different chromatin relaxation levels (Fig 5): the immuno-signal was slightly but significantly more intense in jmjd- Nevertheless, accompanying changes in the transcriptome, possibly involving transcripts of DNA repair proteins, may have also contributed to the ICL-hypersensitivity of the jmjd-1.1 mutant.
Because of the retarded relaxation of heterochromatin structure in jmjd-1.1 cells, our next question was which histone signature is modulated by JMJD-1.1. Since JMJD-1.2 has enzymatic activity on histone H3K9me2 and H3K27me2, we thought that JMJD-1.1 might also target these substrates. However, we were not able to distinguish the levels of these two signatures between wild-type and jmjd-1.1 worms after ICL formation (S5 Fig, data not shown for H3K27me2). Interestingly, a homolog of PHF8 and JMJD-1, EPE1, is an antisilencing factor in S. pombe, as it regulates the extent of heterochromatin domains [30]. The protein is recruited to heterochromatin protein 1 (HP1) bound to methylated histone H3K9 and selectively enriched at heterochromatin boundaries [31].
In an epistasis test, the sensitivity of jmjd-1.1 worms to ICLs reverted to the level of wildtype worms after lig-4 knockdown. This is very similar to the observation that phenotypes of a mutant of fcd-2, which promotes HR on DNA intermediates derived from ICLs, was rescued by a lig-4 mutation [13]. This similarity between jmjd-1.1 and fcd-2 in the genetic interaction with lig-4, agrees well with our argument that JMJD-1.1 functions in HR. This argument was supported by our results showing the effect of JMJD-1.1 on the disappearance of RAD-51 foci after ICL and DSB formations. In contrast to our expectation based on the role of JMJD-1.1 in HR, rad-54 RNAi synergized with the absence of JMJD-1.1 after ICL treatment. This revealed that JMJD-1.1 in its effect on HR is not fully epistatic to RAD-54. One possible explanation could be that both of JMJD-1.1 and RAD-54 regulates HR by chromatin structure modulation. RAD54 is a SNF2/SWI2 family protein with ATPase activity and is thought to be involved in RAD51 loading and its subsequent dissociation, as well as in branch migration in yeasts and mammals [32]. The RAD54 homolog in C. elegans affects RAD-51 dissociation (data not shown) as does JMJD-1.1, suggesting that both proteins probably regulate the same step of HR but via different mechanisms. However, another possibility is that JMJD-1.1 also influences a minor DSB repair pathway such as single-strand annealing (SSA) and alternative end-joining (Alt-EJ) as well as HR, thereby showing synergism with RAD-54.
Our research has generated new insights into the role of JMJD-1.1, a putative histone demethylase, in terms of DNA damage responses. Its novelty resides in the fact that C. elegans JMJD-1.1 and JMJD-1.2 are very similar in terms of amino acid sequence, but have both distinct and overlapping roles in DNA damage responses. Another interesting finding is that JMJD-1.1 affects homologous recombination in response to ICLs and DSBs, probably by modulating chromatin structure in parallel with RAD-54. It will be important to determine which histone proteins JMJD-1.1 acts on to relax chromatin structure and also which repair gene transcripts, if any, it regulates.

Embryonic survival assay
To induce interstrand DNA crosslinks, L4 stage worms were incubated with 100 μM TMP (4,5,8-trimethylpsoralen, Sigma-Aldrich) for 40 min in a dark chamber. The worms were transferred onto NGM plates and exposed to 150 J/m 2 UVA (UVL-28 EL, UVP). For doublestrand break formation, L4 stage worms were irradiated with 60 Gy of γ-rays from a 137 Cs source (IBL 437C, CIS Biointernational). Eggs were collected between 24 and 40 h after induction of DNA damage, and their viability into larvae was scored after 24 h.

Immunostaining
Worms were placed on glass slides covered with 1x PTw, and their gonads were extracted with an ethanol-cleansed blade. They were fixed in 3% paraformaldehyde (0.1 M K 2 HPO 4 , pH 7.2) for 15 min followed by addition of 1 ml 100% methanol (-20°C). They were then washed five times with 1 ml 1x PTw, and left in 200 μl of blocking solution (goat serum mixed with an equal volume of 1x PTw) for 1 h at room temperature. After blocking, the gonads were incubated with the indicated primary antibody at 4°C overnight. They were then washed five times with 1x PTw and incubated with FITC-conjugated goat secondary antibody (Roche, 1:1000 dilution in blocking solution) matching the host (mouse, rat, or rabbit) of the primary antibody at room temperature for 1 h. After three washes with 1x PTw, the gonads were stained with 1 μg/mL DAPI for 15 min followed by 2 washes with 1x PTw. Lastly, anti-fade solution was added and the gonads were observed under a fluorescence microscope (DMR HC, Leica).

Western analysis
L4 stage worms were treated with 200 μM photoactivated TMP for 40 min followed by 150 J/ m 2 UVA. The worms were grown further for 6 h and boiled in reducing SDS sample buffer. Proteins were separated by 12% SDS-PAGE and transferred onto a nitrocellulose membrane.
Supporting Information S1 Fig. A shorter mRNA is made from the jmjd-1.1 deletion allele, and at a greatly reduced level. Total RNA was isolated from mixed-stage worms of wild type and jmjd-1.1(tm3980), and reverse-transcribed. To determine the splicing pattern of jmjd-1.1 gene in the mutant, primers in the first and fifth exons were used in the amplification of a cDNA fragment, which were 5'-CCGGACTTGAGGAATACGAGTACT and 5'-GCCTCCAAAAGCGAGAGAGTC, respectively. The cDNA sequence showed that splicing occurs between the second and the fifth exons in the mutant (Fig 1). Relative amount of the short jmjd-1.1 transcript with respect to the normal transcript in the wild type was measured by real-time PCR using primers in the fifth exon. The primer pairs were 5 0 -TCAGGCCCACGAGACGTC and 5 0 -GCCTCCAAAAGCGAGAGAGTC in the 5th exon. (PDF) The gonads were immuno-stained with antibody against histone H3K9me2 at 3, 6, 9, 12 and 24 h post treatment. Scale bar, 10 μm. (B) Worm extracts were prepared 6 h after ICL formation and separated on a 12% SDSpolyacrylamide gel. After transfer to a nitrocellulose membrane, proteins were probed for histone H3K9me2 and α-tubulin. Band intensities were measured and plotted in the bar graph.
Each bar represents an average of three independent experiments. p values were obtained by Student's t-test. (PDF)