ATP synthase F1 subunits recruited to centromeres by CENP-A are required for male meiosis

The histone H3 variant CENP-A epigenetically defines the centromere and is critical for chromosome segregation. Here we report an interaction between CENP-A and subunits of the mitochondrial ATP synthase complex in the germline of male Drosophila. Furthermore, we report that knockdown of CENP-A, as well as subunits ATPsyn-α, -βlike (a testis-specific paralogue of ATPsyn-β) and -γ disrupts sister centromere cohesion in meiotic prophase I. We find that this disruption is likely independent of reduced ATP levels. We identify that ATPsyn-α and -βlike localise to meiotic centromeres and that this localisation is dependent on the presence of CENP-A. We show that ATPsyn-α directly interacts with the N-terminus of CENP-A in vitro and that truncation of its N terminus perturbs sister centromere cohesion in prophase I. We propose that the CENP-A N-terminus recruits ATPsyn-α and -βlike to centromeres to promote sister centromere cohesion in a nuclear function that is independent of oxidative phosphorylation.

M eiosis is the specialised cell division cycle in which one round of DNA replication precedes two rounds of chromosome segregation that generate haploid gametes (eggs and sperm). Defects in meiosis lead to reduced fertility, sterility or aneuploidy in gametes or resulting zygotes 1 . Centromeres, defined epigenetically by incorporation of the histone H3 variant CENP-A 2 play a key role in coordinating meiotic chromosome segregation. Studies in plants suggest that CENP-A adopts meiosis-specific functions via its highly divergent N terminus [3][4][5][6] . To investigate functions of the CENP-A N terminus in meiosis in an animal, we used biochemical and genetic approaches in testis of Drosophila males. We uncover unexpected functional links between CENP-A, mitochondrial ATP synthase F 1 subunits and sister centromere cohesion in meiosis. We propose that the CENP-A N-terminus recruits ATPsyn-α and -βlike, a testis-specific paralogue of -β, to centromeres to promote sister centromere cohesion in a novel nuclear function that is independent of canonical roles in oxidative phosphorylation.

CENP-A functions in meiotic sister centromere cohesion.
To investigate meiosis-specific requirements for CENP-A in Drosophila, we performed testis-specific knockdown of CENP-A using the UAS-GAL4 RNAi system. We expressed GAL4 under the control of the bag of marbles (bam) promoter knocking down CENP-A in the mitotic divisions immediately prior to meiotic prophase I (Fig. 1a). In Drosophila males, meiotic prophase I lacks the conventional features of synapsis between homologues and instead is divided into sub-stages S1-S6 based on nuclear and spindle morphologies 7 . At S5/6, the four Drosophila chromosomes are separated into three territories; the 2nd and 3rd autosomes each form a large territory and the X-Y chromosomes form a third territory with the 4th chromosome (Fig. 1a). We immuno-stained RNAi-depleted S5/6 spermatocytes for the centromere markers CENP-A and CENP-C, confirming an~30% reduction in CENP-A level at centromeres (Supplementary Fig. 1A). At S5/6, an average of 6.5 centromere foci are normally visible; two each per 2nd/3rd autosomal territories, one or two per 4th chromosomal territory and one or two per X-Y territory 8,9 (Fig. 1a). In S6 nuclei depleted for CENP-A an unexpected increase in the number of centromere foci was observed compared to the control (Fig. 1b). Quantitation of centromere foci per nucleus revealed a significant increase (****p < 0.0001) at late prophase I (S5/6, 7.7 compared to 6.5 in the control) and prometaphase I (7.1 compared to 5.7 in the control) (Fig. 1c). As homologues are normally unpaired at this stage (except the 4th chromosome), these results suggest that the maintenance of sister centromere cohesion at late prophase I is defective upon CENP-A reduction. This phenotype was enhanced with increased RNAi efficiency in crosses performed at 29°C ( Supplementary Fig. 1A, 1B, 1C); an almost 50% reduction in CENP-A at centromeres resulted in 9.1 foci per S5/6 nucleus compared to 6.5 in the control (****p < 0.0001). We also assayed the number of centromere foci at early prophase I (S1/2a). At this stage, on average three CENP-A spots are normally visible as homologues are paired, sister chromatid cohesion is intact and centromeres cluster non-specifically akin to chromocenters 8 (Fig. 1a). Quantitation of centromere foci per nucleus revealed a significant increase (****p < 0.0001) at early prophase (S1/2a, 4.3 foci compared to 3.07 in the control), suggesting defects either in sister centromere cohesion or homologue pairing. The number of centromere foci detected per nucleus at interphase did not differ from the control (3.7 compared to 3.6, p = 0.366). It is possible that these interphase cells derived from prophase I cells with normal cohesion, or that these cells arise due to compensation by additional factors that maintain cohesion at this time. Additionally, it is also possible that the CENP-A depletion was less efficient in these cells.
ATP synthase F 1 subunits co-purify with CENP-A. In parallel, we aimed to identify novel regulators of CENP-A function in germ cells using a biochemical approach. Based on findings in plants in which either the deletion of the CENP-A N terminus or its replacement by that of histone H3.3 resulted in sterility 4,5 , we hypothesised that the CENP-A N terminus is functionally important in Drosophila meiosis. To identify proteins interacting with the CENP-A N terminus, we made soluble extracts from wild-type adult fly testes and performed a pull-down using a recombinantly expressed CENP-A N terminus with a GST tag (GST-Nterm-CENP-A) or GST only as bait (Fig. 1d). Surprisingly, subunits of the F 1 portion of the mitochondrial ATP synthase complex V co-purified with GST-Nterm-CENP-A (Supplementary Data 1). ATP synthase normally functions in oxidative phosphorylation catalysing the synthesis of ATP from ADP and inorganic phosphate 10 . Yet, literature searches revealed previous links between the complex and male fertility in Drosophila. First, ATPsyn-α mutants are male sterile 11 . Second, ATPsyn-β expression is normally repressed in the testis and its derepression impairs fertility 12 . Third, we noted a testis-specific paralogue of ATPsyn-β, ATPsyn-βlike, previously identified in male-sterile screens 11,13 . ATPsyn-βlike is 70% homologous to ATPsyn-β, but harbours unique N and C terminal extensions ( Supplementary Fig. 1D) and recent phylogenetic analysis detected ATPsyn-βlike in insect subgroups Diptera and Lepidoptera 14 . We confirmed the testis-specific expression of ATPsynβlike by RT-PCR ( Supplementary Fig. 1E) and western analysis ( Supplementary Fig. 1F). Finally, we noted that ATP synthase F 1 subunits (ATPsyn-α, -β and -γ) are functionally linked to germ line stem cell differentiation in Drosophila females 15,16 . Moreover, this unexpected function was proposed to be independent of canonical functions in oxidative phosphorylation 16 . Based on these findings, we investigated further the link between ATPsyn-α, -β, -βlike and -γ in male fertility, as well as potential links to meiotic centromere function.
To assay whether observed meiotic phenotypes might be due to reduced ATP supply, we quantified the level of ATP in testes depleted for ATPsyn-α, -β, -βlike and -γ. Direct measurement of ATP concentration in adult testes confirmed a reduction on average to 45% in ATPsyn-α RNAi (****p < 0.0001), 55% in ATPsyn-βlike RNAi (****p < 0.0001) and 66% in ATPsyn-γ RNAi (***p < 0.001) (Fig. 2e). In ATPsyn-β RNAi samples a modest reduction in ATP level was measured (*p = 0.044) and the ATP level was not perturbed in CENP-A RNAi samples (p = 0.725). Importantly, we found that the reduction in ATP in RNAi samples did not correlate with phenotype severities (Fig. 2b) or cell cycle delays (Fig. 1e). For example, the ATP level was similarly reduced in testes depleted for ATPsyn-α and ATPsyn-βlike, yet spermatocytes depleted for ATPsyn-βlike display a severe loss in sister centromere cohesion and a cell cycle arrest at prometaphase I, whereas spermatocytes depleted for ATPsyn-α produce mature sperm. To test whether an acute reduction in ATP supply was sufficient to induce excess centromere foci in spermatocytes, we treated wild-type larval testes ex vivo with ATP synthase inhibitors (2,4-Dinitrophenol, oligomycin A) and with an inhibitor of ATP hydrolysis (Sodium Azide (NaN 3 )) 18 and immuno-stained for CENP-A and CENP-C. Quantitation revealed no significant increase in centromere foci at S5/6 after drug treatments, despite an~70% reduction in ATP ( Supplementary Fig. 2C). Finally, knockdown of an additional ATP synthase F 1 subunit, ATPsyn-b, as well as ATP synthase complex I components ND23 (NADH dehydrogenase (ubiquinone) 23 kDa subunit) and ND51 (NADH dehydrogenase (ubiquinone) 51 kDa subunit), did not result in a premature loss-of-sister centromere cohesion at S5/6 despite a comparable reduction in ATP supply ( Supplementary Fig. 2D, 2E). Taken together, these results suggest that ATP synthase components -α, -βlike and -γ might function in sister centromere cohesion through a mechanism distinct from canonical roles in ATP generation. To explain how defects in sister chromatid cohesion might result in a prometaphase I arrest/delay, we stained testes depleted for ATPsyn-α or -βlike with antibodies recognising MEI-S332 (Drosophila Shugoshin), which localises to and functions at centromeres to protect cohesion at this cell cycle time 19 and may require CENP-A for its localisation 20 . MEI-S332 localised to centromeres at prometaphase I as expected in controls 21 (Fig. 2f). Yet, in 100% of ATPsyn-α-depleted prometaphase I spermatocytes with abnormal nuclei, MEI-S332 did not localise to centromeres and was excluded from the nucleus. Strikingly, in ATPsyn-βlike-depleted prometaphase I arrested spermatocytes, in 100% of cells analysed MEI-S332 at centromeres was reduced and it localised unexpectedly to chromosome arms.

CENP-A N terminus promotes sister centromere cohesion.
Finally, to map the interaction site between CENP-A and ATPsyn-α, we immobilised the CENP-A N terminus (residues 1-126) using peptide array-based techniques and probed arrays with recombinant His-ATPsyn-α. Peptide spots corresponding to N terminal conserved sequence blocks B1 and B2 revealed an interaction 23 (Fig. 5a). To test the importance of the CENP-A N terminus in meiosis we generated a fly line expressing a GFPtagged CENP-A transgene lacking amino acids 1-118 (GFP-CENP-A-Δ118) (Fig. 5b), which removes the B1 and B2 domains but leaves the functional B3 domain intact 24 . As a control, we utilised a line in which GFP is inserted at the identical position, but leaves the N terminus intact (GFP-CENP-A) (Fig. 5b), previously shown to complement lethal cenp-a null alleles 25 . Full length GFP-CENP-A expression had a dominant negative effect (*p = 0.0142) on the number of centromere foci at S5/6 (6.684 compared to 6.412 in wild type). Truncated GFP-CENP-A-Δ118 localised to centromeres, but showed a dominant negative effect (****p < 0.0001) on the number of centromeric foci at S5/6 compared to nuclei expressing full length GFP-CENP-A (Fig. 5c). These results suggest that perturbation of the CENP-A N terminus can disrupt sister centromere cohesion in meiotic prophase I.

Discussion
Here, in addition to an expected function in ATP synthesis, we report a function for ATPsyn-α and ATPsyn-βlike in male meiosis and fertility. We show that in testes depleted for ATPsyn-α orβlike prophase I cells accumulate prior to meiosis I, providing a possible explanation for observed sterility in previously isolated mutants 11,13 . Given that canonical ATPsyn-β expression in testis is reduced compared to whole adults 12 , ATPsyn-βlike might normally compensate for ATPsyn-β function. Moreover, although the expression pattern of ATPsyn-βlike is entirely consistent with a testis-specific function, we note ATPsyn-βlike expression at larval and pupal stages (modENCODE RNA-Seq). This raises the possibility that ATPsyn-βlike adopts additional functions in development, which we have not addressed in this study. In addition to its canonical role, we report a nuclear function for ATPsyn-α and -βlike, in particular at centromeres. We find that ATPsyn-α and -βlike localise to centromeres at meiotic prophase I and that this localisation requires CENP-A. Moreover, CENP-A, ATPsyn-α and ATPsyn-βlike are each required to maintain sister centromere cohesion at this stage. Remarkably, ATPsyn-βlike specifies the enrichment of the cohesion protector MEI-S322 to centromeres at prometaphase I, perhaps comparable to the Chromosome Passenger Complex subunit INCENP 26 . In contrast, ATPsyn-α appears to have a distinct function in the nuclear and centromeric localisation of MEI-S332. MEI-S332 mislocalisation to global chromatin in ATPsyn-βlike-depleted nuclei is particularly striking and might be a consequence of a sustained prometaphase I arrest or indicates a more general function of Fig. 2 Centromere defects upon ATP synthase-α/-β/-βlike/-γ RNAi. a Immuno-fluorescent micrograph of control S5/6 nuclei or nuclei RNAi-depleted of ATPsyn-α, -β, -βlike and -γ (at 25°C) stained with antibodies against CENP-A (red) and CENP-C (green) (n = 3). DNA is stained with DAPI (blue). Numbers indicate centromere foci per nucleus. Scale bar = 10 μm. b Quantitation of centromere foci per control S5/6 nucleus or nucleus RNAi-depleted of ATPsyn-α, -β, -βlike and -γ (at 25°C). For each RNAi sample, p-values were calculated compared to respective TRiP or VDRC isogenic controls using an unpaired Student's t-test. The data (n = 100 nuclei) are pooled from three individual experiments. Error bars = SEM. ****p < 0.0001, NS = not significant, p > 0.05. c Immuno-fluorescent micrograph of control S1/2a nuclei or nuclei RNAi-depleted of ATPsyn-α, -β, -βlike and -γ (at 25°C) stained with antibodies against CENP-A (red) and CENP-C (green) (n = 3). DNA is stained with DAPI (blue). Scale bar = 5 μm. d Line graph showing quantitation of the number of centromere foci per control nucleus or nucleus RNAi-depleted of ATPsyn-α, -β, -βlike and -γ at S1/2a, S5/6, prometaphase (PMI) or interphase stages of meiosis I. The data (n = 100) are pooled from three individual experiments and was analysed using an unpaired Student's t-test, ****p < 0.0001, ***p < 0.001, **p < 0.01 and *p < 0.05. Error bars = SEM. e Relative ATP concentration in control adult testes (isogenic, bam-Gal4) or testes RNAi-depleted (at 25°C) for CENP-A, ATPsyn-α, -β, -βlike and -γ. T-test compares RNAi knockdowns to isogenic control. Experiments were carried out in triplicate and data are pooled from three independent RNAi experiments. Significance was analysed using an unpaired Student's t-test, ****p < 0.0001, *p < 0.05, NS = not significant p > 0.05. Error bars = SEM. f Immuno-fluorescent micrograph of control nuclei at prometaphase I or perturbed prometaphase I nuclei RNAi-depleted for ATPsyn-α or -βlike stained for MEI-S332 (green), CENP-A (red) and tubulin (grey) (n = 2). DNA is stained with DAPI (blue). Scale bar = 10 μm ATPsyn-βlike on chromatin. In Drosophila males, an alternative cohesin complex made up of ORD, SOLO and SUNN maintains meiotic sister centromere cohesion at late prophase I S6 9,17,27 . We find that CENP-A is required for centromere cohesion early in prophase I at S1/2a, prior to ORD, SOLO and SUNN. We suggest that observed defects in cohesion lead to failed progression through meiosis I and ultimately reduced fertility or sterility. Intriguingly, depletion of ATP synthase F 1 subunits also disrupts sister chromatid arm cohesion and 4th homologue pairing/ cohesion, suggesting additional global functions outside of the *** *** *** *** centromere. The ATPsyn-α subunit directly interacts with the CENP-A N terminus, providing a first function for conserved B1 and B2 domains 23 . We propose that ATPsyn-α recruits ATPsynβlike to centromeres. Our functional analyses of flies expressing a GFP-tagged CENP-A lacking amino acids 1-118 show the CENP-A N terminus is not required for meiotic centromere localisation, different from plants [3][4][5] . Instead, the fly CENP-A N terminus appears to be important for meiotic sister centromere cohesion, possibly via the recruitment of ATPsyn-α and ATPsyn-βlike.
Our data support a model in which mitochondrial ATP synthase F 1 subunits adopt nuclear functions that appear to be independent of ATP production. First, ATPsyn-α and ATPsynβlike interact with CENP-A/centromeres. Second, the severity of observed meiotic phenotypes does not correlate with ATP supply. Third, ATP depletion was not sufficient to induce a loss of cohesion. Finally, our findings are in line with an ATP independent requirement for ATPsyn-α, -β and -γ in germ line stem cell differentiation in Drosophila females 16 . In conclusion, we propose that the CENP-A N-terminus recruits ATPsyn-α and ATPsyn-βlike to centromeres to promote sister centromere cohesion in a novel nuclear function that is independent of canonical roles in oxidative phosphorylation.

Methods
Fly stocks and husbandry. Stocks were cultured on standard cornmeal medium (NUTRI-fly) preserved with 0.5% propionic acid and 0.1% Tegosept at 20°C under a 12 h light dark cycle. UAS-RNAi lines were obtained from the Bloomington Stock Centre, the Transgenic RNAi Project (TRiP) or Vienna Drosophila RNAi Centre (VDRC) (Supplementary Table 1). Appropriate isogenic RNAi lines for TRIP or VDRC were used as controls. The testes-specific promoter bam was used to drive GAL4 expression (w;; bam-Gal4-VP16, UAS-dcr2; provided by M.Fuller) and crosses were performed at 25°C or 29°C. Transgenic lines expressing N terminal tagged mCherry-ATPsyn-β, eGFP-ATPsyn-βlike or eGFP-Δ118-CENP-A under respective endogenous promoters were generated by transposable (P) element transformation of pCaSpeR5 vector in w 1118 embryos (injection, selection and balancing by BestGene Inc). ATPsyn-βlike cDNA was amplified from wild type with 900 bp upstream and 600 bp downstream. Δ118 cid (bp 354-678) cDNA was amplified from wild type with 413 bp upstream and 417 bp downstream; a 3× glycine linker was placed between the GFP tag and the cid start codon. Transgenic flies expressing GFP-CENP-A and YFP-CENP-C were gifts from C. Lehner and S. Heidmann 25 . Flies harbouring an insertion in ATPsyn-βlike (w 1118 ; PBac{w[ + mC] = RB}ATPsynbetaLe01800) were obtained from the Bloomington Stock Centre (17989). For fertility tests, two virgin age-matched males/females were crossed, allowed to lay eggs for 2 days and the number of adult progeny was scored after 20 days. For cell cycle analysis, the number of cysts in meiosis I, II or containing spermatids in at least 12 testes pooled from two individual RNAi experiments from adults <5 h old were scored.
Recombinant protein production and in vitro binding assays. GST, GST-Nterm-CENP-A, GST-FL-CENP-A (full-length), His-ATPsyn-α, His-ATPsyn-β and His-ATPsyn-βlike were expressed in BL21 Star TM Codon-Plus-RIL E. coli. Histagged ATPsyn-α, -β, and -βlike were solubilised from inclusion bodies in 5 M urea, purified under denaturing conditions using Ni-NTA HisPur agarose beads and re-natured by stepwise dialysis into 50 mM Tris-HCL, pH 8.0. For tissue protein extracts, wild-type adult testes were digested in 1X PBS containing 1 mg/ml collagenase, 100 μg/ml DNase and 1.5 mM CaCl 2 , passed through a 40 μm sieve and treated with a hypotonic buffer (10 mM HEPES, 1.5 mM NaCl, 1.5 mM MgCl 2 , 0.1 mM EGTA, 1 mM DTT, 0.1% Triton X-100, 1% protease inhibitor cocktail) before lysis in 300 mM NaCl. For GST pull-down, GST or GST-Nterm-CENP-A (amino acids 1-126) was incubated with testes extracts for 3 h at 4°C, followed by the addition of glutathione agarose beads for 1 h. Precipitated proteins were eluted and analysed by silver staining (SilverQuest, Invitrogen). For mass spectrometry (MS), gel lanes were excised and trypsin-digested for analysis by Nano LC-MS/MS (Proteomics Facility, University of Bristol). The CENP-A peptide array of 18-mer overlapping peptides was generated by automatic SPOT synthesis 28 on Whatman 50 cellulose membrane supports using Fmoc (9-fluorenylmethyloxycarbonyl) chemistry with the MultiPep RSi (Intavis Bioanalytical Instruments). Specifically, a library of overlapping peptides 18 amino acids in length, each shifted by four amino acids, and encompassing the sequence of the CENP-A N terminus was SPOT synthesized on nitrocellulose membranes to generate CENP-A arrays 28 . Peptide arrays were challenged with His-ATPsyn-α (5 μg/ml) and binding patterns were revealed by anti-ATPsyn-α western blot.
IF, FISH and Microscopy. For Immunofluorescence (IF), testes from young adult males (<1 days old) or 3rd instar larvae were dissected in 1X PBS, gently squashed onto poly-L-lysine coated slides, snap frozen in liquid nitrogen and fixed in 4% paraformaldehyde for 10 min or in cold methanol for 5 min, followed by cold acetone for 2 min (for anti-tubulin staining). For cytosol extraction, samples were immediately washed in 1X PBS-0.1% Triton X-100 (0.1 PBT). For cytosol preservation and for fluorescence in situ hybridisation (FISH), fixed samples were passed through an ethanol series (75-85-95%) at −20°C and dried prior to permeabilisation in 1X PBS-0.4% Triton X-100 (0.4 PBT) with 0.3% sodium deoxycholate. For IF, samples were blocked in 0.1PBT with 1% BSA for 1 h at room temperature, incubated with primary antibodies overnight at 4°C and with secondary antibodies for 1 h at room temperature. For FISH, prehybridisation was carried out in 2X Saline Sodium Citrate (SSC)-0.1% Tween-20 with 50% formamide for 2 h at 37°C. DNA probes for the 2nd/3rd (AATAACATAG) 3 and 4th (AATAT) 6 chromosomes were directly labelled with Alexa Fluor conjugates (Eurofins). Hybridisation of DNA probes (20 ng) was carried out overnight at 20°C. Imaging was carried out using a DeltaVision Elite wide-field microscope system (Applied Precision). Images were acquired as z-stacks with a step size of 0.2 μm; raw data files were deconvolved using a maximum intensity algorithm. 3D z-stack images were represented in 2D by projection using SoftWorx (Applied Precision). Focal fluorescent intensities were measured as corrected total cellular fluorescence (CTCF) using Image J software (NIH). Pearson Coefficient of Co-localisation was calculated in SoftWorx.
ATP assay. ATP was extracted from 20 adult testes from flies aged 3 to 24 h by homogenisation in a chaotropic buffer and ATP levels were quantified using a luciferase based ATP assay (Molecular Probes) as described 29 . For ATP depletion, testes were treated ex vivo with oligomycin A (50 μg/ml), 2,4-Dinitrophenol (1 mM) or NaN 3 (5 mM) for 1 h.