Rec8 Cohesin-mediated Axis-loop chromatin architecture is required for meiotic recombination

Abstract During meiotic prophase, cohesin-dependent axial structures are formed in the synaptonemal complex (SC). However, the functional correlation between these structures and cohesion remains elusive. Here, we examined the formation of cohesin-dependent axial structures in the fission yeast Schizosaccharomyces pombe. This organism forms atypical SCs composed of linear elements (LinEs) resembling the lateral elements of SC but lacking the transverse filaments. Hi-C analysis using a highly synchronous population of meiotic S. pombe cells revealed that the axis-loop chromatin structure formed in meiotic prophase was dependent on the Rec8 cohesin complex. In contrast, the Rec8-mediated formation of the axis-loop structure occurred in cells lacking components of LinEs. To dissect the functions of Rec8, we identified a rec8-F204S mutant that lost the ability to assemble the axis-loop structure without losing cohesion of sister chromatids. This mutant showed defects in the formation of the axis-loop structure and LinE assembly and thus exhibited reduced meiotic recombination. Collectively, our results demonstrate that the Rec8-dependent axis-loop structure provides a structural platform essential for LinE assembly, facilitating meiotic recombination of homologous chromosomes, independently of its role in sister chromatid cohesion.


INTRODUCTION
Meiosis is a process of cell division unique to sexual reproduction, which halves the chromosome number in two suc-cessive nuclear divisions after a single round of DNA replication to produce haploid gametes. In the first meiotic division (meiosis I), homologous chromosomes (homologs) are segregated to opposite poles. This type of division is called reductional segregation and is essentially different from mitosis or meiosis II, in which sister chromatids are segregated to opposite poles (1). Crossovers formed by reciprocal recombination constitute a physical link between homologs, ensuring reductional segregation (2,3). Failure to establish such interhomolog links leads to the mis-segregation of chromosomes during meiosis I, resulting in aneuploidy of gametes. This failure is a major cause of human miscarriage or developmental disorders caused by chromosome trisomy, such as Down syndrome (4,5).
During meiotic prophase, homologs become closely associated or aligned through an elaborate proteinaceous structure called the synaptonemal complex (SC), followed by crossover recombination. The SC consists of two lateral elements connected by a central region (6,7). The lateral element is formed along the axis joining a pair of sister chromatids and is essential for meiotic recombination between homologs. Formation of these meiosis-specific chromosome structures is mediated by the meiotic cohesin complex (1,8).
The fission yeast Schizosaccharomyces pombe forms an atypical SC composed of a structure called the linear element (LinE). The LinE resembles the lateral element of the SC but lacks a central region composed of transverse filaments (42)(43)(44)(45)(46)(47). The assembly of S. pombe LinEs requires phosphorylated Rec11, a homolog of the meiosisspecific SA3 subunit of the Rec8 cohesin complex; phosphorylated Rec11 interacts with Rec10, a homolog of the lateral element factor SYCP2 in mammals (S. cerevisiae Red1) (48,49). In LinE-defective mutants of S. pombe meiosis, Rec8 cohesin-dependent chromosomal structure is maintained (39,48). Similarly, in SYCP2-depleted mouse spermatocytes defective in lateral element formation, normal cohesin-axis is formed (22,50,51). Therefore, the Rec8dependent chromosomal structure is independent of lateral elements (or LinEs in S. pombe), but the biological significance of this structure remains unclear.
In this study, we performed Hi-C analysis to delineate the meiotic chromosome structures of S. pombe. We then sought to clarify the functional significance of the Rec8dependent axis-loop structures by identifying rec8 mutants that specifically show defects in axis formation but not in cohesion.

Schizosaccharomyces pombe strains
All media and growth conditions were as described previously (52). Complete medium supplemented with uracil, lysine, adenine, histidine, and leucine (YE5S), minimal medium (EMM2), and sporulation-inducing medium (SPA and SSA) were used unless otherwise stated. The strains used in this study are listed in Supplementary Table S1. Deletion of the endogenous genes (rec12 + , rec11 + , rec7 + , rec15 + , mek1 + , rec8 + , rec10 + , hop1 + , wpl1 + , pds5 + , and mei4 + ) and tagging (rec10 + , rec8 + , psm1 + , rec25 + , rad21 + , and rec11 + ) with GFP, mCherry, 3 × hemagglutinin (HA) tag, 3 × Pk (V5), or 3 × Flag were performed according to the PCR-based gene targeting method for S. pombe, using the ura4 + , kanMX6 (kanR), hphMX6 (hygR), natMX6 (natR), and bsdR genes as selection markers (53)(54)(55). Histone H2B tagging by mCherry for time-lapse imaging has been described previously (56). The CK1 shutoff mutant, rec11-5A mutant, rec11-5D mutant, and the strain expressing Rec10-GFP-Rec11 fusion protein have been described previously (48). For the assay of crossing over recombination frequency, the hygR cassette was inserted 1,215 bp upstream from the start codon of the cox7 + gene, and the natR cassette was inserted 1,000 bp upstream from the start codon of vps38 + gene, using a PCR-based gene targeting method. To generate strains of rec8 point mutants with tagging of GFP or 3 × Pk tag, the endogenous rec8 + locus with its promoter and terminator was cloned into pUC119. Using the NEBuilder system (New England Biolabs), a DNA fragment encoding GFP or 3 × Pk tag with kanR or bsdR was introduced immediately before the stop codon of the rec8 gene cloned in pUC119. Next, target sequences of rec8 + were subsequently mutated using the KOD One Master Mix Blue (TOYOBO) to generate an amino acid substitution. In case of rec8-S552P, a C-terminal portion of the rec8 + open reading frame, including the mutation site and marker gene with the rec8 + terminator, was amplified by PCR and transformed into the wild-type rec8 + locus. Alternatively, in the case of N-terminal mutations such as rec8-F204S, the rec8 + gene region spanning the promoter to the terminator, including the mutation and marker gene, was amplified by PCR and transformed into the rec8::ura4 + strain. Correct replacement was confirmed using PCR and direct sequencing. The strain in which the mitotic cohesins, Rad21 and Psc3, were exchanged for meiotic cohesins (Rec8 and Rec11) has been described previously (57). To construct the pat1-as2 diploid with MDR-sup, haploid strains (h − pat1-as2 MDR-sup and h − pat1-as2 MDR-sup mat-Pc) were used (they were generously provided by Masamitsu Sato, Waseda University). After introducing a tag or disrupting the gene of interest in both haploids, protoplast fusion was performed as described previously (58), except for the use of Lysing Enzyme (Sigma-Aldrich) instead of Lallzyme or Zymolyase.

Preparation of meiotic cells
For microscopic observation of Rec8-GFP, Rec11-GFP, Rec10-mCherry, and Rec25-mCherry, h 90 cells cultured on YE plates at 26 • C overnight were streaked on SSA plates or suspended in 20 mg/mL of leucine and spotted on SPA or SSA plates. The cells were incubated at 26 • C for 6-10 h to observe the cells at the horsetail stage. In case of lacO/lacI-GFP observation for cohesion assays ( Figure 5C and S5C), h + and h − cells were mixed with 20 mg/mL of leucine and spotted on SSA or SPA plates. For the immunoprecipitation of Rec8-GFP with Psm1-3HA (Supplementary Figure  S5E), h + /h − diploid cells with mei4Δ were used. The diploid cells were grown in EMM2 liquid medium containing 5 mg/mL NH 4 Cl at a density of 5 × 10 6 cells/mL at 25 • C and then resuspended in EMM2 (1% glucose) medium lacking NH 4 Cl at a density of 1 × 10 7 cells/mL at 25 • C for 15 h. For the immunoprecipitation of Rec8-3 × Flag with Psm1 (Supplementary Figure S5F), ChIP assay ( Supplementary Figure S6B), detection of phosphorylated Rec11 ( Figure 7C), and Hi-C analyses, we used h − /h − diploid cells with pat1-as2 and MDR-sup alleles. Cells grown in YE liquid medium at a density of 5-10 × 10 6 cells/mL at 25 • C were collected by centrifugation and washed twice with EMM2 (1% glucose) medium lacking NH 4 Cl (EMM2-N). Cells were then resuspended in EMM2-N with 50 mg/mL leucine at a density of 5 × 10 6 cells/mL and incubated at 25 • C for 6.5 h to induce G1 arrest. By adding the same volume of EMM2-N with 50 mg/mL leucine and 15 M 1-NM-PP1 (Toronto Research Chemicals Inc.) (5-8 × 10 6 cells/mL), cells can enter meiosis. The culture was further incubated for 2.5 h or more. Before harvesting the meiotic cells, 1 mM phenylmethylsulfonyl fluoride (PMSF; Sigma-Aldrich) was added to reduce protein degradation.

Rec8 mutant screening
The rec8 + cDNA mutant library used for the screening was constructed as follows. The DNA region encoding 3 × HA-ura4 + of the plasmid VP068 (pUC119 base, Padh1-rec8 + cDNA-3 × HA-ura4 + -Trec8) (59) was replaced by the DNA region encoding GFP-BsdR to produce VP068-GFPbsd. Error-prone PCR of rec8 + cDNA in VP068-GFP-BsdR by Ex Taq DNA Polymerase (TaKaRa Bio) was performed using several concentrations of MnCl 2 (final 0.05 mM, 0.15 mM, 0.25 mM, and 0.3 mM) and an additional 1 mM dGTP or dATP with 0.2 mM of dNTP mix (total eight conditions). By using the NEBuilder system, amplified rec8 + cDNA pools (from eight conditions) were then cloned into the fragment, which was amplified by inverse PCR of VP068-GFP-BsdR without the rec8 + cDNA region. In total, approximately 5,300 Escherichia coli transformants (representing library size) were generated. All colonies were recovered from the plates, and plasmids (pRec8 library) were extracted. The pRec8 library was estimated to contain approximately six mutations per kilobase. Using the pRec8 library as a template, the region spanning 620 bp upstream of the adh1 + promoter to 600 bp downstream of the rec8 + terminator was amplified by PCR. These fragments were used to transform the host strains for screening. More than 95% of blasticidin-resistant colonies obtained after transformation showed correct integration, as confirmed by PCR. Rec8-GFP signals during meiotic prophase were observed among the blasticidin-resistant colonies, which showed relatively normal growth.

Fluorescence microscopy
Fluorescence images were obtained using an Axioplan2 microscope (Carl Zeiss) equipped with a Quantix cooled CCD camera (Photometrics) and a DeltaVision microscope system (GE Healthcare Inc.) equipped with a CoolSNAP HQ 2 cooled CCD camera (Photometrics) and a 60 × Plan-ApoN SC oil immersion objective lens (NA = 1.40; Olympus). The brightness of the images was adjusted using Fiji software (60) without changing the gamma settings. The pairing frequency was quantified as described previously (61). For time-lapse live-cell imaging of Rec8-GFP and H2B-mCherry, cells were cultured overnight in liquid EMM2 at 26 • C to reach the stationary phase; thereafter, the cells were spotted onto SPA plates to induce meiosis, followed by incubation at 26 • C for 4-5 h. Zygotes suspended in 1% glucose EMM2 (1% glucose) without nitrogen were sonicated briefly and immobilized on a coverslip coated with lectin in a 35 mm glass-bottomed culture dish (MatTek) and observed at 26 • C in liquid EMM2 (1% glucose) without nitrogen using a DeltaVision microscope in a temperaturecontrolled room. Time-lapse images were collected every 5 min at 9 focal sections with 0.45 m interval at each time point. Projected images were subsequently acquired using Z-sectioning and projected according to the SoftWoRx software program (Applied Precision).

ChIP assay
The procedures were carried out as described previously (62). Anti-FLAG M2 monoclonal antibody (Sigma-Aldrich) and IgG-conjugated Dynabeads (Invitrogen) were used for immunoprecipitation. DNA prepared from the whole-cell extracts or immunoprecipitated fractions was analyzed via quantitative PCR (qPCR) with the StepOnePlus Real-Time PCR system (Thermo Fisher Scientific) using Power SYBR Green PCR Master Mix C Green I Master (Thermo Fisher Scientific). The primers used for PCR have been described previously (48). Immunoprecipitation with control mouse IgG antibody (Abcam) was performed in each experiment to account for nonspecific binding in the ChIP fractions and verify the significance of immunoprecipitation. The percent IP was quantified using three independent qPCR experiments.

Recombination assay
For intergenic recombination, haploid strains containing appropriate markers were mated and sporulated at 26 • C on SSA plates for 30 h. The spores were isolated using ␤-glucuronidase solution (Wako) and plated on YE. The colonies were then replicated in selective media [EMM2uracil, EMM2-lysine, EMM2-lysine, YE5S containing hygromycin B (FUJIFILM) or YE5S containing clonNAT (WERNER BioAgents)] and YE5S containing phloxine B (Wako) to select haploid cells. In the ura1-lys3 interval, the degree of lysine auxotrophy among the uracil prototrophic colonies (n > 700, replicated three times) was assessed. In the cox7 + -vps38 + interval, the degree of hygromycin Bresistant (hygR-containing) colonies among the clonNATresistant (natR-containing) colonies was assessed (n > 500, and replicated three times). In the vps38 + -leu1 + interval, the degree of leucine prototrophy among the clonNATresistant (natR-containing) colonies was assessed (n > 500, and replicated three times). For intragenic recombination, h 90 haploid strains containing the ade6-M26 allele and plasmids harboring the ade6-469 allele according to the ura4 + marker (63) were sporulated at 26 • C on SPA plates for 30 h. The spores were isolated by treatment with glusulase and plated on YE5S containing phloxine B (to count all viable haploid spores) and EMM2-adenine plates. The frequency of intragenic recombination was quantified as the degree of adenine prototrophic colonies among viable haploid spores (n > 7,400, replicated three times).

Flow cytometry
A total of 1 × 10 7 cells were fixed with 70% ethanol and incubated with 50 g/mL RNase A in 500 L of 50 mM sodium citrate (pH 7.5) for 4 h at 37 • C and then stained with 0.5 g/mL propidium iodide. DNA content of 1.5 × 10 5 cells was measured on a FACS Calibur instrument (BD Biosciences).

In situ Hi-C
In situ Hi-C experiments were performed as described previously (64) with minor modifications. Asynchronous or meiotic cell cultures of 4 × 10 8 cells were fixed with 3% paraformaldehyde at 26 • C for 10 min. The fixed cells were disrupted using glass beads and Mini-Beadbeater-16 (BioSpec Products) in lysis buffer [50 mM HEPES/KOH pH 7.5, 140 mM NaCl, 1 mM EDTA, 1% Triton X-100, 0.1% sodium deoxycholate, 1 × Complete (Roche, 11836170001), and 1 mM PMSF]. The lysed cells were collected by centrifugation; pellets were resuspended with 1.11 × NEBuffer 2 (B7002, NEB) containing 0.1% sodium dodecyl sulfate (SDS), and cells were permeabilized by incubation at 62 • C for 7 min. After SDS was quenched by adding 1/9 volume of 10% Triton X-100, the permeabilized cells were treated with 25 units of MboI (R0147, NEB) at 37 • C overnight. The enzyme was inactivated by incubation at 62 • C for 20 min. After centrifugation, the pellet containing the restricted DNA fragments was used in the following procedure. A half volume of fill-in mix [ at 20 • C for 15 min followed by 0.032 units/L USER treatment at 37 • C for 15 min, with two washes with TNB at 55 • C for 2 min after each reaction. The resultant adaptor-ligated Hi-C library was amplified by PCR with five to seven cycles with Q5 master mix (M0544, NEB) and index primers (E7335/E7500/E7710/E7730, NEB) before 37 bp pairedend sequencing using the NextSeq 500 device (Illumina).

In situ Hi-C data analysis
In situ Hi-C data were processed as described previously (64). The mapping statistics of Hi-C data are summarized in Supplementary Table S2. The paired reads were separately aligned to the fission yeast genome (version ASM294v2. 19) using Bowtie2 (version 2.2.9) with the iterative alignment strategy. The aligned reads were assigned to MboI fragments. Redundant paired reads derived from PCR biases, aligned to repetitive sequences or with low mapping quality (MapQ < 30) were removed. Paired reads potentially derived from self-ligation and undigested products were also discarded. The fission yeast genome was divided into nonoverlapped 20, 5, or 2 kb bins. Raw contact matrices were constructed by counting paired reads assigned to two bins, and Hi-C biases in the contact matrices were corrected using the ICE method (65). The ICE normalization was repeated 30 times. ICE-normalized contact scores were used for all heat maps, where the maximum intensity corresponded to the top 5% score. For subtraction heat maps, the top 5% absolute values of difference scores (T) were extracted. The T and -T scores were drawn as the maximum and minimum intensity, respectively.

Loop detection
Knight and Ruiz (KR) normalized matrices (66) in hic format were calculated from raw contact matrices using the 'pre' function of Juicer tools (version 1.22.01) (67). Loop calling was performed for KR-normalized matrices using HiCCUPS ('hiccups' function of Juicer tools) with the following parameters: r 2000 -k KR -f .01 -p 3 -i 7 -d 10000".

Induction of highly synchronized meiosis under physiological conditions
To perform Hi-C analysis of chromosomal configurations during meiotic prophase, we first developed an experimental system to prepare S. pombe cell populations undergoing highly synchronized meiotic progression under physiological conditions. For this purpose, we employed an h − /h − pat1-as2/pat1-as2 diploid strain harboring one copy of the Nucleic Acids Research, 2022, Vol. 50, No. 7 3803 mat-Pc gene cassette (68)(69)(70). In this strain harboring an ATP analog-sensitive allele of the Pat1 kinase (pat1-as2), meiosis can be induced by addition of the ATP analog that inactivates Pat1 kinase. This allele offers advantages over the temperature-sensitive allele (pat1-114), in which raising the temperature causes detrimental effects on meiosis (68,71). In addition, MDR-sup alleles (72,73), which allow efficient cellular uptake of ATP analogs, were also exploited. The meiotic progression of wild-type diploid cells harboring Rec8-GFP was highly synchronous (∼70%), as indicated by the timing of nuclear division in meiosis I (Supplementary Figure S1A). Pre-meiotic DNA replication was almost completed approximately 1.5 h after the induction of meiosis by the addition of an ATP analog after G1 arrest via nitrogen starvation (Supplementary Figure S1A, left panel).

Hi-C analysis recapitulates meiotic bouquet chromosomes in S. pombe
For Hi-C analysis, cells were collected at 0 h (G1) and 2.5 h (meiotic prophase) after meiosis induction. Cells in vegetative growth (VG; mainly in mitotic G2 phase) were also collected for comparison. Notably, even at 0 h, a significant amount of Rec8-GFP protein was expressed and loaded onto chromatin (Figures S1B and S1C). The amount of Rec8-GFP peaked at approximately 2.5-3 h after meiotic induction (Figures S1B and S1C). We detected obvious centromere clustering (indicated by green arrows in Figures 1B and C) in the Hi-C contacts of all three chromosomes in VG cells, suggesting a Rabl orientation ( Figure 1A, left). In contrast, as cells progressed from G1 to meiotic prophase, centromere clustering became vague, while telomere clustering became evident (Figures 1B and C; blue arrows). As cells entered the meiotic prophase, X-shaped inter-and intra-chromosomal contacts appeared ( Figures 1B and C). These X-shaped contact patterns represent chromosomal arms bundled at the telomere, as previously demonstrated in S. pombe meiotic prophase (74,75) ( Figure 1A; right). In S. pombe, alignment of homologs relies on oscillatory movement of the nucleus (horsetail movement) in a bouquet orientation with all telomeres clustered to the spindle pole body (SPB) (42,(75)(76)(77)(78)(79). To systematically evaluate the degree of alignment of intra-chromosomal arms, we devised a measure 'alignment index' (Supplementary Figure S1D). The alignment index was calculated as the shortest distance from the telomere that gives normalized Hi-C contact scores above the threshold (refer to the legend of Supplementary Figure S1D for details). It corresponds to a length of the solid vertical line in Figure 1D. The alignment index of chromosome 1 increased from 1.6 (onset of meiosis at 0 h) to 2.5 (meiotic prophase at 2.5 h) compared with 0.55 in VG ( Figure 1D). A similar structure was observed during meiotic prophase in mouse spermatocytes by Hi-C analysis, although it was not as evident as in S. pombe (80).

Rec8 cohesin complex produces a chromosome shape suitable for alignment
To examine contributions of LinE and meiotic recombination to the chromosome structures, we performed Hi-C analysis in mutant cells lacking three factors that play a central role in meiotic recombination (Rec10, Rec12, and Rec8). Rec10 (a homolog of Red1/SYCP2) is a major component of LinE, and Rec12 is a Spo11 homolog that induces DNA double-strand breaks (81,82). The Rec8 cohesin complex is required for proper assembly of LinE (46,48,49,83). For Hi-C analysis, we first confirmed synchronous progression of meiosis in these mutants (Supplementary Figure S2). Hi-C analysis for rec10Δ and rec12Δ cells showed X-shaped contacts similar to wild-type cells (Figures 2A and B), with an alignment index of 2.8 (rec10Δ), 2.5 (rec12Δ), and 2.5 (wild-type, wt). In contrast, rec8Δ cells showed prominent defects with almost no X-shaped contacts (Figures 2A and  B; alignment index 1.5). These results suggest that Rec8 is required for the alignment of homologs in a manner distinct from LinE formation.
In the absence of Rec8, the bulk of the chromosomes do not follow the horsetail movement at meiotic prophase while only telomere regions follow the leading edge of the moving nucleus (39,84). This chromosome behavior reflects that the traction force generated by the horsetail movement is not transmitted to the chromosomes as proposed previously. The Hi-C results and chromosome movements suggest that the Rec8 cohesin complex forms the axis-loop chromatin structure and has structural properties to counteract the traction force ( Figure 3F). This Rec8-dependent chromatin structure promotes the alignment of homologs without requiring LinE.

Rec8 cohesin complex is crucial to form meiosis-specific chromatin loops
Magnification of Hi-C data (2 kb binning) in chromosome arm regions revealed that punctate Hi-C interactions occurred at the onset of meiosis (0 h) and became more prominent as meiosis progressed (2.5 h) in the wild-type (arrowheads in Figures 3A and S3A). As meiosis progressed in the wild-type, P(s) (the contact frequency versus distance) indicated a 'shoulder,' representing an increase in interactions within a region shorter than that length ( Figure 3C). The positions at the bases of these interactions coincided with previously identified Rec8 accumulation sites (85) (Figures 3B and S3B), suggesting that these punctate Hi-C interactions are generated by Rec8. In fact, punctate Hi-C interactions observed in the wild-type were mostly lost in rec8Δ ( Figure 3A). In contrast, punctate Hi-C interactions were retained in rec10Δ or rec12Δ (Figures 3A and S3A). Consistently, the shoulder in P(s) was lost in the rec8Δ mutant but not in rec10Δ or rec12Δ (upper panel in Figure  3D). This difference in these mutants is also seen in the P(s) derivative plot (lower panel in Figure 3D). In the Hi-C data with 5 kb binning, topologically associating domain (TAD)-like signals were detected in wild-type VG cells (dotted lines in Figure 3E), representing the manifestation of relatively long-range interactions in mitotic chromosomes (27,64,80,86). As these TAD-like signals disappeared, Rec8dependent loop structures appeared during meiosis ( Figure  3E), similar to mammalian spermatocytes (80,87,88). These results suggest that a Rec8-dependent cis-loop structure is formed upon entry into meiosis in the wild-type, leading to the linear compaction of chromosomes ( Figure 3F).

Wpl1 in S. pombe regulates the length of chromatin loops in meiotic prophase
In somatic cells, the Rad21/Scc1 cohesin complex organizes chromatin with TADs, and Wapl depletion lengthens the chromatin loops and thickens the chromatin axis (34,37). However, evidence for Wapl affecting Rec8-dependent chromatin loops during meiosis remains elusive. To reveal involvement of Wapl in meiosis, we analyzed the chromosome structure in wpl1Δ cells. Fluorescence images of Rec8-GFP showed that axial structures in meiotic chromosomes were more prominent in wpl1Δ than in the wild-type (Figure 4A), suggesting that the axial structure becomes thicker in S. pombe during meiosis. To further examine the chromosome structures in wpl1Δ, we performed Hi-C analysis. We first confirmed synchronous progression of meiosis in wpl1Δ as in wild-type (Supplementary Figure S4A). Hi-C contact maps showed that the background in wpl1Δ (upper panel of Figure 4B) was reduced than that in the wild-type (lower panel of Figure 4B). This reduced background probably reflects a reduction of the inter-and intrachromosomal long-range interactions, including X-shaped contacts, in wpl1Δ. This notion is supported by fluorescence microscopic observations showing that alignment of chromosome arms was defective in wpl1Δ ( Supplementary Figure S4B) and that the chromosomes in wpl1Δ rarely showed torsional turning (Supplementary Figure S4C, Supplemental Movies 1, and 2) during horsetail movements that is important for the alignment of homologs (89). These results suggest that Wpl1 plays a role in alignment of homologs through Rec8-dependent formation of axis-loop chromatin structure.
In the amplified view of Hi-C data (2 kb binning) in wpl1Δ, the number of grid-like punctate signals increased and the size of the diagonals of each grid increased upon entry into meiosis ( Figures 4C and S4D), suggesting that loop lengths are longer in wpl1Δ. In fact, compared to the wild-type, the shoulder in P(s) showed a rightward shift in wpl1Δ (upper panel in Figure 4D). The derivative plot of P(s) also showed a rightward shift in wpl1Δ (lower panel in Figure 4D). Consistently, the HiCCUPS analysis showed that the number of longer loops increased in wpl1Δ ( Figure  4E). These results imply that the proximal chromatin loops fuse with each other at their bases to form longer loops in the absence of Wpl1, leading to the linear compaction of chromosomes with thicker axes (Supplementary Figure  S4E).
Fluorescence microscopy showed that the thick Rec8 axis observed in wpl1Δ was maintained in wpl1Δ rec10Δ cells, in which LinE formation was lost (Supplementary Figure  S4F). Moreover, the thick Rec8 axis in the wpl1Δ background was formed in the mutants of Hop1 (a homolog of HORMAD) and Mek1 (Supplementary Figure S4G), which are required for LinE formation (90). The thick Rec8 axis in the wpl1Δ background was also observed in the mutants of Rec12, Rec7, Rec15 (Supplementary Figure S4G), and Mei4 (Supplementary Figure S4H), which are required for recombination (91)(92)(93)(94). These results further support our idea that the Rec8 axis formation is independent of LinE and recombination.

Identification of rec8 mutants showing defects in axis formation
Ectopic expression of Rad21 in meiosis partly compensates for the cohesion defect in the absence of Rec8 (57,59). However, Rad21 does not form a visible axis structure along meiotic chromatin, unlike Rec8 (Supplementary Figure S5A). Therefore, besides the function of cohesion, Rec8 is expected to have an additional function that is more robust than that of Rad21 in chromatin axis formation. Based on this idea, we attempted to identify mutants of rec8 that specifically shows defects in axis formation without affecting its cohesion function. To this end, we constructed the S. pombe strain in which Rec8 and Rec11 were ectopically expressed as the sole source of cohesin during mitosis, replacing Rad21 and Psc3 (57) ( Figure 5A). Using this strain as a screening host, randomly mutagenized rec8 cDNAs fused with GFP-coding DNA were transformed into the host strain. Since Rec8 is the only available kleisin protein in this strain, transformants bearing a rec8 mutation with a loss of cohesion show a lethal or slow-growth phenotype. Furthermore, we used the h 90 mating-type in the mei4Δ background as a host strain to arrest each transformant at the meiotic prophase on agar plates with the sporulation medium (91). We also introduced wpl1Δ in the host strain to produce a prominent axis of Rec8-GFP ( Figure 4A) so that defects in axis formation could be easily evaluated by microscopic observation (Figure 5A).
Among approximately 3,000 viable transformants of the rec8 mutant library, we searched for mutants that showed axis formation defects with normal cohesion function under a fluorescence microscope. We identified a mutant that displayed normal growth (Supplementary Figure S5B) but lost the Rec8 axis structure ( Figure 5B). In this mutant, the amino acid phenylalanine at the 204 th residue of Rec8 was substituted with serine (rec8-F204S). The rec8-F204S mutant maintained sister chromatid cohesion as assessed at the cut3 gene locus ( Figure 5C) and other chromosomal loci (Supplementary Figure S5C); rec8-S552P and rec8Δ, which showed the cohesion defect, were used as a control strain (see Supplementary Figure S5C, S5D, and S5E for details of the rec8-S552P mutant). Moreover, a normal cohesin complex was formed in the rec8-F204S mutant, with comparable expression of Rec11 (Figures S5F and S5G).
To delineate the axis formation defects caused by rec8-F204S, we observed the chromosome morphology during horsetail movement in meiotic prophase. In the rec8-F204S mutant, only the leading edge of the nucleus followed the horsetail movement, while the bulk of chromosomes were left behind, similar to rec8Δ ( Figure 5D, Supplemental Movies 1, and 3). To determine the degree of chromosome compaction more directly, we measured the distance between the telomere and the ade8 gene locus during meiotic prophase using a GFP-tagged telomere protein (Taz1) and a lacO/lacI-GFP tag at the ade8 gene locus ( Figure 5E). As in rec8 , the telomere-ade8 distance was longer in the rec8-F204S mutant than in the wild-type, suggesting that the chromatin of the rec8-F204S mutant was flexible and was abnormally stretched by the traction of the horsetail movement.     rec8-wt In meiosis, the Rec8-Psc3 cohesin complex exists only at the centromeres, while the Rec8-Rec11 cohesin complex is present at the chromosomal arm regions. The right panel shows the scheme for isolation of rec8 mutants that specifically show defects in axis formation but display normal cohesion in meiotic prophase. The promoter of the rec8 gene is replaced by the adh1 promoter, and the promoter of the rec11 gene is replaced by the adh41 (a weaker version of adh1) promoter, to allow gene expression both in mitosis and meiosis. After transformation of a library of GFP-fused rec8 mutants with a selective marker, bsdR (not shown here), the blasticidin-resistant colonies obtained were subsequently induced to undergo meiosis to observe the axial Rec8-GFP signals. Growth rate of the blasticidin-resistant clone acts as an indicator of the functionality of Rec8-Rec11 cohesin. (B) Representative images of Rec8-GFP depicting cell shape (white dotted line) during meiotic prophase in wild-type and the isolated rec8-F204S mutant in this screening. (C) Cohesion defects in the chromosome arm region (cut3 + locus, left arm of chromosome 2) were examined in the indicated cells arrested at prophase I by the mei4Δ mutation. After fixation by methanol, the number of heterozygous cut3-GFP signals was counted (n > 150). A score of 2 or 1.5 cut3-GFP signals represents a complete or partial loss of sister chromatid cohesion, respectively. The rec8-S552P identified in this screening displays a remarkable defect in cohesion that is stronger than that of rec8Δ (see the text). Error bars show standard deviations from three independent experiment. P values were obtained from Chi-square test for statistical significance. (D) Time-lapse observations (5 min intervals) of horsetail nuclear movement in the indicated cells stained with histone H2B-mCherry. The cartoon on the right represents a model of the chromatin state for each strain. In rec8Δ, Rad21 partially compensates for the cohesion defect. (E) The upper left panel shows a schematic diagram of the distance of ade8-lacO inserts from the telomere of chromosome II. The upper right panels represent images of GFP signals for the indicated strains. Bright GFP signals at the edge of the nucleus represent telomeres (Taz1-GFP), and weak GFP signals represent the ade8 locus. The graph at the bottom displays a plot of the distance between the telomere and the ade8 locus in each cell during meiotic prophase for the indicated strains. The average distance is plotted (n = 55).
To evaluate the specificity of the amino acid at the 204 th residue of Rec8 in terms of the axis formation, we examined Rec8 mutants bearing other amino acid substitutions .  F204T, F204N, F204D, F204A, F204R, and F204H mutants showed defects in the Rec8 axis formation, whereas F204C, F204W, and F204L mutants formed the normal Rec8 axis (Supplementary Figure S5H). These results suggest that the hydrophobicity of this residue in Rec8 may be important for its axis-forming function.

rec8-F204S mutant is defective in chromatin loop formation
To examine the chromatin loop formation in the rec8-F204S mutant defective in axis formation, we performed Hi-C analysis in the rec8-F204S mutant as performed for the rec8wt (Supplementary Figure S6A). In the Hi-C contact maps of three chromosomes at meiotic prophase (2.5 h), formation of X-shaped patterns representing the bouquet orientation were evidently impaired in the rec8-F204S mutant, as in rec8Δ ( Figure 6A). The alignment index of chromosome 1 at meiotic prophase (2.5 h) decreased in rec8-F204S (1.6) compared with the wild-type (rec8-wt, 2.6) ( Figure 6B). Moreover, in the Hi-C data with amplified 2 kb binning, the number of punctate Hi-C interactions decreased in rec8-F204S ( Figure 6C), although it was not as prominent as in rec8Δ. In addition, the shoulder in P(s) became flattened in rec8-F204S compared to rec8-wt, both at 0 and 2.5 h after meiosis induction ( Figure 6D). Consistently, chromatin loops with a length of approximately 20 kb that formed in a Rec8-dependent manner were diminished at the meiotic prophase in rec8-F204S ( Figure 6E). These results suggest that in the rec8-F204S mutant, as in rec8Δ, the Rec8dependent meiosis-specific short chromatin loop structures are lost, resulting in a concomitant loss of the structural property of the chromosome required for proper alignment.
Next, we performed a chromatin immunoprecipitation (ChIP) assay to determine the localization pattern of Rec8 on chromatin. The accumulation of Rec8-F204S protein apparently decreased compared with wild-type Rec8 at the loci of known cohesin-enriched regions (Supplementary Figure  S6B). In contrast, the localization in other chromosome arm regions seemed to be normal (Supplementary Figure  S6B). Rec8-enriched regions mostly coincided with each base of the chromatin loops ( Figures 3C and S3B). In the case of rec8-F204S, the amount of Rec8 in the non-enriched region appeared normal. Thus, the reduced accumulation of Rec8 at known enriched regions possibly reflects a noticeable loss of chromatin loop formation (Supplementary Figure S6C).

rec8-F204S mutant is defective in LinE formation and recombination
Next, we examined the formation of LinEs in the rec8-F204S mutant. In wild-type cells, the LinE component Rec10-mCherry localized within the entire nucleus, similar to Rec8-GFP (Figures 7A and S7A; CK1 + , rec8 + , rec11wt). In rec8-F204S cells, however, Rec10-mCherry formed aberrant dotty or filamentous aggregates within the nucleus, similar to rec8Δ and rec11-5A cells in which LinE was completely lost (Figures 7A and S7A) (48). Consistent with the LinE formation defect, rec8-F204S cells showed a decrease in the frequency of crossover recombination at several loci ( Figure 7B). Intragenic recombination between the genomic ade6-M26 allele and the plasmid harboring the ade6-469 allele was also significantly reduced in rec8-F204S (Supplementary Figure S7B), similar to rec8Δ. Rec11 phosphorylation by CK1, which is required for the assembly of LinEs (48), occurred normally in the rec8-F204S mutant, similar to the wild-type ( Figure 7C). Expression of a phosphomimetic mutant of Rec11 (rec11-5D), which can suppress the defects caused by the loss of Rec11 phosphorylation via CK1 (48), did not suppress defects in LinE formation or meiotic recombination in rec8-F204S (Figures S7A and  S7C). These results suggest that axis-loop formation along sister chromatids via the Rec8 cohesin complex is still essential for LinE assembly, even in the presence of phosphorylated Rec11. Collectively, these results demonstrate that Rec8-dependent axis-loop structures facilitate meiotic recombination through the LinE assembly ( Figure 7D).

DISCUSSION
In this study, we have demonstrated the Rec8-dependent formation of axis-loop structures during meiosis in S. pombe, an asynaptic organism. The Rec8-dependent axisloop structure functions as a scaffold for the formation of the LinE axis, thereby ensuring homologous recombination during meiosis.

Pairing of homologous chromosomes in S. pombe meiosis
Our Hi-C analysis revealed the emergence of an intriguing X-shaped contact map representing tight contacts between chromosome arm regions near the telomere in the meiotic prophase of S. pombe. This result is consistent with previous microscopic observations of telomere clusters at the SPB (the bouquet orientation of chromosomes) during meiotic prophase (75,79) ( Figure 3F). The Hi-C data indicate that contacts occur not only between homologous arms but also between non-homologous arms. It should be noted that chromosomes recognize their homologous pair among a bundle of all the chromosomes. Therefore, mechanisms to eliminate non-homologous pairing while simultaneously promoting homologous pairing may exist. Such selective forces between chromosomes can be provided by the oscillatory movement of telomere-bundled chromosomes (89,95). Additionally, noncoding RNAs play a robust role in recognizing homologous pairs (76,78,96). Importantly, the Rec8dependent formation of axis-loop chromatin structure is necessary for optimizing these mechanisms involving telomere clustering, horsetail movement, and noncoding RNAmediated recognition.

Possible mechanisms for formation of the Rec8 axis-loop chromatin
A loop extrusion model has been proposed to form chromatin loops by SMC protein complexes including cohesin (26,28,30,97). In this model, the SMC protein complex interacts with DNA, reels, and extrudes it through its ring structure to form a DNA loop (98)(99)(100)(101)(102). In addition, cislooping of distal chromatin or TAD formation is thought     to be established by the action of the Rad21 cohesin complex anchored on chromatin via the CCCTC-binding factor (CTCF) in mammalian somatic cells (66,(103)(104)(105). CTCF is conserved in vertebrates (106). In S. pombe, CTCF is not present, and proteins that replace the function of CTCF are unknown. In contrast, Scc2 and Scc4 are conserved from yeasts to humans (Mis4 and Ssl3 in S. pombe, respectively); human Scc2 and Scc4 are required for the DNA extruding activity of human Rad21 cohesin complex at least in vitro (98). In S. pombe, chromatin loop structures are already formed at the onset of meiosis (0 h) when the Rec8 cohesin complex is loaded on the chromatin by Mis4 and Ssl3 before pre-meiotic DNA replication ( Figures 3A and S1A). Thus, Mis4 and Ssl3 can be initial factors that create chromatin loops in S. pombe meiosis. Alternatively, some changes in chromatin potential, involving histones, Rec8, or other yetunidentified factors, may occur upon entry into meiosis to form the chromatin axis-loop structure. The rec8-F204S mutant isolated in this study showed a significant defect in the formation of axis-loop chromatin structures, despite maintaining the cohesion function. The F204 residue of Rec8 locates in the kleisin-Scc2 interaction domain, as previously shown by an in vitro binding assay (107). However, this phenylalanine residue is not conserved in vertebrates (Supplementary Figure S5I) despite conservation in neighboring sequences. Thus, Rec8 and Scc2/Mis4 with Scc4/Ssl3 may also play some roles in modulating meiosis-specific chromatin loop structures, although precise mechanisms remain to be elucidated.

Rec8 axis-loop structure as a scaffold for LinE assembly
In the rec8-F204S mutant, which was unable to form the axis-loop structure, LinE formation failed to occur, and the recombination frequency was markedly reduced (Figures 7A, S7A, 7B, and S7B). However, phosphorylation of Rec11, which is required for LinE formation, was normal in the rec8-F204S mutant (Figures 7C), suggesting that mechanisms other than Rec11 phosphorylation are involved in the LinE defects of this mutant. It has been reported that chromosomal localization of Rec8 does not exactly overlap with that of Rec10 (a LinE component) despite the interaction of the Rec8 cohesin complex with Rec10 (48,92,108). These results imply that LinE factors, including Rec10, localize to chromosomes initially through interaction with the Rec8 cohesin complex and then redistribute along the chromosome appropriately to assemble LinEs. Based on these notions, we speculate that oligomerization of LinE factors requires the Rec8 axis as a scaffold, which is lost in the rec8-F204S mutant. Without this Rec8 scaffold, the association between LinE factors and chromatin would not be stabilized, resulting in the formation of aggregates (Figures 7A,  S7A, and 7D), probably due to their propensity to adhere to each other. These aggregates are reminiscent of the polycomplex formed by mutants with defects in SC formation during meiosis of S. cerevisiae (2).
Expression of the Rec11-Rec10 fusion protein as well as the Rec11-5D mutant protein, both of which can suppress the defects caused by the loss of Rec11 phosphorylation (48), did not suppress the LinE formation defect of rec8-F204S ( Figures S7A, S7C and S7D). These results sup-port the notion that the defects in LinE formation in this mutant are due to the unstable interaction between Rec10 and the Rec8 cohesin complex ( Figures 7D). LinEs in S. pombe are crucial in chromatin recruitment of factors that induce DNA double-strand breaks for recombination (92). This fact explains why the rec8-F204S mutant fails to undergo meiotic recombination. Although the F204 residue in mouse is leucine, the phenylalanine-to-leucine substitution in S. pombe Rec8 showed almost no defects in axis formation ( Figures S5H and S5I). Rec8 and Rec11/SA3 are required for lateral element formation in mammals, as well as for LinE formation in S. pombe (18)(19)(20)(21)46,48,49,83). Thus, there may be a mechanism by which meiosis-specific cohesin-dependent axial structures can function as a scaffold in the formation of lateral elements in mammals.
In conclusion, the current results obtained in S. pombe indicate that the Rec8-dependent axis-loop chromatin structures are crucial for meiotic recombination. As meiotic recombination is essential for faithful propagation of chromosomes into the progenies during meiosis in eukaryotes, this chromatin structure will provide a basis for understanding the molecular mechanisms of meiotic chromosome propagation in mammals, potentially related to miscarriages and congenital abnormalities in humans (4,5).

DATA AVAILABILITY
Hi-C data were submitted to Gene Expression Omnibus (GEO) with the accession number: GSE192516. Further data are available from the corresponding authors upon reasonable request.