Calreticulin is required for development of the cumulus oocyte complex and female fertility

Calnexin (CANX) and calreticulin (CALR) chaperones mediate nascent glycoprotein folding in the endoplasmic reticulum. Here we report that these chaperones have distinct roles in male and female fertility. Canx null mice are growth retarded but fertile. Calr null mice die during embryonic development, rendering indeterminate any effect on reproduction. Therefore, we conditionally ablated Calr in male and female germ cells using Stra8 (mcKO) and Zp3 (fcKO) promoter-driven Cre recombinase, respectively. Calr mcKO male mice were fertile, but fcKO female mice were sterile despite normal mating behavior. Strikingly, we found that Calr fcKO female mice had impaired folliculogenesis and decreased ovulatory rates due to defective proliferation of cuboidal granulosa cells. Oocyte-derived, TGF-beta family proteins play a major role in follicular development and molecular analysis revealed that the normal processing of GDF9 and BMP15 was defective in Calr fcKO oocytes. These findings highlight the importance of CALR in female reproduction and demonstrate that compromised CALR function leads to ovarian insufficiency and female infertility.

Proper folding in the endoplasmic reticulum (ER) is a prerequisite for correct localization and function of most secreted and transmembrane proteins 1 . Failures in protein folding and quality control compromise cellular functions and cause disease including amyloidosis, cystic fibrosis, and diabetes 2,3 . Failure in ER quality control also results in male infertility [4][5][6] . In the ER, soluble calreticulin (CALR) and membrane bound calnexin (CANX) were originally discovered as homologous calcium binding proteins and later shown to be lectin like chaperones that chiefly mediate nascent glycoprotein folding [7][8][9] . Despite their extensive homology, CALR and CANX have contrasting functions. For example, proteins in Calr −/− cells have accelerated folding with an accumulation of misfolded proteins whereas folding is significantly impaired in Canx −/− cells 10 . Differences in CALR and CANX function are further reflected in the distinct phenotypes of knock-out mice. Calr −/− mice are embryonic lethal due to defective heart development whereas Canx −/− mice are viable but growth retarded with neurological deficits [11][12][13] . However, the client specificity of calnexin/calreticulin in vivo has not been fully established.
We have previously demonstrated that calmegin (CLGN) and calsperin (CALR3) are testis specific homologues of CANX and CALR, respectively, and are expressed in the ER of spermatogenic cells 4,5,14 . CLGN mediates the heterodimerization of ADAM1A/ADAM2 that is required for the maturation of

Generation of conditional knockout (KO) mice for Canx and Calr genes.
Mice with genetically disrupted Canx and Calr genes are growth retarded or die during embryonic development, respectively [11][12][13] . Thus, we generated male (mcKO) and female (fcKO) germ cell-specific conditional knockout mice to investigate the role of these genes in fertility. We first generated mice carrying floxed alleles (fl/+ ) using gene targeting vectors that floxed exons 3 and 4 for Canx and exons 4-7 for Calr (Supplementary . However, 50% of the homozygous null pups died within 48 hours and very few survived to three months, as reported previously 12 . Similar postnatal lethality was not observed in Canx gene trapped mice despite the complete absence of CANX protein 13 . Paradoxically, truncated CANX protein was present in one of the two lines in the earlier report with the more severe phenotype. Using antibodies against the N-and C-terminal residues, CANX protein was not detected in our −/− mice ( Supplementary Fig. S1D). Therefore the variance in postnatal viability reported earlier may reflect the combined effect of CANX deficiency and other factors, including genetic background and animal husbandry.
Canx −/− females were also fertile, but litter sizes were smaller (3.3 ± 1.5 pups, n = 4 litters from 4 females) possibly due to their smaller body size. We produced oocyte specific Canx fcKO mice by introducing the ZP3-cre transgene (Canx fl/− ; Zp3-cre). Mice were genotyped by PCR analysis and the lack of CANX protein was confirmed by immunoblot ( Supplementary Fig. S1C, related to Fig. 1). Canx fcKO female mice had normal fertility when mated with wild type males (8.4 ± 2.6 pups, n = 9 litters from 3 females) (Fig. 1A,B). All 76 pups carried the knockout allele, which confirmed successful excision of the floxed exons by the Zp3 promoter driven Cre recombinase. Therefore, we conclude that CANX is not required for either male or female reproduction.
Female infertility in Calr conditional knockout mice. It was reported that the Stra8 promoter driven Cre recombinase was expressed at the postnatal day 3 in early-stage spermatogonia and the recombination efficiency was > 95% 26 . In the present study, with our Stra8-cre transgenic line, Calr disruption was confirmed in most (84.2%, 219/260) of the testicular germ cells as determined by immunostaining (Supplementary Fig. S4A and B, related to Fig. 1). When mated with normal female mice, Calr Scientific RepoRts | 5:14254 | DOi: 10.1038/srep14254 mcKO males had comparable fertility (10.1 ± 1.5 pups, n = 26 litters from 6 males) with normal male mice (10.6 ± 1.3 pups, n = 8 litters from 3 males). Whereas some pups inherited the floxed allele, the majority (82.1%, 215/262) of the offspring inherited the knockout allele, which was consistent with the aforementioned observations in testicular germ cells.
We next used ZP3-cre transgenic lines to disrupt the Calr gene during oogenesis (Supplementary Fig.  S2C and Supplementary Fig. S4C, related to Fig. 1). Whereas control female mice (Calr fl/+ ; Zp3-cre) had normal fertility, Calr fcKO female mice had a profound decrease in fecundity (Fig. 1A,B). Only 2 litters with 1 pup each were obtained from 16 copulations with 4 Calr fcKO females, whereas 10 litters from 10 copulations were obtained with 4 control females. The average litter sizes were 0.1 ± 0.3 and 8.4 ± 2.6 pups, respectively. When we superovulated Calr fcKO females with gonadotropins, successful copulation was observed. Thus, the female infertility was not caused by disrupted mating behavior. However, ovulation was severely impaired in the Calr fcKO female mice (Table 1).
To elucidate the cause of infertility in the Calr fcKO females, we superovulated 3 ( Fig. 1C-F) or 12 (Supplementary Fig. S5C-H) week-old female mice and prepared ovarian sections 2 hours before the anticipated time of ovulation. In Calr fcKO mice, the ovary was smaller, possibly due to defective folliculogenesis (Table 2). A few preovulatory follicles appeared, but most of the follicles at the surface of the ovary were immature. Follicular development was arrested at the early antral stage, the number of cumulus cells that surrounded an oocyte after germinal vesicle breakdown in cross sections was reduced (33.1 ± 13.6 in fcKO and 141.6 ± 30.5 in control ) and cumulus expansion was impaired ( Fig. 1C-F and Supplementary Fig. S5A-H). Corpora lutea were rarely observed in Calr fcKO ovaries. Because antral follicles were present, we assayed in vitro maturation and in vitro fertilization. Comparable numbers of oocytes were collected from control and Calr fcKO mice 46-48 hours after PMSG injection. Both control and mutant oocytes underwent germinal vesicle breakdown (GVBD) and matured to the metaphase II (MII) stage in vitro. The oocytes collected from the Calr fcKO had slightly larger diameters than those collected from control mice (77.6 ± 0.3 μ m in fcKO and 73.5 ± 0.4 μ m in control, P < 0.01), as reported in Gdf9 KO mice 23 . Although the efficiency was lower than in controls, these MII eggs could be fertilized and developed to term after transfer into pseudopregnant females (Table 3).
CALR mediated quality control of GDF9 and BMP15. Calr fcKO had impaired ovulation with defects in cumulus expansion. GDF9 and BMP15 are two growth factors that are secreted from oocytes and stimulate cumulus cell expansion 23,28 . To examine the effect of CALR disruption on the production of these proteins, primary mouse embryonic fibroblast (MEF) cells were prepared from Calr +/+ and Calr −/− mice and transfected with mouse Gdf9 or Bmp15 expression vectors. Using co-immunoprecipitation, we confirmed that CALR associates with GDF9 and BMP15 in wild-type cells ( Fig. 2A). In Calr +/+ MEFs, GDF9 and BMP15 were secreted into the extracellular fluid for 4-5 days, whereas in Calr −/− MEFs, secretion stopped within 1-2 days (Fig. 2B). When carefully observed, GDF9 proproteins appeared as a doublet on SDS-PAGE 29 , but the upper band was not present in Calr −/− cell supernatants (Fig. 2C). Recovery of the doublet was observed after co-transfection of a Calr expression vector. Similar results were observed for BMP15. To examine whether GDF9 is expressed and/or secreted in Calr fcKO oocytes, we collected 300 oocytes and examined their lysates by immunoblot (Fig. 2D). Whereas comparable amounts of GDF9 proproteins (57 kDa) were observed in control and fcKO oocytes, mature type GDF9  Complementation of defective GOC (granulosa-oocyte complex) development by recombinant GDF9 and BMP15. External addition of recombinant BMP15 or GDF9 improves cumulus cell proliferation, differentiation, and steroidogenesis in Bmp15/Gdf9 null mice 30,31 . In the present study, we examined whether GDF9 and BMP15 supplements could restore the defective GOC development in Calr fcKO mice using an in vitro follicle culture system. Without supplementation, granulosa cells proliferated and the size of the follicles gradually increased when CALR was present. In Calr fcKO GOC, the follicle size did not increase compared to controls (Fig. 3). When the GOC growth medium was supplemented with either recombinant GDF9 or BMP15, both factors significantly enhanced Calr fcKO GOC growth. Conditioned medium from cells expressing both GDF9 and BMP15 did not show a synergistic effect.

Discussion
Although CANX is not essential for mouse development in vivo 12,13 , it has been difficult to investigate its role in the reproductive system since few Canx KO mice survived to adulthood and they were smaller than their littermates. In the present study, we showed that Canx KO males were able to copulate and successfully impregnate females. Canx KO females showed reduced litter size, but oocyte specific disrupted cKO females were fully fertile. Therefore we conclude that CANX is dispensable for both male and female reproductive systems. We next examined the role of CALR, the soluble homologue of CANX, in the reproductive system. Calr KO mice are embryonic lethal but ectopic expression of calcineurin in the heart enabled Calr KO mice to survive until adulthood 32 . However, the surviving mice exhibited growth retardation, hampering any study of their reproduction. Here, we generated male germ cell specific cKO mice and showed that CALR is dispensable for spermatogenesis and sperm fertilizing ability. Although CANX and CALR have contrasting functions in other tissues 12,32 , their major roles may be redundant and complementary to each other, at least in developing male germ cells. In contrast, the homologues of CANX and CALR, CLGN and CALR3, respectively, have different substrate specificity and both, albeit by different mechanisms, are required for fertilization by controlling ADAM3 presentation on the sperm surface 4,5 . Thus, the present study reinforces the uniqueness and importance of the CLGN/CALR3 ER chaperone system in the male reproductive system.
In the female reproductive system, we have shown that CALR has a novel and indispensable role in COC development and female fertility. Since GDF9 and BMP15 are secreted proteins, soluble CALR might be more accessible to these molecules than membrane-tethered CANX. However it is also reported that soluble CANX and CALR have different substrate specificity 33 and an in-depth analysis of the molecular interactions of these chaperones with these two growth factors requires future investigation. Our data indicated CALR plays an important role in regulating the folding of GDF9 and BMP15 in the ER. The folding defects could cause various effects, such as protein instability, secretion defects, abnormal cleavage or aberrant post-translational modifications of GDF9 and BMP15 at later stages of the secretory pathway.
In the present study we could not detect mature GDF9 and BMP15 proteins in MEF cells. The defect in the processing machinery may cause different behaviors in unfolded GDF9 and BMP15 in MEF and oocytes. It has been reported that the mature recombinant mouse BMP15 is not processed in HEK 293T cells and CHO cells 34 . Further, the post-translational processing of BMP15 is precisely regulated during meiosis and the mature BMP15 can be detected 9 hours after the hCG injection in mice 28 . These studies imply that oocytes may have a unique post-translational processing mechanism for GDF9 and BMP15.
The follicular development in the Gdf9 KO is arrested at the primary stage and the Bmp15 KO showed normal folliculogenesis despite decreased ovulation and rates of fertilization 23,24 . However, follicular development in Calr fcKO mice arrested at the early antral stage. In Calr fcKO mice, GDF9 proprotein is expressed in oocytes, but not cleaved. In rat, macaque and human, GDF9 proprotein or BMP15 is detected in the follicular fluid 35,36 . In addition, mutation in the protease cleavage site of GDF9 causes female infertility in ewes 37 . The hypoplastic ovaries of homozygous mutated ewe lambs contain large numbers of primordial follicles and developing follicles up to the early antral stage. These data suggest that the proprotein of GDF9 accounts for the different phenotypes in Gdf9 KO and Calr fcKO mice.
The mechanism of COC development has been well investigated and many key factors have been identified in both oocyte and surrounding granulosa cells 38,39 . Among these factors, GDF9 and BMP15 are secreted from developing oocytes and stimulate granulosa cell proliferation and expansion 24,31,40 . GDF9 and BMP15 have a unique disulfide bond arrangement among TGF-beta family proteins [41][42][43] , and showed different mobility in SDS electrophoresis under reducing or non-reducing conditions 31,44 , indicating the importance of disulphide bond formation during their proper folding, targeting, and function. Here we showed that CALR interacts with GDF9 as well as BMP15 and the lack of CALR compromised their secretion. This is reminiscent of the misfolding and disappearance of ADAM3 in Calr3 knockout male mice. In testis, CALR3 recruits PDILT (protein disulfide isomerase like in testis) and assists in the quality control of ADAM3 6 . From studies of other nascent glycoproteins folding in somatic cells [45][46][47] , CALR may cooperate with PDIA3 (ERP57) to regulate the quality control of GDF9/BMP15. GDF9 and BMP15 play important roles in energy metabolism and/or cholesterol biosynthesis in granulosa cells 25,48,49 . Because TGF-beta/Smad3 signaling regulates glucose and energy homeostasis, and Smad3 KO mice exhibit improved glucose tolerance and enhanced glucose-stimulated insulin secretion from pancreatic islet beta cells 50,51 , it will be interesting to investigate the role of CALR in the secretion of TGF-beta family proteins in other tissues. That might also explain why Calr KO mice, transgenically rescued from embryonic lethality, still die mainly due to metabolic failure 32 . Of interest, it was also reported that TGF-beta stimulates cells and induces extracellular matrix secretion that depends on CALR mediated Ca 2+ signaling 52 . Thus the ability of CALR to regulate calcium availability might also be important downstream of the TGF-beta signaling pathway. In the present study, we could not fully recover GOC development with recombinant GDF9/BMP15 (Fig. 3), implying that there are other CALR dependent factors required for complete oocyte development.
In conclusion, we document that CALR plays an indispensable role in the COC development by controlling the maturation of the TGF-beta proteins GDF9 and BMP15. Because GDF9 and BMP15 are known to be involved in premature ovarian failure and polycystic ovarian syndrome 22,[53][54][55] , targeting CALR and the ER quality control system in follicular development should be considered as a novel avenue for female infertility treatment. Calr cKO were generated through the same breeding strategy as Canx cKO using Stra8-cre or Zp3-cre transgenic mice. Pr-FlpeF; 5′ -ccacctaaggtcctggttcgtcagtttgtg -3′ and pr-FlpeR; 5-atacaagtggatcgatcctaccccttgcgc -3′ primers were used for the Flpe transgene in addition to the primers indicated in Figure S1 and Figure S2.

Methods
Histology of the ovary. Female mice were injected with pregnant mare serum gonadotropin (PMSG) and human chorionic gonadotropin (hCG) at 48 hour intervals. Ovaries were collected at 10 hours after hCG injection and fixed with 4% paraformaldehyde/PBS overnight. Then the ovaries were embedded into glycol methacrylate (Technobit 8100: Heraeus Kulzer, Germany) after treatment with a graded ethanol series. Plastic thin sections (5 μ m) were stained with hematoxylin and eosin. For Periodic acid Schiff (PAS) staining, sections were rehydrated, and treated with 2% metaperiodic acid for 15 minutes, followed by treatment with Schiff 's reagent (Wako, Japan) for 20 minutes. The sections were stained with hematoxylin prior to imaging.
In vitro maturation and fertilization. For in vitro maturation 58 , immature GV (germinal vesicle) oocytes were collected from ovaries 46 hours after injection with PMSG. Antral follicles were punctured with 26G needles in FHM 59 with 100 μ M dibutyryl-cyclic AMP (Sigma). After dissociating the cumulus cells by pipetting, oocytes were washed 3 times and cultured in Minimum Essential Medium Alpha (GIBCO) with 3 mg BSA (A3311, Sigma). After 14 hours, partial zona dissection of MII oocytes was performed using a piezo-micromanipulator with a glass capillary needle (diameter of 5-10 μ m) 60 and incubated in TYH medium 61 with 2 × 10 5 /ml B6D2F1 capacitated sperm. Fertilization was determined by formation of two-cell embryos, which were then transferred into the oviduct of pseudopregnant mice.
Immunoblot and immunoprecipitation. Oocytes were collected from ovaries 46 hours after PMSG injection. Cumulus or granulosa cells were dissociated using glass pipettes and collected oocytes were boiled in SDS-PAGE sample buffer. Proteins were separated by SDS-PAGE and run under reducing conditions. Immunoprecipitation was performed as described 5 . Antibodies used included rabbit antisera against CAXN and CALR 5 , rabbit antibody against the N-terminus of CAXN (sc-11397, Santa Cruz Biotechnology), penta-His mouse monoclonal antibody (Qiagen), horseradish peroxidase conjugated goat anti-rabbit IgG and goat anti-mouse IgG antibodies (Jackson Immuno Research Laboratories); and a GDF9 monoclonal antibody 62 , kindly provided by Dr. Martin M. Matzuk.

Preparation of Calr deficient mouse embryonic fibroblasts (MEF).
MEFs were isolated from 13.5-14.5 day-old embryos. The head and all internal organs were removed from the embryo. The remaining tissue was minced with scissors and incubated in 0.25% trypsin containing penicillin (100 U/ ml, Nacalai) and streptomycin (100 μ g/ml, Nacalai) for 10-20 min. The cells were pipetted and plated onto a 10 cm tissue culture dish in culture medium [Dulbecco's Modified Eagle Medium (DMEM, GIBCO) with 10% FCS, penicillin (100 U/ml) and streptomycin (100 μ g/ml)]. The next day, the medium was replaced and cells were cultured at 37 °C until confluent, when the cells were harvested and frozen.
Production of recombinant GDF9 and BMP15. The HIV-1-based self-inactivating-type lentiviral vector plasmid pLV-EGFP was constructed by replacing the EGFP cDNA with mouse GDF9 or BMP15 cDNA, respectively, and isolated for infection of HEK293T cells 63 . In vitro follicle culture. Ovaries were collected from 12 day-old mice from which follicles were mechanically isolated prior to in vitro culture 64 . Each follicle was placed in a 30 μ l droplet of DMEM (GIBCO) with 10% FCS, penicillin (100 U/ml) and streptomycin (100 μ g/ml). The droplets were placed in 60 mm uncoated culture dishes (Iwaki) and covered with paraffin oil (Nacalai). After overnight culture, half of the culture medium was replaced by HEK293T cell-conditioned medium containing recombinant GDF9 or BMP15. Culture medium from non-transfected HEK293T cells was used as a negative control. Half of the medium was replaced every other day. Follicle volumes were calculated as described 65 .

Purification of His tagged GDF9 or BMP15 secreted into MEF cell media. Mouse Gdf9 or
Bmp15 cDNA tagged with Flag/Hisx6 at the C-terminus was inserted into a pNCAG vector consisting of the CAG-promoter and rabbit β globin poly-adenylation signal, respectively 56 . Calr +/+ or Calr −/− MEF cells were cultured in 6-well plates and transfected with pNCAG-Gdf9 Flag/His or pNCAG-Bmp15 Flag/ His using lipofectamine LTX (Invitrogen). After 18 hours, cell debris was removed by two washes in PBS (GIBCO) and the medium was replaced [DMEM with 3% FCS, penicillin (100 U/ml) and streptomycin (100 μ g/ml)]. The culture medium was collected every 24 hours for 4 days. Purification of His tagged recombinant protein in the culture medium was performed using μ MACS anti-His Isolation Kit (Miltenyi Biotec). Two milliliters of culture medium were used for purification and the proteins were eluted into 70 μ l.

Statistical analysis.
The values were the means ± standard deviation or standard error of the mean from at least three independent experiments. Statistical analyses were performed using Student's t-test. Differences were considered significant at P < 0.05.