Glucose-6-phosphate isomerase is necessary for embryo implantation in the domestic ferret

Abstract

The mechanism of implantation in carnivores is poorly understood. However, a previously unidentified 60-kDa protein has been shown to be necessary for embryo implantation in ferrets. Here we identify this protein as glucose-6-phosphate isomerase (GPI). GPI is expressed by the corpus luteum on days 6–9 of pregnancy, the time at which implantation-promoting activity has been found in corpora lutea. Passive immunization against GPI reduced the number of implantation sites in pregnant ferrets in a dose-dependent manner. GPI is a multifunctional protein. Although first identified for its role in glycolysis, GPI has since been implicated in neural growth, lymphocyte maturation, and metastasis. This study demonstrates a previously uncharacterized function of this protein that may represent the natural motility-stimulating activity that has been co-opted by tumor cells.

Embryo implantation is the first step in the establishment of pregnancy in marsupial and eutherian mammals. Despite its fundamental importance in mammalian reproduction, the process of implantation varies widely among species, as demonstrated by the rapid evolution of trophoblast proteins (1). One of the important variations is the nature of ovarian support for implantation. Mice ovariectomized after fertilization require only progesterone to retain embryos and estrogen to induce implantation (2). In mustelid carnivores, this steroid combination is not sufficient to induce embryo implantation in the absence of an ovary, as shown in the ferret, mink, ermine, spotted skunk, and European badger (3–5). Instead, a luteal protein is required during implantation in mustelid carnivores (6, 7).

Thus, a carnivore animal model is needed to characterize the luteal protein necessary for implantation in mustelid carnivores. Knowledge of reproductive mechanisms such as implantation is needed for species conservation (8), and the Mustelidae family includes several threatened or endangered species such as the black-footed ferret (Mustela nigripes). In addition, although it is poorly understood, the implantation process in other arctoidean carnivores including bears, seals, sea lions, and walruses seems to be similar to implantation in mustelids (5). We have selected the ferret as a domesticated mustelid model in which to study carnivore reproductive physiology. In ferrets ovulation is induced by mating (day 0) and occurs 30 h later. If ovulation occurs without successful fertilization, ferrets will experience a pseudopregnancy, forming corpora lutea (CL) that produce progesterone at the same concentration and for the same duration as the CL of pregnancy (9). Embryo implantation occurs on day 12 of pregnancy; total gestation length is 42 days (10).

In ferrets ovariectomized before day 8 and given progesterone, implantation is delayed (6). A similar delay of implantation occurs naturally in several mustelid, bear, and seal species (5). Injection of luteal protein homogenate triggers implantation in ovariectomized ferrets, terminating the delay (6). This implantation-triggering activity is present in ferret CL from days 6–9 of pregnancy or pseudopregnancy (11). Previous studies implicated a 60-kDa protein with an isoelectric point (pI) of 8.5, but the protein had not been identified (12). The purpose of this study was to identify the ferret luteal implantation protein.

Materials and Methods

Animals. All animal procedures were approved by the University of Illinois Animal Care Advisory Committee. Ferrets (Marshall Farms, Rose, NY) were maintained on a 16-h light/8-h dark cycle and provided with feed and water ad libitum. Determination of estrus was made by the animal supplier on the basis of vulval swelling.

Chromatofocusing Sample Preparation. Ferrets in estrus were injected by the supplier with 100 units of human chorionic gonadotrophin to induce pseudopregnancy (11). The day of injection was designated day 0 of pseudopregnancy. On day 8, ovaries were surgically removed from 24 ferrets. CL were isolated from the excised ovaries and frozen in liquid nitrogen. The frozen luteal tissue was sonicated on ice in 0.75 M Tris-acetate buffer, pH 9.3, containing pepstatin, leupeptin, and phenylmethylsulfonyl fluoride as protease inhibitors in three 10-sec bursts. Lipids were precipitated by adding 62.5 μl of 5% (wt/vol) sodium dextran sulfate per ml of homogenate followed by dropwise addition of 112.5 μl of 11.1% (wt/vol) CaCl_2. After the homogenate was incubated overnight at 4°C, it was centrifuged at 50,000 × g for 15 min. The supernatant was dialyzed in 0.75 M Tris-acetate overnight at 4°C to remove excess salts and equilibrated with chromatofocusing buffer.

Chromatofocusing Chromatography. The tissue homogenate was loaded onto a preequilibrated (0.75 M Tris-acetate, pH 9.3) chromatofocusing column (MonoP, Amersham Pharmacia) attached to a fast protein liquid chromatograph. Proteins were eluted with Polybuffer 96 (Amersham Pharmacia) at a 1:10 dilution in 0.075 M Tris-acetate, pH 6.0, at a flow rate of 0.5 ml/min for 1 h. Fractions were collected automatically at 1-min intervals. An internal pH meter and a UV detector, set to 280-nm wavelength, were used to follow the elution process. Chromatofocusing accuracy was confirmed by chromatofocusing of proteins of known pI values.

Protein Analysis. Column eluate was concentrated by using a micrometer filter with a low molecular mass cut-off. Protein concentrations then were determined by using the MicroBCA assay kit (Pierce) according to the manufacturer's microtiter plate protocol. Each sample was boiled in 5% (wt/vol) SDS for 3 min, and then 0.5 μg of protein was run in a 12.5% polyacrylamide gel in a Phast gel system (Pierce). The gel was stained in 20% (wt/vol) AgNO₃, developed in a formaldehyde buffer followed by 1% (vol/vol) acetic acid, and preserved in a 5% (vol/vol) glycerol/10% (vol/vol) acetic acid solution, and then dried for computer scanning. For further study, polybuffer was removed from the eluate by microdialysis against Tris-acetate buffer. Samples then were lyophilized for shipping to Kendrick Laboratories (Madison, WI) for SDS/PAGE separation of proteins for analysis. Proteins were eluted from gel fragments and subjected to limited protease digestion (trypsin), and 7.5% of each digest was analyzed by matrix-assisted laser desorption/ionization-mass spectrometry (Protein Chemistry Core Facility, Columbia University, New York). The remaining tryptic digest was separated by HPLC, and internal sequencing was performed on two of the fragments by Edman degradation.

RT-PCR. Total RNA was extracted from frozen day-6 CL by using TRIzol reagent (GIBCO/BRL) according to manufacturer protocol. RNA was reverse-transcribed by using the RetroScript kit (Ambion, Austin, TX) with random decamer primers. All cDNA was stored at -20°C before use for subsequent PCR. PCR primers were designed on the basis of conserved sequences of the pig (GenBank accession no. X07382), mouse (GenBank accession no. NM008155), and Drosophila (GenBank accession no. NM078939) glucose-6-phosphate isomerase (GPI) genes by using oligo 4.0 software (National Biosciences, Plymouth, MN) to identify primer annealing temperatures and avoid self-annealing or primer dimer formation. Forward primer sequences were 5′-AAGACCTTCACCACCCAGGAGACCATC-3′ and 5′-GCACCAAGATGATACCCTGTGACTT-3′ corresponding to nucleotides 664–692 and 1228–1252 of the pig GPI sequence. Reverse primers were 5′-CAGGAAGTCACAGGGTATCATCTTG-3′ and 5′-ACTGGTCAAAGCTGTTGATGTCCC-3′ corresponding to nucleotides 1232–1256 and 1549–1572. PCR was performed in a 50-μl solution containing 2.5 μl of reverse-transcription reaction product, 0.5 μl of each primer, 25 mM MgCl₂, 200 μM of each dNTP, and one TaqBead (Promega) in associated 1× reaction buffer. Reactions began with a 5-min hot start at 95°C followed by 35 cycles of 1 min at 94°C, 2 min at 48°C, and 3 min at 72°C and then 7 min at 72°C and storage at 4°C.

Cloning and Nucleotide Sequencing. PCR products were purified by using the Strataprep PCR purification kit and then cloned by using the PCR-Script Amp cloning kit (Stratagene) according to manufacturer instructions. Plasmids were isolated from bacterial pellets by using the Qiagen plasmid mini kit (Qiagen, Valencia, CA) and submitted for sequencing. Oligonucleotide synthesis and DNA sequencing were performed at the University of Illinois Biotechnology Center (W. M. Keck Center for Comparative and Functional Genomics, Urbana, IL).

Ig Preparations. Antibodies were generated in white leghorn hens against a synthetic peptide conjugated to keyhole limpet hemocyanin. Control Ig preparations were obtained from untreated hens and hens injected with keyhole limpet hemocyanin alone. The synthetic peptide (Quality Control Biochemicals, Hopkington, MA) NH₂-GDMESNGKYITKSGTRVDHQTGPILWGEPGTNGC-COO^-, based on the ferret GPI cDNA sequence, was chosen for its predicted antigenicity (Immunology Resource Center, University of Illinois Biotechnology Center).

Ig preparations were extracted from eggs according to the procedure of Polson et al. (13) with slight modification. Cheesecloth-filtered yolks were dissolved in 2 volumes of 0.1 M PBS. After the addition of 1 volume of chloroform, samples were mixed thoroughly and then centrifuged for 1 h at 1,000 × g. The aqueous phase was transferred to a new tube, and an equal volume of 24% (wt/vol) polyethylene glycol was added. After centrifugation for 1 h at 5,000 × g, the supernatant was discarded, and the pellet was resuspended in PBS to the original yolk volume. The polyethylene glycol precipitation was repeated, and the remaining pellet was lyophilized for storage.

Immunohistochemistry. Immunostaining for GPI was performed on ovaries and uteri taken from four ferrets each on days 3, 6, 9, and 12 of pregnancy. Tissues were fixed in 10% (vol/vol) neutral buffered formalin and then dehydrated overnight in an alcohol–xylene series in a Tissue-Tek processor (Sakura, Torrance, CA) and embedded in paraffin. Four-micrometer sections were placed on slides, deparaffinized, and rehydrated in a xylene–alcohol series. Slides were incubated in 0.3% (vol/vol) hydrogen peroxide in methanol to eliminate endogenous peroxide activity. After three 5-min washes in 0.01 M PBS, sections were blocked with normal goat serum for 10 min. Slides then were incubated overnight in 67 μg/ml chicken anti-GPI Ig preparation. After three PBS washes, goat anti-chicken Ig (Kirkegaard & Perry Laboratories) was applied to slides for 60 min. Slides then were stained with 0.20% (wt/vol) diaminobenzidine solution and rinsed in tap water. After counterstaining in hematoxylin for 1 min, slides were dehydrated in an alcohol–xylene series and coverslipped with permount (Sigma).

Western Blotting. Samples containing type IX rabbit GPI (Sigma), day-9 ferret CL homogenate, or day-15 ferret uterine homogenate were denatured at 95°C for 3 min in reducing sample buffer and then loaded onto 10% polyacrylamide gels and run on a Bio-Rad MiniProtean II electrophoresis unit at 150 V in 0.1% (wt/vol) SDS running buffer. Proteins were transferred to a nitrocellulose membrane in a MiniTransblot cell (Bio-Rad) in transfer buffer at 50 mA overnight or 250 mA for 1.5 h. After a brief rinse in distilled water, the blot was incubated in 5% (wt/vol) milk and 1% BSA (wt/vol) (tissue homogenates only) blocking buffer for 90 min. It then was incubated for 90 min with 10 μg/ml GPI Ig preparation in 2.5% (wt/vol) milk blocking buffer. The blot was washed three times for 15 min each in 0.1% (vol/vol) Tween 20 wash buffer. Blots were incubated with horseradish peroxidase-conjugated goat anti-chicken antibody (Kirkegaard & Perry Laboratories) diluted 1:5,000 for 60 min and then washed. Two milliliters of Pico chemiluminescence reagent (Pierce) then were applied to the blot for 2–5 min. After removal of the luminescent reagent, the blot was enclosed in plastic wrap and exposed to x-ray film for 30 sec.

Passive Immunization. Twenty ferrets in estrus were mated to two males each. This was designated pregnancy day 0. Ig preparations from two chickens, H12 and B2, were used for passive immunization. From days 5–15, ferrets were injected s.c. once daily with 2 ml of sterile saline containing (i) 210 mg of anti-GPI Ig preparation (H12), (ii) 70 mg of anti-GPI Ig preparation (B2), (iii) 70 mg of control anti-keyhole limpet hemocyanin Ig preparation, or (iv) 210 mg of control nonimmunized hen Ig preparation. Results from the control groups were combined, as no differences were observed. Animals were anesthetized and plasma samples were taken by cardiac puncture on day 9, 24 h after the previous Ig preparation injection. On day 16, animals were killed by sodium pentobarbital overdose, and uteri were examined for implantation sites and fixed in 10% (vol/vol) neutral buffered formalin for histological analysis. At this time, ovaries were examined, and only animals with viable CL were used for analysis of implantation success. Results were analyzed by using the general linear model procedure in sas (SAS Institute, Cary, NC).

Plasma GPI Activity. The total GPI enzymatic activity in plasma was determined by using the phosphohexose isomerase kit (Sigma Diagnostics) according to manufacturer instructions. Three milliliters of Tris buffer were added to each vial of phosphohexose isomerase substrate and inverted to mix. One hundred microliters of ferret plasma were added to the vial and poured into 3-ml disposable cuvettes and then read in a spectrophotometer for 7 min. The change in absorbance between 2 and 7 min (slope) was determined by using internal spectrophotometer software. All reactions were performed in duplicate. To validate the phosphohexose isomerase enzyme activity kit, the activity of a standard of known activity (enzyme control 2-E, Sigma Diagnostics) was measured accurately. Predicted activity levels of this standard were also measured when known amounts were added to ferret plasma. Serial dilutions of ferret plasma resulted in proportionally decreased activity levels. Because the assay measures a change in absorbance over time, the reaction was followed for 30 min to confirm continued linearity. Activity assay results were log-transformed for normality and homogeneity of variance and compared by two-tailed Student's t test.

Results and Discussion

Luteal Protein Identification. The ferret luteal implantation protein was isolated by chromatofocusing and identified by internal peptide sequencing. Chromatofocusing of CL from day 8 of pseudopregnancy yielded four to six major peaks between pH 8.6 and 8.0. Subsequent electrophoresis of the proteins that eluted in this range revealed at least 11 different bands on a silver-stained gel (Fig. 1). These bands included a 60-kDa protein present in fractions eluted with a pH of 8.3–8.5. The 60-kDa band was excised, subjected to trypsin digest, and analyzed by matrix-assisted laser desorption/ionization-mass spectrometry. Because the mass spectrometry results did not reveal sufficient similarity to any protein database spectra, two of the tryptic peptides were sequenced. The two peptides, KIEPELDGSSPVTSHD and VWFVSNIDGT, were found to be identical to amino acids 524–539 and 180–189, respectively, of pig (Sus scrofa) GPI and highly similar to the hamster, mouse, and human GPI sequences as well.

Fig. 1.

Silver-stained one-dimensional SDS/PAGE of chromatofocusing fractions eluting between pH 8.6 and 8.0. The right lane contains molecular mass standards (in kilodaltons), and arrows indicate the 60-kDa protein of interest.

Nucleotide Sequencing. Protein methods allowed identification of the unknown 60-kDa protein but provided limited sequence information. Thus, RT-PCR was used to obtain a partial GPI nucleotide sequence and a predicted protein sequence. The regions corresponding to nucleotides 728–1,255 and 1,264–1,572 of the pig GPI gene were PCR-amplified and sequenced (GenBank accession no. AY154742). The ferret and pig GPI sequences show 90.4% sequence identity, and ferret and rabbit GPI show 90.3% sequence identity in these regions (Fig. 2).

Fig. 2.

Partial sequence of ferret GPI cDNA and its predicted amino acid sequence, aligned with the rabbit GPI cDNA sequence. Italicized letters indicate sequence differences, and the shaded bar marks the sequence of the synthetic peptide antigen used to generate anti-GPI antibodies.

GPI Expression. Previous studies indicated that implantation-promoting activity was present in the CL removed from ferrets on days 6–9 of pregnancy or pseudopregnancy (14). Thus, we examined GPI expression in the CL and uterus during this preimplantation period. A dramatic increase in GPI immunostaining was observed in CL (Fig. 3), but only weak staining was present in the uterus during the preimplantation period (Fig. 4). In day-3 CL, little GPI immunostaining was visible. In CL from days 6 and 9, however, strong positive staining could be seen in luteal cells. Some staining occurred in day-12 CL but with less intensity and in fewer cells than on days 6 and 9. Immunostaining was GPI-specific, because preincubation of the anti-GPI Ig preparation with rabbit GPI abolished this staining.

Fig. 3.

Immunohistochemical staining of luteal tissue with anti-GPI Ig preparation (brown) and hematoxylin nuclear counterstain (purple). (a) Day 3 CL. (b) Day 6. (c) Day 9. (d) Day 12. (e) No primary. (f) Primary Ig preparation preincubated with rabbit GPI. Intense staining is present in CL from days 6 and 9 but not days 3 and 12. No staining is visible in controls (e and f).

Fig. 4.

Immunohistochemical staining of uterine tissue with anti-GPI Ig preparation (brown) and hematoxylin nuclear counterstain (purple). (a) Day 3. (b) Day 6. (c) Day 9. (d) Day 12. Light staining occurred only within uterine glands. Staining was consistent at all days.

Passive Immunization. Protein isolation and immunostaining demonstrated that GPI fits the protein characteristics and expression profile of the implantation protein. Passive immunization then was used to determine whether GPI is required for embryo implantation in the ferret. An anti-GPI Ig mixture was prepared, and its ability to bind GPI was confirmed by Western blot (Fig. 5a). Administration of 70 mg/day of the Ig preparation from days 5–16 of pregnancy reduced the number of implantation sites in pregnant ferrets to half that in control animals (Fig. 6a). This dose was not sufficient to significantly decrease plasma GPI activity levels relative to plasma activity in control animals (Fig. 6b), although this was somewhat due to variation within the control group. Thus, additional Igs were prepared to administer a higher dose (210 mg/day). Western blotting confirmed the specificity of the Ig preparation, although the presence of a faint, lower molecular mass band in both CL homogenate and a commercially available GPI preparation suggests a GPI break-down product (Fig. 5 b–d). The higher dose, which significantly decreased plasma GPI activity (p < 0.05), completely prevented implantation in two ferrets and reduced the number of implantation sites by approximately one-third in the remaining ferret (Fig. 6). Overall, passive immunization to GPI significantly decreased (P = 0.01) the number of implantation sites in pregnant ferrets in a dose-dependent manner.

Fig. 5.

Western blotting. Antibodies B2 (a) and H12 (b–d) recognize purified GPI (from rabbit muscle, Sigma) (a and b) and a 60-kDa band in ferret luteal (c) and uterine (d) tissues.

Fig. 6.

(a) Number of implantation sites in anti-GPI-treated and control animals on pregnancy day 16. Columns with different letters are significantly different (Tukey test). (b) Plasma GPI activity on pregnancy day 9 in ferrets treated with either control egg Ig preparation, 70 mg/day B2, or 210 mg/day H12 Ig preparation. ^*, significant difference from controls (Student's t test); circles, data points from individual animals; bars, standard error.

The results of this study indicate that GPI is the ferret implantation protein. GPI fits the characteristics of the previously unidentified implantation protein, is expressed by the ferret CL during the critical preimplantation period, and is required for implantation. At first glance, GPI seems an unlikely candidate for an implantation protein. It is a highly conserved cytoplasmic enzyme involved in glycolysis in all organisms from bacteria to humans (15).

However, GPI also functions as neuroleukin, maturation factor, and autocrine motility factor (AMF), a metastasis-promoting factor (16). GPI/AMF is secreted from tumor cells and acts through a receptor, gp78 (17, 18). Metastatic processes are highly similar to the invasive process of implantation, suggesting a potential mechanism for GPI during implantation (19). Several mediators of GPI/AMF function in metastasis have been implicated previously in embryo implantation. These mediators include integrins (20, 21), the lipoxygenase pathway (22, 23), cathepsin B (24), and matrix metalloproteinase 2 (21, 25). Finally, GPI/AMF promotes angiogenesis, a key step in both metastasis and implantation (26). Thus, embryo implantation may represent the natural motility-stimulating function of GPI/AMF that has been co-opted by tumor cells. Further study is needed to determine whether mediators of GPI/AMF action, such as gp78, are common to both embryo implantation and metastasis.

The receptor-mediated action of GPI/AMF during metastasis raises the possibility of a similar mechanism during implantation. In addition, the finding that GPI is expressed by the CL but not the uterus, where implantation occurs, suggests that GPI may be acting in an endocrine fashion, traveling through the circulation from the CL to the uterus. Further, injection of the Ig preparation, which can only affect extracellular GPI, prevented implantation. Additional study will be needed to test this hypothesis directly. Whatever the mechanism of action, embryo implantation represents a previously uncharacterized function of GPI. Evolution is conservative; changes in existing genes and their expression bring about new functions. There are many related hormone and hormone-receptor families. Several glycolytic enzymes serve multiple functions such as acting as lens crystallins, structural proteins in the eye (27). The enzyme glyceraldehyde-3-phosphate dehydrogenase has multiple functions including roles in membrane fusion and microtubule bundling (28). The recruitment of an existing cytoplasmic enzyme as an implantation protein represents a previously uncharacterized variation on this evolutionary theme.

Current understanding of embryo implantation in carnivores is limited. As in ferrets, a luteal protein extract triggers implantation in mink, the only other carnivore species in which it has been studied (7). Further study will be needed to determine the importance of GPI in implantation in other species. However, the similarities in implantation among the mustelids, bears, and seals and the failure of hormones such as estrogen to trigger implantation in these species (5) suggest that the role of GPI in implantation is not limited to ferrets.