Rapid, Efficient Isolation of Murine Gonadotropes and Their Use in Revealing Control of Follicle-Stimulating Hormone by Paracrine Pituitary Factors

Abstract

FSH and LH are produced only in gonadotropes, which are reported to comprise 3–12% of mammalian pituitaries. Factors made within the pituitary are powerful regulators of FSH and also influence LH expression, but their identities and cellular origins are unknown because it is impossible to isolate and individually analyze different pituitary cell types. In this study FSH-producing gonadotropes were specifically tagged in vivo with a transgenic cell surface antigen (H-2K^k) so they could be purified in vitro using paramagnetic anti-H-2K^k-microbeads. After enzymatic dispersion of pituitary cells, it took 1 h or less to extract 55 ± 5% of FSH-producing gonadotropes at 95 ± 0.5% purity, as judged by immunostaining for FSH or prolactin. Although this procedure selected for FSH expression, the isolated gonadotropes were also enriched 22-fold for LH-containing cells. For studies aimed at understanding factors that control FSH transcription, the purified gonadotropes were treated with activin A, which increased FSH expression 480% above basal levels (d 3 of culture). Coincubation of purified gonadotropes with pituitary nongonado-tropes increased FSH expression 800% (d 3 of culture). Follistatin, an activin-binding protein, decreased FSH expression 35–50%, suggesting that gonadotropes make some activin and/or other follistatin-sensitive molecule(s) that induce FSH. These data show that paracrine factors from pituitary non-gonadotropes can play a major role in controlling FSHβ at the pituitary level. The study presented here describes a rapid, reliable, and efficient method for isolating any specialized cell type, including all cells that produce endocrine hormones.

FSH AND LH ARE produced in primary pituitary cell cultures of many species (1, 2) and/or the transformed LβT2 gonadotrope cell line (3, 4). Such synthesis depends on follistatin-sensitive hormones thought to be members of the TGFβ family such as activins (5, 6), bone morphogenetic proteins (BMPs) (7), and follistatin itself (8, 9) made by cells within the pituitary. In fact, most specialized mammalian cells and cell products are partly regulated by paracrine and/or autocrine factors. To understand this type of regulation, it is necessary to separate the specialized cell from its surrounding cells. This is relatively easy if the cell type of interest has unique physical properties, but most specialized mammalian cells do not have unique physical attributes that facilitate separation. Given the plethora of cells that make specific products for critical life functions, it is important to have an approach that can be used broadly to isolate these cells for study. Such a method was devised in this study for isolating gonadotropes, but the method should be applicable to any cell type that produces a unique product.

Gonadotropes, defined by positive immunostaining for LH and/or FSH, are reported to comprise 3–12% of all pituitary cells (10–15), so up to 33-fold purification is required to obtain pure gonadotropin-producing cells. To date, enrichment strategies have worked in the rat, in which most gonadotropes are larger than other pituitary cells. Thus, differential sedimentation at unit gravity (16) was first used for purification and then refined by using greater gravitational force (elutriation) to produce high yields of gonadotropes (1 million gonadotropes from 10 million dispersed rat pituitary cells) that are 95% pure (17). Cells purified by elutriation have been used to study the effects of endocrine and paracrine factors on rat gonadotropes (18), but the procedure is time consuming and has not been extended to other mammals.

Transformed murine gonadotropes (αT3 and LβT2) provide a potential substitute for primary gonadotropes, but these cells are derived from embryonic transformants. α-T3 cells express many proteins specifically associated with gonadotropes (19), but cannot make β-subunits for either of the two mammalian gonadotropins. LβT2 cells were transformed later in development and can produce both gonadotropin β-subunits (20), but like α-T3 cells, they express the simian virus 40 large T antigen, which originally transformed them during embryogenesis and continues to stimulate their cell division. Thus, neither αT3 nor LβT2 cells can be considered fully mature or normal gonadotropes (21, 22).

A novel scheme for purifying gonadotropes recently emerged from work showing that 4.7 kb of the ovine FSHβ promoter specifically express luciferase in gonadotropes (2). In this study (2) the expression of oFSHβLuc was 98 times higher in the pituitary than in any other tissue (activity/mg protein), and it was regulated just like endogenous mouse FSH. In the study reported here, transgenic mice were produced carrying an analogous oFSHβH-2K^k transgene designed to express H-2K^k primarily in FSH-producing gonadotropes. H-2K^k is a major histocompatibility protein absent in most mouse strains used for transgenic work (www.informatics.jax.org) (23). It lacks protease-sensitive sites on its extracellular amino terminus, so it is not digested by enzymes used to disperse mammalian cells. Furthermore, the particular H-2K^k used here lacks an intracellular carboxyl terminus, so it has no intracellular signaling ability to interfere with normal cell functions. Another advantage of using H-2K^k is the recent development of a commercial technique that uses magnetic immuno-microbeads to rapidly and efficiently isolate whole cells that express H-2K^k on their cell surface.

Reported here are the details of isolating primary mouse gonadotropes that actively produce FSH and their initial use to probe the complex processes involved in regulating FSH by follistatin, TGFβ family members, and/or unknown paracrine factors produced in the pituitary.

Materials and Methods

The oFSHβH-2K^k construct

Ovine (o) FSHβH-2K^k was made by substituting the H-2K^k gene for luciferase in oFSHβLuc (2). The H-2K^k cDNA was amplified by PCR from the pMACs Kk.II expression plasmid (2472–3575 bp; Miltenyi Biotec, Inc., Auburn, CA), and XbaI and NcoI restriction sites were created on the 5′ and 3′ ends of this gene, respectively, using the PCR oligonucleotide primers: 5′-cggtaccccacTCTAGAaaccaacacac-3′ (XbaI site in capital letters) and 5′-cgcCCATGGcgatggcaccctgcatgc-3′ (NcoI site in capital letters). The PCR product was ligated into pCR-Blunt II-TOPO vector (Invitrogen Life Technologies, Inc., Carlsbad, CA), cut from it with XbaI and NcoI restriction enzymes, and then ligated into a pGL3 vector to replace the luciferase gene that was cut out using XbaI and NcoI restriction enzymes (Promega Corp., Madison, WI). Finally, the pGL3-H-2K^k construct was opened at its polylinker with KpnI and SmaI restriction enzymes, and the 4.7-kb promoter plus the first intron of the oFSH β-subunit was directionally ligated into this site, as previously reported for oFSHβLuc (2). The H-2K^k cDNA does not encode the carboxyl terminus of H-2K^k involved in signal transduction, but does contain 326 amino acids of the amino terminus, which extend from the cell surface (Miltenyi Biotec). Finally, there are no tryptic cleavage sites in the H-2K^k amino terminus, and we experimentally showed that treatment with either collagenase (type I; Sigma-Aldrich Corp., St. Louis, MO) or pancreatin (porcine pancrease; Sigma-Aldrich Corp.) did not destroy the ability of H-2K^k to act as an effective cell surface antigenic hook (unpublished observations of Dr. J. Sebastian, our laboratory).

Generation and screening of transgenic mice

The oFSHβH-2K^k fragment was linearized by digestion with BamHI and KpnI restriction nucleases, purified, and microinjected into the pronuclei of fertilized B6SJL mouse eggs as described previously (2). Genomic DNA was purified from mouse tails as reported and was tested for the presence of oFSHβH-2K^k using PCR to create a diagnostic fragment 339 bp long using the following primers specific for H-2K^k: 5′-caatagtcactggagctgtggtggc-3′ and 5′-ctcccacacctccccctgaacctgaaac-3′. Sixteen founder mice were obtained and mated with CD-1 mice (Charles River Laboratories, Raleigh, NC). Only two of the best founder lines were kept and bred to each other to obtain homozygous oFSHβH-2K^k mice or to CD-1 mice to produce large numbers of hemizygous oFSHβH-2K^k mice.

Mice carrying two transgenes (oFSHβH-2K^k and oFSHβLuc) for the expression of both H-2K^k and luciferase in their gonadotropes were obtained by breeding homozygous oFSHβH-2K^k mice with homozygous oFSHβLuc mice (founder line 7913 in Ref. 24).

Finally, mice were housed in the Biological Resource Facility of North Carolina State University. Gonadotropes that were isolated from the pituitaries of ovariectomized females were isolated 2 wk after ovariectomy. All mice were cared for according to the rules and regulations of the institutional animal care and use committee of North Carolina State University.

Purification of gonadotropes

Pituitaries from two to 21 mice were dispersed using collagenase and pancreatin as previously reported (2) and were filtered through a 27-μm pore size nylon mesh to remove undigested tissue. Cells were then collected by centrifugation (155 × g) and suspended in 100 μl freshly degassed PBS buffer (complete degassing is absolutely necessary) containing 4 mM EDTA, 0.5% BSA, and 10 μl biotin antimouse H-2K^k (BD Pharmingen, San Diego, CA). Cells were gently rotated for 10 min at 4 C, and then 2 ml PBS (EDTA/BSA) buffer were added before the cells were centrifuged again at 155 × g. Cells were incubated with 100 μl PBS (EDTA/BSA) buffer plus 20 μl antibiotin paramagnetic microbeads (Miltenyi Biotec), rotated for 15 min at 4 C, washed with 2 ml buffer as before, and suspended in 1 ml PBS (EDTA/BSA) buffer. The cells were added to and eluted from a 0.2 × 2-cm magnetic separation column (high gradient MS⁺, Miltenyi Biotec) that had been prewashed with 0.5 ml buffer according to instructions by Miltenyi Biotec. Cells flowing directly through the column while it was in the magnetic field were analyzed as flow-through cells (not magnetically attracted to the column). Cells attracted to the column were subsequently eluted after removal of the magnetic field and were analyzed as enriched gonadotropes having undergone one cycle of purification (see Fig. 1B). Cell counts for all dispersed cells were performed in triplicate using a hemocytometer. For the studies summarized in Figs. 2 and 3, cells were enriched to 95% purity using two consecutive cycles through two separate MS⁺columns (see Fig. 1C).

FIG. 1.

Gonadotrope purification from a representative preparation of dispersed pituitary cells. A, Some 3.2% ± 0.2% of freshly dispersed cells stained for FSH (top row shows FSH, green; bottom row shows nuclei, red/orange; magnification, ×125). B, About the same number of cells was analyzed as shown in A, but almost half the cells stained green for FSH (top row) after one cycle of enrichment (magnification, ×125). C, Highly enriched gonadotropes after two cycles of purification (magnification, ×500). Note that gonadotropes are large, medium, and small in size.

FIG. 2.

Expression of LH, FSH, and luciferase in unpurified cell preparations and purified gonadotrope fractions. A, Some 24,000 dispersed unpurified pituitary cells or purified gonadotropes were cultured in triplicate for 2 d, and the media were assayed for LH by RIA. B, Twenty thousand dispersed unpurified pituitary cells or purified gonadotropes were cultured in triplicate for 3 d, and the media were assayed for FSH by RIA. C, Fifteen thousand freshly dispersed unpurified pituitary cells or purified gonadotropes were cultured in triplicate for 3 d, and the cells were assayed for luciferase activity. Pituitaries in C were obtained from mice harboring the oFSHβLuc transgene.

FIG. 3.

Expression of FSHβ in 95% pure gonadotropes exposed to gonadotrope-depleted pituitary cells, follistatin, or activin. Purified gonadotropes (n = 2,000) were incubated in 384-multiwell culture plates for 2, 3, 4, or 5 d before analysis. Cultures were divided into the following groups on d 2: A, follistatin-treated (100 ng/ml); B, no treatment; and C, activin-treated (100 ng/ml). A fourth group, started on d 0, involved coincubation of 2,000 purified gonadotropes with 13,500 dispersed pituitary cells from the flow-through fraction of magnetic bead separation, which were significantly depleted of gonadotropes (labeled pituitary non-gonadotropes). Data from this fourth group are shown only in B. Because culture medium was not changed after cell plating, the FSH results (A) represent accumulated hormone from d 0. The data in B indicate luciferase activities [as relative light units (RLU)] associated with each culture well. Luciferase activities and FSH values were from the same cultures, and the data represent the mean ± SEM of three separate experiments, each performed in triplicate.

Analysis of gonadotrope purity

Dispersed pituitary cells were centrifuged onto polylysine-coated slides (10,000–20,000 cells/slide) using a Cytospin centrifuge (Shandon Southern, Pittsburgh, PA) at 1,000 rpm for 10 min, fixed for 1 h with freshly prepared 2% paraformaldehyde, washed three times with PBS (5 min each), and then stored in PBS until use. Cells were immunostained for FSH, LH, prolactin (PRL), or GH and chemically stained for nuclei. The nuclear stain (Vectashield with propidium iodide, Vector Laboratories, Inc., Burlingame, CA) was applied in the coverslip mounting solution to permit an accurate cell count on all slides (see Fig. 1, red/orange nuclear stain on the bottom row; the stain was strong enough to be visualized in the top row also). Rabbit antimouse polyclonal antibodies for FSH (H1426) or LH (H5346) were used at a 1:200 dilution for 30 min at 37 C to specifically label mouse FSH or LH (Accurate Chemical & Scientific Corp., Westbury, NY; no. A581/RH4), followed by a 30-min incubation at 25 C with a 1:40 dilution of fluorescein isothiocyanate-labeled goat antirabbit antibody (H + L) as second antibody (Zymed Laboratories, San Francisco, CA). Cells containing PRL or GH were identified using first antibodies from the NIDDK (PRL, AFP-131078; GH, AFP5641801) at dilutions of 1:20 and 1:200, respectively, followed by the second antibody noted above. Before incubation with either the first or second antibody, cells were washed with PBS and incubated for 20 min with blocking solution (10% charcoal-treated sheep serum plus 10% brain-heart infusion, BD Biosciences, Cockeysville, MD).

Gonadotropes in culture

Purified gonadotropes and/or flow-through cells were cultured in medium 199 (Invitrogen Life Technologies, Inc., Gaithersburg, MD) with 10% charcoal-treated sheep serum and antibiotics/antimycotics as previously reported (2). The data in Fig. 2 were obtained from 15,000–24,000 cells cultured in 100–200 μl medium in 96-well Primaria culture plates (BD Biosciences, Franklin Lakes, NJ). The data in Fig. 3 were obtained by incubating 2,000 cells in 80 μl medium using 384 well plates coated with poly-D-lysine (781940P) from Greiner Bio-One (Longwood, FL).

RIA for FSH and LH

The levels of FSH and LH in culture medium were measured with reagents provided by the National Pituitary and Hormone Program of the NIDDK using a double antibody method previously described (1, 2). All samples were assayed in duplicate from each medium sample obtained from triplicate culture wells; the intraassay variation was 8% or less. Culture media were collected and frozen at −20 C before RIA. For the FSH RIA, rabbit anti-oFSH antiserum (AFP-C5288113) was used as the first antibody, rat FSH (AFP-11454B) was used as iodinated tracer, and mouse FSH (AFP-5308D) was used as the reference protein. For the LH RIA, rabbit antirat LH antiserum (AFPC697071P) was used as the first antibody, rat LH (AFP-115368) was used as iodinated tracer, and mouse LH (AFP-5306A) was used as the reference preparation. The second antibody was sheep antirabbit antiserum prepared in our laboratory and used as previously reported (1, 2).

Real-time RT-PCR (RT-rtPCR)

Total mouse RNA was isolated using Tri-Reagent (Molecular Research Center, Inc., Cincinnati, OH) according to Tri-Reagent instructions. Triplicate wells of a 96-well culture plate were plated with 6000 purified gonadotropes and treated in the same way as cells shown in Fig. 3 for up to 3 d. Then media were removed, and cells were treated with 0.8 ml Tri-Reagent along with 4 μl Polacryl Carrier (Molecular Research Center). Total RNA was converted to cDNA using the iScript cDNA Synthesis kit from Bio-Rad Laboratories (Hercules, CA). The PCR probes for mouse FSHβ cDNA were 5′-AGAGAAG-GAAGAGTGCCGTTTCTG-3′ (forward) and 5′-ACATACTTTCT-GGGTATTGGGCCG-3′ (reverse), and the TaqMan probe was (6-carboxy fluorescein) 5′-ATCAATACCACTTGGTGTGCGGGCTA-3′. The internal standard was mouse 18S ribosomal RNA, which was measured as cDNA using the following oligonucleotides: 5′-GAAACTGCGAATGGCTCATTAA-3′ (forward; 966 –987 bp), 5′-GAATCACCACAGTTATCCAAGTAGGA (reverse; 1046 –1021 bp), and (6-carboxy fluorescein) 5′-ATGGTTCCTTTGGTCGCTCGCTCC-3′ (995–1018 bp). rtPCR was performed according to Bio-Rad Laboratories in the iCycler, and values, relative to the control, were calculated using the 2^−ΔΔCt method (25).

Luciferase assay

Luciferase activity was quantified on a Monolight 2010 luminometer (Analytical Luminescence Laboratory, San Diego, CA) using the Promega luciferase assay system as previously reported (2).

Statistics

Statistical calculations were performed using PRISM (version 4, GraphPad, Inc., San Diego, CA). The mean ± SEM in Fig. 1 were obtained by counting 1000 cells or more from each of three separate dispersed cell preparations. The data in Fig. 2, A–C, were obtained from three separate gonadotrope isolations. Significant differences between means were calculated using one-way ANOVA, followed by Tukey’s multiple comparison testing for Fig. 1 and rtPCR data associated with Fig. 3. Differences in Fig. 1, A–C, gonadotrope purifications were significant, with P < 0.05. A t test was used for analyzing the data in Fig. 2, where differences between unpurified cells and purified gonadotropes had a P < 0.001 in A–C.

Results

Four transgenic founder lines were shown to support significant gonadotrope purification, and one line (H-2K^k-7) was chosen for all of the studies reported here because it provided the best purification and was the most prolific. A second line gave similar purification results (H-2K^k-5). Studies with these two founder lines showed that gonadotropes were efficiently purified only from the pituitaries of males or ovariectomized females. Careful attention to cycle timing for intact females might find suitable gonadotrope purification associated with estrus, but this type of experiment has not yet been performed.

Figure 1 shows sequential pictures for the purification of dispersed male pituitary cells. All cells were stained for both FSH (green) and nuclei (red/orange). Figure 1A shows cells that were dispersed and prepared for purification, but did not go through a magnetic bead column. The percentage of gonadotropes in these unpurified cells was 3.2 ± 0.2%. This percentage is not significantly different from that found in CD-1 mice treated the same way (2.6 ± 0.1%; n = 3), showing that the expression of H-2K^k in gonadotropes had no significant effect on the number of gonadotropes in dispersed mouse pituitary preparations.

Figure 1B shows that after one cycle of purification, gonadotropes comprised 47 ± 3.5% of the cells. The percentage of gonadotropes in the flow-through fraction from the first cycle of column purification was 0.7 ± 0.06. Comparison with the 3.2 ± 0.2% population of gonadotropes entering the column indicated a 22% loss (78% recovery) of gonadotropes due to cells not binding to the column. The second column showed a flow-through loss of 16 ± 3% (84% recovery), giving a cumulative loss of 38 ± 5% for both purification cycles, which is not statistically different from the observed loss of 45 ± 5% (55 ± 5% recovery). Finally, Fig. 1C shows gonadotropes that underwent two cycles of enrichment. They were considered approximately 95% pure based on immunostaining for FSH.

The purity of gonadotropes after two cycles of isolation was difficult to quantify because small numbers of nuclei were present that had lost their cyctoplasm, and their original cell type was undefined (see white arrows in Fig. 1C). Also, occasional cells contaminated the gonadotrope fraction, but two to four cells would make a large difference in calculating purity. To obtain an independent estimate of gonadotrope purity, lactotropes were stained and counted in the original mixed cell preparation and in two-cycle enriched gonadotrope fractions. Lactotropes are small, durable, and easily stained cells. Furthermore, PRL has never been reported to coexist with FSH in male gonadotropes. The percentage of lactotropes in the unpurified dispersed pituitary cells varied from 25.6% to 35.6% in three separate preparations (mean, 31 ± 4%), but the level of lactotropes consistently dropped by 95% in each purified gonadotrope preparation (i.e. there was an average of only 1.5% lactotropes in highly purified gonadotrope fractions; data not shown). Based on the difference in lactotrope abundance in the unpurified and two-cycle enriched gonadotropes, it was calculated that the purified gonadotrope fraction contained 95.3 ± 0.5% gonadotropes.

The data in Table 1 show the number of purified gonadotropes obtained from four separate preparations of dispersed male pituitaries (two to 21 pituitaries/preparation). The average number of total dispersed pituitary cells obtained per mouse was 0.5 ± 0.04 million, and the average number of purified gonadotropes was 8800 ± 900/pituitary. Assuming that gonadotropes represented 3.2% of each preparation (see Fig. 1A), the overall recovery of enriched gonadotropes was estimated to be 55 ± 5% after two cycles of purification.

TABLE 1.

Yields of dispersed mouse pituitary cells and purified gonadotropes

Pituitaries	Dispersed cells total	FSH cells totala	Yield of FSH cells	% Yield of FSH cells
2	1,000,000	32,000	22,000	69
20	10,000,000	320,000	150,000	47
21	13,000,000	390,000	200,000	48
18	7,200,000	130,000	130,000	56
Average/pituitary	0.5 ± 0.04 × 106	16,000	8,800 ± 900	55 ± 5

^aCalculation based on data in Fig. 1, showing that 3.2% of dispersed cells contain FSH.

The purified gonadotropes (95% FSH-staining cells) were also stained for mouse LH to quantify cells containing FSH/LH or FSH only (Fig. 2A). In contrast to the intense and uniform immunostaining observed for mouse FSH (Fig. 1), LH immunostaining varied widely, from very intense to barely visible above background, which made accurate quantitation by immunostaining impossible (data not shown). As an alternative to immunostaining, RIA was used to quantify LH in the media of an equal number of purified and unpurified gonadotropes (24,000 cells/well; see Fig. 2A). The data in Fig. 2A indicate that purified gonadotropes contained 22 times more LH per cell than unpurified cells (64 ± 5.2 vs. 2.8 ± 0.2), indicating a 22-fold enrichment of LH-containing cells in the gonadotrope fraction.

In Fig. 2B, mouse FSH was quantified by RIA in cultures of unpurified and purified gonadotropes (20,000 cells/well) that were cultured for 3 d in a 96-well plate. The FSH produced by 20,000 purified gonadotropes was 75 times greater than that produced by 20,000 unpurified cells cultured under identical conditions (111 ± 18 and 1.5 ± 0.6 ng/well, respectively). These results are from one experiment, but are typical of three separate experiments.

In a third experiment, 15,000 purified gonadotropes were cultured as described above, and the expression of the oFSHβLuc transgene was measured. In this study luciferase activity was 74 times greater in cultures of purified gonadotropes than in unpurified cells (Fig. 2C). That is, highly enriched gonadotropes expressed 84 ± 7 × 10³ relative light units, whereas unpurified cells expressed only 1.1 ± 0.06 × 10³ relative light units. The data in Fig. 2, B and C, show that the expression of both endogenous mouse FSH and oFSHβLuc copurified to the same extent in the enriched gonadotropes.

The effects of follistatin, activin, and unidentified pituitary paracrine factors were tested on FSH expression in purified gonadotropes (Fig. 3). Double-purified gonadotropes that expressed oFSHβLuc along with endogenous mouse FSH were incubated for 5 d in 384-well culture plates (2000 cells/well). Untreated cultures expressed relatively high levels of FSH (Fig. 3A) and luciferase (Fig. 3B) in the absence of any added hormones. Addition of follistatin on d 2 partially decreased both FSH secretion and luciferase activity by approximately 35% (on d 5). By contrast, activin treatment on d 2 increased the rate of FSH accumulation and luciferase activity by 480% on d 3 compared with those after control or follistatin treatment. The data in Fig. 3, A and B, came from the same cultures, but appear different because FSH is an accumulated value (accumulation of mouse endogenous FSH from d 0), whereas luciferase activity represents steady state luciferase expression at the time of analysis.

To determine the effect of paracrine factors on FSH expression, 13,500 pituitary cells from the flow-through fraction of purification were coincubated with 2,000 purified gonadotropes from d 0 of culture. After correcting for residual luciferase activity in the flow-through fraction, it was found that the 2,000 purified gonadotropes expressed 8 times more oFSHβLuc than cultures treated with follistatin. By d 5 of culture, however, luciferase activity dropped dramatically to a level somewhere between that of control and follistatin-treated cultures (Fig. 3B). These data show that paracrine factors that act like activins can play a dominant role in inducing the expression of FSH, but other factors, such as follistatin or other inhibitory substances, are eventually made by nongonadotropes that cause a precipitous fall in FSHβ expression by d 5 (Fig. 3B).

To confirm that the oFSHβLuc activity shown in Fig. 3B accurately reflected steady state levels of endogenous mouse FSHβ mRNA, rtPCR was used to quantify mouse FSHβ mRNA. Total RNA was isolated from two independent gonadotrope cultures on d 3 that were treated as described in Fig. 3B. Ribosomal subunit 18S was used as the internal control, which did not change with treatment and had an average threshold cycle value of about 15. Based on changes in threshold cycle values for mouse FSHβ mRNA (ct), it was determined that activin significantly increased mouse FSHβ mRNA by 5- or 6.9-fold in the two preparations (P < 0.01), whereas follistatin decreased mouse FSHβ mRNA by 49% (P < 0.05).

Finally, the coexistence of GH and FSH in double-purified gonadotropes was tested by immunostaining highly enriched gonadotropes for GH. Cells that stained for GH in unpurified pituitary cells comprised approximately 39%. Therefore, if gonadotropes were totally devoid of GH, there should have been about 1.9% of GH cells in the highly purified gonadotrope preparation. There were, in fact, 12 ± 2% of highly enriched gonadotropes that stained for GH, indicating that approximately 10% ± 2% of the male mouse gonadotropes contained both FSH and GH.

Discussion

As reviewed in the introduction, it is important to study gonadotrope regulation by paracrine and/or autocrine factors made in the pituitary, but this cannot be done effectively without separating gonadotropes from their surrounding cells. Separating gonadotropes has been very difficult in the past, and this report describes a rapid, efficient, highly reliable, and relatively inexpensive method for isolating primary gonadotropes that may be of considerable use in the future not only for gonadotropes, but for any cell type that expresses a unique gene product. The only criterion that discriminates between cells that are isolated and those left behind (flow-through) is the appearance of H-2K^k on the gonadotrope cell membrane that is proportional to FSHβ expression in the cell. Therefore, this new method selects specifically for gonadotropes that are actively producing FSH. Because LH is often made along with FSH (≥85% of gonadotropes are usually FSH/LH cells), a majority of LH-containing cells would also be expected to be isolated, and there was, in fact, a 22-fold enrichment for LH in the gonadotrope fraction. Some LH-producing cells do not make FSH (5–15% in other species) (10–15), so it is also likely that some LH-producing cells were not isolated in the gonadotrope fraction. Nevertheless, 22-fold enrichment predicts 100% pure LH cells if LH cells originally comprised just 4.5% of the pituitary cells, suggesting that LH enrichment was quite high in the gonadotrope fraction.

It was surprising to find, however, that the overall percentage of gonadotropes in the male mouse pituitary was only 3.2%, because 10% or more of rat pituitary cells have been identified as gonadotropes (17). The low percentage might result from selective destruction of a majority of mouse gonadotropes during the dispersion process, but this should lead to widely varying percentages of gonadotropes in different mouse pituitary preparations, which was not observed. A more likely reason for low gonadotrope numbers is that a majority of mouse gonadotropes might be prescient gonadotropes that carry GnRH receptors but produce very little FSH or LH and are not easily detected by immunostaining or RIA. Of course, it may be that the mouse can function with fewer gonadotropes than other mammals. Immunostaining and quantitative analysis of whole mouse pituitary tissue will eventually resolve this issue.

The purification procedure described here yielded 95% pure primary FSH-containing gonadotropes at 55 ± 5% yield (~8,800 gonadotropes/mouse pituitary). As noted in Results, there was an average loss of 22 ± 2% in the flow-through fraction for each cycle of the two-cycle purification. This accounts entirely for the losses that result in the observed 55% overall yield. During manuscript preparation, the length of collagenase incubation was increased to 2 h from 1.5 h to produce an average of 1,000,000 pituitary cells/mouse, which is a 2-fold increase over that shown in Table 1. Altering column parameters to capture more gonadotropes also has a potential for doubling the yield, which could produce up to 35,000 gonadotropes/mouse pituitary.

The purified gonadotropes were easily cultured and analyzed for FSHβ expression by measuring mouse FSH mRNA (rtPCR), mouse FSH (RIA), or oFSHβLuc activity that reflected expression from 4.7 kb of the ovine FSHβ promoter. As few as 2000 highly enriched gonadotropes were cultured per well in a 384-multiwell culture plate, meaning that 11 mouse pituitaries were required to obtain data from 50 individual singlet treatments. Luciferase activity was particularly useful because it provided a rapid assay that reflected steady state FSHβ mRNA concentrations (measured by real-time rtPCR) at any moment in time and is often used as an estimate for transcriptional activity. An entire luciferase assay (60 wells) takes less than 0.5 h to complete, which makes it 100–200 times faster than either real-time rtPCR or RIA. The FSH RIA measured accumulation of FSH over time and was valuable because it monitored a combination of processes that encompassed transcription, mRNA stability, translation, protein processing, and secretion. Because data from the FSH RIA and rtPCR were similar, it was concluded that follistatin or activin only altered FSHβ mRNA levels and not mRNA translation, protein processing, or secretion.

Because only 2000 cells were required per well in a 384-well plate, a clear advantage of this experimental model is its use of only 30–80 μl culture medium/treatment. This means that the consumption of hormones during treatments is relatively small (activin, follistatin, BMPs, inhibin, GnRH, and others). This is important, because commercially available recombinant follistatin and TGFβ family members are costly. Even more important, the bioassay provided by 95% pure primary gonadotropes may permit the rapid screening of minute quantities of unknown paracrine factors that alter FSH expression.

It is instructive to compare FSH production from transformed LβT2 gonadotropes and the purified primary gonadotropes reported here. Purified primary gonadotropes, without activin treatment, produced 2.8–3.7 ng FSH/d·2000 cells. This rate of FSH production is 1000 times greater than that found for an equivalent number of LβT2 cells in culture without activin (3, 4). Even with activin treatment, LβT2 cells produce only 0.007 times the amount of FSH produced by an equivalent number of primary pituitary gonadotropes treated with activin. One cause of such a large difference might be the rapid rate of LβT2 cell division; perhaps only a small number of LβT2 cells are in G₁ producing FSH at any given time. A second possibility is that LβT2 cells may benefit from factors that primary gonadotropes are exposed to in vivo (paracrine or endocrine), which they “remember” in tissue culture. Finally, it is possible that LβT2 cells produce less FSH because the large T antigen interferes with important cellular functions (21, 22). Comparative studies between LβT2 cells and pure primary gonadotropes should help determine which of the above possibilities is correct.

Follistatin inhibits FSH expression by binding and incapacitating activin or other TGFβ family members that may stimulate FSH expression (26, 27) and are sensitive to follistatin. Because FSH is inhibited 35–50% after treatment with follistatin in both purified primary gonadotropes and LβT2 cells (4), it seems likely that both cell types produce significant amounts of either activin- or follistatin-sensitive TGFβ family member(s) as autocrine factors that stimulate FSH expression. It is worth noting in Fig. 3B that purified gonadotropes naturally increase luciferase expression from d 2–5, suggesting that they produce autocrine substances that may or may not be TGFβ family members because follistatin blocks only some of the increase. These data are consistent with studies that have identified activin and/or BMP subunit mRNAs in LβT2 cells, pituitary cultures, or primary gonadotropes (4, 26–28), but the relatively low level of FSH induction compared with FSHβ induction by paracrine factors from nongonadotropes (see discussion below) suggests that these autocrine factors may not be the primary inducers of FSH in the pituitary.

Activin treatment caused a significant increase in FSH expression (480%) as expected, but exposure to pituitary nongonadotropes increased FSH expression even further (800%) compared with expression in the presence of follistatin on d 3 of culture. These data suggest that the expression of FSH is stimulated most dramatically by paracrine factors from nongonadotropes. It is not known whether these factors are TGFβ family members like activin. They may be general growth factors that stimulate gonadotropes to produce TGFβ family members, but the answer to this possibility awaits further study. Interestingly, it is clear in Fig. 3 that basal expression of FSH increased even when follistatin was present, so gonadotropes may produce autocrine stimulators of FSH production that are not TGFβ family members.

Coculture of purified gonadotropes with nongonado-tropes on d 5 expressed much less FSH than on d 3. This was not due to down-regulation by activin action, because constant stimulation by activin alone produced a continuous high level of oFSHβluc expression, but it could have been due to synthesis of follistatin (29) or down-regulation by factors other than activins. It is not yet clear what pituitary-made paracrine and/or autocrine factors are most influential in regulating FSH, but the use of primary gonadotropes to detect these factors has the potential for isolating and identifying the molecules that are so important in regulating FSH, follicular development, and fertility in mammals.

For years it has been known that GH and FSH can coexist in the same pituitary cell in female rats. Furthermore, there is significant evidence showing that a portion of GH cells can cycle in and out of producing gonadotropins and supplement gonadotrope populations when required (30). This indicates significant plasticity in the GH and gonadotrope cell types, which may now be studied in tissue culture in the absence of large numbers of GH-only cells. Because GH expression has not been reported for LβT2 cells, purified primary gonadotropes represent a unique opportunity for studying hormones that control the transitions between somatotropes and gonadotropes. The abundance of GH/FSH-containing cells has not yet been determined for gonadectomized animals or females during estrous; the numbers of these dual-secreting cells may be significantly greater in these populations, which would make it even easier to study the transitions from somatotrope to somato-gonadotrope to gonadotrope.

Finally, the procedures used in this study can be applied to numerous mammalian cell types that express unique proteins. These cell types include the majority of pituitary cells as well as pancreatic, kidney, adrenal, ovary, testicular, mammary, pineal, ocular, cochlear, neural, and immunological cells; all endocrine cells; and many other cell types. Most of these cell types are likely to be heavily influenced by paracrine and/or autocrine hormones, making it essential to study them in isolation from surrounding cells to understand their most critical regulation. The methods presented here can be used to isolate any of these cell types as long as a promoter can be found to express H-2K^k specifically in the cell type of interest. Even the use of luciferase transgenes can be applied to all cell types for studies in many fields of endocrinology once a suitable promoter is identified.

In summary, an inexpensive, rapid, efficient, and reproducible method for isolating 95% pure primary gonadotropes is reported. The method is broadly applicable to all specialized cell types. Production of FSH and expression of FSHβ, as measured by expression of an oFSHβLuc transgene, were easily measured using as few as 2000 purified gonadotropes. Production of FSH in purified primary gonadotropes was induced 1) about 35% by follistatin-sensitive factors thought to be activin-like hormones produced within gonadotropes; 2) 480% by recombinant activin A; or 3) 800% by factors from pituitary nongonadotropes, which might play a dominant role in FSH expression in vivo. Finally, about 10% of purified primary gonadotropes contained both GH and FSH. These cells offer a model for studying the factors that interconvert somatotropes and somato-gonadotropes.