|Home | About | Journals | Submit | Contact Us | Français|
Pelvic inflammatory disease and metritis are important causes of infertility in humans and domestic animals. Uterine infection with Escherichia coli in cattle is associated with reduced ovarian follicle growth and decreased estradiol secretion. We hypothesized that this effect could be mediated by the bacterial lipopolysaccharide (LPS) or cytokines such as tumor necrosis factor alpha (TNFα).
In vitro, bovine ovarian theca and granulosa cells were treated with LPS or TNFα and steroid secretion measured. In vivo, the effect of LPS or TNFα intrauterine infusion was determined by ovarian ultrasonography and measurement of hormones in cattle.
LPS reduced granulosa cell estradiol secretion, whilst TNFα decreased theca and granulosa cell androstenedione and estradiol production, respectively. In vivo, fewer animals ovulated following intrauterine infusion with LPS or TNFα.
LPS and TNFα suppress ovarian cell function, supporting the concept that pelvic inflammatory disease and metritis are detrimental for bovine ovarian health.
Pelvic inflammatory disease (PID) and metritis are important causes of serious disease and infertility in humans and domestic animals. Each year in the United States, more than 1 million women experience an episode of acute PID, more than 100,000 become infertile as a result and more than 150 women die from PID or its complications1. Metritis affects 40% of dairy cattle after parturition, with each case costing an estimated $250 for reduced milk production, delayed conception and treatment2. Much of the infertility following a case of PID or metritis is associated with damage to the genital tract and reduced embryo survival3,4,5. However, there is increasing evidence that uterine disease also affects ovarian function.
Metritis is associated with slower growth of the dominant follicle in the ovary, fewer ovulations and lower peripheral plasma estradiol concentrations compared with clinically normal cattle6,7,8. The most numerous pathogenic bacteria in the bovine uterus is Escherichia coli and its presence is specifically associated with ovarian dysfunction8. The effects of E. coli are likely mediated directly through the endotoxin, lipopolysaccharide (LPS), or indirectly through the inflammatory mediators associated with E. coli infection including cytokines such as tumor necrosis factor alpha (TNFα)9. Indeed, there are increased concentrations of LPS and TNFα in the peripheral plasma of animals with uterine infection6,8,10,11. More importantly, LPS concentrations are increased in the ovarian follicular fluid of animals with uterine disease12.
Studies exploring the effect of LPS on reproductive biology in the whole animal have focused on suppression of GnRH and LH from the hypothalamus and pituitary, respectively, rather than on ovarian follicle function13,14,15. However in sheep, there is evidence that LPS was associated with reduced estradiol secretion independently of LH pulse secretion16. There is in vitro evidence that theca and granulosa cell function may be perturbed by LPS in the rat17. Ovarian granulosa cells express the innate immune receptor complex for detection of LPS, and treatment with LPS modulates their endocrine function12. Alternatively, cytokines associated with uterine inflammation may affect ovarian function as they appear to suppress ovarian cell steroidogenesis, although serum-free culture methods were not always used in previous experiments18,19.
In the present study, cattle have been used to investigate the effect of uterine disease on ovarian function because the disease is biologically relevant and granulosa cells can be isolated free of immune cell contamination12. Furthermore, unlike humans, ovarian tissue is readily available from normal animals post-mortem and intervention studies can readily be preformed in vivo20,21. We use pure populations of ovarian cells in vitro and uterine infusion in vivo to test the hypothesis that LPS directly, or indirectly via TNFα, perturbs ovarian function.
Granulosa cells were obtained and cultured separately in serum-free medium as previously described12,22. Briefly, bovine ovaries were collected at a local abattoir immediately post-mortem and returned to the laboratory within 1 h. Follicles were isolated manually by dissection and selected for isolation of cells if they had a translucent appearance, a well vascularised theca and clear follicular fluid with no visible debris or blood. Follicles were measured using a grid or calipers and classed by external diameter as small (< 4 mm diameter), medium (4-8 mm diameter) or large (> 8 mm diameter), reflecting their gonadotropin dependence and changes in the expression of steroidogenic enzymes and LH receptors23,24. At 4 mm diameter, follicles are recruited into follicle waves in cattle and become responsive to FSH, with increased expression of aromatase25. From 8 mm diameter, granulosa cells express LH receptors and these selected dominant follicles require pulsatile LH stimulation to continue growing25. Follicles were cut in half and granulosa cells obtained by flushing the hemisected shells and collecting the cell-rich supernatant22. Theca cells were then obtained by manually peeling the basal lamina from the hemisected follicular shells and digesting the tissue in M199 (Sigma-Aldrich) containing 1 mg/ml collagenase (Sigma-Aldrich) and 100 μg/ml trypsin inhibitor, in a moving water bath for 45 minutes at 37°C and collecting the cell rich supernatant26.
As determined by Trypan Blue exclusion, granulosa and theca cells were > 80% viable and were seeded separately in 96 well plates (Nunc) at 1.5 × 106 cells/ml. Cells were incubated at 37°C in a 5% CO2 atmosphere in serum-free medium, with granulosa cell medium supplemented with 10-7 M androstenedione22.
After an initial 48 h establishment period, cell culture media was removed and replaced with fresh media containing treatments. In the first experiment, cells were treated with media containing either 0, 0.1, 1 or 10 μg/ml LPS (Sigma: E.coli serotype 055.B5). These concentrations are similar to the concentrations of LPS in the follicular fluid of animals with clinical disease, and are also similar to concentrations of LPS used in previous studies to explore immune cell function12,27,28,29. In a further experiment, cells were treated with 0, 1, 10 or 100 ng/ml TNFα as these concentrations have been shown to affect steroidogenesis18. The media were changed after 48 h and then collected at 96 h to identify any longer term effects on ovarian cell steroidogenesis, reflecting the whole animal clinical context7,8. At the end of the incubation period, the number of viable cells was determined by neutral red dye uptake, as previously described30.
Culture supernatants were analyzed by radioimmunoassay as previously described31, adapted for estradiol or androstenedione. Samples were diluted in 0.05 M Tris buffer containing 0.1% gelatin and 0.01% sodium azide. Standards, antiserum and tritiated tracer were purchased from Sigma (Sigma, UK), Biogenesis (Biogenesis, UK) and Amersham International PLC (Amersham, UK), respectively. The limit of detection for estradiol was 80 pg/ml. The respective intra- and inter-assay coefficients of variation were 8.8% and 9.9% for estradiol and 3.1% and 12.6% for androstenedione.
Eight nulliparous post-pubertal Holstein heifers aged between 13 and 15 months at the start of the study were assigned to control or treatment groups in a randomized crossover design (n = 8 per treatment). Animals were housed in a straw-bedded yard in a naturally ventilated shed and fed a diet formulated according to standard guidelines32. The diet comprised of ad libitum grass hay and concentrates (Growergrain Nuts, BOCM Pauls Ltd, Ipswich) and the animals had ad libitum access to water. To limit any influence of infectious disease, the heifers were tested free of BVD, Leptospirosis, BHV-1, Tuberculosis, Brucella and EBL before experiments began. Maiden heifers were chosen for the study as they were predicted to have a sterile uterine environment, which was confirmed by regular uterine bacterial swabs, collected as previously described33. No bacteria were isolated from any of the animals at any point during the study. In addition, before the study began, the heifers were clinically assessed to ensure normal reproductive anatomy and ovarian function, and animals were acclimatized to the handling and housing facilities for 3 weeks. All procedures were carried out under Home Office authorization in compliance with the Animals (Scientific Procedures) Act 1986, and experimental protocols were approved by the Royal Veterinary College Ethical Review Committee.
The experimental protocol was as previously described34. Briefly, estrus cycles were synchronized in the heifers by two intramuscular injections of a prostaglandin F2α analogue (cloprostenol, Estrumate, Scherring Plough Animal Health) 11 days apart (PG1 and PG2). For 3 days following the PG2 injection, animals were observed for 30 min every 2 h to detect estrus, as determined by the first time an animal stood to be mounted. On the morning of the seventh day after first observed standing oestrus, each heifer was given a third injection of prostaglandin F2α analogue (PG3) to induce luteal regression and permit ovulation of the expected dominant follicle.
In the first experiment investigating the effects of LPS, animals were randomly administered a control infusion of 10 ml sterile PBS or 10 ml sterile PBS containing 3.0 μg/kg E. coli serotype 055:B5 LPS (Sigma-Aldrich). In the second experiment investigating the effects of TNFα, animals were randomly administered a control infusion of 10 ml sterile PBS, or 10 ml sterile PBS containing 0.1 μg rhTNFα (Sigma-Aldrich). Treatments were administered in a crossover design such that each animal received both control and LPS treatments in the first experiment, followed by control and TNFα treatments in the second experiment, with a recovery period of one spontaneous estrous cycle between treatment periods to minimize carry-over effects from the previous treatment.
Infusions began 24 h after first observed standing estrus following PG2 and were administered every 6 h for 9 days. Infusions were carried out by passing a sterile, disposable bovine uterine catheter (Arnolds, UK) through the cervix and into the uterine lumen guided by transrectal palpation. The treatment was drawn into a sterile 10 ml syringe (BD) which was then attached to the end of the catheter and the contents infused into the uterus.
Reproductive function was monitored as previously described7,8. Briefly, the genital tract of each cow was examined by transrectal palpation and ultrasonography using a 7.5 MHz linear array transducer (Honda HS-2000; Honda Electronics, Aichi, Japan). Ovarian follicles ≥ 4 mm in diameter and corpora lutea in each ovary were identified, and the maximum diameter of each structure was estimated by averaging measurements made using the instrument’s internal calipers in two dimensions at 90 degrees7,35. A follicle was defined as a non-echogenic structure with a clear demarcation between the follicular wall and the antrum; and, the day of dominance was defined as the day the largest follicle on the ovary achieved an internal diameter of ≥ 8.5 mm which corresponds to the onset of deviation36. The day of ovulation was classed as the last day a dominant follicle was scanned before it disappeared and a CL subsequently formed in its place35. A CL was characterized by a grainy echogenic structure with a well-defined border within the ovarian stroma, often with a non-echogenic lacuna.
General health was monitored every 6 h throughout the study period, including measurement of rectal temperature using a standard digital thermometer (National Veterinary Supplies, Stoke-on-Trent, UK). Body weight was estimated using a proprietary weigh-band (Ascott Ltd, Powys, UK) placed around the chest and shoulder area. Body condition was scored by visual assessment and palpation of areas of each animals body and assigning a score based on the fat distribution37.
Blood samples were collected at regular intervals from the jugular vein. Blood samples were collected following PG2 every 3 h for 6 days, every 12 h for five days and finally every 3 h until the end of the study. In the LPS experiment only, blood samples were collected via jugular catheter every 12 minutes during an 8 h time period. This allowed measurement of pulsatile LH release from the pituitary as there is evidence that LPS can suppress LH pulsatility13.
Plasma concentrations of estradiol were analyzed using the Estradiol MAIA radioimmunoassay kit (Biogenesis, Poole, UK) by the method previously described by Prendiville et al38, with some modifications as described by Williams et al8. The intra- and inter- assay coefficients of variation were 20.8 and 21.6 %, respectively. Plasma concentrations of progesterone were measured in duplicate using a commercial ELISA kit (Ridgeway Science, Gloucester, UK) following the manufacturers guidelines. The intra- and inter- assay coefficients of variation were 2.7 and 12.2%, respectively.
The concentrations of LH and FSH were measured in duplicate in peripheral plasma as described previously39. Standards were purchased from the National Hormone and Pituitary Programme (NHPP, California, USA) and iodinated tracers were purchased from Amersham International PLC (Amersham, UK). For FSH the internal recovery was 95%, the intra-assay coefficient of variation was 14.0% and the inter-assay coefficient of variation was 10.5%. For LH, internal recovery was 95%, the intra-assay coefficient of variation was 14.0% and the inter-assay coefficient of variation was 10.4%. LH concentrations were analyzed using the computer algorithm PULSAR determine the mean, maximum and minimum LH concentrations, number of pulses, mean pulse amplitude, mean pulse length, mean frequency and mean inter-pulse interval40. The onset of LH peak was defined as the time at which LH concentration exceeded 5 ng/ml.
Concentrations of the acute phase protein α1-acid glycoprotein were measured using a previously described method adapted for 96-well plates (Life Technologies, Invitrogen, UK)41,42. The intra-assay and inter-assay coefficients of variation were 12 and 18%, respectively.
All data were analyzed using the statistics program SAS 9.143. Results are reported as the arithmetic mean ± SEM, and significance ascribed when P <0.05. The effects of treatments in vitro were explored using Mixed Model ANOVA as described previously7. In vivo data were normalized to first day of uterine infusion, which was defined as day 0. Data from control and treated animals were compared by ANOVA and post hoc tests performed with Dunnet’s adjustment, except for proportions, which were compared using the Fisher exact test.
Treatment with LPS did not affect cell numbers or androstenedione production by theca cells isolated from small, medium or large ovarian follicles (Fig 1a). However, LPS treatment decreased estradiol production by granulosa cells isolated from small (P < 0.001: mixed model analysis), medium (P < 0.001) or large (P < 0.001) ovarian follicles (Fig 1b).
Treatment with TNFα reduced androstenedione production by theca cells isolated from all follicle sizes despite increasing the number of cells in some cases (P< 0.001; Fig 2a). In addition, TNFα suppressed estradiol production by granulosa cells isolated from small (P < 0.01), medium (P < 0.001) and large (P < 0.01) ovarian follicles supplied with 10-7M androstenedione. Granulosa cell numbers were not affected by TNFα (Fig 2b).
In the LPS and TNFα studies, rectal temperatures were consistently within the normal range of 38.0 to 38.5°C during each study period, and peripheral concentrations of the acute phase protein AGP were not different between treatments. In the LPS study, AGP concentrations ranged from 0.07 – 2.25 mg/ml and 0.04 – 2.28 mg/ml, for control and treated animals, respectively. In the TNFα study, AGP concentrations ranged from 0.30 – 4.35 mg/ml and 0.33 – 4.34 mg/ml, for control and treated animals, respectively. One animal was removed prior to the TNFα study because of an unrelated clinical disease.
All animals had an active corpus luteum at the time of induction of luteolysis (PG2), as determined by ultrasonography and peripheral plasma progesterone concentrations > 1 ng/ml, displayed standing estrus and the dominant follicle ovulated. An LH surge was observed at the time of first standing oestrus ± 6 h and there was no significant difference between treatment groups. Similarly, FSH concentrations did not differ between treatment groups (Fig. 3).
Following estrus a new wave of ovarian follicles emerged in all animals and dominant follicle diameter increased over time (P < 0.001). No significant differences were observed in dominant follicle diameter or oestradiol concentrations between treated and control animals (Fig 4). Plasma progesterone concentrations increased over time (P < 0.001) concomitant with formation of the CL after estrus in all animals (Fig 5ai). Progesterone concentrations were lower in animals treated with LPS between days 3 and 9 (P < 0.05: mixed model analysis). A prostaglandin F2α analogue injection was administered on day 6 to permit ovulation of the new dominant follicle (PG3), but fewer LPS treated animals did ovulate (3/8 vs. 7/8, P < 0.05).
Following PG3, pulsatile LH secretion was observed in all animals and no difference in mean, maximum or minimum LH concentrations was observed between LPS or control treatments. The number of peaks observed was the same for each treatment and there was no difference in peak amplitude, peak frequency, peak length or inter-peak interval between treatments (Table 1). Representative patterns of pulsatile LH secretion during the control or LPS infusion are shown in Figure 6.
Following estrus a new wave of ovarian follicles emerged in all animals and as expected dominant follicle diameter increased over time (P < 0.001). Dominant follicle diameter and plasma estradiol concentrations did not differ significantly between treatments (Fig. 4). Plasma progesterone concentrations increased over time (P < 0.001) concomitant with formation of the CL after estrus in all animals, but did not differ significantly between treatments (Fig. 5). A prostaglandin F2α analogue injection was administered on day 6 to permit ovulation of the new dominant follicle (PG3), but fewer animals ovulated when treated with TNFα as compared with control (4/7 vs. 6/7, P = 0.09).
Following PG3, an LH surge was observed in all animals at the time of estrus ± 9.1 h (corresponding to 41.4 ± 1.7 h after PG3); LH concentration was 14.1 ± 1.7 ng/ml with no significant difference between control and TNFα treatment.
Animals with metritis, particularly those associated with E. coli infection, have reduced ovarian follicle growth and function and are less likely to ovulate7,8,44. These effects could be mediated directly by E. coli LPS or indirectly by cytokines such as TNFα in response to uterine infection. In the present study, ovarian cell steroidogenesis was inhibited by treatment with LPS or TNFα in vitro. Granulosa cell estradiol secretion was reduced by LPS treatment, whilst TNFα reduced theca and granulosa cell androstenedione and estradiol production, respectively. The effects of LPS or TNFα on ovarian function were more subtle in vivo when infused into animals during a defined period of ovarian follicle and corpus luteum development. However, the dominant follicle formed during LPS or TNFα treatment was less likely to ovulate than in control animals. Taken together these studies support the hypothesis that LPS or TNFα can directly impair ovarian function. Thus, it is important for clinicians treating patients with pelvic inflammatory disease or metritis, to be aware that uterine infection can affect ovarian as well as uterine health.
Microbial infection of the uterus is a common and costly cause of disease and infertility in humans and domestic animals. Dairy cows appear to have a particular propensity for uterine disease after parturition, with up to 40% of animals affected2. Uterine disease and other systemic diseases are associated with perturbation of ovarian follicle function in cattle7,45,46. In metritis, the most important pathogen is E. coli8,10. Many of the effects of E. coli in disease processes are mediated via the endotoxin, LPS, or by cytokines such as TNFα associated with the inflammation that occurs during infection9. These products are found in the peripheral plasma and ovarian follicular fluid during uterine infection6,8,10,12.
Ovarian follicle growth and function are characterized by secretion of estradiol which is produced in the ovary by granulosa cell aromatisation of theca-derived androstenedione24. Estradiol regulates two positive feedback loops to maintain the dominant follicle and induce ovulation. At ovarian level, oestradiol enhances gonadotropic induction of LH receptors and more FSH receptors in granulosa cells and in synergy with FSH, increases its own production by stimulating aromatase activity and the expression of its own receptors in granulosa and theca cells47,48,49,50. At pituitary level, estradiol secreted by the dominant follicle feeds back in the absence of progesterone to enhance gonadotropin secretion, thus ensuring the preovulatory LH surge which induces ovulation51.
The cell surface receptor that mediates most of the effects of LPS is Toll-like receptor 4 (TLR4) in association with CD14 and MD-29,27, and this receptor complex is present in the ovary of mice and cattle12,52. In the present study, estradiol production by granulosa cells in an immune cell-free culture was inhibited in response to LPS, despite FSH and androstenedione concentrations remaining constant. Furthermore, in the theca cell cultures where LH remained constant, androstenedione production did not change in response to LPS. Together these results indicate that rather than affecting pituitary FSH production or secretion, LPS reduces the ability of granulosa cells to respond to FSH or to aromatise androstenedione to oestradiol. Similarly, in the rat, LPS decreases the LH-induced aromatisation of androgens to oestrogen resulting in an inhibition of oestradiol production53. A dose dependant decrease in LH receptor formation was observed in rat granulosa cells treated with LPS. It could, therefore, be postulated that LPS reduces LH receptors on granulosa cells thus inhibiting the response to LH and thereby blocking ovulation. This may explain why, in the present study, animals treated with LPS failed to ovulate in the presence of normal LH concentrations.
It is well established that granulosa cells have receptors for TNFα54,55,56. Estradiol secretion is suppressed following treatment of granulosa cells with TNFα in the rat and human53,54; and the expression of 17βHSD and P450arom mRNA is decreased57. In the cow, FSH-induced estradiol production in granulosa cells from small follicles, and LH-induced androstenedione production by theca cells from large follicles, are both inhibited following treatment with TNFα19. In the present study, TNFα treatment also inhibited FSH-induced estradiol production by bovine granulosa cells from small, medium and large follicles. Furthermore, LH-induced androstenedione production by cells from small, medium and large follicles, was also inhibited. This inhibition of ovarian cell steroidogenesis was observed 96 h after treatment, thus, the results of the present study suggest that it may be prolonged exposure to TNFα, which results in decreased steroidogenic function of these cells.
Infusion of LPS into the uterine lumen blunted the pre-ovulatory LH surge in heifers leading to failure of ovulation and the formation of cystic follicles44. Administration of LPS intravenously results in the disruption of neuroendocrine activity and interference with the oestrous cycle of sheep. Hypothalamic GnRH secretion is suppressed, pulsatile LH secretion from the pituitary is inhibited, and there is a reduction in pituitary responsiveness to GnRH16. However, these effects are seen at concentrations of LPS which induce systemic illness in the animals treated. In contrast to these previous studies, a feature of the present study was the administration of treatment at concentrations which do not cause systemic sickness in the animals. Indeed, the animals were clinically normal; rectal temperature and acute phase protein concentrations were in the normal range. Furthermore, there was no evidence that LPS or TNFα affected the secretion of LH allowing the direct effect on the ovary to be evaluated. Thus, our results provide evidence of a direct utero-ovarian pathway via which LPS or TNFα directly modulate ovarian function. The in vivo responses were subtle perhaps because of the limited systemic effect. However, treatment was associated with fewer ovulations even in the face of normal LH pulse profiles, as observed in sheep16.
Smaller ovulatory follicles result in less effective luteal structures58, however, in the present study, the CL that was affected by LPS or TNFα developed from a follicle which ovulated prior to treatment. This suggests that LPS or TNFα may affect luteal cells directly. Indeed, TNFα has recently been shown to play bifunctional roles in the regulation of CL function during the estrus cycle59. In heifers infused with TNFα, peak progesterone concentrations tended to be lower, although differences were not significant. Receptors for TNFα are present in the bovine CL and low concentrations of TNFα similar to those used in the present study, cause luteolysis59,60. Additionally, TNFα stimulates the release of PGF2α from CL cells, and also endometrial cells, which may also result in luteolysis59. Therefore, intrauterine infusion of TNFα may affect progesterone production directly via actions on luteal cells or indirectly via the induction of prostaglandin synthesis in the bovine endometrium and CL.
In conclusion, ovarian cell steroidogenesis was decreased by treatment with LPS or TNFα in vitro. Granulosa cell estradiol secretion was reduced by LPS treatment, whilst TNFα reduced theca and granulosa cells androstenedione and estradiol production, respectively. Although, the effects of LPS or TNFα on ovarian structures and steroidogenesis were more subtle in vivo, ovulation of the dominant follicle was less likely than in control animals. The response was localised as peripheral immune markers were not increased and pituitary gonadotropin concentrations did not differ between treatment groups. Taken together the present study supports the hypothesis that LPS or TNFα can directly impact ovarian function. Thus, it is important to be aware that pelvic inflammatory disease or metritis can affect ovarian as well as uterine health.
This study was funded by the Wellcome Trust. The authors thank Deborah Fischer, Hilary Purcell and Jean Routley for technical assistance and Dr Claire Glister for advice on theca cell culture. We also thank staff and students at the RVC who helped with collection of clinical samples and staff at Boltons Park Farm for animal husbandry duties.