A Mathematical Model of Estrogen Production and Diffusion within Ovarian Follicles Zachary Miller, Mentor: Karin Leiderman

Abstract A mathematical model has been developed describing the production and diffusion of estrogen within ovarian follicles. The model was based on a series of time-dependent reaction-diffusion equations that accounted for several physiological mechanisms: diffusion and reaction kinetics of the pituitary gonadatropins, androgen production within the theca cell layer, and the ultimate result of these reactions: production and diffusion of estrogen within the granulosa. Preliminary results are not yet robust enough to match data from the biological literature, however it appears that the approach used in developing this model crudely fits the overall behavior of the hypothalamo-pituitary gonadal axis. Further model refinement and incorporation of more detailed reaction kinetics will allow not only an accurate description of normal pre-ovulatory estrogen production within ovarian follicles, but also should capture follicular behavior in pathological amenorrheic states. Introduction The human menstrual cycle is a 28 day periodic cycle spanning the female reproductive years: beginning at puberty around age 13 and ending at menopause at age 50 (1). The inherent periodicity of this cycle can be viewed as an emergent result of communication via hormones between the pituitary gland located near the brain and the ovaries. The pituitary gland communicates with the ovaries mainly through the gonadotropins: FSH and LH (2) collectively leading to the up-regulation of estrogen production within the ovaries (2). Estrogen via the blood plasma feeds-back on the pituitary and leads to down regulation in pituitary secretion of FSH and LH (2). This system lends itself to a control theoretic description, that is, a system whose components are capable of regulating their production, and thus their effects on the rest of the system and body. Not surprisingly, several

well established models exist in the literature that view the menstrual cycle in these terms (3,4,5).

This approach, however, ignores the complex cell-level dynamics in play during the production/secretion of the gonadatropins within the pituitary and estrogen in the ovary. In this paper, estrogen production and diffusion was examined in detail. The choice to examine estrogen dynamics instead of gonadotropin production was a result of two factors. First, several mathematical models already exist in the literature that provide a fairly complete description of gonadotropin production/secretion; this complete mathematical description simply does not exist for estrogen dynamics. Second, while the gonadatropins are important in the context of the menstrual cycle, they do not exert as broad of a systemic effect as estrogen-the primary female sex hormone does (6). Estrogen, in addition to its role in reducing gonadotropin production in the menstrual cycle is extensively involved in the development and maintenance of secondary sex characteristics, mood, bone growth, and body fat content. In particular, improper supplementation of estrogen ie too much estrogen particularly after menopause has been shown to increase the risk of estrogen receptor positive breast and uterine cancers due to estrogens ability to stimulate breast and uterine hyperplasia . At the same time, low levels of estrogen leading to amenorrhea seen in post-menopausal women, cases of female anorexia or athletic overtraining, and various other pathologies are implicated in increased incidence of osteoporosis, depression, and infertility. Thus a thorough understanding of estrogen dynamics within ovarian follicles is key to understanding many of these illnesses (6). The basic unit of the ovary is the ovarian follicle, a spherically shaped structure containing a female oocyte and several layers of cells involved in both the maturation of the oocyte and the production of estrogen. These layers of cells surrounding the oocyte consist of two layers: the outer theca cell layer and the inner granulosa region. These cell regions show differential sensitivity to the

pituitary gonadatropins: the theca only responds to LH while the granulosa is sensitive only to FSH. This distinct gonadotropin responsiveness reflects the compartmentalization of estrogen biosynthesis within the ovarian follicle. Several groups have shown that the granulosa layer is exclusively responsible for estrogen production, but almost paradoxically, removal of the theca results in a rapid decline in estrogen production. This paradox was resolved by showing experimentally that the granulosa is capable only of aromatizing pre-made androgen to estrogen. FSH receptor binding and its concomitant c-AMP cascade only activates transcription of enzymes late in the estrogen biosynthesis pathway, rendering the granulosa incapable of synthesizing androgens or any of its precursors de novo. The body offers an elegant solution to the androgen production problem; LH responsive theca cells transcribe enzymes capable of hydrolyzing stored cholesterol esters and synthesizing androgens which then diffuse into the granulosa allowing estrogen production. Similar to the granulosa, the theca is not capable of synthesizing estrogen alone. Thus the picture that emerges from the physiological details is estrogen production is a result of the collective efforts of both the granulosa and theca cell layers regulated by the gonadatropins (6). 2. Model 2.1 Introduction In this section, a mathematical model for estrogen production and diffusion is developed. It is important to emphasize that this model does not intend to capture the intricacies involved in estrogen production and diffusion, rather its goal is to develop a crude, but intuitive mathematical understanding of estrogen dynamics within the follicle. The model, however, was developed with the following constraints in mind. First, the model must account for the diffusion and differential receptor binding of LH and FSH to the theca and granulosa layers respectively. Additionally, this model must describe the kinetics that lead to the production of androgen within the theca, and its subsequent diffusion and consumption within the granulosa cell layer. Finally, this model must

accurately describe the production and diffusion of estrogen within the follicle. Any mathematical approach taken examining estrogen production/diffusion dynamics will be novel as no known literature exists describing this system despite its clinical importance. A reaction-diffusion approach was taken in this paper, however the menstrual cycle (from other perspectives than cell-level dynamics) has been modeled using classical ODEs (3) and dynamical systems (3).

2.2 Reaction-Diffusion Equations and Assumptions

The reaction-diffusion equation is a classic parabolic PDE that not only tracks the random movement of substances ie steroids, hormones through its diffusion term, but also allows for the description of the consumption and production ie reaction kinetics of these diffusing terms. These equations are a perfect description of substance dynamics within the ovarian follicle for several reasons. First, the follicle is small enough that most substances move purely through diffusion. Steroids in particular are fatty (lipophilic) and thus do not rely on active transport; they are capable of easily passing through lipid membranes. Second, the reaction term allows estrogen biosynthesis to be described in detail. Thus a set of four one-dimensional reaction-diffusion equations in axially symmetric polar coordinates were written:

(diffusion term)

(reaction term)

With the following boundary and continuity (spatial) conditions and initial (temporal) condition:

1.

Robin boundary condition

2.

No-Flux Continuity Condition

3.

Initial Condition

Where i=FSH, LH, estrogen, androgen and is the outer diameter of the follicle, is the diffusivity constant, h and k are permeability constants, assume D,h,k are 1

These equations rest on several assumptions. First, the concentration profiles (the output u(r,t) of the reaction-diffusion equations) are defined on a circular cross section of this idealized follicle in axially symmetric polar coordinates. The radial axis extends from 0<r< where =10 and the angular domain sweeps out . Axial symmetry on a circular domain means the concentration profile u(r,t) is allowed to vary along the radial axis preserving the radial component of the reaction-diffusion equation. However, the angular component is assumed not to vary, that is, diffusion looks the same along the radial axis at 0, the radial axis at , and every angle in between meaning the angular component of the diffusion equation drops out. Thus a harder 2-dimensional problem is simplified to 1-dimension. This assumption is physiologically justified. Concentrations of various substances should change along the radial axis because different cell types/reactions are radially distributed. On the other hand, if a radial location in the follicle is selected, say the granulosa, and the follicle is rotated sixty degrees, this new location after rotation would still be the granulosa with the same cell dynamics/reactions. Second, it is assumed only the gonadatropins and steroidal hormones: androgen and estrogen are

involved in estrogen biosynthesis and diffusion. This assumption reduces the sheer number of tracked substances, focusing only on the essential reactions meaning aromatase, FSH receptors, LH receptors and myriad other substances are not considered variables at this point due to the difficulty in mathematically describing their biological behavior and the lack of detailed literature on their behavior within the follicle. Third, two spatial conditions are defined: a no flux continuity condition at the center of the follicle r=0, and a Robin boundary condition at the follicles outer edge . A single initial condition assuming there is no effective concentration of any tracked substance within the follicle is also defined. The no flux continuity condition at the center of the follicle means there is no diffusion across this boundary. This condition follows directly from axial symmetry because if there was diffusion across the center boundary, concentrations of substances on one side of the follicle would be different (due to diffusion) from the other side breaking the axial symmetry condition. The Robin boundary condition relates the flux of substance across the outer boundary to the difference in concentration inside the follicle versus outside the follicle. The concentrations outside the follicle are assumed to be time dependent for the gonadatropins: a periodic pulse where small, but fixed at 0 for estrogen and androgen. The steroidal hormones are produced primarily within the ovarian follicles, thus their outside concentrations should be negligible, and their flux out the follicle proportional only to their inner concentration. On the other hand, the gonadatropins are much more concentrated outside the follicle than within the follicle. Physiologically, the Robin boundary condition makes sense because materials are exchanged across the outer membrane in time. The initial condition is simply the time point before the gonadatropins have entered the cell and initiated any reactions thus all substance concentrations are zero. Finally, the follicle is assumed to be divided into three layers: the follicular fluid containing the oocyte from 0<r<6, the granulosa from 6r<8, and the theca from 8r10. Reactions occur in the granulosa and theca, and are described by consumption/production first

order kinetics where all rate constants are assumed to be one due to lack of ovarian follicle kinetic studies in the literature. The reactions are as follows:

This purpose of this set of reactions is not to give a precise picture of estrogen biosynthesis kinetics within the follicle, but rather to give an intuitive picture of the physiology. As discussed in the introduction, estrogen biosynthesis within the follicle is highly compartmentalized, and this is reflected within the piecewise nature of the equations. LH only reacts within the theca which leads to the production of androgen, androgens diffusion into the granulosa, and its consumption to form estrogen. FSH reactions were not included because this hormone does not directly stimulate the production of estrogen, rather it leads to the production of aromatase. This

enzyme was not accounted for within the follicle, thus FSH reactions were also not included. Thus it is assumed that LH stimulates androgen production which stimulates estrogen production.
2.3 Analytic and Numerical Methods A multistep process involving both exact and numerical solutions was used to investigate the physiological validity of the reaction-diffusion approach. The scheme established was to first examine the diffusion PDE of one substance u(r,t) with given boundary, continuity and initial conditions alone without the reaction terms. Briefly, this PDE has an analytic solution in the form of a transient Bessel function coupled to a steady state solution (the steady state emerges from an eulerform ODE) (the solution process is a variation on many found in common introductory PDE texts). The analytic solution was found in order to establish a point of comparison for any future numerical analysis. Additionally, the Robin boundary condition was compared to a fixed boundary condition because both numerically and analytically, the Robin boundary condition is more involved relative to the fixed condition. However, upon graphical analysis it was clear that the Robin boundary condition fit reality better as concentrations should be variable in time along the outer boundary of the follicle. After evaluation of the analytic solution, a forward euler numerical scheme implemented in Matlab was compared to the analytic solution. After several more trials, a final numerical scheme was established examining the reaction-diffusion equations with the assumptions in 2.2. This method accounts for the diffusion of LH into the theca from the outer membrane in a pulsatile manner, the consumption of LH within the theca, subsequent production of androgen and androgens diffusion into the granulosa, finally leading to the production and diffusion of estrogen 3.Results/Discussion Forward Euler numerical solution using assumptions in 2.2 is shown below at several different time points up to t=1000. Three concentration profiles are shown on the following time stepped sequence.

Yellow corresponds to LH, Blue to androgen, and Green to estrogen. The x-axis represents the radial distance from the center of the follicle at r=0 to the outer membrane of the follicle at r=10. The y-axis corresponds to the concentration profile.

Figure 1 At t=10, LH pulses (yellow) across the outer membrane, and has just begun to be consumed to form androgen in the region 8<r<10. Following from the initial condition, there is no androgen or estrogen within the follicle, these substances are produced through the reaction of LH within the theca. Later at t=50, the production of androgen (blue) can clearly be seen within the theca layer, however androgen has not diffused far enough yet to initiate production of estrogen.

Figure 2 At t=400, it is clear that the LH pulse across the outer membrane has stopped, and remaining LH is consumed to form androgen. Both androgen and LH have diffused into the granulosa region: 6<r<8 , this is seen by comparing the radial width of these substances at the different time points. The larger the width, the more dispersed the products are. This diffusion of androgen has resulted in the production of estrogen (green).

Figure 3 At t=1000, estrogen concentrations have increased significantly, and estrogen is beginning to diffuse out of the follicle. Both androgen and LH have diffused significantly, and without any more LH boundary pulses, both androgen and LH will eventually go to 0 while estrogen concentration will peak then diffuse out of the follicle.

Figure 5 4.Conclusions The mathematical model of estrogen production and diffusion developed in this paper provides a crude, but intuitive understanding of the underlying follicle physiology. It portrays the pulsatile movement of LH into the follicle, its reaction within the theca to produce androgen, androgens reactions and diffusion within both the theca and granulosa, and estrogens production and diffusion. Despite these successes, a more accurate model is badly needed that will capture both normal menstrual cycle estrogen dynamics, but also pathological states. In order to create this model, several barriers must be crossed. First, rate constants for estrogen biosynthesis must be made available within the biological literature. Second, experimentalists must investigate FSH/LH binding in more depth, and carefully track follicle size throughout the menstrual cycle. This data is badly needed in order to produce a model that is clinically relevant. The author hopes to continue pursuing questions in this field next year.

5. Acknowledgements I would like to thank everyone at the REU program: Professor Leiderman, Professor A. Layton, and

Professor Reed for taking the time to teach me about doing research and loving math. Also, I would like to thank the NSF for funding this research. 6. References
1. Regulation of primordial follicle assembly and development, Hum Reprod

Update September 1, 2005 11 (5) 461-471

2. Control of ovulation number in a model of ovarian follicular maturation

HM Lacker Mathematical Questions in Biology, 1981

3. A.R. Baerwald, G.P. Adams and R.A. Pierson, Characterization of ovarian follicular wave dynamics in women, Biol. Reprod. 69 (2003), p. 1035. 4. R.A. Pierson and D.R. Chizen, Transvaginal ultrasonography in the evaluation and management of infertility, J. Obstet. Gynaecol. Can. 13 (1991), p. 37. 5. R.A. Pierson, D.R. Chizen and O.A. Olatunbosun, Ultrasonographic assessment of ovulation induction. In: R. Jaffe, R.A. Pierson and J.S. Abramowicz, Editors, Imaging in Infertility and Reproductive Endocrinology, J.B. Lippincott, Philadelphia USA (1994), p. 155. 6. J.C. Mariana, F. Corpet and C. Chevalet, control of follicular growth and ovulation in domestic species, Acta Biotheor. 42 (1994), p. 245-248.