You are on page 1of 10

Stability Analysis of an Enzymatic Model of

Rhodopsin Kinetics
Pananos, Corless, Lamb
March 29, 2014
Abstract
We provide details of a computer-algebra assisted study of the stability of the
implicit analytical solution of a nonlinear enzymatic model for the regeneration
of rhodopsin and the recovery of visual sensitivity following exposure of the
human eye to intense illumination. The study identies two parametrically
dened algebraic surfaces where the solution may be sensitive, thus indicating
good regions to experiment for parameter identication.
Begin as you mean to go on...
Thank you to The Doc , without whom, I am sure, my education would not be
the same.
Thank you to Rachel, for listening to me whine about math for 5+ years.
...and go on as you began... - Charles H. Spurgeon
Introduction
Readers who have walked into a building from a sunny winter day are well ac-
quainted with postbleach dark adaptation the sudden loss, and gradual regain,
of visual sensitivity following exposure to intense illumination. The biochem-
istry of dark adaptation has been written on extensively in Dark Adaptation
and the Retinoid Cycle of Vision [?]. In brief, when a photon strikes a molecule
of rhodopsin in the outer segments of the eye, the rhodopsin is broken into its
constituent molecules, opsin and 11-cis-retinal, of which the latter is isomerized
into 11-trans-retinal. The newly isomerized retinal is transported from the outer
segments, across a membrane, to the retinal pigment epithelium, where it can be
converted back into its cis state. Following the end of the visual bleach, the 11-
cis-retinal is pumped across the membrane, back into the outer segments, where
it can combine with free opsin to form rhodopsin, and restore visual sensitivity.
Lamb has derived a model for the concentration of opsin as a function of
time [?]. Though the model is corroborated by experimental data, the authors
realize that the model lacks biophysical realism. To rectify this problem, a new
model has been derived directly from the kinetic equations for the opsin cycle
of vision [?], namely
S +E
k
SI

k
IS
I E
k
IP

k
PI
P +E . (1)
Here, S is the concentration of substrate, I is the concentration of enzyme,
E I is the concentration of enzyme-substrate complex, P is the concentration
of product, and the variables above the reaction arrows indicate the rates of
conversion for their respective directions. The rate of forward reaction can be
written as
r = k
SI
SE k
IS
I = k
IP
I k
PI
PE . (2)
The enzyme is merely a catalyst in the reaction, and so it is conserved over time
(i.e E
tot
= E+I). The concentration of free enzyme can be obtained by solving
the two simultaneous equations in (??), along with the conservation relation, to
yield
E =
E
tot
1 +S/K
S
+P/K
P
. (3)
Where K
S
and K
P
are given by
K
S
=
k
IS
+k
IP
k
SI
and
K
P
=
k
IS
+k
IP
k
PI
.
Substituting (??) into (??) yields
r = E
_
k
IP
S
K
S
k
IS
P
K
P
_
. (4)
1
We now posit that the concentration of substrate is held constant (S = S
0
),
and denote the corresponding concentration of product as P
0
. This is the con-
centration of product that elicits net zero ux, or no reaction. By setting the
rate of reaction in (??) equal to zero, the equilibrium concentration of product
satises
P
0
S
0
=
k
IP
K
P
k
IS
K
S
= K . (5)
Here, K is dened as the equilibrium constant for the conversion of substrate
into product. After substituting (??) into (??), the forward rate of reaction is
r =
k
IP
E
tot
S0
K
S
(1
P
P0
)
1 +
S0
P0
+
P0
K
P
P
P0
. (6)
Note that the (1
P
P0
) term dictates the direction of the reaction. If the con-
centration of product is less than the equilibrium concentration, then the rate is
positive, and more product will be synthesized. If the concentration of product
is greater than the equilibrium concentration, then the rate is negative, and
product will be broken down into substrate.
The rate of binding 11-cis-retinal to free opsin is given by
dOps(t)
dt
= kP(t)Ops(t) . (7)
The Quassi Steady State Assumption allows us to posit that the two rates are
equal:
r =
dOps(t)
dt
= kP(t)Ops(t) . (8)
The solution of this dierential equation is shown in the following sections. Here,
we seek to assess the solutions stability to realistic perturbations of the form
r = r+ cos(t), which can be interpreted as waves of delivery of 11-cis-retinal,
rather than a constant rate of delivery.
We employ traditional perturbation methods, positing that the solution can
be expressed as a power series in the perturbation parameter, and use Maple to
obtain solutions for the resulting dierential equations.
Perturbation
After non-dimensionalizing time, and making the substitution u = K
1
/Ops(t),
the following dierential equation is obtained
d
d
u() = u()
_
u()
1
()K
2

_
(u()
1
() K
2
)
2
4u()
2
()
_
. (9)
2
Where the functions
1
() and
2
() are given by

1
() = 1 cos
_
2

kP
0
_
K
b
1
(10)

2
() = 1 +K
2
cos
_
2

kP
0
_
K
b
1
. (11)
We posit that the solution to this dierential equation can be written as a power
series in the perturbation parameter
u() =

j=0

j
u
j
() . (12)
We focus on terms up to rst order in
u() u
0
() +u
1
() +O(
2
) . (13)
We substitute (??) into (??) to obtain
d
d
(u
0
() +u
1
()) = (u
0
() +u
1
())
_
(u
0
() +u
1
())
1
()
K
2

_
((u
0
() +u
1
())
1
() K
2
)
2
4 (u
0
() +u
1
())
2
()
_
.
(14)
It is evident that doing the perturbation by hand could easily lead to multiple
mistakes. Instead, we leverage the power of Computer Algebra, in this case
Maple, to do the computational and algebraic work for us.
Our rst step in the perturbation is to collect all terms involving . This
will result in two equations since we are working up to rst order in . Using
Maple, this is done in two steps:
new_ODE := series ( ODE , epsilon , 2 )
This command takes the ode in (??) and expresses it as a series in , where the
coecients of the n
th
term are the terms which are multiplied by
n
. Now, we
collect all terms of order zero using the command
coeff ( new_ODE , epsilon , 0 )
We set what is returned from the previous command to zero. This will be the
rst dierential equation to solve. The previous command returns the following
dierential equation
d
d
u
0
() u
0
()
_
u
0
() K
2

_
(u
0
())
2
2 u
0
() K
2
+K
2
2
4 u
0
()
_
= 0 .
(15)
3
The ODE in (??) is solved in The Kinetics of Regeneration of Rhodopsin Under
Enzyme-Limited Availability of 11-Cis Retinoid [?] . Though not beyond human
computation, this separable ODE is much more easily solved in Maple. The
solution is given by
1
4
(L
1
(v) +L
2
(v))
K
2
+ 2
4K
2
A(v) +
K
2

v
2
4 4K
2
4(v + 2 +K
2
)
= (16)
with
v = u 2 K
2
L
1
(v) = ln(v + 2 +K
2
)
L
2
(v) = ln(v +
_
v
2
4 4K
2
)
A(v) = arctanh
_
(2 +K
2
)v + 4 + 4K
2
K
2

v
2
4 4K
2
_
.
Equation (??) is the implicit solution to u
0
() in the series expansion. We now
require a solution for u
1
().
We go about obtaining the dierential equation for u
1
() in the same way we
obtained the dierential equation for u
0
(): express the perturbed dierential
equation as a series in epsilon, collect the coecients of the n
th
term in the
series, and solve the resulting dierential equation. However, we run into a
small problem this time. When we collect the coecients of we obtain the
following
d
d
u
1
() u
0
()
_
u
0
() cos
_
2

kP
0
_
K
b
1
+u
1
() +
_
u
0
() K
2
cos
_
2

kP
0
_
+ 2 u
1
() K
b
+ (u
0
())
2
cos
_
2

kP
0
_
u
0
() u
1
() K
b
+K
2
u
1
() K
b
_
1
_
(u
0
())
2
2 u
0
() K
2
+K
2
2
4 u
0
()
K
b
1
_
u
1
()
_
u
0
() K
2

_
(u
0
())
2
2 u
0
() K
2
+K
2
2
4 u
0
()
_
= 0 .
In order to solve this dierential equation, we require an explicit expression
for u
0
(). This is not possible. We instead note that we may write (??) as
G(u
0
) = , allowing us to write d/du = G

(u
0
). This will prove essential in
our search for a solution.
With a little algebra, we may write the dierential equation for u
1
() in the
following way
du
1
d
+F
17
(u
0
) u
1
= F
14
(, u
0
) . (17)
4
This looks encouragingly similar to a rst order linear dierential equation
solved by rst year calculus students. We leverage this similarity by multiplying
(??) by G

(u
0
). The following results
G

(u
0
)
du
1
d
+G

(u
0
) F
17
(u
0
) u
1
= G

(u
0
) F
14
(, u
0
)
d
du
0
du
1
d
+G

(u
0
) F
17
(u
0
) u
1
= G

(u
0
) F
14
(, u
0
)
du
1
du
0
+G

(u
0
) F
17
(u
0
) u
1
= G

(u
0
) F
14
(u
0
) . (18)
Through this manipulation, we obtain a rst order linear dierential equation
for u
1
in terms of u
0
! This maneuver is called a Hodograph Transformation.
Note that the dependence vanishes in (??), as we can make the substitution
= G(u
0
). We may now employ our rst year techniques, with the assistance
of Maple, to compute the solution. Surprisingly, Maple can compute a closed
for expression for the integrating factor. The integrating factor is
I = exp (G

(u
0
)F
17
(u
0
)) .
So far, we have reduced (??) to
d
du
0
(I u
1
) = I F
14
, (19)
which, by the Fundamental Theorem of Calculus, is
I u
1
=
_
u1
1
I() F
14
()d
u
1
=
__
u1
1
I() F
14
()d
_

1
I
. (20)
Thus far, we have been very fortunate, insofar as Maple has been able to give
us nice, closed form expressions for many of the terms we require. It seems
our luck has run out; Maple is unable to nd an elementary antiderivative for
(??). Indeed, via Maples implementation of the Risch algorithm, this proves
that the integral is not elementary. One may use quadrature to numerically
integrate (??), if one wishes to do so. Although we can not obtain a closed form
expression for u
1
, it suces to nd areas where the integrand is sensitive.
The integrand in (??) is

K
2
2
uK
2
2 u +
_
u
2
2 uK
2
+K
2
2
4 uK
2
u
(21)

u
_
2
_
cos
_

kP
0
__
2
1
_
__
u
2
2 uK
2
+K
2
2
4 u +K
2
+u
_
1
_
K
2
2 +u +
_
u
2
2 uK
2
+K
2
2
4 u
1
_
u
2
2 uK
2
+K
2
2
4 u
K
b
1
.
5
Looking primarily at the denominator of (??) will give us sensitive areas. To
avoid sensitivity, we require that the following hold:
_
(u K
2
)
2
4u = 0 (22)
_
K
2
2 +u +
_
(u K
2
)
2
4u = 0 . (23)
Making use of the substitution v = u 2 K
2
, solving equation (??) yields the
restriction that K
2
= 1. Solving (??) yields u = 2 +K
2

4 + 4K
2
.
Conclusion
We sought to analyze the stability of the enzymatic model for rhodopsin kinet-
ics under realistic perturbation. The model derived by Lamb posited that the
delivery of 11-cis-retinal from the outer segments to the retinal pigment epithe-
lium was constant. We assume that the delivery rate is not constant and that
the 11-cis-retinal is delivered in waves .
A classical perturbation was done on the new system, and although a closed
form solution could not be obtained for the new model, two parametrically
dened algebraic surfaces where the solution is sensitive were obtainable.
Although this study does not shed direct light on the study of postbleach
dark adaptation, it does provide areas for parameter investigation. The al-
gebraic surfaces we have indicated lead to singularities under realistic pertur-
bation, therefore these algebraic surfaces rule out certain parameters for the
model.
More sophisticated techniques may prove useful for nding closed form so-
lutions to the dierential equation for u
1
. A closed form solution for u
1
would
allow us to study the kinetics for a variety of frequencies, which would prove
invaluable in the study of the kinetics.
6
Bibliography
[1] T.D. Lamb and E.N. Pugh Jr. Dark adaptation and the retinoid cycle of
vision. Progress in Retinal and Eye Research, 23(3):307 380, 2004.
[2] Trevor D. Lamb and Edward N. Pugh. Phototransduction, dark adaptation,
and rhodopsin regeneration: The Proctor Lecture. Investigative Ophthalmol-
ogy & Visual Science, 47(12):51385152, 2006.
[3] R. Corless T.D. Lamb, D. Pananos. The kinetics of regeneration of rhodopsin
under enzyme-limited availability of 11-cis retinoid. Journal of Mathematical
Physiology, 23(3):307 380, 2004.
7

You might also like