Professional Documents
Culture Documents
Optimal Regulation of Protein Degradation To Schedule Cellular Events With Precision
Optimal Regulation of Protein Degradation To Schedule Cellular Events With Precision
I. I NTRODUCTION
The process of gene expression is central to a cells
function: the genetic information is used to make mRNA
molecules which are further translated into proteins, the
molecular machines to carry out different tasks. An important
step in execution of many cellular functions is when a
protein attains an effective level [1]. For example, in gene
regulatory pathways, the protein expressed by one gene
regulates expression of a downstream gene once its level
crosses a certain threshold [2][5]. Other examples include
cell-fate decisions which are made when specific regulatory
proteins cross their respective critical levels [5], [6], and
temporal order of gene activation which involves different
thresholds of a given regulatory protein [7][10].
The probabilistic nature of biochemical reactions and the
fact that copy number of the constituents is small make
expression of a gene a stochastic process [11][16]. As a
result, a population of cells induced at the same time sees
difference in the times at which a critical protein level is
attained in individual cells. While the stochasticity in timing
does offer an advantage in terms of phenotype variability
before a cell commits to an irreversible cell fate [17], in some
Khem Raj Ghusinga is with the Department of Electrical and Computer Engineering, University of Delaware, Newark, DE 19716, USA
khem@udel.edu
Abhyudai Singh is with Faculty of Electrical and Computer Engineering,
Biomedical Engineering, Mathematical Sciences, University of Delaware,
Newark, DE, 19716 USA absingh@udel.edu
AS is supported by the National Science Foundation Grant DMS1312926, University of Delaware Research Foundation (UDRF) and Oak
Ridge Associated Universities (ORAU).
Protein
Fig. 1.
Gene expression model with feedback regulation of protein
degradation. The gene transcribes to make mRNAs which are further
translated to proteins. The protein is assumed to regulate its own degradation
in order to achieve precision in the time at which a certain protein level is
achieved.
(1a)
(1b)
(2)
kp
1
= ,
(3)
X1
fFPT (t)dt =
i=0
(6a)
X1
(6b)
(6c)
i=0
X1
i=0
(7)
where the row vector U and the column vector P(t) are given
as follows
U = k (1 )X k (1 )X1 k (1 ) ,
(8)
T
P(t) = p0 (t) p1 (t) pX1 (t) .
(9)
In the expression for P(t) we have used the notation pi (t)
which stands for P (x(t) = i). To determine P(t), we write
the chemical master equation (or the forward Kolmogorov
equation) for the bursty birth-death process with absorbing
state [26], [27], [37]:
p0 (t) = k(1 )p0 (t) + 1 p1 (t),
(10a)
+ k (1 )in pn (t),
1 i X 2, (10b)
n=0
(10c)
n=0
(4)
0,
(i 1) ,
i1
ai j =
k
(1
) (i 1)i1 ,
k (1 )i j ,
P(t)
= AP(t) where
j > i+1
j = i+1
.
j=i
j<i
(11)
(12)
T
where P(0) = 1 0 0
is the initial probability
distribution (as x(t) = 0 at t = 0). Using the expression of
P(t), the distribution of FPT can be written as
fFPT (t) = U exp(At)P(0).
..
.
..
.
0
j1
ll
1
1
1
k+l
l
k
j1
ll
k
1
1
. ,
..
..
.
1
k
k
k+(X1)X1
(18a)
(13)
1
1
k+1 1
k
k+1
1 ..
.
0
j1
ll
1
1
1
X
k +k 1 + j=2 l=1
k+ll 1
j1
ll
1
1
X
k+1 1 + j=3 l=2 k+ll 1
..
.
1
k+(X1)X1
(18b)
By equation (39d), we have UA1
0 = 1 1 1 .
Therefore the mean FPT is negative of summation of the
elements of the vector E1 A1
0 P(0) and is given by
1
+
k 1
!
X j1
1
1
ll
1
+
.
i=1 k + (i 1)i1
j=i+1 l=i k + ll (1 )
(19)
X
t m U exp(At)P(0)dt,
Z
t m exp(At)dt P(0).
=U
(14a)
0
(14b)
1
FPT 2 = 2
k + (i 1)i1 i +
1 i=1
!!
j1
X
1
ll
1
j k + (i 1)i1 k + ll (1 ) , (20a)
j=i+1
l=i
!
1 i1
i
where i :=
(20b)
j + , and
k j=1
the terms denoted by i are given by
1 m+1
P(0),
(15)
A1 = E1 A1
0
1+
j1
j=i+1 l=i
1
ll
k + ll (1 )
!
.
(20c)
A. Mean FPT
1
1 1
i1 +
k
k + (i 1)i1
(17)
0.8
we have
hFPT i =
0.6
FPT CV2
1 i1
ki1 l
i=1
l=0
!2
1
X 2 2
k
i=1
i1
i1
!2
.
l=0
(23)
0.4
1
2
k2 i1
0.1
0.2
0.3
0.4
0.5
1 i1
ki
l , 0 = 1, i =
,
i!1 2 i
i=1 ki1 l=0
!2
j1
X
2 i1 1
var (FPT ) =
j1
i=1 ki1 j=1 k j1
l=0
!2
i1
1
+ 2 2
l .
k i1 l=0
hFPT i =
(21)
(22)
i1
l
l=0
!2
=
1
k2 2j1
j1
!2
, i, j {1, , X 1}
l=0
(24)
!2
j1
X
i1
2
1
= hFPT i2 < X
j1
i=1 ki1 j=1 k j1
l=0
!2
i1
1
+X 2 2
l (25b)
k i1 l=0
= hFPT i2 < X var(FPT )
var(FPT )
1
=
= CV 2 > .
2
X
hFPT i
(25c)
(25d)
= 0.3 1
= 0.2 1
1) The diagonal
elements aii < 0 for i = 1, 2, , X,
X
ai j
2) max < 1.
1 jX i=1 a j j
FPT CV2
= 0.1 1
= 0 1
0.1
j6=i
0.01
ai j
( j 1) j1
a j j = k(1 ) + ( j 1)
i=1
j6=i
0.001
0.01
0.02
0.03
0.04
0.05
(1 )i j
(27a)
i= j+1 k(1 ) + ( j 1) j1
k(1 ) 1 (1 )X j + ( j 1) j1
=
< 1. (27b)
k(1 ) + ( j 1) j1
+ k
k(1 )
0
k(1 )
k(1 )
k(1 )2
k(1 )
A0 =
,
..
..
..
..
.
.
.
.
0
0
0
Ae =
.
..
0
0
1
1
0
..
.
0
0
..
.
..
.
0
0
0
0
0
..
.
0
..
.
(X 2)X2
0
(X 1)X1
(X 1)X1
(29)
1
A1
0 .
(30)
The term A1
0 is easy to determine and is given by
1
k
1
k
1
k
A1
0 =
.
1
..
1
k
1
k
1
k
1
k
..
.
..
.
0
0
0
..
.
1
k
1
k
1
k
1
k
0
0
0
..
.
.
1
k
(31)
Further, we also need inverse of E := I + A1
0 Ae in order
to determine A1 . To this end, let us compute A01 Ae :
1
0 0
0
1
1
0
0
k
k
1
1
0
0
k.
(32a)
.
.
.
.
..
..
..
..
1
..
1
1
1k
k
0
1
1
1
1k
k
k
k
0 1
0
0
0 1
0
0
0
.
0
0
0
.
..
..
..
..
..
.
.
.
.
0
(X 2)X2 (X 1)X1
0
0
0
(X 1)X1
(32b)
1
0 (1)k
0
0
1
0
0
0
..
0
.
0
0
0
= .
. (32c)
.
.
.
.
..
..
..
..
..
(X2)X2
(X1)X1
0
(1)k
0
k
0
l=i
0
22
22
0
..
.
0
0
0
0
33
33
..
.
0
0
..
.
0
0
0
0
..
.
(X 2)X2
0
0
0
0
0
(X 1)X1
(X 1)X1
, (35)
while H is a X 1 X matrix
0
0
H = ...
0
0
1
0
..
.
0
1
..
.
..
.
0
0
..
.
0
0
0
0
1
0
0
0
0
1
(36)
(X1)X1
k
if i > j,
0
k
0
if i = j,
(33)
ei, j = k+(i1)i1 ,
j1
ll
1
k
, if i < j.
k+(i1)i1
1
1
0
0
G=
.
.
.
0
0
k+ll (1)
j1
ll
1
1
1
1 k+
l=1 k+l
1 1
l 1
j1
ll
k
k
1
0
k+1
k+1 l=2 k+ll 1
E1 =
. (34)
..
..
..
..
.
.
.
k
0
0
k+(X1)
X1
E1
and
A1
0
which
C. Expression of UA1
In [27], where both production and degradation of protein
do not have feedback regulation, we show that UA1 simplifies to a row vector with all of its elements being 1. It turns
out that even in the case when the protein degradation has
a feedback, the same result is true. The calculations are as
follows. Consider two matrices G and H such that Ae = GH
where G is a X X 1 matrix
A1 = (A0 + GH)1
= A0
A1
0 G
(37a)
1
I + HA1
HA0 1 .
0 G
(37b)
This implies
1
1
UA1 = UA1
0 UA0 G I + HA0 G
1
HA0 1 .
(37c)
T
1
1
0
1 1 22
1 0
2
1 0
0
.. ..
..
. .
.
1 0
0
0
0
1
= 0 0 0 .
..
.
0
0
0
0
..
.
(X 1)X1
X 1X1
(38a)
(38b)
shown below.
UA1
0 =
1
T 1
0 0
0
k
k (1 )X
1
1
0
0
X1
k
k
k (1 )
.
X2 1
.. 0
1
k (1 )
0
k
k
..
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
2
1
1
k (1 ) 1
k 0
k
k
1
1
1
k (1 )
1k
k
k
k
T
(1)X
+ X1
(1 )l
l=1
(1)X1
l
X2
+
(1
l=1
(1)X2
l
X3
+ l=1 (1 )
..
2
(1)
+ (1 )
(39a)
(39b)
h 1 1
=
1
= 1 1 1 1
1 .
(39c)
(39d)
1
ll
2
FPT = 2
r + r k , where
i=1 k r=1
j=i+1 r=1
l=i
(40b)
!
X j1
1
ll
i =
1+
.
(40c)
k
j=i+1 l=i k
Though we have skipped the detailed steps here, these results
are equivalent to the results derived in [26] and [38]. More
specifically, if i = , we get the results of [26]. The results
in [38] are more general and simplify to our results if the
birth rate is made constant while the death rate is taken as
i instead of ii .
R EFERENCES
[1] H. H. McAdams and A. Arkin, Stochastic mechanisms in gene
expression, Proceedings of the National Academy of Sciences, vol. 94,
pp. 814819, 1997.
[2] T. Lamb, Gain and kinetics of activation in the G-protein cascade of
phototransduction, Proceedings of the National Academy of Sciences,
vol. 93, no. 2, pp. 566570, 1996.
[3] M. C. Gustin, J. Albertyn, M. Alexander, and K. Davenport, Map
kinase pathways in the yeast saccharomyces cerevisiae, Microbiology
and Molecular biology reviews, vol. 62, no. 4, pp. 12641300, 1998.
[4] E. H. Davidson, J. P. Rast, P. Oliveri, A. Ransick, C. Calestani, C.-H.
Yuh, T. Minokawa, G. Amore, V. Hinman, C. Arenas-Mena, et al., A
genomic regulatory network for development, science, vol. 295, no.
5560, pp. 16691678, 2002.
[5] J. E. Ferrell and E. M. Machleder, The biochemical basis of an allor-none cell fate switch in xenopus oocytes, Science, vol. 280, no.
5365, pp. 895898, 1998.
[6] P. Ingham and J. Smith, Crossing the threshold, Current Biology,
vol. 2, no. 9, pp. 465 467, 1992.
[7] A. Zaslaver, A. E. Mayo, R. Rosenberg, P. Bashkin, H. Sberro,
M. Tsalyuk, M. G. Surette, and U. Alon, Just-in-time transcription
program in metabolic pathways, Nature genetics, vol. 36, no. 5, pp.
486491, 2004.
[8] M. Fujita, J. E. Gonzalez-Pastor, and R. Losick, High-and lowthreshold genes in the spo0a regulon of bacillus subtilis, Journal of
bacteriology, vol. 187, no. 4, pp. 13571368, 2005.
[9] U. Alon, An introduction to systems biology: design principles of
biological circuits. CRC press, 2006.
[10] A. Amir, O. Kobiler, A. Rokney, A. B. Oppenheim, and J. Stavans,
Noise in timing and precision of gene activities in a genetic cascade,
Molecular Systems Biology, vol. 3, 2007.
[11] W. J. Blake, M. Krn, C. R. Cantor, and J. J. Collins, Noise in
eukaryotic gene expression, Nature, vol. 422, pp. 633637, 2003.
[12] J. M. Raser and E. K. OShea, Noise in gene expression: origins,
consequences, and control, Science, vol. 309, pp. 20102013, 2005.
[13] J. Yu, J. Xiao, X. Ren, K. Lao, and X. S. Xie, Probing gene expression
in live cells, one protein molecule at a time, Science, vol. 311, pp.
16001603, 2006.
[14] A. Raj and A. van Oudenaarden, Nature, nurture, or chance: stochastic gene expression and its consequences, Cell, vol. 135, pp. 216226,
2008.
[15] M. Krn, T. C. Elston, W. J. Blake, and J. J. Collins, Stochasticity
in gene expression: from theories to phenotypes, Nature Reviews
Genetics, vol. 6, pp. 451464, 2005.
[16] A. Singh and M. Soltani, Quantifying intrinsic and extrinsic variability in stochastic gene expression models, PloS One, vol. 8, p. e84301,
2013.
[17] E. Yurkovsky and I. Nachman, Event timing at the singlecell level,
Briefings in Functional Genomics, vol. 12, pp. 9098, 2013.
[18] A. Becskei and L. Serrano, Engineering stability in gene networks
by autoregulation, Nature, vol. 405, pp. 590593, 2000.
[19] H. El-Samad and M. Khammash, Regulated degradation is a mechanism for suppressing stochastic fluctuations in gene regulatory networks, Biophysical journal, vol. 90, pp. 37493761, 2006.
[20] U. Alon, Network motifs: theory and experimental approaches,
Nature Reviews Genetics, vol. 8, pp. 450461, 2007.
[21] A. Singh and J. P. Hespanha, Evolution of gene auto-regulation in the
presence of noise, Systems Biology, IET, vol. 3, pp. 368378, 2009.
[22] Y. Tao, X. Zheng, and Y. Sun, Effect of feedback regulation on
stochastic gene expression, Journal of Theoretical Biology, vol. 247,
pp. 827836, 2007.
[23] A. Singh and J. P. Hespanha, Optimal feedback strength for noise
suppression in autoregulatory gene networks, Biophysical Journal,
vol. 96, pp. 40134023, 2009.
[24] G. Chalancon, C. N. Ravarani, S. Balaji, A. Martinez-Arias, L. Aravind, R. Jothi, and M. M. Babu, Interplay between gene expression
noise and regulatory network architecture, Trends in genetics, vol. 28,
pp. 221232, 2012.
[25] K. R. Ghusinga and A. Singh, First-passage time calculations for a
gene expression model, in IEEE 53rd Annual Conference on Decision
and Control (CDC), 2014, pp. 30473052.
[26] K. R. Ghusinga, P.-W. Fok, and A. Singh, Optimal auto-regulation to
minimize first-passage time variability in protein level, in American
Control Conference (ACC), 2015, pp. 44114416.
[27] K. R. Ghusinga and A. Singh, Theoretical predictions on the firstpassage time for a gene expression model, in IEEE 54th Annual
Conference on Decision and Control (CDC) (accepted), 2015.
[28] J. Paulsson, Models of stochastic gene expression, Physics of Life
Reviews, vol. 2, pp. 157175, 2005.
[29] A. Singh and P. Bokes, Consequences of mrna transport on stochastic
variability in protein levels, Biophysical journal, vol. 103, pp. 1087
1096, 2012.
[30] A. Singh, B. S. Razooky, R. D. Dar, and L. S. Weinberger, Dynamics
of protein noise can distinguish between alternate sources of geneexpression variability, Molecular systems biology, vol. 8, p. 607,
2012.
[31] A. Singh, B. Razooky, C. D. Cox, M. L. Simpson, and L. S.
Weinberger, Transcriptional bursting from the hiv-1 promoter is
[32]
[33]
[34]
[35]
[36]
[37]
[38]
[39]
[40]
[41]