Spatial organization of bacterial transcription and translation

Michele Castellana1, 2 and Ned S. Wingreen2

1
Joseph Henry Laboratories of Physics, Princeton University, Princeton, New Jersey 08544, United States
2
Lewis-Sigler Institute for Integrative Genomics, Princeton University, Princeton, New Jersey 08544, United States
In bacteria such as Escherichia coli, DNA is compacted into a nucleoid near the cell center,
while ribosomes—molecular complexes that translate messenger RNAs (mRNAs) into proteins—are
localized at the poles. We study the impact of this spatial organization using a minimal reaction-
diffusion model for the cellular transcriptional-translational machinery. Our model predicts that
∼ 90% of mRNAs are segregated to the poles, and reveals a “circulation” of ribosomes driven by
the non-equilibrium mRNA flux, from synthesis in the nucleoid to degradation at the poles. This
circulation increases ribosome efficiency, albeit a larger increase of ribosome efficiency stems directly
from the localization of the nucleoid at midcell.

The cytoplasm of many bacterial cells exhibits a strik- the right half, from x = 0 at the cell center to x = l at
ing spatial organization: rather than filling the cell vol- the right cell pole (Fig. 1). The 1D concentration of free
ume, the DNA forms a condensed structure called a (F) ribosomes, cF (x), denotes the number of F ribosomes
“nucleoid” that is generally localized near midcell (Fig. per unit length in an infinitesimal slice of the cell perpen-
1) [1]. Moreover, ribosomes—large molecular complexes dicular to the x axis. Similarly, ρm,n (x) is the 1D con-
that translate messenger RNAs (mRNAs) into proteins— centration of mRNAs with m transiently bound (B) [7]
are observed to be anti-localized from the nucleoid [2]. ribosomes and n translating (T) ribosomes; consistently
These observations raise two natural questions: (1) What with the experimental evidence presented below, we fix
physical processes are responsible for this subcellular or- the maximal number of allowed B and T ribosomes per
ganization? (2) How does this internal structure influence mRNA at M = 8 and N = 20, respectively. The resulting
the basic processes of mRNA transcription and protein reaction-diffusion equation for the F-ribosome density is
translation in the cell?
 2 
In the model bacterium Escherichia coli, ∼1.5 mm of ∂cF (x, t) ∂ cF (x, t) d2 vF (x)
= DF v (x) − c (x, t)
DNA are compacted into a ∼1 µm3 nucleoid volume [2], ∂t ∂x2
F F
dx2
thus forming a dense DNA mesh with average pore di- M
X −1 X
N M X
X N
ameter ∼ 50 nm. As a result, free ribosomes, with di- B
−kon cF (x, t) B
ρm,n (x) + koff m ρm,n (x)
ameter ∼ 20 nm, can readily diffuse into the nucleoid [3], m=0 n=0 m=1 n=0
while polysomes, molecular complexes composed of mR- M N
X X −1 M X
X N
T T
NAs with multiple bound ribosomes and having an ef- −kon cF (x, t) ρm,n (x) + koff n ρm,n (x)
fective diameter & 50 nm, anti-localize from the nucleoid m=0 n=0 m=0 n=1
due to excluded-volume effects. In vivo measurements of M X
X N
mRNA mobility suggest a typical diffusion coefficient of + β (m + n) ρm,n (x, t). (1)
D ∼ 0.05 µm2 /s, implying that mRNAs formed in the nu- m=0 n=0
cleoid by transcription from DNA can diffuse out of the
nucleoid to the ribosome-rich regions in a few seconds—a The first terms on the RHS represent diffusion including
time significantly shorter than the typical mRNA lifetime excluded-volume effects due to the condensed DNA (Fig.
of ∼5 min [2, 4]. These observations suggest that most 1); DF is the diffusion coefficient for F ribosomes, and
mRNAs formed in the nucleoid diffuse out of the nucleoid vF (x) is the fractional volume available to an F ribosome
to the ribosome-rich regions where ribosomes and mR- within the DNA mesh at position x (Fig. 2), see Supple-
NAs are colocalized, and where the bulk of translation mental Material Section S1 for details. The first term in
occurs. However, experimental measurements of mRNA the second line represents an F ribosome binding to an
localization in bacteria are challenging [5], and so far lim- mRNA and becoming a B ribosome, and is thus propor-
ited to specific mRNAs [6]: as a result, ribosome-mRNA tional to the F → B transition rate kon B
and to the total
colocalization still lacks experimental validation. In this local density of mRNA. The next term describes a B ribo-
Letter, we employ known properties of mRNAs, ribo- some unbinding from an mRNA, where koff B
denotes the
somes, and the nucleoid to predict the extent of mRNA unbinding rate and the multiplicity factor m accounts for
localization. Our approach also provides quantitative in- multiple B ribosomes on the mRNA. Similarly, the third
sights into the effect of mRNA degradation on ribosome line represents transitions between F and T ribosomes:
efficiency—a major determinant of the overall rate of cell since measurements suggest that the lifetime of the B
growth. state is significantly shorter than that of the T state [7],
We describe the coupled dynamics of ribosomes and here the F → B → T transition is incorporated into the
mRNAs in an E. coli cell using a minimal, 1D reaction- F → T transition, with an effective rate konT
. Finally, the
diffusion model. We introduce a coordinate x running last line represents B and T ribosomes being freed from
along the long cell axis, and, given the approximate left- mRNA molecules as these are degraded at a rate β [8].
right symmetry of a typical E. coli cell, we consider only The 1D reaction-diffusion equations for the mRNA
 2

as follows: The axial length was set at l = 1.5 µm, half

the length of a typical E. coli cell. The F ribosome diffu-
sion coefficient was taken to be DF = 0.4 µm2 /s [2], while
the diffusion coefficient of mRNAs was set at the average
diffusion coefficient of polysomes, D = 0.05 µm2 /s [3].
The F-ribosome available volume vF (x) was estimated
from the concentration profiles of free ribosomal subunits
[3], and similarly for the mRNA available-volume pro-
x
files vm+n (x) (see Fig. 2 and Section S2). The average
0 l time for a ribosome to complete protein translation is
FIG. 1: Schematic of the spatial organization of transcrip- estimated to be τT = 40 s [9], so we took the T → F
tion and translation in E. coli. mRNAs (black solid curves) transition rate to be koffT
= 1/τT = 2.5 × 10−2 s. To
are transcribed into the nucleoid—the condensed DNA chro- set the other transition rates we used the observation
mosome (gray solid curve) at the cell center. Ribosomes (red that ∼80% of ribosomes are T ribosomes [2], with non-
circles) bind to mRNAs forming polysomes, i.e. mRNAs with T ribosomes estimated to spend ∼ 90% of their time as
multiple bound ribosomes. Polysomes diffuse preferentially B ribosomes and ∼ 10% of their time as F ribosomes:
out of the nucleoid due to excluded-volume effects. Eventu-
here, the 90% : 10% division of non-T ribosomes be-
ally the mRNA molecules are degraded (black dashed lines).
In our 1D model, the coordinate x runs along the long cell tween B and F ribosomes is inferred from ribosome diffu-
axis and, assuming symmetry, we model only the right half of sion rates in Caulobacter crescentus [7], but it has lit-
the cell (0 ≤ x ≤ l). tle effect on the overall results, see Fig. S1. Global
steady state between T and F ribosomes requires the
F → T transition rate kon T
to satisfy NF ρ kon
T
= NT koff
T
,
densities are where NF and NT are the total number of F and T ri-
∂ρm,n (x, t) bosomes, respectively, and ρ ' 1.7 × 103 /µm is the av-
=
∂t  erage total mRNA axial density [8]. According to the
 2 2
∂ ρm,n (x, t) d vm+n (x) estimate above for the average number of F, T, and B
D 2
vm+n (x) − ρm,n (x, t) ribosomes, we have NT /NF = 80%/(20% × 10%), which
∂x dx2
B B yields kon
T
= 6 × 10−4 µm/s. Since the F → B rate kon B
− kon cF (x, t)ρm,n (x, t) − m koff ρm,n (x, t)
T T
represents rapid, transient, non-productive bindings of F
− kon cF (x, t)ρm,n (x, t) − n koff ρm,n (x, t) ribosomes to mRNAs, we chose this rate to be signifi-
B B
+ kon cF (x, t)ρm−1,n (x, t) + koff (m + 1) ρm+1,n (x, t) cantly larger than the F → T rate; in practice we set
T
+ kon T
cF (x, t)ρm,n−1 (x, t) + koff (n + 1) ρm,n+1 (x, t) kon
B
= 10 kon
T
= 6 × 10−3 µm/s. The B → F rate koff B

was estimated assuming rapid F ↔ B equilibrium, i.e.

+δm,0 δn,0 α(x) − β ρm,n (x). (2)
NF ρ kon
B
= NB koff
B
. According to the estimate above for
Here D is the average mRNA diffusion coefficient, and the average number of F, T, and B ribosomes, one has
vm+n (x) is the fractional available volume within the nu- NF /NB = 2%/18%, thus koff B
= 1.1/s. The mRNA degra-
cleoid for an mRNA with m + n attached ribosomes (Fig. dation rate was taken to be β = 3 × 10−3 /s, correspond-
2), see Section S2 for details. The third and fourth lines ing to a mean mRNA lifetime of ∼ 5 mins [4]. The total
Rl
represent binding and unbinding of B and T ribosomes mRNA production rate αtot = 2 0 dxα(x) was obtained
from mRNAs of species m, n, while the fifth line repre- from the total number of mRNAs per cell via the relation
sents B ribosomes binding to mRNAs of species m − 1, αtot /β = NmRNA = 5×103 [8]. The profile of mRNA pro-
n or unbinding from mRNAs of species m + 1, n. Sim- duction α(x) was chosen to reproduce the observed DNA
ilarly, the terms in the sixth line represent T ribosomes axial density distribution [2]. Finally, the total number
binding and unbinding from mRNAs of species m, n − 1 of ribosome in the cell was taken to be Ntot = 6 × 104 [2].
and m, n + 1 (only allowed mRNA species 0 ≤ m ≤ M ,
0 ≤ n ≤ N are considered). Finally, the last line repre- We numerically solved Eqs. (1), (2) at steady state.
sents transcription of initially ribosome-free mRNAs ac- The resulting mRNA profiles and ribosome concentra-
cording to the nucleoid profile α(x) (Fig. 2), and mRNA tions are shown in Figs. 3 and 4, respectively. As shown
degradation at rate β. in Fig. 3, the total mRNA profile ρtot is markedly local-
ized outside the nucleoid region—most of the mRNAs are
At steady state, we enforced in Eq. (1) segregated from the nucleoid because of excluded volume.
a constraint on the total number of ribosomes, Indeed, the density profiles ρm,n (x) show that the more
Rl P
2 0 dx[cF (x) + m,n (m + n)ρm,n (x)] = Ntot , and ribosomes an mRNA has bound, the more segregated the
we set a no-flux boundary condition at the cell pole, mRNA is from the nucleoid, see Fig. S2 for details. Since
[dcF (x)/dx vF (x) − cF (x) dv(x)/dx]|x=l = 0, see Section mRNAs are generated at the nucleoid, the strong segre-
S1. Similarly, in Eq. (2) we imposed no-flux bound- gation of total mRNA away from the nucleoid at steady
ary conditions at the cell pole and at the cell center, state must result from the majority of mRNAs being
[vm+n (x) dρm,n (x)/dx − ρm,n (x) dvm+n (x)/dx]|x=0,l = loaded with ribosomes such that excluded-volume biases
0, the latter reflecting the left-right symmetry of the cell. their diffusion away from the nucleoid. This conclusion is
We fixed the model parameters from experimental data confirmed in the inset of Fig. 3, which shows a heat map
 3

1 5 5,000
ρtot (x)
8 ρm,n (x)
10 60 ρ0,0 (x)
4 6 4,000
0.75 ρ8,20 (x)
4
30 Nm,n

m
α(x) (1/µm s)

ρm,n (x) (1/µm)

2

ρtot (x) (1/µm)

3 3,000
0 0

v(x)
0.5 0 5 10 15 20
5 n
2 2,000
α(x)
vF (x) 0.25
vm+n (x) 1 1,000

0 0
0 0
0 0.25 0.5 0.75 1 0 0.25 0.5 0.75 1
x/l x/l

FIG. 2: mRNA production rate and available-volume pro- FIG. 3: Steady-state mRNA and polysome distributions. To-
files in 1D model of the E. coli transcriptional-translational tal mRNA density ρtot (x) (red) and density ρm,n (x) of mR-
machinery. Local rate of mRNA production α(x), available- NAs with m transiently bound (B) ribosomes and n translat-
volume profile for free ribosomes vF (x) = 1 − κ πr2 ϕ(x), ing (T) ribosomes (gray). The density ρ0,0 (x) of ribosome-free
and available-volume profiles for polysomes with m tran- mRNAs (green) and the density ρ8,20 (x) of mRNAs with the
siently bound and n translating ribosomes vm+n (x) = largest number of B, T ribosomes considered (blue) are also
2
exp[−κ πrm+n ϕ(x)] for right half of cell (cf. Fig. 1), where the shown. The profiles ρm,n (x), ρ0,0 (x), and ρ8,20 (x) are nor-
arrow indicates the direction of increasing m+n. Here the con- malized to unit area. Inset: distribution of mRNA species,
stant κ = 0.25 has been estimated by fitting the observed con- shown as a heat map of the number Nm,n of mRNAs with m
centration profile of freely diffusing ribosomesp [3], r = 10 nm B ribosomes and n T ribosomes in the right half of the cell.
2
is the ribosome radius [10], and rm+n = rR + (m + n)r2 ,
rR = 20 nm are the polysome effective radius and the mRNA
radius of gyration [11], respectively. Finally, ϕ(x) denotes the
axial density of DNA length, see Supplemental Material Sec-
As expected from the observed segregation of mRNAs,
tion S2 for details. the T and B ribosomes are markedly excluded from the
nucleoid, and there is a net poleward flux of T and B
Rl ribosomes. Notably, the effect of excluded volume in the
of the total number Nm,n = 0 dxρm,n (x) of mRNAs of nucleoid is so strong that mRNAs and their associated ri-
species m, n. Most mRNAs are loaded with ∼ 10 T ribo- bosomes flow from a low- to a high-concentration region.
somes and ∼ 2 B ribosomes. These loading numbers can By contrast, F ribosomes are small enough to freely pen-
be understood as follows: The number of T ribosomes per etrate the nucleoid, and a flux of F ribosomes is driven
mRNA is NT /NmRNA ∼ 80%Ntot /NmRNA ∼ 10; a similar by the gradient of these ribosomes from the poles to the
estimate yields ∼ 2 B ribosomes per mRNA. Since each nucleoid. Overall, these results illustrate and quantify a
ribosome has a linear size of ∼ 20 nm [10], the effective “circular” flux within the ensemble of T, B, and F ribo-
size of an mRNA molecule with ∼ 10 bound ribosomes is somes [13], cf. Fig. 1: First, multiple F ribosomes bind
significantly larger than the pore size of the DNA mesh to mRNAs made in the cell nucleoid. Each mRNA is thus
in the nucleoid, which we estimate to be ∼ 50 nm, see loaded with ∼ 10 actively translating T ribosomes, and
Section S2 for details. Thus, the majority of mRNAs ex- ∼ 2 transiently bound B ribosomes to become a polysome.
perience strong excluded-volume effects which push them Second, the effects of excluded volume in the nucleoid re-
out of the nucleoid region. sult in a net flow of these polysomes to the cell poles.
Since mRNAs are created by transcription in the nu- Once polysomes reach the poles, they ultimately decay
cleoid but end up segregated away from the nucleoid, and free their ribosomes. This “pumping” of T and B
ribosomes from the nucleoid to the poles results in an ex-
there must be a flux of mRNAs toward the cell poles.
cess of F ribosomes at the poles, and thus in a diffusive
Given that new mRNAs are rapidly loaded with T ribo-
somes at a rate konT
NF /(2l) = 6×10−4 µm/s×1200/3µm ' return flux of F ribosomes to the nucleoid.
0.24/s, implying full occupation by ∼ 10 T ribosomes The existence of a steady ribosome circulation implies
in ∼ 3 s, the poleward flux of mRNAs carries with it a that there must be an external source of energy driving
poleward flux of ribosomes. Since ribosomes are con- these circular fluxes. There are two possible candidates
served in our model, reflecting the long half-life of ri- within our model: Process (1) is the non-equilibrium
bosomal components [12], there must be a compensating creation and degradation of mRNAs, and Process (2) is
flux of F ribosomes from the poles toward the nucleoid. mRNA and F-ribosome binding in the nucleoid and sub-
In Fig. 4 we show the steady-state concentration of F ri- sequent expulsion from nucleoid by excluded-volume ef-
bosomes cP F (x), the concentrations of T and B ribosomes,
P fects. Process (1) should be strictly dependent on new
cT (x) = m,n n ρm,n (x), cB (x) = m,n m ρm,n (x), the mRNA production, whereas Process (2) should persist
flux JF = −DF [dcF (x)/dx vF (x)−cF (x) dvF (x)/dx] of F ri- even in the limit where the mRNA production and degra-
bosomes, and the fluxes JT and JB of T and B ribosomes. dation rates are both low, with the total number of mR-
 4

demonstrating that the total mRNA profile closely re-

cT (x)
cB (x) sembles that of T ribosomes (cf. Figs. 3, 4). In principle,
104 cF (x) measurement of mRNA localization is possible in living
cells via engineered mRNA constructs that bind to flu-
c(x) (1/µm)

orescent proteins [5] or in fixed cells using fluorescence

in situ hybridization [6]. While so far these approaches
have only been applied to specific mRNA transcripts,
103
our model makes clear predictions for strong, genome-
wide mRNA localization away from the nucleoid. Be-
sides mRNA localization at the cell poles, other localiza-
tion patterns may include concentration of mRNAs near
102 the cell membrane throughout the cell and between the
0 0.25 0.5 0.75 1
x/l two nucleoid lobes during cell division, both places where
slow-diffusing T ribosomes have been observed [2].
JT (x)
JB (x) Our model also reveals a “circulation” of ribosomes
JF (x) within the cell driven by the flux of newly synthesized
mRNAs from the nucleoid to the poles. The most rele-
FIG. 4: Steady-state ribosome concentrations and ribosome vant biological implication of this circulation would be its
fluxes. Top: Concentrations cT (x), cB (x), and cF (x) of trans- effect on ribosome efficiency, since the growth rate of E.
lating (T), transiently bound (B), and free (F) ribosomes in coli under a wide range of conditions is limited by protein
the right half of the cell (cf. Fig. 1). Bottom: Fluxes of T, translation [15, 16]. In this regard, a reasonable definition
B, and F ribosomes along the cell’s long axis depicted in the of ribosome efficiency is the average number of proteins
top panel. The arrow length is proportional to local ribosome translated per unit time per ribosome. To extract this
flux, and the arrows in the legends correspond to a flux of
quantity from our model, we have to consider the fate of T
20/s.
ribosomes that are released when their mRNA degrades.
Since mRNAs are predominantly degraded from their 5’
end to their 3’ end [17], it is natural to assume that these
NAs fixed and equal to NmRNA . Therefore, we varied T ribosomes are able to complete translation, which also
the mRNA production and degradation rates together, occurs from the 5’ end to 3’ end. To this end, we estimate
keeping the total mRNA number constant: The circula- the total rate of protein production per cell in our model
tion vanished as the mRNA rates slowed (Fig. S3), thus as the total rate of F →T ribosome binding, i.e. Γ =
identifying Process (1), the flux of new mRNAs from nu- PM PN −1 R l ∗
2 kon
T
m=0 n=0 0 dxcF (x)[ρm,n (x)+ρm,n (x)], and de-
cleoid to pole, as the driver of ribosome circulation. This
fine the ribosome efficiency as ε = Γ/Ntot = 2.2 × 10−2 /s.
conclusion is confirmed by an analytical estimate for the
Interestingly, if mRNA degradation was instead predomi-
poleward flux of T and B ribosomes, which is shown to be
nantly 3’ to 5’, then degraded mRNAs would release their
proportional to the mRNA production rate, see Section
ribosomes prior to completion of translation; the ribo-
S5 for details.
some efficiency ε in our model would then be reduced
PM PN Rl
Before discussing the biological implications of our re- by 2β m=0 n=0 n 0 dx ρm,n (x)/Ntot = 2.6 × 10−3 /s,
sults, it is worth considering that mRNA transcription implying a ∼ 12% decrease in efficiency.
takes a finite amount of time ∼ 1 min, during which
For the biologically relevant case of dominant 5’ to 3’
nascent mRNAs are bound to DNA while being elongated
mRNA degradation, Fig. 5 shows the effect of ribosome
[14]. We therefore extended our model to include these
circulation on ribosome efficiency. The circulation is am-
nucleoid-bound mRNAs, whose axial densities we denote
plified by simultaneously increasing mRNA production
by ρ∗m,n (x): these mRNAs turn into free mRNAs at a rate
and degradation rates to increase the mRNA flux, see
γ = 1/(1 min) (see Section S6 for details). The extended
Section S5. This mRNA flux increases ribosome efficiency
model confirms the picture obtained by considering only
by carrying T and B ribosomes to the cell poles, then re-
one mRNA species: in particular, the solution still dis-
leasing them exactly where the mRNA density is highest,
plays strong mRNA localization away from the nucleoid
and the ribosomes can rebind to mRNAs most quickly.
(cf. Fig. S4) and a net circular flux for T, B, and F
ribosomes (cf. Fig. S5). However, a more significant contribution to ribosome
efficiency stems directly from the existence of the nu-
Discussion—In this Letter, we presented a minimal
cleoid. Compared to a cell with no nucleoid, i.e. with
reaction-diffusion model for the transcriptional and trans-
the DNA uniformly distributed, the overall ribosome ef-
lational machinery in E. coli. The experimental obser-
ficiency is increased by 0.2%. This increase of efficiency
vation that ribosomes are markedly localized at the cell
arises from the enhanced density of F ribosomes at the
poles, and thus segregated from the nucleoid located at
poles due to the excluded-volume effects of the nucleoid
the cell center [2], suggests that mRNAs are also likely
(cf. Figs. 2 and S2), and it persists even in the absence
to be segregated from the nucleoid. Our model confirms
of an mRNA flux. Overall, our analysis shows that the
and quantifies this prediction, indicating that ∼ 90% of
intracellular organization of transcription and translation
mRNAs are typically located outside the nucleoid, and
 5

β0 [7] P. Montero Llopis, O. Sliusarenko, J. Heinritz, and

0.04 C. Jacobs-Wagner. In vivo biochemistry in bacterial cells
0.04 using FRAP: Insight into the translation cycle. Biophys.
Ribosome efficiency ε (1/s)

J., 103(9):1848, 2012.

ε (1/s) [8] H. Chen, K. Shiroguchi, H. Ge, and X. S. Xie. Genome-
0.03 wide study of mRNA degradation and transcript elonga-
tion in Escherichia coli. Mol. Syst. Biol., 11(1):781, 2015.
0.03 0.02 [9] M. Siwiak and P. Zielenkiewicz. Transimulation - Protein
0 0.02 0.04 Biosynthesis Web Service. PLoS ONE, 8(9):e73943, 2013.
β (1/s) [10] M. Kaczanowska and M. Rydén-Aulin. Ribosome bio-
genesis and the translation process in Escherichia coli.
Microbiol. Mol. Biol. R., 71(3):477, 2007.
ε0 [11] A. Gopal, Z. H. Zhou, C. M. Knobler, and W. M. Gel-
0.02 bart. Visualizing large RNA molecules in solution. Rna,
18(2):284, 2012.
10−3 10−2 [12] K. Piir, A. Paier, A. Liiv, T. Tenson, and Ü. Maiväli.
mRNA degradation rate β (1/s) Ribosome degradation in growing bacteria. EMBO Rep.,
12(5):458, 2011.
[13] J. C. Weisshaar. Private communication, 2014.
FIG. 5: Dependence of the ribosome efficiency ε on the mRNA
[14] I. Golding. Private communication, 2015.
degradation rate β. Efficiency is estimated from the total rate
[15] M. Scott, C. W. Gunderson, E. M. Mateescu, Z. Zhang,
of ribosome translation initiation (see text). Here, the mRNA
and T. Hwa. Interdependence of cell growth and
degradation rate and the overall mRNA production rate are
gene expression: Origins and consequences. Science,
increased together to keep the total mRNA number constant.
330(6007):1099, 2010.
The β axis is in logarithmic scale, and the wild-type mRNA
[16] M. Scott, S. Klumpp, E. M. Mateescu, and T. Hwa.
degradation β0 = 1/(5 min) and the corresponding efficiency
Emergence of robust growth laws from optimal regula-
ε0 = 2.2 × 10−2 /s are marked both on the axes and as a dot.
tion of ribosome synthesis. Mol. Syst. Biol., 10(8):747,
Inset: same plot as in the main panel, with the β axis in linear
2014.
scale to illustrate the linear dependence of ε on β.
[17] D. W. Selinger, R. M. Saxena, K. J. Cheung, G. M.
Church, and C. Rosenow. Global RNA half-life analysis in
Escherichia coli reveals positional patterns of transcript
degradation. Genome Res., 13(2):216, 2003.
contributes significantly to ribosome efficiency.
Acknowledgments We thank E. Bonomi, Z. Gitai, I.
Golding, J. C. Weisshaar, and M. Wilson for valu-
able conversations and suggestions. Research supported
by National Science Foundation grant PHY-1305525
and GRFP DGE-1148900, and by US National Insti-
tutes of Health grants NIDA DP1DA026192 and NIAID
R21AI102187.

[1] P. J. Lewis. Bacterial subcellular architecture: recent ad-

vances and future prospects. Mol. Microbiol., 54(5):1135,
2004.
[2] S. Bakshi, A. Siryaporn, M. Goulian, and J. C. Weis-
shaar. Superresolution imaging of ribosomes and RNA
polymerase in live Escherichia coli cells. Mol. Microbiol.,
85(1):21, 2012.
[3] A. Sanamrad, F. Persson, E. G. Lundius, D. Fange, A. H.
Gynnå, and J. Elf. Single-particle tracking reveals that
free ribosomal subunits are not excluded from the Es-
cherichia coli nucleoid. Proc. Natl. Acad. Sci. U.S.A.,
111(31):11413, 2014.
[4] J. A. Bernstein, P.-H. Lin, S. N. Cohen, and S. Lin-Chao.
Global analysis of Escherichia coli RNA degradosome
function using DNA microarrays. Proc. Nat. Acad. Sci.
USA, 101(9):2758, 2004.
[5] I. Golding and E. C. Cox. RNA dynamics in live
Escherichia coli cells. Proc. Natl. Acad. Sci. U.S.A,
101(31):11310, 2004.
[6] M. Pilhofer, M. Pavlekovic, N. M. Lee, W. Ludwig, and
K.-H. Schleifer. Fluorescence in situ hybridization for
intracellular localization of nifH mRNA. Syst. Appl. Mi-
crobiol., 32(3):186, 2009.
 Supplemental Material for “Spatial organization of bacterial transcription and
translation”
Michele Castellana1, 2 and Ned S. Wingreen2
1
Joseph Henry Laboratories of Physics, Princeton University, Princeton, New Jersey 08544, United States
2
Lewis-Sigler Institute for Integrative Genomics, Princeton University, Princeton, New Jersey 08544, United States

Contents

S1. Diffusion equations in the presence of excluded volume 1

S2. Available-volume profiles 2

S3. Model with no transiently bound ribosomes 2

S4. mRNA species inside versus outside the nucleoid 2

S5. Estimate of ribosome flux 3

S6. Model including nucleoid-bound mRNAs 4

References 6

S1 Diffusion equations in the presence of excluded volume

In this Section we derive the diffusion equation for a population of F ribosomes, taking into account the effects of
excluded volume due to the nucleoid. The continuum limit of this equation can be derived starting from a discrete
version [1]. We divide the interval 0 ≤ x ≤ l, from midcell to the cell pole, into bins of width ∆x, where xi denotes the
position of bin i: Here, each bin represents a projection on the one-dimensional x axis of a three-dimensional section
of the cell obtained by slicing the cell perpendicularly to its long axis. The master equation for the F ribosome density
in bin i, cF (xi , t), can then be obtained by considering hopping of ribosomes between adjacent bins:

∂cF (xi , t)
= dF {[cF (xi−1 , t) + cF (xi+1 , t)]vF (xi ) − cF (xi , t)[vF (xi−1 ) + vF (xi+1 )]} , (S1)
∂t
where dF is the bare rate at which a ribosome hops from one site to another, and vF (x) is the fraction of available
volume at position x. The continuum limit of Eq. (S1) can be obtained by expanding cF (x, t) and vF (x) around x = xi ,
which yields
 2 
∂cF (x, t) ∂ cF (x, t) d2 vF (x)
= DF vF (x) − cF (x, t) , (S2)
∂t ∂x2 dx2

where DF = dF ∆x2 is the bare diffusion coefficient. The 1D free-ribosome flux at position x is given by
 
dcF (x) dvF (x)
JF (x) = −DF vF (x) − cF (x) . (S3)
dx dx

For steady state we assume no-flux boundary conditions at midcell and the cell pole, JF (0) = JF (l) = 0. Note that
in the absence of reaction terms, the steady-state solution of Eq. (S1) is

cF (x) = C vF (x), (S4)

where the constant C depends on the total number of F ribosomes.

 2

S2 Available-volume profiles

In this Section we estimate the available volume for ribosomes, mRNAs, and polysomes inside the nucleoid. To
estimate the DNA pore size inside the nucleoid, consider a DNA polymer of length L confined in a volume V . For
simplicity, suppose that the DNA is arranged to occupy the edges of a three-dimensional cubic lattice of total volume
3
V . The volume is divided into M cubic pores each with edge length a, so that
p V = M a . In addition, the DNA length
is related to the pore size by L = 3 a M , thus providing the estimate a = 3V /L. For a DNA length L = 1.5 mm and
nucleoid volume V = 1.2 µm3 [2], the estimated pore size is a ' 50 nm.
We now provide a rough, dimensional estimate of the available volume for an F ribosome inside the nucleoid.
Continuing to use the simple cubic lattice model for the nucleoid, we estimate the excluded volume inside each
cubic pore of DNA: as the ribosome approaches one of the lattice edges (which, by assumption, is occupied by
DNA), the excluded volume is given by the edge length a times the area within the pore from which the center
of mass of the ribosome is excluded when it approaches the edge, i.e. 1/4 of a circle with the ribosome’s radius
r ' 10 nm [3]. Multiplying by the total number of edges (12), we obtain an estimate for the excluded volume,
vexcl = 12 a πr2 /4 = 3 a πr2 . The available-volume fraction in the nucleoid, which we will denote by vFin , is then
expressed in terms of the ratio between the excluded volume and the pore volume as
vexcl L
vFin = 1 − κ 3
= 1 − κ πr2 . (S5)
a V
The numerical correction factor κ is introduced to improve our dimensional estimate, and is set by fitting our model to
the experimentally observed concentration profile of freely diffusing ribosomes. If we imagine dividing the cell into two
regions along the axial direction, the nucleoid region (in) and the polar region outside the nucleoid (out), then according
to Section S1 the concentration of freely diffusing ribosomes satisfies cin out
F /cF = vFin /vFout . Since the concentration of
freely diffusing ribosomes is roughly 10% larger outside the nucleoid than inside [4], we have vFin /vFout = vFin = 90%,
where vFout = 1 because there is no DNA in the out region. As a result, we obtain κ = 0.25 in Eq. (S5).
The available-volume fraction for mRNAs loaded with m B ribosomes and n T ribosomes can be estimated along
the same lines. Since the excluded volume for a single ribosome depends on its two-dimensional cross section πr2
rather than on its volume, we introduce an effective radius rm+n for an mRNA of species m, n, where rm+n is chosen
2
to reproduce the overall cross section of the mRNA along with its ribosomes: πrm+n = πrR2 + (m + n)πr2 , where
in
rR = 20 nm is a typical mRNA radius of gyration [5]. The mRNA excluded volume in the nucleoid vm+n is then
determined by replacing r with rm+n in Eq. (S5).
Finally, the smooth excluded-volume profiles vF (x), vm+n (x) shown in Fig. 2 were obtained from vFin , vm+nin
as
follows. We replaced L/V in Eq. (S5) with the local 1D density of DNA length ϕ(x) = dL/dV . The profile of
ϕ(x) ∝ 1/{1+exp[µ(x/l−1/2)]}, with µ = 20, was chosen to reproduce the observed axial DNA density and normalized
Rl
to the total DNA length per volume, (1/l) 0 dxϕ(x) = L/V [2]. Given that for mRNA species with large m + n the
excluded volume may be so large that vm+n (x) is negative in the nucleoid, whereas the flexibility of polysomes should
2
allow some penetration into the nucleoid region, we replaced the available-volume profile vm+n (x) = 1 − κ πrm+n ϕ(x)
2
with the linearly-equivalent, positive-definite form vm+n (x) = exp[−κ πrm+n ϕ(x)].

S3 Model with no transiently bound ribosomes

To show that the division of non-T ribosomes into F and B ribosomes has little effect on the results, in Fig. S1 we
show the solution of the reaction-diffusion equations (1), (2) with no B ribosomes: In the left panel we show the mRNA
densities ρn (x), labeled uniquely by the number n of T ribosomes, while in the right panel we show the concentrations
cT (x), cF (x) of T and F ribosomes, and the ribosome fluxes.

S4 mRNA species inside versus outside the nucleoid

In Fig. S2 we show the distribution of mRNA species in the nucleoid region and in the polar region as heat maps
in out
for the numbers Nm,n , Nm,n of mRNAs of species m, n for 0 ≤ x ≤ l/2 and l/2 ≤ x ≤ l, respectively.
 3

5 5,000 4 × 104
ρtot (x) cT (x)
200 ρn (x) cF (x)
4 ρ0 (x) 4,000
Nn

ρ20 (x)
100

ρtot (x) (1/µm)

ρn (x) (1/µm)

3 3,000

c(x) (1/µm)
0
0 5 10 15 20 104
n
2 2,000

4 × 103
1 1,000

0 0
0 0.25 0.5 0.75 1 0 0.25 0.5 0.75 1
x/l x/l
JT (x)
JF (x)

FIG. S1: Steady-state mRNA and ribosome distributions for the model with no transiently bound ribosomes. Left: Total
mRNA density ρtot (x) (red), and density ρn (x) of mRNAs with n translating (T) ribosomes (gray). The density ρ0 (x) of
ribosome-free mRNAs (green) and the density ρ20 (x) of mRNAs with the largest number of T ribosomes considered (blue) are
Rl
also shown. The profiles ρn (x), ρ0 (x), and ρ20 (x) are normalized to unit area. Inset: distribution Nn = 0 dxρn (x) of mRNAs
with n T ribosomes in the right half of the cell as a function of n. Right: On the top, we show the concentrations cT (x), cF (x) of
T and free (F) ribosomes, for the right half of the cell. Bottom: Fluxes of T and F ribosomes along the cell’s long axis depicted
in the top panel, where the arrow length is proportional to local ribosome flux, and the arrows in the legends correspond to a
flux of 14/s.

8 8 80
8
6 6 60
6
in out
4 Nm,n 4 40 Nm,n
m

2 2 2 20

0 0 0 0
0 5 10 15 20 0 5 10 15 20
n n

FIG. S2: Comparison of mRNA species inside versus outside the nucleoid. Left: heat map of the number of mRNAs inside the
in
R l/2
nucleoid Nm,n = 0 ρm,n (x) in the right half cell for mRNAs of species m, n as a function of the numbers m, n of transiently
bound and translating ribosomes, respectively. Here ρm,n (x) is the solution of Eqs. (1) and (2), l/2 is the position of the edge
of the nucleoid along the cell’s long axis (cf. Figs. 1, 2), and the model parameters are discussed in the main text. Right: same
out
Rl
plot as in the left panel for the number of mRNAs outside the nucleoid Nm,n = l/2 ρm,n (x).

S5 Estimate of ribosome flux

In this Section, we present an analytical estimate of the flux of mRNA-bound ribosomes flowing from the nucleoid
region toward one cell pole. The ribosome flux can be estimated by assuming that every newly synthesized mRNA
migrates from the nucleoid to the pole, and carries an average load of ribosomes, yielding an estimated ribosome flux

est αtot NT + NB
JT+B = , (S6)
2 NmRNA
Rl
where αtot = 2 0 dxα(x) ∼ 17/s is the total mRNA transcription rate in the nucleoid region, and the second factor is
the average number of ribosomes carried by each mRNA, with NT = 4.8 × 104 , NB = 1.1 × 104 , NmRNA = 5 × 103 .
est
Equation (S6) predicts a flow of Jrib ∼ 100/s, in rough agreement with our numerical result JT+B ∼ 20/s for the
fluxes of T, B ribosomes at the nucleoid edge shown in Figs. 4, S5. The factor of ∼ 5 discrepancy is not due to our
 4

JT (x)
JB (x) β = 1/(5 min)
JF (x)

β = 1/(10 min)

β = 1/(20 min)

FIG. S3: Fluxes of translating (T), transiently bound (B) and free (F) ribosomes along the cell’s long axis: The arrow length
is proportional to local ribosome flux, and the arrows in the legends correspond to a flux of 20/s. The fluxes are shown for
different values of mRNA degradation rate β, where β and the total mRNA production rate αtot are varied together, keeping
the total mRNA number constant.

est
estimate of the mRNA flux JmRNA ' αtot /2 ' 8.5/s, which agrees well with the numerical result JmRNA ' 7/s at the
edge of the nucleoid. Rather, the difference is due to the fact that mRNAs with fewer ribosomes diffuse faster due
to the strong dependence of the available volume factor vm,n on m + n, which means that most of the diffusive flux
of mRNAs is due to polysomes with a number of ribosomes substantially smaller than average. Thus our use of the
average ribosome loading in Eq. (S6) substantially overestimates the ribosome flux carried by mRNAs. However, the
est
estimate JT+B in Eq. (S6) does correctly capture the full numerical result that the ribosome flux is proportional to
the total mRNA production rate αtot . This proportionality leads us to conclude that ribosome circulation is driven
by the flux of new mRNAs from the nucleoid to the poles, and not by binding of mature mRNAs and F ribosomes in
the nucleoid with subsequent expulsion from the nucleoid due to excluded-volume effects—see the main text and Fig.
S3.

S6 Model including nucleoid-bound mRNAs

In this Section, we extend our model to include the fact that mRNAs remain bound to the nucleoid while being
transcribed. We introduce an additional set of mRNA species with m B ribosomes and n T ribosomes, whose densities
we denote by ρ∗m,n (x, t). The 1D equations for the nucleoid-bound mRNA densities are

∂ρ∗m,n (x, t) B
= − kon cF (x, t)ρ∗m,n (x, t) − m koff
B
ρ∗m,n (x, t) − kon
T
cF (x, t)ρ∗m,n (x, t) − n koff
T
ρ∗m,n (x, t)
∂t
B
+ kon cF (x, t)ρ∗m−1,n (x, t)+ koff
B
(m + 1) ρ∗m+1,n (x, t)+ kon T
cF (x, t)ρ∗m,n−1 (x, t)+ koffT
(n + 1) ρ∗m,n+1 (x, t)
+δm,0 δn,0 α(x) − β ρ∗m,n (x, t) − γ ρ∗m,n (x, t). (S7)

Since the mRNA species ρ∗m,n are bound to the nucleoid, there is no diffusion term. The terms in the first and second
lines in the right-hand side correspond to ribosome binding and unbinding as in Eq. (2). Finally, the first two terms
in the last line represent mRNA transcription and degradation, while the last term represents unbinding of an mRNA
from the nucleoid at rate γ. The dynamics of free mRNA species is still described by Eq. (2), where the mRNA-
transcription term in the last line is replaced by the source term γ ρ∗m,n (x), which represents a nucleoid-bound mRNA
becoming a free mRNA. We take the unbinding rate of nucleoid-bound mRNAs to be γ = 1/min, i.e. the inverse of
the average time for mRNA transcription [6].
We numerically solved the equations for ρm,n and ρ∗m,n at steady state. The resulting mRNA profiles and ribosome
concentrations are shown in Figs. S4 and S5, respectively. The results confirm the qualitative behavior of the model
with only free mRNAs, see Figs. 3, 4. The plots of ρtot (x) and ρ∗tot (x) in Fig. S4 show that the large majority
of mRNAs are free, and that free mRNAs are strongly segregated from the nucleoid, while nucleoid-bound mRNAs
are by construction localized in the nucleoid region. For free mRNAs, the larger m, n, the stronger the segregation
away from the nucleoid, and the overall free polysome distribution is still peaked around ∼ 10 T ribosomes and ∼ 2
B ribosomes per mRNA as in Fig. 3. For nucleoid-bound mRNAs, the vast majority of mRNAs have low m, n, with
the distribution peaked at m ∼ 1 and n ∼ 3 − 4, reflecting the somewhat lower concentration of F ribosomes in the
nucleoid (Fig. S5), and, more importantly, the limited time for ribosomes to bind before the mRNAs become free.
The localization pattern of these typical bound mRNAs matches the profile of the nucleoid, i.e. the profile of mRNA
production. By contrast, the spatial distributions of the few bound mRNA species with large m, n are strongly peaked
at the edge of the nucleoid, reflecting the higher density of free ribosomes outside versus inside the nucleoid.
Figure S5 shows the steady-state concentrations, cT (x) and cB (x), of T and B ribosomes loaded on free mRNAs,
as well as the concentration of F ribosomes cF (x), and the fluxes JT (x), JB (x), and JF (x). The figure also shows
 5

5 5,000 12 1,000
ρtot (x) 8 16
8 60 ρm,n (x) ρ∗m,n (x)
ρ0,0 (x) ρ∗0,0 (x) 6
4 6 4,000 ρ∗8,20 (x) 800
40 ρ8,20 (x) 4 ∗
8 Nm,n

m
4 Nm,n ρ∗tot (x)
m

20 8 2
ρm,n (x) (1/µm)

ρ∗m,n (x) (1/µm)

2

ρtot (x) (1/µm)

ρ∗tot (x) (1/µm)

3 3,000 0 0 600
0 0 0 5 10 15 20
0 5 10 15 20 n
n
2 2,000 400
4

1 1,000 200

0 0 0 0
0 0.25 0.5 0.75 1 0 0.25 0.5 0.75 1
x/l x/l

FIG. S4: Steady-state mRNA and polysome distributions for the extended model including both free and nucleoid-bound
mRNAs. Left: mRNA and polysome distribution for free mRNAs. Total mRNA density ρtot (x) (red) and density ρm,n (x) of
mRNAs with m transiently bound (B) ribosomes and n translating (T) ribosomes (gray). The density ρ0,0 (x) of ribosome-free
mRNAs (green) and the density ρ8,20 (x) of mRNAs with the largest number of B, T ribosomes considered (blue) are also shown.
The profiles ρm,n (x), ρ0,0 (x), and ρ8,20 (x) are normalized to unit area. Inset: distribution of mRNA species, shown as a heat
Rl
map of the number Nm,n = 0 dxρm,n (x) of mRNAs in the right half of the cell with m B ribosomes and n T ribosomes. Right:
∗
Rl
same as left panel for nucleoid-bound mRNAs, with Nm,n = 0 dxρ∗m,n (x) .

cT (x)
cB (x) 103
10 4 cF (x)
102
c∗ (x) (1/µm)
c(x) (1/µm)

101
3
10
100
c∗T (x)
10−1 c∗B (x)
102
0 0.25 0.5 0.75 1 0 0.25 0.5 0.75 1
x/l x/l
JT (x)
JB (x)
JF (x)

FIG. S5: Steady-state ribosome concentrations and ribosome fluxes for the extended model including both free and nucleoid-
bound mRNAs. Left: concentrations and fluxes for translating (T) and transiently bound (B) ribosomes loaded on free mRNAs,
and for free (F) ribosomes. On the top, we show the concentrations cT (x), cB (x), and cF (x) of T, B, and F ribosomes, for the
right half of the cell. Bottom: Fluxes of T, B, and F ribosomes along the cell’s long axis depicted in the top panel, where
the arrow length is proportional to local ribosome flux, and the arrows in the legends correspond to a flux of 35/s. Right:
Concentrations c∗T (x), c∗B (x) of T and B ribosomes loaded on ribosome-bound mRNAs.

P
the concentrations of T and B ribosomes loaded on nucleoid-bound mRNAs, c∗T (x) = m,n n ρ∗m,n (x) and c∗B (x) =
P ∗
m,n m ρm,n (x), respectively. While T and B ribosomes loaded on free mRNAs are strongly excluded from the
nucleoid, F ribosomes can penetrate the nucleoid region. In addition, there is a net poleward flux of T and B
ribosomes loaded on free mRNAs and a net flux of F ribosomes toward the cell center, in agreement with the model
with only free mRNAs. Finally, T and B ribosomes loaded on nucleoid-bound mRNAs are localized approximately
uniformly throughout the nucleoid region.
 6

[1] L. Brieger and E. Bonomi. A stochastic cellular automation simulation of the non-linear diffusion equation. Physica D,
47(1):159, 1991.
[2] S. Bakshi, A. Siryaporn, M. Goulian, and J. C. Weisshaar. Superresolution imaging of ribosomes and RNA polymerase in
live Escherichia coli cells. Mol. Microbiol., 85(1):21, 2012.
[3] M. Kaczanowska and M. Rydén-Aulin. Ribosome biogenesis and the translation process in Escherichia coli. Microbiol. Mol.
Biol. R., 71(3):477, 2007.
[4] A. Sanamrad, F. Persson, E. G. Lundius, D. Fange, A. H. Gynnå, and J. Elf. Single-particle tracking reveals that free
ribosomal subunits are not excluded from the Escherichia coli nucleoid. Proc. Natl. Acad. Sci. U.S.A., 111(31):11413, 2014.
[5] A. Gopal, Z. H. Zhou, C. M. Knobler, and W. M. Gelbart. Visualizing large RNA molecules in solution. Rna, 18(2):284,
2012.
[6] I. Golding. Private communication, 2015.