Cytokine xxx (xxxx) xxx

Contents lists available at ScienceDirect

Cytokine
journal homepage: www.elsevier.com/locate/cytokine

Regulation of STAT3 signaling in IFNγ and IL10 pathways and in their

cross-talk
U. Sarma *, 1, M. Maiti 1, A. Nair, S. Bhadange, Y. Bansode, A. Srivastava, B. Saha , D. Mukherjee *
National Centre for Cell Science, NCCS Complex, Ganeshkhind, SP Pune University Campus, Pune 411007, India

A R T I C L E I N F O A B S T R A C T

Keywords: The pro-inflammatory IFNγ-STAT1 pathway and anti-inflammatory IL10-STAT3 pathway elicit cellular responses
IFNγ primarily utilizing their canonical STATs. However IL10 mediated STAT1 and IFNγ mediated STAT3 activation is
IL10 also observed, suggesting crosstalk of these functionally opposing signaling pathways can potentially reshape the
STAT1
canonical dynamics both STATs and alter the expression of their target genes. Herein, we measured the dynamics
STAT3
SOCS1
of STATs in response to different doses of IL10 or IFNγ and in their co-stimulation and employed quantitative
modeling to understand the regulatory mechanisms controlling signal responses in individual and co-simulation
scenarios. Our experiments show, STAT3 in particular, exhibits a bell-shaped dose–response while treated with
IFNγ or IL10 and our model quantiatively captured the dose-dependent dynamics of both the STATs in both
pathways. The model next predicted and subsequent experiments validated that STAT3 dynamics would robustly
remain IL10 specific when subjected to a co-stimulation of both IFNγ and IL10. Genes common to both pathways
also exhibited IL10 specific expression during the co-stimulation. The findings thus uncover anovel feature of the
IL10-STAT3 signaling axis during pathway crosstalk. Finally, parameter sampling coupled to information theory
based analysis showed that bell-shaped signal-response of STAT3 in both pathways is primarily dependent on
receptor concentration whereas robustness of IL10-STAT3 signaling axis in co-stimulation results from the
negative regulation of the IFNγ pathway.

1. Introduction of activation and inhibition of canonical signaling axis is well-

Cytokines are small proteins or polypeptides that orchestrate cellular
communication and intracellular signaling to elicit a plethora of func­
tionally distinct cell fate decisions [1]. Their primary response is often
mediated through canonical signaling pathways, for instance,
interferon-gamma (IFNγ) and the Interleukin 10 (IL10) pathways that
trigger pro-inflammatory and anti-inflammatory cellular responses,
respectively, where the IFNγ-STAT1 axis and IL10-STAT3 signaling axis
primarily trigger the hallmark responses [2,3]. Studies link dynamics of
STATs in response to the cytokine signaling to distinct phenotypic out­
comes; IL10-activated sustained STAT3 dynamics triggers anti-
inflammatory (AIF) responses whereas transient IFNγ-STAT1 signaling
elicits pro-inflammatory (PIF) responses [4,5]. Although the regulation
characterized, the implication of signaling cross-talk [3,6] between
the functionally opposing pathways is less understood, for instance, IL10
stimulation also activates STAT1 and IFNγ stimulation activates STAT3
[2,3]; it is not clear how the quantitative dynamics of STATs, as well as
expression of their downstream target genes are regulated during
crosstalk where functionally opposing cues can potentially cross-
activate the canonical axis of each other.
In this study, our experiments on Balb/c derived peritoneal macro­
phages revealed distinct dynamics of STAT1 and STAT3 (also referred to
as S/1/3 in the next sections) in response to different doses of IFNγ or
IL10 stimulus. Generally, dose-dependent dynamics of signaling path­
ways offer regulatory insights and leads to the identification of non-
intuitive emergent features of signaling pathways [7], serving as

Abbreviations: IFNγ, Interferon gamma; IL10, Interleukin 10; STAT, Signal transducer and activator of transcription; SOCS, Suppressor of cytokine signaling;
TGFß, Transforming growth factor-beta; NF-ƙƙβ, Nuclear factor kappa-light-chain-enhancer of activated B cells; SHP-2, Src homology region 2 domain-containing
phosphatase-2; IL10R1, Interleukin 10 receptor 1; AUC, Area under curve; JAK, Janus kinase; IFN-LR, Interferon-ligand receptor; Tyk2, Tyrosine kinase 2; IL6,
Interleukin 6.
* Corresponding author.
E-mail addresses: uddipans@gmail.com (U. Sarma), mukherjee.debasri@gmail.com (D. Mukherjee).
1
Equal contribution.

https://doi.org/10.1016/j.cyto.2021.155665
Received 15 March 2021; Received in revised form 20 July 2021; Accepted 23 July 2021
1043-4666/© 2021 Published by Elsevier Ltd.

Please cite this article as: U. Sarma, Cytokine, https://doi.org/10.1016/j.cyto.2021.155665

U. Sarma et al.
excellent training and testing datasets for calibrating quantitative
mathematical models [8–10] aiming to identify regulatory processes
controlling the observed response [8–12]. Here, we built a simplified
mathematical model of IFNγ and IL10 signaling pathways and calibrated
the model to the STATs dynamics at different doses of both the stimuli.
We show STAT3 amplitude and dynamic range (area under curve
[AUC]), in particular, was highest for a medium (M) dose of IFNγ or
IL10, and at both low (L) and high (H) doses STAT3 responses were
inhibited, resembling a bell-shaped response. Analyzing the model we
identified the parameters critically determining the bell-shaped STAT3
responses in both pathways. To investigate the consequence of signaling
cross-talk we next predicted the effect of co-stimulation of IFNγ and IL10
on S/1/3 dynamics; the model predicted and later experiments validated
that STAT3 dynamics would robustly remain IL10 driven during the co-
stimulation. In fact, the dynamics of SOCS1, transcriptionally induced
negative feedback loop of IFNγ pathway also remained IL10 driven in
the co-stimulation. Additionally, in co-stimulation expression of target
gene common to both pathways either showed robust IL10-STAT3
driven expression as observed for interferon regulatory factor 7 (IRF7)
or, the effect of IFNγ-STAT1 signaling is significantly reduced which is
observed in the expression of Interferon-stimulated gene 15 (ISG15).
Analyzing the model, we could underpin the regulatory elements con­
trolling robust IL10-STAT3 signaling in co-stimulation. Together,
through data-driven mathematical modeling and validation experi­
ments, we show distinct dynamics of STAT1 and STAT3 and their
regulation in response to different doses of functionally opposing stim­
uli, uncovered predominance of IL10-STAT3 pathway over the IFNγ-
STAT1 pathway in co-stimulation and its effect on downstream gene
expression.
2. Materials and Methods:
Experimental protocol: Balb/c-derived macrophages were treated
with increasing doses of IL10 and IFNγ. The cells were then lysed and
processed for immunoblotting. Dose-response studies were the basis for
selection of the high (20 ng/ml), medium (5 ng/ml), and low (1 ng/ml)
doses of both cytokines, and kinetic studies were performed at these
three selected doses.
Immunoblotting and quantification of immunoblot data: Cells
were treated with the indicated cytokines. After stimulation, the cells
were washed twice with chilled PBS and lysed in NP-40 cell lysis buffer
[20 mM Tris (pH 7.5), 150 mM NaCl, 10% glycerol, 1 mM EDTA, 1 mM
EGTA, 1% Nonidet P-40, protease inhibitor mixture (Roche Applied
Science, Mannheim, Germany) and phosphatase inhibitor mixture
(Pierce)]. The lysates were centrifuged (10,500 rpm, 10 mins) and su­
pernatants were collected. Quantification of protein was performed
using the Bradford reagent (Pierce) and an equal concentration of pro­
tein in the Laemmli sample buffer was loaded on SDS–PAGE. The
resolved proteins were transferred onto the PVDF (Millipore) mem­
brane. The membranes were blocked using 5% non-fat dried milk in
TBST [25 mM Tris (pH 7.6), 137 mM NaCl, and 0.2% Tween 20].
Membranes were incubated with primary antibody overnight at 4˚C,
followed by washing with TBST. This is followed by incubation of
membranes with HRP-conjugated secondary antibodies. Immuno-
reactive bands were visualized with the luminol reagent (Santa Cruz
Biotechnology). The STAT1 antibody detected both splice variants of
-STAT1 (Tyr701), p91 STAT1α, and p84 STAT1β, here we detected
STAT1α. The STAT3 antibody we used is bound to tyrosine-
phosphorylated STAT3 molecules of both isoforms STAT3α(86 kDa)
and STAT3β (79 kDa). To ensure that the number of cells used for each
experiment is comparable across all experiments an approximate 3 ×
106 cells were plated per experiment. Signal obtained in response to
each ligand stimulated cell population is compared with an unstimu­
lated/control population and the ratio of signal intensity from stimu­
lated by unstimulated cell population is calculated. The label ‘control’
refers to the unstimulated control population(Figure S1a). The total
2
Cytokine xxx (xxxx) xxx
protein(phosphorylated + unphosphorylated) is also measured (for
instance total STAT1 is shown as T STAT1, Figure S1a or S1b) from the
same cell population, a protocol of quantification we have also used
elsewhere [13,14]. Next, each stimulated lane in a blot is divided by its
respective Total ([p STAT1time = ti]/[T STAT1time = ti ; ti = ith experi­
mental time point, i = [0, 3, 7, 15, 30, 60, 120] minutes and this ratio is
normalized to 1 for the unstimulated/control lane. Fold change over
control was used to calculate the phosphorylation amplitude at a
measured time point in response to IFNγ or IL10. For SOCS1 and SOCS3
quantification is done by dividing the SOCS concentration at various
time points post-stimulation by their respective unstimulated/control
values.
Gene expression analysis by qRT-PCR: RNA was isolated from the
treated macrophages using Tri reagent. RNA was checked for quantity
and quality by measuring the absorbance at 260/280 nm using a
nanodrop spectrophotometer. cDNA was prepared using RNA (2 μg)
which was used as a template for gene expression analysis by quanti­
tative real-time PCR. SYBR (TAKARA) green fluorescent dye was used to
check the real-time amplification, the program used for PCR was Step1:
Initial denaturation 95 ◦ C, 30 s. Step2: 45 cycles (denaturation, 95 ◦ C, 5
s, annealing and amplification 60 ◦ C 35 s), step3: melt curve. GAPDH
was used as a control gene to calculate the relative fold change in the
gene expression compared to the untreated control. The primers used in
this study are given in supplementary table TS1.
Reagents: Antibodies specific for p-STATI (Tyr-701), STAT1, p-
STAT3 (Tyr-705), and STAT3 were purchased from Cell Signaling
Technology (Danvers, MA) and those for SOCS1, SOCS3 and β-actin
were from Santa Cruz Biotechnology (Santa Cruz, CA). Soluble mouse
recombinant IL10 and IFNγ were procured from BD Biosciences (San
Diego, CA). RPMI 1640 medium, penicillin–streptomycin, and fetal calf
serum were purchased from Gibco®-ThermoFisher Scientific ((Life
Technologies BRL, Grand Island, NY). All other chemicals were of
analytical grade.
Animals and cell culture: BALB/c mice originally obtained from
Jackson Laboratories (Bar Harbor, ME) were bred in the experimental
animal facility of the National Centre for Cell Science. All animal usage
protocols were approved by the Institutional Animal Care and Use
Committee. Studies were performed using 6 to8 weeks old mice. 3%
thioglycollate-elicited peritoneal macrophages were isolated from Balb/
c mice and cultured in RPMI supplemented with 10% fetal bovine serum
(FBS). After the cells adhered, they were washed with PBS to remove the
non-adherent population and maintained for 48 hrs in a humidified CO2
incubator at 37 ◦ C. The cells were serum-starved (by addition of media
with 0.2% FBS) before stimulation.
Mathematical modeling and analysis of IFNγ and IL10 pathway:
Both FNγ and IL10 pathways were built as one model such that prefer­
entially either the IFNγ or IL10 pathways could be turned on (individual
ligand stimulation) or both pathways could be turned on simultaneously
(co-stimulation). Details of the model development, calibration, Monte-
Carlo sampling, entropy calculation, model validation steps are elabo­
rated in Supplementary file 1.
3. Results
3.1. STAT1 and STAT3 phosphorylation in response to different doses of
IL10 and IFNγ:
Peritoneal macrophages obtained from BALBc mice were stimulated
with increasing doses (0.5, 1.0, 2.5, 5, 10, 20, and 40 ng/ml) of either
IL10 or IFNγ ligands. We observed that STAT1 phosphorylation
increased proportionally to the increasing dose of IFNγ stimulation till
the 20 ng/ml, followed by a decrease; STAT3 phosphorylation in
response to IFNγ treatment was observed to be highest at 10 ng/ml
(Fig. 1a). IL10 induced STAT1 phosphorylation significantly increased
only at 2.5 and 5 ng/ml of IL10 and STAT3 phosphorylation showed
comparable phosphorylation profiles for higher doses of IL10 (Fig. 1b).
U. Sarma et al. Cytokine xxx (xxxx) xxx

Fig. 1. Dose-response and kinetic studies of STAT1 and STAT3 phosphorylation upon IL10 or IFNγ stimulation. 48 hrs rested Balb/c derived peritoneal
macrophages, cultured in RPMI 1640 with 10% fetal bovine serum were subjected to serum starvation for 3hrs. The cells were stimulated with increasing con­
centrations (0.5 ng/ml, 1.0 ng/ml, 2.5 ng/ml, 5.0 ng/ml, 10.0 ng/ml, 20.0 ng/ml, 40.0 ng/ml) of (a) recombinant IFNγ protein or (b) recombinant IL10 protein for
15 min then washed with ice-cold phosphate-buffered saline and lysed with RIPA lysis buffer containing protease and phosphatase inhibitors. The cell lysates were
further processed for immunoblotting and probed for phosphorylated and total STAT1 and STAT3 proteins. (c) Bar plots show STAT1 phosphorylation at low (1 ng/
ml), medium (5 ng/ml), and high (20 ng/ml) doses of IFNγ stimulation(L, M, and H respectively) at 6 different time points ranging from 3 min to 120 min. ‘Fold
change’ in the y axis represents the change in STAT1 phosphorylation corresponding to its baseline value where the baseline is normalized to 1. (d) STAT3 phos­
phorylation at L, M, H doses of IFNγ stimulation at different timepoints. (e) Temporal profile of STAT1 phosphorylation shown for L, M, H doses of IL10 stimulation.
(f) STAT3 phosphorylation at L, M, H doses of IL10 stimulation. (g) Fold changes in SOCS1 expression w.r.t its corresponding baseline value are shown in response to
M dose of IFNγ and IL10. (h) Fold change in SOCS3 expression in response to M dose of IFNγ and IL10. Data in (c-h) represents mean ± s.d. from three replicates.
From (c)-(g) the “Fold change” in the y-axis represents the relative change observed corresponding to the respective baseline values which is equal to 1 at time = 0
min, shown as Control in the bar plots.

For both IFNγ and IL10 stimulations of 0.5 ng/ml and 1 ng/ml, weak
induction of their respective non-canonical STATs (STAT1 in IL10 and
STAT3 in IFNγ stimulation) was observed; at 5 ng/ml both canonical and
non-canonical STATs have high phosphorylation, and, at 20 ng/ml non-
canonical STATs’ amplitude are inhibited in both pathway. Guided by
this dose–response behavior we next investigated the dynamics of the
STATs in both the functionally opposing pathways. For this, we selected
three doses of IFNγ and IL10: low (L), medium (M), and high (H), which
are respectively 1 ng/ml, 5 ng/ml, and 20 ng/ml.
Cells were stimulated with L, M, and H doses of IFNγ and IL10, and S/
1/3 dynamics were captured at different time points. For IFNγ treatment
STAT1 phosphorylation increases with dose strength (Fig. 1c). However,
STAT3 phosphorylation in response to IFNγ stimulation exhibits a
distinct dose-dependent behavior- at M dose STAT3 is rapidly phos­
phorylated to a high amplitude coupled to faster dephosphorylation
whereas at L and H doses STAT3 phosphorylation amplitude is much less
compared to the M dose (Fig. 1d). In IL10 stimulation, STAT1 phos­
phorylation is comparable in both M and H doses, especially at the late
time (120 min, Fig. 1e). A bell-shaped dose–response in STAT3
phosphorylation, although less pronounced compared to IFNγ stimula­
tion, was also observed in response to IL10 (Fig. 1f, maximum phos­
phorylation is observed distinctly for M dose from 15 min).
Representative immunoblots are shown in Supplementary Figure S1.
Such dose-dependent inhibition of response in signaling pathways in­
dicates the presence of negative regulator(s) that can potentially inhibit
signaling at various stages of signaling [15,16]. For instance, SOCS1 is a
commonly observed negative regulator of the IFNγ-STAT1 pathway
which is transcriptionally induced and it inhibits IFNγ signaling at the
receptor level by sequestering the receptor and blocking its access to the
signaling pathways downstreams [18,20], hence we also captured dy­
namics of SOCS1 expression for M dose of IFNγ (Fig. 1g). We observed,
upon IFNγ stimulation, SOCS1 induction remains close to its basal value
(Fig. 1g), but IL10 stimulation leads to a relatively stronger SOCS1 in­
duction (Fig. 1g). Studies show, SOCS3 can potentially inhibit IFNγ
signaling when SOCS1 is silenced, but the inhibitory strength of SOCS3
is negligible compared to SOCS1 when both the SOCS are present in the
system [33]. The dynamics of SOCS3 induction were measured as a
target gene in both pathways (Fig. 1h).

3
U. Sarma et al. Cytokine xxx (xxxx) xxx

3.2. The quantitative model captures IL10 and IFNγ dose-dependent
dynamics of STAT3 and STAT1
To understand the regulatory processes controlling the observed S/
1/3 dynamics in response to the pro-inflammatory (IFNγ) and anti-
inflammatory (IL10) stimuli, we built a mathematical model
comprising both the pathways. The kinetic data of S/1/3 phosphoryla­
tion at different doses of IL10 and IFNγ was used for model calibration.
Both the pathways comprised of three modules
I. Receptor activation module
II. STAT1 and STAT3 phosphorylation module
III. Transcriptional modules where SOCS1 and SOCS3 are induced
upon IFNγ and IL10 stimulation.
3.3. IFNγ pathway
Fig. 2a schematically shows the structure of the IFNγ pathway as a
minimal model. The model has a simplified step of receptor activation
that lumps the details of the interaction between IFNγ Receptor (IFN-R)
and the Janus kinases JAK1 and JAK2 [17] to a one-step binding and
activation process [19]. As receptor ligation and complex formation are
frequently observed as reversible [17,19] we modeled the receptor
activation step as a reversible process. Following the ligand-receptor
binding, the activated receptor complex (IFN-LR) phosphorylates the
transcription factors STAT1 and STAT3. Both the STATs undergo
dimerization and form transcriptionally active complexes [17,21],
which in turn induces target genes such as SOCS1 and SOCS3. Tran­
scriptional induction of target genes was implemented using Hill

Fig. 2. Quantitative modeling of STAT1 and STAT3 dynamics. (a) Schematic representation of the IFNγ pathway. The blunt heads with solid lines represent
catalysis; arrowheads with solid lines represent binding, unbinding, phosphorylation, and dephosphorylation reactions; blunt-headed dashed lines represent tran­
scriptional induction. In IFNγ pathway ligand (IFN) binding the receptor (IFNR) forms an active signaling complex (IFN-LR) which triggers the activation of STAT1
(STAT1 → STATp) and STAT3 (STAT3 → STAT3p) through phosphorylation. STAT1p and STAT3p undergo dimerization to become STAT1p_Dm and STAT3p_Dm,
respectively. Transcription induction of target genes such as SOCS1 and SOCS3 is shown. SOCS1 is a negative feedback regulator of IFNγ signaling which forms a
functionally inactive complex [SOCS1.IFN-LR] and inhibits the input signal at the receptor level. Tyrosine phosphatases such as SHP-2 which can dephosphorylate
both STAT1 and STAT3 are represented in our model as Phos. (b) schematics of IL10 signaling pathways in the model are shown. Notation of the arrows is kept the
same as described in the IFNγ pathway. Upon ligand (IL10) binding, the receptor (IL10R) forms an active complex (IL10-LR) which activates both STATs. At the
transcriptional level induction of SOCS1 and SOCS3 are represented. IL10Ri represents an inhibitor of IL10 signaling which acts by sequestrating and subsequently
degrading the active IL10 receptor. (c) IFNγ pathway in the model was calibrated to STAT1 and STAT3 activation dynamics using three different doses of both
stimuli- 1 ng/ml(L), 5 ng/ml(M), and 20 ng/ml(H). The pathway was also calibrated to the dynamics of SOCS1 and SOCS3 induction at M dose. (d) IL10 pathway in
the model was calibrated to STAT1, STAT3 phosphorylation dynamics at L, M, and H doses and SOCS1, SOCS3 expression dynamics at M dose of IL10. In both (c) and
(d) the lines represent model trajectories and blank circles represent respective experimental measurements. As indicated, the y axis of figures (c) and (d) shows the
fold change in phosphorylation (for STAT1 and STAT3) w.r.t the total value (represented as STAT3T for STAT3, for instance) at a measured time point. For the SOCS,
SOCS1B for instance represents the SOCS1 expression at time = 0 min, so at other time points post-stimulation, expression change in SOCS1 is calculated by
normalizing it to SOCS1B.

4
U. Sarma et al.
functions [10]. Dephosphorylation of S/1/3 was assumed to be carried
out by constitutively present phosphatases such as SHP2 [22,23] which
is represented as “Phos’’ in our model.
3.4. IL10 pathway
Fig. 2b shows the schematics of IL10 receptor-mediated phosphor­
ylation of STATs and the transcriptional induction of the SOCS. Similar
to the IFNγ pathway, the steps of IL10 receptor 1 (IL10R1) and receptor
2 (IL10R2) binding to JAK1, Tyk2 kinases leading to the formation of an
active signaling complex [24] were simplified to one step activation-
deactivation process. S/1/3 activation was carried out by the active
IL10 receptor and their dephosphorylation was assumed to be carried
out by constitutive phosphatases such as SHP2 [23]. Negative regulation
of IL10 signaling at the receptor level was considered to be carried out
by a constitutive negative regulator of IL10 receptor like the Beta-TrCP-
Containing Ubiquitin E3 ligase that binds to and promotes degradation
of the active IL10 receptor [25].
3.5. Model calibration to IFNγ and IL10 pathways
We studied the responses of STAT1 and STAT3 to L, M, and H doses
of IFNγ and IL10 by calibrating the model to the measured dynamics of
the STATs in each pathway. Details of model calibration steps can be
found in Supplementary file 1. The calibrated model quantitatively
captures both S/1/3 dynamics in response to IFNγ (Fig. 2c, 1st, and 2nd
row, respectively) and IL10 (Fig. 2d, 1st, and 2nd row, respectively)
stimulation. The bell-shaped dose–response of STAT3 phosphorylation,
where STAT3 activation is maximum at M dose and phosphorylation is
suppressed in both L and H doses of both IFNγ (Fig. 2c, 2nd row) and
IL10 (Fig. 2d, 2nd row), was quantitatively captured by the model. The
model also captured SOCS1 and SOCS3 induction dynamics in response
to both stimuli (in the 3rd row of both Fig. 2c, and Fig. 2d). We measured
the SOCS dynamics in M dose as both the canonical and non-canonical
STATs are optimally activated in the M doses of both stimuli. Notably,
in response to IFNγ stimuli transcriptional induction of SOCS1 was
negligible (Fig. 2c, 3rd row), however, basal SOCS1 (denoted as
SOCS1B) was required for model calibration where SOCS1 negatively
regulates IFNγ signaling by blocking the receptor access to its substrate
[17]. SOCS1 expression in response to IL10 was stronger compared to
IFNγ stimulation (Fig. 2d, 3rd row). SOCS3 expression was modeled in
both pathway simulations scenarios as a target gene. Earlier studies
comparing IL10 and IL6 pathway show despite closely comparable early
STAT3 dynamics in response to IL10 or IL6 stimulation (until 45 min
post-stimulation) the expression level of multiple target genes exhibited
pathway specific expression [4], where some target genes were
expressed relatively strongly upon IL10 but modestly upon IL6 stimu­
lation whereas some other genes showed the reciprocal expression
profile [4]. This indicates common target genes of STAT3 in different
pathways can have pathway-specific induction rates. Our model
captured the IFNγ and IL10 pathways-specific dynamics of SOCS3 by
allowing differences in SOCS3 induction rates in both pathways during
the calibration process; a relatively higher SOCS3 induction rate in the
IL10 pathway (details in table TS2) captures the higher amplitude
SOCS3 expression (Fig. 2d, 3rd row) in IL10 stimulation.
During model calibration, the common signaling intermediates such
as STAT3 concentration a biochemical parameter (common between
both pathways) were constrained to have one common value in both
pathways such that the shared parameter/species concentration with
one best-fit value can simultaneously capture the STATs and SOCSs
trajectories in response to both stimulations. Next, we analyzed the
calibrated model to understand the emergence of bell-shaped STAT3
responses in both the functionally opposing pathways.
5
Cytokine xxx (xxxx) xxx
3.6. Regulators of bell-shaped STAT3 responses in IFNγ and IL10
pathways
The canonical IFNγ-STAT1 and the IL10-STAT3 pathway exhibited
distinct dynamics; STAT1 peak/maximum amplitude in M and H dose
are comparable, but STAT3 showed a decline in response at H dose of
IL10 and maximum STAT3 activation was observed at M dose. A bell-
shaped STAT3 activation with a much stronger peak was also
observed during its non-canonical activation upon IFNγ stimulation. To
understand how different variables in the model contribute to the
emergence of the observed bell-shaped STAT3 responses, and in rela­
tion, to identify the key regulator(s), we took the hypercube of best-fit
model parameters and generated a set of 20,000 distinct parameter
vectors (in the [0.2–5] fold range of best fit parameters, see method
details in Supplementary file 1) using Monte Carlo sampling [26,27].
Next, simulating the resulting models we captured the maximum (or
peak) amplitude as well as the area under the S/1/3 trajectories
[quantified as area under curve (AUC)], especially focusing on the
STAT3 trajectories in response to both IFNγ and IL10 stimulation. The
simulations and analysis reveal STAT3 can potentially exhibit a different
class of signal response in both pathways in the right parametric com­
binations, where instead of bell-shape response to increasing dose
STAT3 activation proportionally increases as a function of signal
strength. Fig. 3a and 3b show the relative distribution of representative
model components such as concentrations of receptors, STATs as well as
the induction rate of the negative regulator from the sampled parameter
population exhibiting a bell-shaped STAT3 responses in both IFNγ
(Fig. 3a) and IL10 (Fig. 3b) pathways. Similarly, Fig. 3c and 3d show the
distribution of the same model variables where STAT3 response is pro­
portional. Next, to understand the relative contribution of different
parameters determining either of the response types we next calculated
the entropy and subsequently the information content [26] of each
parameter (details in supplementary file 1, see information content
calculation section). This led to the identification of the critical pa­
rameters whose values are constrained in a specific range to generate a
specific response, in this case, bell-shaped or proportional response . The
analysis shows concentrations of IFNγ receptor (IFNR in our model) in
the IFNγ pathway (Fig. 3e) and IL10R in the IL10 pathway (Fig. 3f)
contain the maximum information for bell-shaped STAT3 responses. The
contribution of other key variables, such as STAT3 concentration itself,
or, the contribution of negative regulators of receptor signaling is rela­
tively smaller. Also, the maximum information of the single most sen­
sitive variable in the IFNγ pathway(IFNR, ~ 0.8 bits, Fig. 3e) is more
than two-fold than its counterpart (IL10R, ~ 0.38 bits, Fig. 3f) in the
IL10 pathway, implying, perturbations of IFNR can change the STAT3
response phenotype more dramatically compared to perturbation of
IL10R in IL10 pathway where the latter would require simultaneous
perturbation of multiple parameters to exhibit the similar extent of
changes in STAT3 response. This was observed in our simulations where
randomly replacing only IFNR values from an ensemble of proportional
responders to bell-shaped responders, or vice-versa, we could change
STAT3 response from bell-shaped to proportional, or vice-versa, but
with a higher frequency in the IFNγ pathway (data not shown).
3.7. Prediction and validation of the effect of co-stimulation
Both IFNγ and IL10 stimulation leads to phosphorylation of STAT1
and STAT3 and SOCS1 transcriptional induction was observed in both
pathways although SOCS1 transcriptional induction is negligible in the
IFNγ pathway. However, basal SOCS1 acting as a stoichiometric inhib­
itor of the IFNγ receptor was necessary for successful model calibration
to the IFNγ pathway. SOCS1 was strongly induced upon IL10 stimulation
(compare SOCS1 induction in IFNγ and IL10 stimulation in 1 g). Hence,
to understand the consequences of IL10 induced SOCS1 on the IFNγ
pathway when cells are subjected to co-stimulation we used the cali­
brated model to predict S/1/3 signaling and SOCS1 induction dynamics
U. Sarma et al. Cytokine xxx (xxxx) xxx

Fig. 3. Regulators of bell-shaped STAT3 responses in IL10 and IFNγ pathways. Representative model variables corresponding to bell-shaped and proportional
responder classes are shown as box plots and information content in these parameters is shown as bar plots. Bell-shaped (a) and proportional (b) STAT3 responses to
IFNγ primarily vary in the expression level of IFNγ receptor (IFNR). (c) Information content analysis quantitatively shows the relative significance of the variables
shown in (a) and (b). In the IFNγ pathway, IFNR contains the maximum information. In the IL10 pathway, the bell-shaped responders (d) and proportional responders
(e) also show distinct ranges of IL10 receptor concentration (IL10R). (f) Information content analysis shows IL10R contains the highest information. The Y-axis from
(a)-(d) shows the normalized distribution of each parameter value where the sampled values ([Sampled]) of a parameter from 5000 simulations is divided by the
[Best-fit] value of that parameter.

in co-stimulation. We chose the M dose of both the ligand types as both
the canonical and non-canonical STATs were strongly activated in the M
dose. As many model parameters are often non-identifiable[28], we
used the approach of making predictions using multiple independently
fitted parameter vectors with comparable goodness of fit [14], which is
shown as the shaded area in Fig. 4a-c. The simulations predict and co-
stimulation experiments(experimental data are shown as circles in
Fig. 4a-c; representative immunoblots are shown in Supplementary
Figure S2A and S2B) validate that STAT3 signaling would remain IL10
driven when both IL10 and IFNγ stimuli are applied simultaneously
(Fig. 4a). We found a difference in STAT1 amplitudes between model
prediction and validation datasets (Supplementary Figure S2C), despite
the differences in absolute amplitude, the shape of the STAT1 trajectory
remained closely comparable to the observed data as a strong quanti­
tative match between model and data was obtained only by adjusting
the height of model trajectory using a scaling factor (Fig. 4b shows the
height adjusted model trajectory) without changing any of the fitted
model parameter values. SOCS1 induction dynamics in co-stimulation is
also predicted and validated to be IL1 0 driven (compare Fig. 4c with
SOCS1 dynamics in IL10-only stimulation-Fig. 2d, 3rd row). Next, we
hypothesized that the observed robust IL10 driven information encoding
at the signaling level might also influence downstream gene expression.
To investigate this we first chose a group of genes activated strongly in
response to IFNγ. We initially selected the genes IRF3, IRF7, IRF9 and
ISG15, which are frequently observed to be IFNγ stimulation dependent,
to validate this hypothesis. We observe strong expressions of IRF7
(Fig. 4d), and ISG15 (Fig. 4e) in response to IFNγ but IRF3 and IRF9
expressions were close to baseline in response to both stimuli (Supple­
mentary figure S2d). Hence, guided by the individual stimulation data
we selected IRF7 and ISG15 for the co-simulation experiment. Fig. 4d
and 4e respectively show fold change in expression of IRF7 and ISG15
during individual and co-stimulation scenarios. We found the expression
of IRF7 remains completely IL10 driven during the co-stimulation
(Fig. 4d) with a significant reduction in its amplitude compared to IFNγ
stimulation. ISG15 expression was also much stronger in IFNγ-only
stimulation compared to IL10-only or co-stimulation(Fig. 4e) where
costimulation resulted in the reduction of ISG15 expression to less than
half its IFNγ mediated expression. Although a significant inhibition in
ISG15 expression is observed in costimulation it still remains approxi­
mately 2 fold higher compared to IL10 only stimulation. This suggests a
gene-specific regulation of expression can be achieved by the cells via
co-stimulation. Regulatory processes controlling distinct inhibition of
different IFNγ target genes in co-stimulation can be explored in further
studies.
Next to underpin the key regulators in the model controlling the IL10
driven dynamics of STAT3 signaling in co-stimulation we sampled the
parameter space (Supplementary file 1, Monte-Carlo sampling) and
searched for potential combinations when STAT3 dynamics is no longer
robustly IL10 dependent during co-stimulation. Specifically, using five
thousand sampled parameter vectors we evaluated the maximum
amplitude and AUC of STAT3 in both IL10 and co-stimulation. Indeed,
we could identify parameter vectors that resulted in STAT3 signaling
unique to both IL10 and co-stimulation. Here STAT3 dynamics was
labeled as robust if 0.8 < STAT3AUC[IL10] /STAT3AUC[co-stimulation] < 1.2,
and 0.8 < STAT3max[IL10] /STAT3max[co-stimulation] < 1.2; otherwise
STAT3 response was considered as non-robust. Each boxplot in Fig. 5a
and 5b shows the normalized distribution of a parameter from a set of
5000 parameter vectors where the normalized values were obtained by

6
U. Sarma et al. Cytokine xxx (xxxx) xxx

Fig. 4. IL10 driven signaling dynamics and gene expression in co-stimulation of IL10 and IFNγ. Signaling dynamics of STAT3 (a), STAT1 (b) and SOCS1
induction dynamics (c) upon costimulation are predicted using multiple best fit models with comparable goodness of fit. The shaded area shows a prediction range
from 40 independent best fits and the black dots represent experimental observation. STAT3p and STAT3T in the y axis of (a) and (b) represent phosphorylated and
total STAT3. In the y-axis of (c) SOCS1B represents basal SOCS1 at time t = 0 min. The representative western blots can be found in Figure S2. (d)-(e)Expression of
IRF7 (d) and ISG15(e) in response to medium doses (5 ng/ml) of both IFNγ and IL10, individually, and in their co-stimulation is shown. “Fold change” represents the
change w.r.t to the respective pre-stimulated values.

Fig. 5. Regulators of robust STAT3 response in co-stimulation. Model variables subjected to Monte-Carlo sampling are shown. Each of the best fit model pa­
rameters is normalized to 1, so the range in the y-axis shows the normalized fold change for each parameter. Parameter distribution corresponding to robust(a) and
non-robust (b) STAT3 dynamics in co-stimulation is shown. (c) The top 5 parameters containing maximum information for STAT3 robustness are labelled. Also, for
comparison, parameters with maximum information for bell-shaped responses in individual stimulation(IFNR and IL10R; refer to Fig. 3e and 3f for details) are also
shown. (d) The parameters with maximum information that regulates robustness of STAT3 dynamics in co-stimulation. Randomly selecting and replacing only
“kf_feedback_IFNg”, the parameter with maximum information, from a set of 5000 parameter vectors with robust STAT3 response, to a set of another 5000 parameter
vectors exhibiting non-robust STAT3 response while keeping the rest of the parameters unchanged in the latter, we could achieve ~ 35% shift from non-robust to
robust responder types in the later models. Simultaneously replacing the top three parameters related to SOCS1 (kf_feedback_IFNg, SOCS1_Basal, SOCS1_induc­
tion_rate_IL10) resulted in more than 50% change from non-robust to robust STAT3 responses in co-stimulation.

7
U. Sarma et al.
dividing with their respective best-fit values. Subsequently, the entropy
of individual model variables was calculated for parameter sets gener­
ating either robust or non-robust STAT3 dynamics and the information
content of each parameter was subsequently calculated (method in
supplementary file 1). The parameters in the box plot (Fig. 5a and 5b)
corresponding to robust and non-robust STAT3 responses in co-
stimulation are shown in ascending order of their information content.
Fig. 5c shows the information content of all the model variables ar­
ranged in an ascending order. This analysis uncovered basal SOCS1
concentration (SOCS1_Basal), the binding rate for SOCS1 and activated
IFNγ receptor (kf_feedback_IFNg) together with SOCS1 induction by
IL10 pathway (SOCS1_induction_rate_IL10) comprises 3 of the top 5
parameters with the highest information content. Indeed, when we
replaced only these three variables from 5000 robust into 5000 non-
robust responder parameter vectors while keeping the rest of the
model variables unchanged in the latter (a random selection and
replacement protocol from the former to the latter group of parameters),
we could obtain more than 50% conversion of the non-robust responder
types to robust responder types (Fig. 5d).
Mechanistically, the reduction or increase in the basal concentration
of the IFNγ inhibitor (SOCS1) has a dramatic effect on the peak ampli­
tude and AUC of the STAT3 trajectory during the co-stimulation. During
co-stimulation, the induced SOCS1 inhibits the IFNγ pathway only after
the transcriptional time lag, but the early changes in STAT3 activation
resulting from variation in the concentration of SOCS1 (SOCS1_Basal)
dramatically alter the AUC of STAT3 trajectories. The binding rate of
SOCS1 and activated receptor complex IFN-LR determines the rate at
which active IFNγ ligand-receptor complex (IFN-LR) is sequestered
away as an inactive complex, initially by basal SOCS1, and later by its
induced counterpart which is a small fraction in IFNγ-only stimulation.
Thus, this analysis suggests, although the IL10 pathway transcription­
ally induces SOCS1 that can inhibit IFNγ signaling, the SOCS1_Basal
amount is a key determinant of the early signaling and the observed
robustness of IL10-STAT3 signaling axis; induced SOCS1 ensures the
maintenance of the IL10 pathway in the later times (greater than30
min), but, as implied in the information content comparison (Fig. 5c),
the contribution of IL10 induced SOCS1 in shaping the STAT3 robust­
ness is significantly less compared to basal SOCS1.
4. Discussion
The information encoded in the dynamics of signaling pathways
shapes growth, proliferation, apoptosis, or developmental lineage
commitments [29–38]. Recent studies show cytokine-induced signaling
networks can prioritize dynamic range over signal strength [39]
uncovering the significance of dynamic features of signaling pathways
and their relation to regulation of downstream phenotypic processes.
The STAT1 and STAT3 transcription factors studied here regulate a
range of cell fate decisions, many of which are functionally opposing.
For instance, activation of macrophages is enhanced by STAT1 and
inhibited by STAT3, cell proliferation is inhibited by STAT1 and pro­
moted by STAT3 and they act antagonistically to each other in T-helper
cell differentiation [40–46]. STAT1 induces the expression of death re­
ceptors to promote apoptosis and negatively regulates the expression of
several oncogenes [47–49] while STAT3 activation facilitates survival of
many primary tumor cells [47] and activates anti-apoptotic genes that
can promote proliferation of tumor cells [48,50]. Although such func­
tionally opposing behavior of S/1/3 is well established in various
physiological contexts, a quantitative data-driven understanding of how
STAT1 and STAT3 dynamics are regulated in response to functionally
opposing cues was not investigated from a systems-level perspective. In
complex scenarios like cross-talk involving pathways activated by
functionally opposing stimuli, it is necessary to better understand the
control mechanism underlying the distinct phenotypic outcomes in
health and disease so that better systems-level interventions can be
designed.
8
Cytokine xxx (xxxx) xxx
The dynamics of proinflammatory IFNγ-STAT1 or IL6-STAT3 path­
ways are transient and whereas IL10 induced STAT3 dynamics are
sustained and anti-inflammatory in nature [4]. Although both the STATs
are shown to be activated in response to IFNγ as well as IL10 stimulation
[2,3], quantitative differences between S/1/3 activation dynamics in
both pathways, relative robustness of the canonical axis during crosstalk
scenarios as well as the impact of such crosstalk in downstream gene
expression remained unexplored. In this study using a combination of
mathematical modeling coupled to experiments, we have addressed the
following questions-
1. How are the dynamics of canonical and non-canonical STATs regu­
lated in the IFNγ and IL10 pathways?
2. How functionally opposing cues applied simultaneously affect the
signaling dynamics in each pathway, and further, how the expression
of genes common to both pathways will get affected by such co-
stimulation?
We conducted experiments in Balb/c derived peritoneal macro­
phages where we captured the phosphorylation dynamics of STAT1 and
STAT3 at different doses of IFNγ and IL10 stimuli. STAT1 responses to
different doses of both the stimuli were proportional (or saturating), but
STAT3 in response to both stimuli exhibited a bell-shaped response with
intermediate (M) doses yielding maximum amplitude and activity
(measured as AUC). To capture the STATs dynamics and to underpin the
plausible regulatory mechanisms shaping the observed signal responses
we next constructed a simplified mathematical model and calibrated the
model to experimental data. The model quantitatively captured the
dynamics of S/1/3 in different doses of both stimuli.
We analyzed the calibrated model to identify the key variables
regulating the bell-shaped STAT3 responses in both pathways. To this
end, we generated thousands of parameter vectors around the best fit
parameter vector and searched for STAT3 responses that are not bell-
shaped. Indeed our analysis identified parameter vectors in both path­
ways, where the bell-shaped response is no longer obtained. Subse­
quently to determine the key differences between both the observed
(bell-shaped) and alternate (proportional) responder types we calcu­
lated the information content of the model variables, which unraveled,
in both pathways receptors concentration(IFNR in IFNγ pathway and
IL10R in IL10 pathway) has maximum information and they critically
determine the STAT3 responses characteristics. Earlier studies show
bell-shaped responses depend on receptor multimerization(dimeriza­
tion), where, at high doses ligand saturates all binding sites preventing
the multimerization and signal is inhibited [51,52]. Our simplified
model without explicit implementation of such receptor multi­
merization steps also identified, in similar lines, that receptor concen­
tration in both pathways but not the concentration of downstream
signaling elements or the concentration of feedback element, is the key
regulator of the observed bell-shape responses. Our analysis shows bell-
shaped response is observed at relatively low receptor concentration
(high ligand concentration favors ligand-receptor saturation) compared
to proportional dose–response responses that require significantly high
receptor concentration to avoid ligand saturation (compare IFNR con­
centration between Fig. 3a and 3c and IL10R concentration between
Fig. 3b and 3d, respectively).
Since STAT1 and STAT3 are shared between both IFNγ and IL10
pathways, we next studied the relative robustness of the canonical
signaling axis of each pathway during a co-stimulation and the effect of
such costimulation on the expression of target genes. Our model pre­
dicted and subsequent experiments validated that STAT3 signaling pri­
marily remains IL10 driven in co-stimulation. SOCS1 induction
dynamics also remained IL10 specific. At the level of gene expression,
we show the expression of two target genes, IRF7 and ISG15, expressed
uniquely in response to each stimulus. Functionally, IRF7 is a master
regulator of interferon signaling [53], the mediator of antiviral and
innate immunity [54], and necessary for apoptotic response mediated by
U. Sarma et al.
the tumor suppressor BRCA1 via the IFNγ pathway [55]. ISG15 is a
ubiquitin-like cellular gatekeeper that prevents the spread of cellular
threats via a process called ISGylation of target proteins, ensures inhi­
bition of viral or bacterial infection, inhibits protein translation under
cellular stress, and plays a dual role in cancer progression which often
involves interferon signaling [56–58]. We observed, both IRF7 and
ISG15 exhibited significantly higher expression in response to IFNγ only
but co-stimulation ensured an IL10 driven IRF7 expression; ISG15
expression is also reduced by more than 2 fold in co-stimulation
compared to IFNγ-only stimulation. This indicates, in various physio­
logical and pathophysiological settings a range of functional outcomes
can be controlled/altered due to such cross-talks. Next, analyzing the
mathematical model we could identify important regulators facilitating
robust IL10-STAT3 signaling in co-stimulation. We found that basal
concentration of SOCS1 and binding rate constant of IFN receptor to
SOCS1 (leading to the formation of inhibited signaling complex) pri­
marily facilitates robust IL10-driven STAT3 dynamics during the co-
stimulation.
We have further compared results from our model and experiments
to data generated elsewhere. Earlier studies comparing STAT3 phos­
phorylation dynamics in IL6 and IL10 stimulation in dendritic cells show
IL6 signaling returns to baseline by 120 min where IL10 signaling is
reduced to 53% of its maximum value in the same time (in [4] see
Fig. 3B-data for 5 ng/ml). Comparing our data for the same ligand dose,
in peritoneal macrophages we observed a 30% drop from the maximum
in STAT3 phosphorylation at 120 min (Fig. 2d, 2nd row, 2nd column).
With small parameterization steps without any structural changes, the
model can quantitatively capture the dynamics of STAT3 in dendritic
cells. IFNγ-stimulated STAT1 phosphorylation is transient and STAT1
returning to baseline transpires in a longer time scale [5] compared to
our experimental time scale. For instance, a study with bone marrow-
derived macrophages shows STAT1 phosphorylation returns to base­
line by 240 min post-stimulation [5]. Simulating the best-fit model using
the same IFNγ dose (125 ng/ml) used in [5] we observed a signal
termination time of 300 min for STAT1 (Supplementary figure S3a,
lower panel). Additionally, the impact of SOCS1 KO on STAT1 phos­
phorylation dynamics in IFNγ stimulation is also comparable between
our model and data derived from bone marrow derived macrophages [5]
(Supplementary figure S3a, compare upper panel (data) and lower panel
(model simulations)). Both WT phosphoSTAT1 dynamics and its change
in response to SOCS1 KO are closely comparable between the model and
data, so, we next conducted in-silico experiments with SOCS1 KO to
understand its implication on IL10 driven STAT3 signaling in co-
stimulation. The simulations show SOCS1 KO leads to significant
changes in STAT3p amplitude and dynamics (Supplementary figure
S3b). Mechanistically, although SOCS1 KO does not change STAT3 dy­
namics for IL10 only stimulation(Figure S3b), in co-stimulation, a
stronger STAT3 phosphorylation occurs due to the absence of SOCS1
mediated inhibition of IFNγ pathway at the receptor level, which results
in additional STAT3 phosphorylation via the IFNγ receptor, and,
STAT3p amplitude no longer remains only IL10 driven. The additional
comparisons of our calibrated model to data from Balb/c derived peri­
toneal macrophages to independently published studies affirms the
models general predictive power and it’s future potential to reproduce
and predict signal responses involving similar pathways in different
cellular or pathophysiological contexts.
In summary, our study explored how two functionally opposing
pathways with shared signaling intermediates encode input information
in the dynamics of their canonical and non-canonical signaling elements
during individual and in co-stimulation scenarios. We could identify
novel features of STAT3 signaling dynamics in both individual (IFNγ or
IL10) and co-stimulation conditions. During the cross-talks, modifica­
tions in the expression of some key IFNγ target genes were found to be
IL10 driven, suggesting cells can also potentially use the individual and
co-stimulation strategies to achieve a spectrum of expression profiles of
target genes specific to various cell types. Cross-talks and resource
9
Cytokine xxx (xxxx) xxx
sharing are frequently observed in the biological systems and advan­
tages of such sharing strategies as an evolutionary strategy have also
been studied[41]. Systems-level mechanistic studies in the future can
further uncover signaling specificity during crosstalk of such function­
ally opposing pathways and identify off-target physiological implica­
tions of perturbing components of such pathways and use the derived
knowledge to help design therapies targeting modulation of such
pathways.
CRediT authorship contribution statement
U. Sarma: Conceptualization, Methodology, Software. M. Maiti: . A.
Nair: . S. Bhadange: . Y. Bansode: . A. Srivastava: . B. Saha:
Conceptualization, Methodology, Software. D. Mukherjee: Conceptu­
alization, Methodology, Software.
Declaration of Competing Interest
The authors declare that they have no known competing financial
interests or personal relationships that could have appeared to influence
the work reported in this paper.
Acknowledgments
We would like to thank Bhaswar Ghosh from Max Planck Institute for
Terrestrial Microbiology, Marburg, Germany for critical comments on
the manuscript. DM would like to thank the Department of Science and
Technology, Gov. of India for supporting the study(grant number: LSBM-
37).
Data availability : Supporting data are included within the main
article and its supplementary files.
Appendix A. Supplementary material
Supplementary data to this article can be found online at https://doi.
org/10.1016/j.cyto.2021.155665.
References
[1] J.X. Lin, W.J. Leonard, Fine-Tuning Cytokine Signals, Annu Rev Immunol 26 (37)
(2019) 295–324.
[2] J. Wehinger, F. Gouilleux, B. Groner, J. Finke, R. Mertelsmann, R.M. Weber-Nordt,
IL10 induces DNA binding activity of three STAT proteins (Stat1, Stat3, and Stat5)
and their distinct combinatorial assembly in the promoters of selected genes, FEBS
Lett 394 (1996) 365–370, https://doi.org/10.1016/0014-5793(96)00990-8.
[3] Y. Qing, G.R. Stark, Alternative activation of STAT1 and STAT3 in response to
interferon-gamma, J Biol Chem 279 (2004) 41679–41685, https://doi.org/
10.1074/jbc.M406413200.
[4] D.A. Braun, M. Fribourg, S.C. Sealfon, Cytokine Response Is Determined by
Duration of Receptor and Signal Transducers and Activators of Transcription 3
(STAT3) Activation, J Biol Chem 288 (5) (2013) 2986–2993, https://doi.org/
10.1074/jbc.M112.386573.
[5] S. Wormald, J.-G. Zhang, D.L. Krebs, L.A. Mielke, J. Silver, W.S. Alexander, T.
P. Speed, N.A. Nicola, D.J. Hilton, The comparative roles of suppressor of cytokine
signaling-1 and -3 in the inhibition and desensitization of cytokine signaling, J Biol
Chem 281 (16) (2006) 11135–11143, https://doi.org/10.1074/jbc.M509595200.
[6] D.J. Klinke, N. Cheng, E. Chambers, Quantifying crosstalk among interferon-γ,
interleukin-12, and tumor necrosis factor signaling pathways within a TH1 cell
model, Sci Signal 5 (ra32) (2012), https://doi.org/10.1126/scisignal.2002657.
[7] L. Gaspers, P. Bartlett, A. Politi, P. Burnett, W. Metzger, J. Johnston, S. Joseph,
T. Höfer, A. Thomas, Hormone-induced calcium oscillations depend on cross-
coupling with inositol 1,4,5-trisphosphate oscillations, Cell Rep 9 (4) (2014)
1209–1218, https://doi.org/10.1016/j.celrep.2014.10.033.
[8] P. Vizán, D.S. Miller, I. Gori, D. Das, B. Schmierer, C.S. Hill, Controlling long-term
signaling: receptor dynamics determine attenuation and refractory behavior of the
TGF-β pathway, Sci Signal 6 (ra106) (2013), https://doi.org/10.1126/
scisignal.2004416.
[9] S. Ahmed, K.G. Grant, L.E. Edwards, A. Rahman, M. Cirit, M.B. Goshe, Data-driven
modeling reconciles kinetics of ERK phosphorylation, localization, and activity
states, Mol Syst Biol 10 (2014) 718, https://doi.org/10.1002/msb.134708.
[10] J. Strasen, U. Sarma, M. Jentsch, S. Bohn, C. Sheng, D. Horbelt, P. Knaus,
S. Legewie, A. Loewer, Cell-specific responses to the cytokine TGFβ are determined
by variability in protein levels, Mol Syst Biol 14 (2018), e7733, https://doi.org/
10.15252/msb.20177733.
U. Sarma et al.
[11] S.L. Spencer, S. Gaudet, J.G. Albeck, J.M. Burke, P.K. Sorger, Non-genetic origins of
cell-to-cell variability in TRAIL-induced apoptosis, Nature 459 (7245) (2009)
428–432, https://doi.org/10.1038/nature08012.
[12] Kallenberger SM, Beaudouin J, Claus J, Fischer C, Sorger PK, Legewie S, Elis R
(2014) Intra- and interdimeric caspase-8 self-cleavage controls strength and timing
of CD95-induced apoptosis. Sci Signal 7 :ra23. doi: 10.1126/scisignal.2004738.
[13] U. Sarma, A. Sareen, M. Maiti, V. Kamat, R. Sudan, S. Pahari, N. Srivastava, S. Roy,
S. Sinha, I. Ghosh, A.G. Chande, R. Mukhopadhyaya, B. Saha, M. Cebecauer,
Modeling and experimental analyses reveals signaling plasticity in a bi-modular
assembly of CD40 receptor activated kinases, PLoS One 7 (7) (2012) e39898,
https://doi.org/10.1371/journal.pone.003989810.1371/journal.pone.0039898.
g00110.1371/journal.pone.0039898.g00210.1371/journal.pone.0039898.
g00310.1371/journal.pone.0039898.g00410.1371/journal.pone.0039898.
g00510.1371/journal.pone.0039898.g00610.1371/journal.pone.0039898.
g00710.1371/journal.pone.0039898.g00810.1371/journal.pone.0039898.
s00110.1371/journal.pone.0039898.s00210.1371/journal.pone.0039898.
s00310.1371/journal.pone.0039898.s00410.1371/journal.pone.0039898.
s00510.1371/journal.pone.0039898.s00610.1371/journal.pone.0039898.
s00710.1371/journal.pone.0039898.s00810.1371/journal.pone.0039898.
s00910.1371/journal.pone.0039898.s01010.1371/journal.pone.0039898.
s01110.1371/journal.pone.0039898.s01210.1371/journal.pone.0039898.s013.
[14] R.K. Mathur, A. Awasthi, P. Wadhone, B. Ramanamurthy, B. Saha, Reciprocal
CD40 signals through p38MAPK and ERK-1/2 induce counteracting immune
responses, Nat Med 10 (5) (2004) 540–544, https://doi.org/10.1038/nm1045.
[15] Westermarck J, Li SP, Kallunki T, Han J, Kähäri VM (2001) p38 mitogen-activated
protein kinase-dependent activation of protein phosphatases 1 and 2A inhibits
MEK1 and MEK2 activity and collagenase 1 (MMP-1) gene expression. Mol Cell
Biol 21 :2373-83. doi:10.1128/MCB.21.7.2373-2383.2001.
[16] N. Srivastava, R. Sudan, B. Saha, CD40-modulated dual-specificity phosphatases
MAPK phosphatase (MKP)-1 and MKP-3 reciprocally regulate Leishmania major
infection, J Immunol 186 (10) (2011) 5863–5872, https://doi.org/10.4049/
jimmunol.1003957.
[17] S. Yamada, S. Shiono, A. Joo, A. Yoshimura, Control mechanism of JAK/STAT
signal transduction pathway, FEBS Lett 534 (2003) 190–196, https://doi.org/
10.1016/s0014-5793(02)03842-5.
[18] A. Gambin, A. Charzyńska, A. Ellert-Miklaszewska, M. Rybiński, Computational
models of the JAK1/2-STAT1 signaling, JAKSTAT 2 (3) (2013) e24672, https://
doi.org/10.4161/jkst.24672.
[19] K. Rateitschak, A. Karger, B. Fitzner, F. Lange, O. Wolkenhauer, R. Jaster,
Mathematical modelling of interferon-gamma signaling in pancreatic stellate cells
reflects and predicts the dynamics of STAT1 pathway activity, Cell Signal 22
(2010) 97–105, https://doi.org/10.1016/j.cellsig.2009.09.019.
[20] Z. Zi, K.H. Cho, M.H. Sung, X. Xia, J. Zheng, Z. Sun, In silico identification of the
key components and steps in IFN-gamma induced JAK-STAT signaling pathway,
FEBS Lett 579 (2005) 1101–1108, https://doi.org/10.1016/j.febslet.2005.01.009.
[21] Moya C, Huang Z, Cheng P, Jayaraman A, Hahn J (2011) Investigation of IL-6 and
IL10 signalling via mathematical modelling.IET Syst Biol 5:15.doi:10.1049/iet-
syb.2009.0060.
[22] M. Baron, J.L. Davignon, Inhibition of IFN-gamma-induced STAT1 tyrosine
phosphorylation by human CMV is mediated by SHP2, J Immunol 181 (2008)
5530–5536, https://doi.org/10.4049/jimmunol.181.8.5530.
[23] D. Xu, C.K. Qu, Protein tyrosine phosphatases in the JAK/STAT pathway, Front
Biosci 13 (2008) 4925–4932, https://doi.org/10.2741/3051.
[24] J.K. Riley, K. Takeda, S. Akira, R.D. Schreiber, Interleukin-10 receptor signaling
through the JAK-STAT pathway. Requirement for two distinct receptor-derived
signals for anti-inflammatory action, J Biol Chem 274 (23) (1999) 16513–16521,
https://doi.org/10.1074/jbc.274.23.16513.
[25] H. Jiang, Y.i. Lu, L. Yuan, J. Liu, L. Buday, Regulation of interleukin-10 receptor
ubiquitination and stability by beta-TrCP-containing ubiquitin E3 ligase, PLoS One
6 (11) (2011) e27464, https://doi.org/10.1371/journal.pone.002746410.1371/
journal.pone.0027464.s00110.1371/journal.pone.0027464.s00210.1371/journal.
pone.0027464.s00310.1371/journal.pone.0027464.s00410.1371/journal.
pone.0027464.s005.
[26] B. Ghosh, U. Sarma, V. Sourjik, S. Legewie, Sharing of Phosphatases Promotes
Response Plasticity in Phosphorylation Cascades, Biophys J 114 (1) (2018)
223–236, https://doi.org/10.1016/j.bpj.2017.10.037.
[27] L. Qiao, R.B. Nachbar, I.G. Kevrekidis, S.Y. Shvartsman, A. Asthagiri, Bistability
and Oscillations in the Huang-Ferrell Model of MAPK Signaling, PLoS Comput Biol.
3 (9) (2007) e184, https://doi.org/10.1371/journal.pcbi.003018410.1371/
journal.pcbi.0030184.g00110.1371/journal.pcbi.0030184.g00210.1371/journal.
pcbi.0030184.g00310.1371/journal.pcbi.0030184.g00410.1371/journal.
pcbi.0030184.g00510.1371/journal.pcbi.0030184.t00110.1371/journal.
pcbi.0030184.t00210.1371/journal.pcbi.0030184.t00310.1371/journal.
pcbi.0030184.t00410.1371/journal.pcbi.0030184.sg00110.1371/journal.
pcbi.0030184.sg00210.1371/journal.pcbi.0030184.sg00310.1371/journal.
pcbi.0030184.sg00410.1371/journal.pcbi.0030184.st00110.1371/journal.
pcbi.0030184.st00210.1371/journal.pcbi.0030184.sd00110.1371/journal.
pcbi.0030184.sd00210.1371/journal.pcbi.0030184.sd003.
[28] A. Raue, C. Kreutz, T. Maiwald, J. Bachmann, M. Schilling, U. Klingmüller,
J. Timmer, Structural and practical identifiability analysis of partially observed
dynamical models by exploiting the profile likelihood, Bioinformatics 25 (15)
(2009) 1923–1929, https://doi.org/10.1093/bioinformatics/btp358.
[29] Y.R. Chen, X.P. Wang, D. Templeton, R.J. Davis, T.H. Tan, The role of c-Jun N-
terminal kinase (JNK) in apoptosis induced by ultraviolet C and gamma radiation –
Duration of JNK activation may determine cell death and proliferation, J Biol
Chem 271 (1996) 31929–31936, https://doi.org/10.1074/jbc.271.50.31929.
10
Cytokine xxx (xxxx) xxx
[30] R.E. Dolmetsch, R.S. Lewis, C.C. Goodnow, J.I. Healy, Differential activation of
transcription factors induced by Ca2+ response amplitude and duration, Nature
386 (6627) (1997) 855–858, https://doi.org/10.1038/386855a0.
[31] L.O. Murphy, J. Blenis, MAPK signal specificity: the right place at the right time,
Trends Biochem Sci 31 (5) (2006) 268–275, https://doi.org/10.1016/j.
tibs.2006.03.009.
[32] L.O. Murphy, S. Smith, R.-H. Chen, D.C. Fingar, J. Blenis, Molecular interpretation
of ERK signal duration by immediate early gene products, Nat Cell Biol 4 (8)
(2002) 556–564, https://doi.org/10.1038/ncb822.
[33] S.D.M. Santos, P.J. Verveer, P.I.H. Bastiaens, Growth factor-induced MAPK
network topology shapes Erk response determining PC-12 cell fate, Nat Cell Biol 9
(3) (2007) 324–330, https://doi.org/10.1038/ncb1543.
[34] S. Sasagawa, Y.-I. Ozaki, K. Fujita, S. Kuroda, Prediction and validation of the
distinct dynamics of transient and sustained ERK activation, Nat Cell Biol 7 (4)
(2005) 365–373, https://doi.org/10.1038/ncb1233.
[35] C.J. Marshall, Specificity of Receptor Tyrosine Kinase Signaling – Transient Versus
Sustained Extracellular Signal-Regulated Kinase Activation, Cell 80 (2) (1995)
179–185, https://doi.org/10.1016/0092-8674(95)90401-8.
[36] I. Amit, A. Citri, T. Shay, Y. Lu, M. Katz, F. Zhang, G. Tarcic, D. Siwak, J. Lahad,
J. Jacob-Hirsch, N. Amariglio, N. Vaisman, E. Segal, G. Rechavi, U. Alon, G.
B. Mills, E. Domany, Y. Yarden, A module of negative feedback regulators defines
growth factor signaling, Nat Genet 39 (4) (2007) 503–512, https://doi.org/
10.1038/ng1987.
[37] J.W. Tullai, M.E. Schaffer, S. Mullenbrock, G. Sholder, S. Kasif, G.M. Cooper,
Immediate-early and delayed primary response genes are distinct in function and
genomic architecture, J Biol Chem 282 (33) (2007) 23981–23995, https://doi.org/
10.1074/jbc.M702044200.
[38] J. Purvis, G. Lahav, Encoding and decoding cellular information through signaling
dynamics, Cell 152 (5) (2013) 945–956, https://doi.org/10.1016/j.
cell.2013.02.005.
[39] K.A. Janes, H.C. Reinhardt, M.B. Yaffe, Cytokine-induced signaling networks
prioritize dynamic range over signal strength, Cell 135 (2) (2008) 343–354,
https://doi.org/10.1016/j.cell.2008.08.034.
[40] C. Herrero, X. Hu, W.P. Li, S. Samuels, M.N. Sharif, S. Kotenko, L.B. Ivashkiv,
Reprogramming of IL10 activity and signaling by IFN-gamma, J Immunol 171
(2003) 5034–5041, https://doi.org/10.4049/jimmunol.171.10.5034.
[41] A.P. Hutchins, D. Diez, D. Miranda-Saavedra, The IL10/STAT3-mediated anti-
inflammatory response: recent developments and future challenges, Brief Funct
Genomics 12 (6) (2013) 489–498, https://doi.org/10.1093/bfgp/elt028.
[42] Murray PJ (2005) The primary mechanism of the IL10-regulated antiinflammatory
response is to selectively inhibit transcription.Proc Natl Acad Sci U S A 102 :8686-
8691. doi:10.1073/pnas.0500419102.
[43] X. Hu, L.B. Ivashkiv, Cross-regulation of signaling pathways by interferon-gamma:
implications for immune responses and autoimmune diseases, Immunity 31 (2009)
539–550, https://doi.org/10.1016/j.immuni.2009.09.002.
[44] I. Rauch, M. Müller, T. Decker, The regulation of inflammation by interferons and
their STATs, JAKSTAT 2 (1) (2013) e23820, https://doi.org/10.4161/jkst.23820.
[45] J.J. O’Shea, P.J. Murray, Cytokine signaling modules in inflammatory responses,
Immunity 28 (4) (2008) 477–487, https://doi.org/10.1016/j.
immuni.2008.03.002.
[46] H.A. Purvis, A.E. Anderson, D.A. Young, J.D. Isaacs, C.M.U. Hilkens, A negative
feedback loop mediated by STAT3 limits human Th17 responses, J Immunol 193
(3) (2014) 1142–1150, https://doi.org/10.4049/jimmunol.1302467.
[47] S. Pensa, G. Regis, D. Boselli, F. Novelli, V. Poli, STAT1 and STAT3 in
tumorigenesis: two sides of the same coin. Landes, Bioscience (2013) 100–121.
[48] S.L. Kominsky, A.C. Hobeika, F.A. Lake, B.A. Torres, H.M. Johnson, Down-
regulation of neu/HER-2 by interferon-gamma in prostate cancer cells, Cancer Res
60 (2000) 3904–3908.
[49] H. Yu, M. Kortylewski, D. Pardoll, Crosstalk between cancer and immune cells: role
of STAT3 in the tumour microenvironment, Nat Rev Immunol 7 (1) (2007) 41–51,
https://doi.org/10.1038/nri1995.
[50] Y.S. Choi, D. Eto, J.A. Yang, C. Lao, S. Crotty, Cutting edge: STAT1 is required for
IL-6-mediated Bcl6 induction for early follicular helper cell differentiation,
J Immunol 190 (7) (2013) 3049–3053, https://doi.org/10.4049/
jimmunol.1203032.
[51] D.J. Matthews, R.S. Topping, R.T. Cass, L.B. Giebel, A sequential dimerization
mechanism for erythropoietin receptor activation, Proc Natl Acad Sci U S A 93 (18)
(1996) 9471–9476, https://doi.org/10.1073/pnas.93.18.9471.
[52] J.A. Wells, Binding in the growth hormone receptor complex, Proc Natl Acad Sci U
S A 93 (1) (1996) 1–6, https://doi.org/10.1073/pnas.93.1.1.
[53] K. Honda, H. Yanai, H. Negishi, M. Asagiri, M. Sato, T. Mizutani, N. Shimada,
Y. Ohba, A. Takaoka, N. Yoshida, T. Taniguchi, IRF-7 is the master regulator of
type-I interferon-dependent immune responses, Nature 434 (7034) (2005)
772–777, https://doi.org/10.1038/nature03464.
[54] L. Zhang, J.S. Pagano, Structure and function of IRF-7, J Interferon Cytokine Res 22
(1) (2002) 95–101, https://doi.org/10.1089/107999002753452700.
[55] H.N. Andrews, P.B. Mullan, S. McWilliams, S. Sebelova, J.E. Quinn, P.M. Gilmore,
N. McCabe, A. Pace, B. Koller, P.G. Johnston, D.A. Haber, D.P. Harkin, BRCA1
regulates the interferon gamma-mediated apoptotic response, J Biol Chem 277 (29)
(2002) 26225–26232, https://doi.org/10.1074/jbc.M201316200.
[56] F. Okumura, A.J. Okumura, K. Uematsu, S. Hatakeyama, D.-E. Zhang, T. Kamura,
Activation of double-stranded RNA-activated protein kinase (PKR) by interferon-
U. Sarma et al. Cytokine xxx (xxxx) xxx

stimulated gene 15 (ISG15) modification down-regulates protein translation, J Biol
Chem 288 (4) (2013) 2839–2847, https://doi.org/10.1074/jbc.M112.401851.
[57] B. Sainz, B. Martín, M. Tatari, C. Heeschen, S. Guerra, ISG15 is a critical
microenvironmental factor for pancreatic cancer stem cells, Cancer Res. 74 (24)
(2014) 7309–7320, https://doi.org/10.1158/0008-5472.CAN-14-1354.
[58] V.B. Carolina, S. Guerra, S.M. Francisco, ISGylation - a key to lock the cell gates for
preventing the spread of threats, J Cell Sci 130 (18) (2017) 2961–2969, https://
doi.org/10.1242/jcs.205468.

11