Available via license: CC BY 4.0
Content may be subject to copyright.
ARTICLE OPEN
Cellular and Molecular Biology
3D collagen migration patterns reveal a SMAD3-dependent
and TGF-β1-independent mechanism of recruitment for
tumour-associated fibroblasts in lung adenocarcinoma
Yago Juste-Lanas
1,2
, Natalia Díaz-Valdivia
3
, Alejandro Llorente
3,4
, Rafael Ikemori
3
, Alejandro Bernardo
3
,
Marselina Arshakyan
3
, Carlos Borau
1
, Josep Ramírez
5,6
, José Carlos Ruffinelli
7,8
, Ernest Nadal
7,8
, Noemí Reguart
6,9
,
José M. García-Aznar
1
and Jordi Alcaraz
3,4,6
✉
© The Author(s) 2022
BACKGROUND: The TGF-β1 transcription factor SMAD3 is epigenetically repressed in tumour-associated fibroblasts (TAFs) from
lung squamous cell carcinoma (SCC) but not adenocarcinoma (ADC) patients, which elicits a compensatory increase in SMAD2 that
renders SCC-TAFs less fibrotic. Here we examined the effects of altered SMAD2/3 in fibroblast migration and its impact on the
desmoplastic stroma formation in lung cancer.
METHODS: We used a microfluidic device to examine descriptors of early protrusions and subsequent migration in 3D collagen
gels upon knocking down SMAD2 or SMAD3 by shRNA in control fibroblasts and TAFs.
RESULTS: High SMAD3 conditions as in shSMAD2 fibroblasts and ADC-TAFs exhibited a migratory advantage in terms of
protrusions (fewer and longer) and migration (faster and more directional) selectively without TGF-β1 along with Erk1/2
hyperactivation. This enhanced migration was abrogated by TGF-β1 as well as low glucose medium and the MEK inhibitor
Trametinib. In contrast, high SMAD2 fibroblasts were poorly responsive to TGF-β1, high glucose and Trametinib, exhibiting impaired
migration in all conditions.
CONCLUSIONS: The basal migration advantage of high SMAD3 fibroblasts provides a straightforward mechanism underlying the
larger accumulation of TAFs previously reported in ADC compared to SCC. Moreover, our results encourage using MEK inhibitors in
ADC-TAFs but not SCC-TAFs.
British Journal of Cancer; https://doi.org/10.1038/s41416-022-02093-x
INTRODUCTION
Lung cancer is the leading cause of cancer mortality worldwide,
with a 5-year survival rate of ~19% [1]. Histologically, non-small
cell lung cancer (NSCLC) is diagnosed in ~85% of lung cancer
patients, and is classified into adenocarcinoma (ADC; ~50%),
squamous cell carcinoma (SCC; ~40%) and other less frequent
subtypes [2]. SCC tumours are strongly associated with smoking
and are commonly located in proximal airways, whereas ADC
typically arise in distal pulmonary sites [2]. Although both ADC
and SCC are epithelial in origin, it is increasingly recognised that
the desmoplastic/fibrotic stroma rich in tumour-associated
fibroblasts (TAFs, also referred to as cancer-associated fibro-
blasts or CAFs), play a key role in tumour progression and
therapy resistance [3,4]. Accordingly, there is growing interest
in understanding the aberrant behaviour of fibroblasts in solid
tumours [5].
Most lung TAFs exhibit an activated/myofibroblast-like pheno-
type [6,7], which is roughly characterised by the intracellular
expression of α-smooth muscle actin (α-SMA) and the abundant
extracellular deposition of fibrillar collagens [3]. Transforming
growth factor-β1 (TGF-β1) is the most potent fibroblast activator
known to date and is frequently upregulated in NSCLC. Moreover,
both TGF-β1 and TAF activation markers are associated with poor
prognosis in NSCLC [7,8]. Intriguingly, we recently reported that
TAF activation and associated fibrosis is higher in ADC compared
to SCC, owing to the larger epigenetic repression of the important
pro-fibrotic transcription factor SMAD3 of the canonical TGF-β
pathway selectively in SCC-TAFs, caused by their increased
Received: 11 March 2022 Revised: 19 November 2022 Accepted: 25 November 2022
1
Department of Mechanical Engineering, Aragón Institute of Engineering Research (I3A), University of Zaragoza, 50018 Zaragoza, Spain.
2
Department of Biochemistry and
Molecular and Cellular Biology, University of Zaragoza, 50009 Zaragoza, Spain.
3
Unit of Biophysics and Bioengineering, Department of Biomedicine, School of Medicine and
Health Sciences, Universitat de Barcelona, 08036 Barcelona, Spain.
4
Institute for Bioengineering of Catalonia (IBEC), The Barcelona Institute for Science and Technology (BIST),
08028 Barcelona, Spain.
5
Pathology Service, Hospital Clínic de Barcelona, 08036 Barcelona, Spain.
6
Thoracic Oncology Unit, Hospital Clinic Barcelona, 08036 Barcelona, Spain.
7
Department of Medical Oncology, Catalan Institute of Oncology, L’Hospitalet de Llobregat (Barcelona), 08908 Barcelona, Spain.
8
Preclinical and Experimental Research in
Thoracic Tumors (PrETT) group, Oncobell Program, Bellvitge Biomedical Research Institute (IDIBELL), L’Hospitalet de Llobregat (Barcelona), 08908 Barcelona, Spain.
9
Institut
d’Investigacions Biomèdiques August Pi i Sunyer (IDIBAPS), Barcelona 08036, Spain. ✉email: jalcaraz@ub.edu
www.nature.com/bjc
British Journal of Cancer
1234567890();,:
exposure to cigarette smoke particles. We also showed that the
epigenetic repression of SMAD3 elicited a compensatory increase
in the expression and activity of its closely related homologue
SMAD2 in SCC-TAFs, which is weakly associated with fibrosis [9].
Consistently, ADC-TAFs but not SCC-TAFs exhibited a positive
response to the antifibrotic drug nintedanib in culture [10],
thereby mimicking the therapeutic benefits reported by ninteda-
nib in ADC but not in SCC patients in the LUME-1 clinical trial [11].
Likewise, both the lower expression of fibrosis markers and the
poor nintedanib response of SCC-TAFs could be reproduced in
normal fibroblasts upon knocking down SMAD3 with shRNA,
whereas knocking down SMAD2 had opposite effects, supporting
that shSMAD2 and shSMAD3 fibroblasts exhibit ADC-like and SCC-
like phenotypes, respectively [9,12].
SMADs 2 and 3 (referred to as SMAD2/3 thereafter) are direct
mediators of canonical TGF-β1 signalling and exhibit some
overlapping functions; however, they also regulate distinct
processes, as reported in knock-out mice [13]. In fibroblasts,
SMAD2/3 not only differentially regulate fibrosis and response to
antifibrotic drugs [9,14] but they may also control cell migration,
which may be relevant for the formation of the fibrotic tumour
microenvironment (TME). However, our knowledge of how
SMAD2/3 control cell migration is very limited [15], and their
impact on the formation of the desmoplastic TME in lung cancer is
unknown. Moreover, the few available fibroblast-specific analyses
of migration regulation by SMAD2/3 were performed in two
dimension (2D) or transwells [15,16], whereas fibroblast migration
in vivo occurs within a three-dimensional (3D) microenvironment
rich in type I collagen [17]. To address this gap of knowledge, we
knocked down either SMAD2 or SMAD3 in primary human
pulmonary fibroblasts as surrogate models of ADC-TAFs or SCC-
TAFs, respectively. Fibroblasts were cultured in dense 3D collagen
gels in the absence or presence of TGF-β1 to mimic the
progression of the desmoplastic tumour stroma [18], and the
formation of pro-migratory protrusions and subsequent migration
was analysed. For this purpose, 3D cultures were prepared within
a microfluidic device to assess a panel of biophysical descriptors of
protrusions and migration by multidimensional microscopy, and
key findings were validated with ADC-TAFs.
METHODS
Patient-derived tissue samples and pulmonary fibroblasts
Primary fibroblasts were previously obtained from a cohort of 20 NSCLC
surgical patients [6]. Fibroblasts were derived from either tumour or
patient-matched uninvolved pulmonary tissue (referred to as control
fibroblasts thereafter) using protocols approved by the Ethics Committees
of the Hospital Clinic de Barcelona and the Universitat de Barcelona.
Selected patients were male, chemo-naïve, Caucasian, >55 years old and
current/former smokers. Tumour tissue samples for histological analysis
were obtained from the Hospital de Bellvitge (10 ADC, 9 SCC) with the
approval of the Ethics Committee. The study was performed in accordance
with the Declaration of Helsinki and written informed consent was
obtained from all patients. Further clinical characteristics are shown in
Supplementary Table 1.
Histologic analysis
Tumour samples were processed as described [9], counterstained with
haematoxylin and stained for either cleaved microtubule-associated
protein 1 light chain 3 (LC3A) (#Ap1805a, Abgent), which is largely
negative in fibroblasts [19], eosin or α-SMA as reported [6]. Fibroblast
nuclear density was assessed by image analysis of haematoxylin staining
using the QuPath software [20] under the guidance of our pathologist (JR).
Further details are provided in Supplementary Materials.
2D cell culture and fibroblast immortalisation
Control fibroblasts and ADC-TAFs from randomly selected patients (#5, #13,
#37) were immortalised with hTERT as reported [9]. Unless otherwise
indicated, all fibroblast experiments were performed in culture medium
containing serum-free high-glucose (4.5 g/l) DMEM supplemented with 1%
insulin–transferrin–selenium (ITS) and antibiotics as described [10]. In some
experiments, fibroblasts were stimulated with 2.5 ng/ml recombinant
human TGF-β1 (Miltenyi Biotec) at different time points as indicated, which
is similar to the average TGF-β1 concentration reported in the
bronchoalveolar lavage fluid of lung cancer patients [21].
SMAD2 and SMAD3 knock down with shRNA and siRNA
SMAD2 or SMAD3 were stably knocked down in immortalised primary
control fibroblasts and ADC-TAFs with lentiviral vectors derived from
Sigma MISSION collection as reported [9]. A nonmammalian targeting
shRNA vector was used as control (shControl). Alternatively, SMAD3 was
transiently knocked down by siRNA as described [9]. Further details are
provided in Supplementary Materials.
qRT-PCR
RNA extraction and reverse transcription were conducted as reported
[10,22]. SMAD2/3and MMP1 mRNA levels were assessed using specific
primers, with ACTB or POLR2A as endogenous controls, respectively.
Further details are provided in Supplementary Materials.
Western blot (WB) analysis
WB analysis of SMAD2/3 and Erk1/2 was conducted as described [6,9],
using primary antibodies against total SMAD2/3 (#3102, Cell Signaling),
pSMAD2 (#3104, Cell Signaling), pSMAD3 (#07-1389, Merck Millipore), Erk1/
2, pErk1/2 (#9102 and #9101; Cell Signaling Technology), β-actin (#A1978,
Sigma-Aldrich) and α-tubulin (#2144; Cell Signaling Technology). The latter
two were used as loading controls. Additional details are provided in
Supplementary Materials.
Fabrication of the microfluidic device
Microfluidic devices were fabricated as described [23], using photomasks
as previously reported [24,25]. In brief, masks were used to fabricate
positive 300 µm high SU8 masters (Stanford University). Polydimethylsilox-
ane (PDMS) (Sylgard 184, Dow Corning) was mixed at a 10:1 weight ratio of
base to curing agent and poured on the SU8 master until the desired
thickness (4 mm) was obtained. The PDMS solution was cured in an oven,
cut out and removed from the wafer, perforated and autocleaved. PDMS
microdevices were plasma-bonded to 35 mm glass-bottom petri dishes
(Ibidi) and coated with 1 mg/ml poly-D-lysine (PDL) (Sigma-Aldrich) to
enhance surface-collagen gel attachment. The geometry of the micro-
device was based on [26] and included a 300 µm high central chamber to
allocate the 3D collagen culture and two parallel liquid channels located
on each side of the central chamber that were in direct contact with the
gel for hydration and transport of nutrients and other factors [25]. Further
details are provided in Supplementary Materials.
3D collagen cell culture within the microfluidic device
Collagen hydrogels were prepared as reported [23,25]. Briefly, type I
collagen solution (BD Bioscience) was mixed with DPBS (Thermo Fisher
Sci.) and neutralised with NaOH (Sigma-Aldrich) to pH 7.4. For 3D culture
experiments, cells suspended in culture medium were mixed with the
collagen solution to a final concentration of 4 mg/ml and cell dilution of
2×10
5
cells/ml, which enables local matrix remodelling but not global gel
contraction [17]. The collagen and cells solution were loaded into the
central chamber of the microfluidic device using the auxiliary inlet
channels and attached to it through surface tension. The device containing
the collagen and cells solution was placed into an incubator to allow
collagen polymerisation for 30 min. Next, the 3D culture was hydrated with
culture medium and kept in the incubator before experiments.
Analysis of the protrusions of single fibroblasts in 3D collagen
cultures
Protrusion analysis was performed by adapting our previous protocol [27,28].
Fibroblasts were kept with culture medium with or without 2.5 ng/ml TGF-β1
for 72 h in 2D culture, trypsinised and used to prepare 3D collagen cultures
within the microfluidic device. Protrusions were imaged after collagen
polymerisation and up to 4 h based on previous observations [17]as
reported [27,28], using a phase contrast Nikon Eclipse Ti-E inverted
microscope (Nikon, Japan) provided with an incubator. Imaging was
conducted at least 100 µm away from the glass and PDMS surface to avoid
potential edge effects [29]. The whole gel thickness (300μm) was imaged at
Y. Juste-Lanas et al.
2
British Journal of Cancer
5μm intervals (Zaxis) every 5 min at ×200 magnification with a ×20 objective
(CFI Plan Fluor ELWD ADM, NA 0.45; Nikon), eliciting 49 time points and 2940
images/fibroblast. Best Zplane was chosen for each image, and both the cell
body and the protrusions of randomly selected fibroblasts were manually
outlined [27,28]. The aspect ratio was computed as major axis/minor axis as
reported [30] (Supplementary Fig. 1). A panel of descriptors was analysed for
each “mother”protrusion that stemmed from the cell body (referred to as
protrusions thereafter) [27,31], including length, number, and protrusion and
retraction growth rates, using in-house MATLAB (MathWorks) algorithms. For
each fibroblast, the absolute values for the protrusion growth and retraction
###
shControl
shSMAD2
(ADC-like)
shSMAD3
(SCC-like)
shControl
shSMAD2
(ADC-like)
shSMAD3
(SCC-like)
shControl
shSMAD2
(ADC-like)
shSMAD3
(SCC-like)
shControl
shSMAD2
(ADC-like)
shSMAD3
(SCC-like)
✱✱
###
– TGF-
β
1 (basal)– TGF-
β
1 (basal) + TGF-
β
1
TGF-
β
1
cba
fed
ih
g
jlk
SMAD2
SMAD3
β-actin
m
TGF-β1+++
SCC-TAFs
(#22)
β-actin
ADC-TAFs
(#12)
0 min 30 min0 min 30 min
SMAD2
SMAD3
– TGF-
β
1
(0 min)
+ TGF-
β
1
(30 min)
ADC-TAFs
SCC-TAFs
ADC-TAFs
SCC-TAFs
0
1
2
3
0
1
2
3
Fold SMAD3 mRNA
Fold SMAD2 mRNAFold SMAD2 mRNA
Fold SMAD3 mRNA
Fold SMAD3 mRNA
Fold SMAD3/SMAD2 mRNAFold SMAD3/SMAD2 mRNAFold SMAD3/SMAD2
##
###
ADC-TAFs
SCC-TAFs
ADC-TAFs
SCC-TAFs
ADC-TAFs
SCC-TAFs
ADC-TAFs
SCC-TAFs
ADC-TAFs
SCC-TAFs
ADC-TAFs
SCC-TAFs
0.0
0.5
1.0
1.5
2.0
2.5
###
0
1
2
3
0
1
2
3
p = 0.1
0.0
0.5
1.0
1.5
2.0
2.5 #
#
p = 0.09
0.0
0.5
1.0
1.5
Fold SMAD3/SMAD2
0.0
0.5
1.0
1.5 p = 0.08
0.0
0.5
1.0
1.5
Fold SMAD2 mRNA
0.0
0.5
1.0
1.5
Fold SMAD3/SMAD2
0.0
0.5
1.0
1.5
Y. Juste-Lanas et al.
3
British Journal of Cancer
rates were averaged to indicate protrusion evolution rate. A total of 4
fibroblasts/condition were examined from 3 independent experiments,
which required analysing >8000 images in total.
Analysis of the migration of single fibroblasts in 3D collagen
cultures
Random fibroblast migration was analysed using our previous protocols
[24,25,32]. Fibroblasts were kept with culture medium with or without
2.5 ng/ml TGF-β1 for 48 h in 2D culture, trypsinised and used in 3D
collagen cultures within the microfluidic device. 3D cultures were kept with
or without TGF-β1 for 24 h within the device, and subsequently imaged for
24 h with a phase contrast inverted microscope provided with an
incubator. Images were acquired every 20 minutes in a manually selected
Zplane at ×100 magnification using a ×10 air objective (CFI Plan Fluor DLL,
NA 0.30, WD 16.0 mm; Nikon) as described [24,33]. Imaging was
conducted at least 100 µm away from the glass and PDMS surface to
avoid potential artifacts. The trajectories of randomly selected fibroblasts
were tracked and used to compute a panel of migration descriptors with
in-house MATLAB algorithms [24,34], including average cell speed, net
displacement and cell persistence. On average, 35 fibroblasts were
analysed for each device, and 190 per condition.
Boyden Transwell migration assay
Fibroblasts migration was performed using the Transwell Boyden assay as
reported [35]. In brief, fibroblasts were maintained for 3 days in serum-free
culture medium with or without 2.5 ng/ml TGF-β1 before seeding them on
Transwell inserts. Culture medium alone or supplemented with 2.5 ng/ml
TGF-β1 was added to the lower Transwell compartment, and cells that
migrated into the lower insert membrane side after 16 h were fixed,
stained with crystal violet and imaged by phase-contrast microscopy with
a ×10 objective. Migration was assessed as percentage of positively stained
image area with Image J. In some experiments, fibroblasts were treated
with 100 nM Trametinib (Selleckchem).
Fibroblast number density
Cell number density of TAFs in 2D cultures was assessed as reported [6]. In
brief, TAFs were cultured in serum-free medium with or without 2.5 ng/ml
TGF-β1 for 5 days and their nuclei were stained with Hoechst 33342
(Molecular Probes) and imaged with a ×10 objective. Number density in the
same 3D cultures used for 3D migration analysis were assessed by manually
counting cells imaged at the end of the experiment. For each experiment,
number density was determined as the average cell density/image.
Statistical analysis
Two-group comparisons were performed with either two-tailed Student’st
test or Mann–Whitney test for non-parametric data (GraphPad Prism v9.0.).
Statistical significance was assumed at p< 0.05. All experiments were
conducted at least in triplicates (≥3 microfluidic devices). All data shown
are mean ± s.e.m.
RESULTS
The relative differences in SMAD2/3 expression between ADC-
TAFs and SCC-TAFs can be mimicked through shRNA in
control pulmonary fibroblasts
To characterise the differences in SMAD2/3 in ADC-TAFs and SCC-
TAFs in more detail, we examined SMAD2/3 expression in TAFs
from randomly selected patients in the absence or presence of
TGF-β1 and without normalising by patient-matched control
fibroblasts as in our previous study [9]. In basal conditions (i.e.
absence of TGF-β1) SMAD3 mRNA was significantly higher in ADC-
TAFs than SCC-TAFs (Fig. 1a), whereas SMAD2 mRNA exhibited the
opposite trend (Fig. 1b), eliciting a markedly higher SMAD3/SMAD2
mRNA ratio in ADC-TAFs (Fig. 1c). These histotype differences were
maintained in response to TGF-β1 (Fig. 1d–f) and are in agreement
with the larger epigenetic repression of SMAD3 in SCC-TAFs [9].
Similar histotype differences in SMAD3/SMAD2 ratio were found at
the protein level (Fig. 1g–i and Supplementary Fig. 2), and
collectively reveal that ADC-TAFs exhibit high SMAD3 mRNA and
SMAD3/SMAD2 expression ratio, whereas SCC-TAFs exhibit high
SMAD2 mRNA and lower SMAD3/SMAD2 expression ratio. Con-
sistently, we previously showed that the response to TGF-β1is
dominated by the activation through phosphorylation of either
SMAD3 in ADC-TAFs or SMAD2 in SCC-TAFs, concomitantly with a
higher expression of fibrosis markers in ADC-TAFs compared to
SCC-TAFs [9].
To model the SMAD2/3 differences observed in TAFs, we stably
knocked down SMAD2 or SMAD3 by shRNA in control fibroblasts
derived from uninvolved pulmonary tissue of a randomly selected
surgical lung cancer patient (#5). In basal conditions, shSMAD2
fibroblasts exhibited higher SMAD3 mRNA (Fig. 1j) and protein
expression (Fig. 1m) than shSMAD3 fibroblasts as in ADC-TAFs.
Conversely, shSMAD3 fibroblasts exhibited the largest SMAD2
mRNA (Fig. 1k) and protein levels (Fig. 1m) as in SCC-TAFs, which
elicited the largest SMAD3/SMAD2 mRNA ratio in shSMAD2
fibroblasts (Fig. 1l). In further agreement with TAFs, the response
of shSMAD2 fibroblasts to exogenous TGF-β1 was dominated by
the activation through phosphorylation of SMAD3 as in ADC-TAFs,
whereas that of shSMAD3 fibroblasts was dominated by the
phosphorylation of SMAD2 (Supplementary Fig. 2) as in SCC-TAFs
[9]. Accordingly, and in agreement with previous studies [9,10],
shSMAD2 and shSMAD3 fibroblasts were used as ADC-like and
SCC-like models henceforth, respectively.
Analysis of protrusions in 3D collagen gels reveals that high
SMAD3 conditions as in ADC-TAFs primes fibroblasts for
migration in the absence of exogenous TGF-β1
Membrane protrusions are pointed as critical regulators of cell
migration in 3D [31,36]. We used a microdevice-based assay [27]
to monitor a panel of protrusion descriptors in single fibroblasts
embedded in a dense 3D collagen gel within the first 4 h as
outlined in Fig. 2a. Protrusion analysis was limited to 4 fibroblasts
per condition owing to the large number of images involved in
each fibroblast analysis. All fibroblasts initially exhibited a round
morphology with a dendritic network of protusions (Fig. 2b, h and
Supplementary Fig. 3A, B). In basal conditions, the average
number of protrusions fluctuated around 4–7, with a modest
increase within the first 1 h followed by a slow decline in shSMAD2
(ADC-like) and shControl fibroblasts down to 4, whereas they
remained stable around 5 in shSMAD3 (SCC-like) fibroblasts
(Fig. 2c, e). In contrast, protrusion length increased with time
Fig. 1 Genetic models to mimic SMAD2/3 alterations in patient-derived ADC-TAFs and SCC-TAFs. a–cFold SMAD3 (a) and SMAD2 (b) mRNA
and corresponding ratio (c) in primary lung TAFs cultured in 2D for 3 days in basal conditions (i.e. without exogenous TGF-β1) (3 ADC, 3 SCC).
d–fFold SMAD3 (d) and SMAD2 (e) mRNA and corresponding ratio (f) in primary lung TAFs cultured in 2D for 3 days in the presence of 2.5 ng/
ml TGF-β1 (7 ADC, 5 SCC). gRepresentative Western blot for total SMAD2, SMAD3 and β-actin of ADC-TAFs and SCC-TAFs from randomly
selected patients at 0 min or 30 min after stimulation with TGF-β1. h,iDensitometry analysis of total SMAD3/SMAD2 ratio in TAFs from
randomly selected patients (3 ADC, 3 SCC) at 0 min (h) or 30 min (i) after stimulation with TGF-β1. j–lFold SMAD3 (j), SMAD2 (k) and SMAD3/
SMAD2 mRNA ratio (l) of shControl, shSMAD2 and shSMAD3 control fibroblasts from patient #5 cultured in 2D for 3 days in basal conditions.
mRepresentative Western blot for total SMAD2, SMAD3 and β-actin of shControl, shSMAD2 and shSMAD3 control fibroblasts (#5) in basal
conditions. Error bars represent mean ± s.e.m. Each dot corresponds to a different patient (a–i).
#
p< 0.05;
##
p< 0.01;
###
p< 0.005 comparing
either ADC-TAFs and SCC-TAFs or shSMAD2 and shSMAD3. **p< 0.01; ***p< 0.005 with respect to shControl. Statistical comparisons were
done using Student’sttest.
Y. Juste-Lanas et al.
4
British Journal of Cancer
selectively in shSMAD2 fibroblasts up to 50 μm, whereas it
reached a plateau in the last 1–2 h that was the lowest in
shSMAD3 fibroblasts (Fig. 2d). Accordingly, protrusion number and
length were averaged within the last hour (3–4 h) henceforth and
were found consistently higher in shSMAD2 compared to
shSMAD3 fibroblasts in terms of length (Fig. 2f) and growth/
retraction evolution rate (Fig. 2g), with average values of ~40 μm
and 0.015 μm/s, respectively. Conversely, protrusion number was
0
2
4
6
Aspect ratio
0
2
4
6
Aspect ratio
p = 0.09
0
2
4
6
8
10
p = 0.1
0
2
4
6
8
10
Avg
basal
Avg
basal
Avg
basal
a
bc d
shControl
shSMAD2
(ADC-like)
shSMAD3
(SCC-like)
shControl
shSMAD2
(ADC-like)
shSMAD3
(SCC-like)
– TGF-
β
1 (basal)+ TGF-
β
1
efg
hji
klm
npo
– TGF-
β
1 + TGF-
β
1
– TGF-
β
1 + TGF-
β
1
4h
±TGFβ1±TGFβ1
72h [ 2D ] 4h [ 3D ]
Imaging
20 mm Culture medium
Best-focused z-plane
Culture medium
Collagen & cell
input channels
4h
50 μm
40 μmprot./ret
rates
6
73
5
4
1
2
Protrusion
length
50 μm
0
10
20
30
40
50
60
Protrusions length (
μ
m)
Protrusions length (
μ
m)Protrusions length (
μ
m)
Protrusions length (
μ
m)
#
0.000
0.005
0.010
0.015
0.020 #
0
10
20
30
40
50
60
p = 0.08
0.000
0.005
0.010
0.015
0.020
Protrusions evolution
rate (
μ
m/s)
Protrusions evolution
rate (
μ
m/s)
#
#
01234
0
10
20
30
40
50
60
Time (h)
01234
0
2
4
6
8
10
Time (h)
Number of protrusionNumber of protrusionsNumber of protrusionsNumber of protrusions
shControl
shSMAD2
(ADC-like)
shSMAD3
(SCC-like)
shControl
shSMAD2
(ADC-like)
shSMAD3
(SCC-like)
shControl
shSMAD2
(ADC-like)
shSMAD3
(SCC-like)
shControl
shSMAD2
(ADC-like)
shSMAD3
(SCC-like)
01234
0
2
4
6
8
10
Time (h)
01234
0
10
20
30
40
50
60
Time (h)
shSMAD2
(ADC-like)
shControl
shSMAD2
(ADC-like)
shSMAD3
(SCC-like)
shControl
shSMAD2
(ADC-like)
shSMAD3
(SCC-like)
shControl
shSMAD2
(ADC-like)
shSMAD3
(SCC-like)
shControl
shSMAD2
(ADC-like)
shSMAD3
(SCC-like)
shControl
shSMAD2
(ADC-like)
shSMAD3
(SCC-like)
shControl
shSMAD2
(ADC-like)
shSMAD3
(SCC-like)
shControl
shSMAD2
(ADC-like)
shSMAD3
(SCC-like)
shControl
shSMAD2
(ADC-like)
shSMAD3
(SCC-like)
50 μm
Y. Juste-Lanas et al.
5
British Journal of Cancer
lower in shSMAD2 than shSMAD3 fibroblasts with marginal
significance (Fig. 2e).
Unlike basal conditions, all TGF-β1-preactivated fibroblasts
exhibited a low initial number of protrusions (1-3) that increased
within the first 1 h up to 7, although at different rates, in which
shSMAD3 fibroblasts were the fastest (Fig. 2i). In contrast,
protrusion length increased over time at a much lower rate than
in basal conditions, particularly in shSMAD2 fibroblasts, which
attained a protrusion length of ~10 μm (Fig. 2j). Globally, TGF-β1
elicited an average ≥50% increase of protrusion number
compared to basal conditions (Fig. 2k and Supplementary Fig. 3C),
whereas it reduced both protrusion length and evolution rate by
>50% in shSMAD2 conditions, and these descriptors remained
stable in shSMAD3 fibroblasts (Fig. 2l, m and Supplementary
Fig. 3D, E), suggesting that shSMAD3 fibroblasts are poorly
responsive to TGF-β1. Since longer and fewer protrusions have
been previously associated with enhanced migration [31], these
results suggest that shSMAD2 (ADC-like) fibroblasts may be
primed for migration selectively in basal conditions. Consistently,
the aspect ratio at 4 h, which is indicative of polarisation along a
major protrusion (Supplementary Fig. 1) and has been positively
associated with migration [36], was the highest in shSMAD2
fibroblasts in basal conditions (Fig. 2n, o), whereas it was the
lowest upon TGF-β1 preactivation (Fig. 2p).
Analysis of migration in 3D collagen gels confirms that high
SMAD3 as in ADC-TAFs enhances migration selectively in
basal conditions
Next, we adapted the microfluidic device assay to analyse single
fibroblast migration within 24–48 h in 3D culture as outlined in
Fig. 3a. All fibroblasts exhibited the archetypical elongated
morphology found in histologic sections [6,17] (Fig. 3b). In
agreement with protrusion analysis, shSMAD2 (ADC-like) fibro-
blasts exhibited significantly enhanced migration in basal condi-
tions, as illustrated by their larger trajectories (Fig. 3c) and
significantly higher values of all migration descriptors compared
to shSMAD3 and shControl fibroblasts (Fig. 3d–f), with an average
speed of ~0.10 μm/min (Fig. 3d), total net displacement of ~50 μm
(Fig. 3e) and directional persistence of ~0.4 (Fig. 3f) that were
~30% (speed), ~90% (displacement) and ~40% (persistence) larger
than the corresponding values of shSMAD3 fibroblasts. To validate
our observations, we analysed ADC-TAFs from patient #37 upon
knocking down SMAD3 with shRNA (Fig. 3g) in the absence of
exogenous TGF-β1. Although migration descriptors in ADC-TAFs
(#37) (Fig. 3h–l) were lower than those found in control fibroblasts
(Fig. 3d–f), shControl ADC-TAFs exhibited in average a larger
effective speed (Fig. 3j), net displacement (Fig. 3k) and persistence
(Fig. 3l) than shSMAD3 ADC-TAFs (#37).
The basal migratory advantage of shSMAD2 fibroblasts was
abrogated upon TGF-β1 stimulation, and fibroblasts in all
conditions exhibited similar trajectories (Fig. 4a, b), speed (Fig. 4c),
net displacement (Fig. 4d) and persistence (Fig. 4e). Moreover,
even though TGF-β1 barely reduced the average cell speed in
shSMAD2 with respect to basal conditions (Fig. 4f), it elicited a
marked reduction in the average persistence and associated net
displacement (Fig. 4g, h), revealing that TGF-β1 promotes the
spatial confinement of ADC-like fibroblasts. In contrast, shSMAD3
fibroblasts were poorly responsive to TGF-β1, exhibiting a small
increase in speed and displacement compared to basal settings
(Fig. 4f, g) in further agreement with protrusion analysis. A similar
trend was observed in shControl compared to shSMAD3 ADC-TAFs
(#37) in terms of speed and persistence (Fig. 4k–m); however,
these differences did not attain statistical significance due to the
large variability associated with their globally low speed and
directionality. These results suggest that the spatial confinement
of ADC-TAFs is evident even in the absence of exogenous TGF-β1.
The strong qualitative agreement between early protrusions
(0–4 h) and subsequent migration data (24–48 h) encouraged us
to conduct a correlation analysis between descriptors of both
processes. Both migration persistence and net displacement were
strongly and positively correlated (R
2
> 0.7) with all protrusion
descriptors but protrusion number (Supplementary Fig. 4E–L),
with the highest correlations consistently observed with aspect
ratio (Supplementary Fig. 4K, L). In contrast, we observed a poor
correlation between migration speed and all protrusions descrip-
tors (Supplementary Fig. 4A, D, G, J). These results further support
the notion that those conditions that elicit fewer and longer
protrusions may help polarise the cell along a major protrusion
and facilitate directed movements, whereas increased number of
protrusions is associated with reduced migration and subsequent
spatial confinement [31,36].
The enhanced migration of high SMAD3 fibroblasts in basal
conditions is independent of collagen degradation
Because our migration analysis was conducted in fibroblasts
embedded in a dense collagen matrix with a expected range of
pore sizes comparable or even lower than the typical width of
elongated fibroblasts [37], it is possible that the enhanced basal
migration of high SMAD3 fibroblasts is driven by increased
collagen degradation rather than intrinsic migratory priming. To
address this question, we analysed migration in the complete
absence of exogenous extracellular matrix (ECM) degradation
using the Boyden Transwell assay without any ECM coating in the
porous Transwell insert membrane. In agreement with our 3D
migration data, the percentage of cells that migrated through the
insert membrane was markedly higher in shSMAD2 fibroblasts
Fig. 2 Impact of altered SMAD2/3 in 3D collagen protrusions of lung fibroblasts and ADC-TAFs with or without exogenous TGF-β1.
aOutline of the microdevice-based analysis of protrusions (0–4 h) of single fibroblasts cultured in dense 3D collagen gels. bRepresentative
phase contrast images of single control fibroblasts (#5) for each group (shControl, shSMAD2, shSMAD3) cultured in 3D collagen gels within
the microdevice for 4 h in basal conditions (−TGF-β1). Scale bar, 50 μm. Representative images at other time points are shown in
Supplementary Fig. 3A, B. c,dTime-course of the average number of protrusions (c) and protrusion length (d)of4fibroblasts for each group
cultured in 3D in basal conditions. e–gNumber of protrusions (e), protrusion length (f) and evolution rate (g) averaged for 4 fibroblasts per
group within the last 1 h (3–4 h) of the experimental time-window (e,f) or the full experiment (0–4h) (g) in basal conditions (shown as bars).
Each dot indicates the average of a single fibroblast henceforth. hRepresentative phase contrast images of single control fibroblasts (#5) for
each group (shControl, shSMAD2, shSMAD3) cultured in 3D collagen gels within the microdevice for 4 h in the presence of TGF-β1. Scale bar,
50 μm. Representative images at other time points are shown in Supplementary Fig. 3A, B. i,jTime-course of the average number of
protrusions (i) and protrusion length (j)of4fibroblasts for each group cultured in 3D in the presence of TGF-β1. k–mNumber of protrusions
(k), protrusion length (l) and evolution rate (m) averaged for 4 fibroblasts per group as in e–gin the presence of TGF-β1. nRepresentative
phase contrast images of the aspect ratio analysis of single shSMAD2 fibroblasts cultured in 3D collagen gels for 4 h without or with TGF-β1.
Scale bar, 50 μm. Further details on the assessment of the aspect ratio shown in Supplementary Fig. 1. o,pAspect ratio averaged for 4
fibroblasts per group at the end of the experimental time-window (4 h) in basal conditions (o) or in the presence of TGF-β1(p). Error bars
represent mean ± s.e.m. Each dot corresponds to the average of a different fibroblast (e–g,k–m,o,p) examined within 3 independent
microdevices.
#
p< 0.05 comparing shSMAD2 and shSMAD3. Other p-values comparing to shControl. Statistical comparisons were done using
Student’sttest.
Y. Juste-Lanas et al.
6
British Journal of Cancer
compared to shSMAD3 in basal settings (Fig. 5a and Supplemen-
tary Fig. 5A). Likewise, the migration advantage of shSMAD2
fibroblasts was attenuated by TGF-β1, whereas shSMAD3 fibro-
blasts were poorly responsive (Fig. 5b, c and Supplementary
Fig. 5B). Consistent differences were observed in ADC-TAFs from
randomly selected patients (#13, #37) in which SMAD3 had been
knocked down by shRNA (#37) (Fig. 3g) or siRNA (#13)
(Supplementary Fig. 5C) compared to parental cells in Transwells
0.00
0.05
0.10
0.15
0.20
0.25
Cell speed (μ
μ
m/min)
###
a
b
c
f
de
– TGF-
E
1 (basal) – TGF-
E
1 (basal)
– TGF-
E
1 (basal)
– TGF-
E
1 (basal)
shControl shSMAD2
(ADC-like)
shSMAD3
(SCC-like)
24 h
36 h
48 h
0
50
100
150
200
250
300
Net displacement (
μ
m)
###
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
Persistence ratio
###
✱
ADC-TAF (#37)
shControl
ADC-TAF (#37)
shSMAD3
h
24 h 36 h 48 h
50 μm
50 μm
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
Persistence ratio
0
50
100
150
200
250
300
Net displacement (
μ
m)
p = 0.1
0.00
0.05
0.10
0.15
0.20
0.25
Cell speed (
μ
m/min)
0.00
0.25
0.50
0.75
1.00
1.25
1.50
Fold SMAD3 mRNA
ADC-TAFs (#37)
shSMAD3
ADC-TAFs (#37)
shControl
ADC-TAFs (#37)
shSMAD3
ADC-TAFs (#37)
shControl
ADC-TAFs (#37)
shSMAD3
ADC-TAFs (#37)
shControl
ADC-TAFs (#37)
shSMAD3
ADC-TAFs (#37)
shControl
g
jkl
i
– TGF-
E
1 (basal)
250
–250
–250 250
250
–250
–250 250
250
–250
–250 250
shControl
shSMAD2
(ADC-like)
shSMAD3
(SCC-like)
shControl
shSMAD2
(ADC-like)
shSMAD3
(SCC-like)
shControl
shSMAD2
(ADC-like)
shSMAD3
(SCC-like)
100 μm
100–100
–100
100 μm
–100
100–100
±TGFβ1 ±TGFβ1 ±TGFβ1
48h [ 2D ] 24h [ 2D ] 24h [ 3D ]
20 mm
Collagen & cell
input channels
Culture medium
Trajectory = rd
rd
nd
Net displacement = nd
Persistence = nd / rd
Cell speed = rd / time
Culture medium
Imaging
Y. Juste-Lanas et al.
7
British Journal of Cancer
in the absence of TGF-β1 (Fig. 5d, e). Likewise, migration of ADC-
TAFs in Transwells was globally lower than control fibroblasts as in
3D. In contrast, we did not observe significant differences in the
mRNA levels of the major collagenase MMP1 between either ADC-
TAFs and SCC-TAFs from randomly selected patients (Supplemen-
tary Fig. 5D) or between shSMAD2 and shSMAD3 fibroblasts
(Supplementary Fig. 5E). These findings reveal that the enhanced
basal migration of high SMAD3 fibroblasts is largely independent
of collagen degradation but rather due to an intrinsic priming of
their migratory properties.
The larger migration of high SMAD3 fibroblasts in basal
conditions requires high glucose-dependent Erk1/2
hyperactivation
Prompted by the consistent migratory advantage observed in
high SMAD3 conditions in both 3D collagen gels and Transwell
assays selectively in the absence of TGF-β1, we began to explore
the underlying mechanisms. A common exogenous factor in both
assays was the presence of high glucose in the culture medium,
which are standard conditions for the cell culture of fibroblasts
and TAFs [38,39]. Moreover, Erk1/2 can be activated by both high
glucose and 3D collagen [40,41], have been strongly implicated in
the regulation of migration in numerous cell types including
fibroblasts [42], and may interact with SMAD3 in the absence of
TGF-β1[43]. To examine the potential involvement of high
glucose and/or Erk1/2 activation, we first analysed basal Transwell
migration with either standard high glucose (4.5 g/l) or low
glucose (1 g/l) medium and found a significant increased
migration with high glucose in shSMAD2 but not in shSMAD3
conditions. In contrast, low glucose medium abrogated the
migration differences between shSMAD2 and shSMAD3 fibroblasts
(Fig. 5f). Likewise, Transwell migration was significantly increased
in siControl ADC-TAFs (#13) in high versus low glucose conditions,
whereas no differences were observed in siSMAD3 ADC-TAFs
(Fig. 5g). Another supplemented soluble factor present in all
assays was insulin-transferrin-selenium (ITS). Because a major
function of insulin in cell culture is to stimulate glucose entry [44],
we also analysed Transwell migration in the presence or absence
of ITS and found a significant migration increase (~35%) with ITS
in low glucose conditions selectively in shSMAD2 fibroblasts,
whereas such migration increase was attenuated in high glucose
conditions (~12%) (Supplementary Fig. 5F, G), further under-
scoring the requirement of high glucose in the basal migration
priming of high SMAD3 fibroblasts. In contrast, shSMAD3
fibroblasts were consistently poorly responsive to both high
glucose and ITS in terms of migration (Fig. 5f and Supplementary
Fig. 5F, G).
Regarding Erk1/2 activation, we found that phosphorylated
Erk1/2 (pErk1/2) levels normalised by α-tubulin were increased by
~70% in standard high glucose in shSMAD2 compared to
shSMAD3 fibroblasts, whereas such increase was reduced to
~20% in low glucose conditions (Fig. 5h, i and Supplementary
Fig. 5H, I). Likewise, normalised pErk1/2 levels were ~35% higher
in siControl compared to siSMAD3 ADC-TAFs (#13) in our standard
high glucose medium (Fig. 5j); however, this difference was
smaller than that found between shSMAD2 and shSMAD3
fibroblasts, possibly due to the high endogenous expression of
TGF-β1 in ADC-TAFs [9], which may alter pErk1/2 [45] and
therefore bias the response to high glucose. In further qualitative
agreement with migration data, normalised pErk1/2 levels were
higher in shSMAD2 compared to shSMAD3 fibroblasts even in the
absence of ITS (Supplementary Fig. 5J). On the other hand, total
Erk1/2 levels remained fairly stable in all conditions (Fig. 5h, j and
Supplementary Fig. 5I). These results unveil a SMAD3-dependent
hyperactivation of Erk1/2 in the presence of high glucose.
To assess whether pErk1/2 hyperactivation is required for the
basal migration priming of high SMAD3 fibroblasts, we analysed
basal Transwell migration in the presence of 100 nM Trametinib, a
clinically approved inhibitor of MEK1/2 MAP kinases in NSCLC and
melanoma that acts right upstream of Erk1/2 [46]. Of note,
Trametinib significantly downregulated basal migration in stan-
dard high glucose medium in shSMAD2 but not shSMAD3
fibroblasts (Fig. 5k). Consistently, Trametinib elicited a drop in
basal Transwell migration in ADC-TAFs (#37, #13) in control
conditions but not upon knocking down SMAD3 in the presence
of standard high glucose (Fig. 5l, m). Yet, we noticed that the
migration reduction elicited by Trametinib varied between ~20-
80% depending on the cell model, even though Erk1/2 activation
was strongly abrogated in all cases (Supplementary Fig. 5K, L),
suggesting that additional molecular events other than pErk1/2
hyperactivation contribute to the migratory differences between
high SMAD3 and high SMAD2 fibroblasts in high glucose
conditions. Collectively, these results implicate high glucose-
dependent Erk1/2 hyperactivation in the migratory advantage of
high SMAD3 fibroblasts in the absence of TGF-β1, and reveal that
high SMAD2 fibroblasts are poorly responsive to both high
glucose/ITS and MEK1/2 inhibition.
The enhanced basal migration of high SMAD3 fibroblasts is
consistent with the larger accumulation of TAFs observed in
ADC compared to SCC at early stages
Finally, we examined the potential contribution of our observed
relationship between altered SMAD2/3 expression and migration
to the excessive accumulation of TAFs in lung cancer, since we
previously reported a larger TAF density in histologic samples in
ADC compared to SCC patients [6] in a small patient cohort (5
ADC, 5 SCC) (re-analysed as number density/image field in
Supplementary Fig. 6a). We confirmed this observation by
Fig. 3 Impact of altered SMAD2/3 in 3D collagen migration in lung fibroblasts and ADC-TAFs in basal conditions (no exogenous TGF-β1).
aOutline of the microdevice-based analysis of migration (24–48 h) of single fibroblasts cultured in dense 3D collagen gels. bRepresentative
phase contrast images of single control fibroblasts (#5) for each group (shControl, shSMAD2, shSMAD3) cultured in 3D collagen gels within
the microdevice in basal conditions during the experimental time-window (24–48 h). Yellow, green and red dots indicate the position of the
fibroblast centre at 24, 36 and 48 h, respectively. Scale bar, 50 μm. cTrajectory maps corresponding to the tracking of the centre of fibroblasts
from each group in basal conditions throughout the experimental time-window. The starting of each fibroblast was shifted to the origin of
coordinates for clarity here and thereafter. d–fAverage cell speed (d), net displacement (e) and persistence ratio (f) in basal conditions of
single control fibroblasts (#5) for each group (shControl, shSMAD2, shSMAD3) gathered from 3 independent microdevices per condition
(shown as bars). Each dot indicates the average of a single fibroblast henceforth. Note that persistence values range between 0 and 1, where 1
indicates migration without changing direction [30]. gFold SMAD3 mRNA of shControl and shSMAD3 ADC-TAFs (#37) cultured as in Fig. 1j.
hRepresentative phase contrast images of single shControl or shSMAD3 ADC-TAFs (#37) cultured as in b. Yellow, green and red dots indicate
the position of the fibroblast centre at 24, 36 and 48 h, respectively. Scale bar, 50 μm. iTrajectory maps corresponding to the tracking of the
centre of ADC-TAFs from each group in basal conditions throughout the experimental time-window as in c.j–lAverage cell speed (j), net
displacement (k) and persistence ratio (l) of single ADC-TAFs (#37) for each group (shown as bars) cultured as in b. Error bars represent
mean ± s.e.m.
###
p< 0.005 comparing shSMAD2 and shSMAD3. *p< 0.05; ***p< 0.005 with respect to shControl or comparing shControl and
shSMAD3 ADC-TAFs. Statistical comparisons were done using Student’sttest or Mann–Whitney (d–f). Mean values correspond to three
independent microdevices.
Y. Juste-Lanas et al.
8
British Journal of Cancer
assessing the number density of TAFs in an independent patient
cohort (Hospital de Bellvitge; 10 ADC, 9 SCC), identified by their
elongated nuclei (Fig. 6a top) and confirmed by their α-SMA
expression (Fig. 6a bottom), and found consistent results (Fig. 6b).
TAF accumulation is thought to arise largely from the
recruitment and/or proliferation of local resident fibroblasts
[5,47]. From the recruitment perspective, our observed migratory
advantage without TGF-β1 in high SMAD3 fibroblasts is consistent
0.0
0.5
1.0
1.5
Fold cell speed
(+TGF-E
E
1/–TGF-
E
1)
###
0.0
0.5
1.0
1.5
0.0
0.5
1.0
1.5
Fold net displacement
(+TGF-
E
1/–TGF-
E
1)
###
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
Persistence ratio
a
b
e
cd
+ TGF-
β
1 + TGF-
β
1
+ TGF-
β
1
+ TGF-
β
1
+ TGF-
β
1
shControl shSMAD2
(ADC-like)
shSMAD3
(SCC-like)
24 h
36 h
48 h
0.00
0.05
0.10
0.15
0.20
0.25
Cell speed (
μ
m/min)
0
50
100
150
200
250
300
Net displacement (
μ
m)
fgh
ADC-TAFs (#37)
shControl
ADC-TAFs (#37)
shSMAD3
ADC-TAFs (#37)
shControl
ADC-TAFs (#37)
shSMAD3
ADC-TAFs (#37)
shControl
ADC-TAFs (#37)
shSMAD3
kl m
ADC-TAF (#37)
shControl
ADC-TAF (#37)
shSMAD3
i
24 h 36 h 48 h
Fold persistence ratio
(+TGF-
E
1/–TGF-
E
1)
0.0
0.5
1.0
1.5
Fold cell speed
(+TGF-
E
1/–TGF-
E
1)
0.0
0.5
1.0
1.5
0.0
0.5
1.0
1.5
Fold net displacement
(+TGF-
E
1/–TGF-
E
1)
Fold persistence ratio
(+TGF-
E
1/ –TGF-
E
1)
###
50 μm
50 μm
j
250 μm
–250
–250 250
250 μm
–250
–250 250
250 μm
–250
–250 250
shControl
shSMAD2
(ADC-like)
shSMAD3
(SCC-like)
shControl
shSMAD2
(ADC-like)
shSMAD3
(SCC-like)
shControl
shSMAD2
(ADC-like)
shSMAD3
(SCC-like)
shControl
shSMAD2
(ADC-like)
shSMAD3
(SCC-like)
shControl
shSMAD2
(ADC-like)
shSMAD3
(SCC-like)
shControl
shSMAD2
(ADC-like)
shSMAD3
(SCC-like)
100 μm
100–100
–100
100 μm
100–100
–100
Y. Juste-Lanas et al.
9
British Journal of Cancer
with a larger recruitment of TAFs in ADC patients selectively at
early stages, when active TGF-β1 is expected to be low [18]. To
assess the potential contribution of differential proliferation, we
examined cell number density—which is a common growth
metric [5]—in different culture settings. Unlike histologic sections,
ADC-TAFs exhibited a significantly lower number density in basal
conditions compared to SCC-TAFs from randomly selected
patients in 2D cultures (Fig. 6c). Likewise, basal number density
was lower in shSMAD2 (ADC-like) fibroblasts compared to
shSMAD3 both in 2D (Supplementary Fig. 6B) and in the same
3D cultures used in our migration studies (Fig. 6e). In contrast, all
these differences were abrogated in the presence of TGF-β1
(Fig. 6d, f and Supplementary Fig. 6C). Moreover TGF-β1
consistently elicited a 10-20% increase in number density in
ADC-TAFs (Fig. 6d) and shSMAD2 fibroblasts (Fig. 6f) compared to
basal conditions, whereas such increase was not observed in SCC-
TAFs, in agreement with our previously observed poor response of
SCC-TAFs to soluble mitogenic cues [6]. Since we previously
reported that differences in number density between ADC-TAFs
and SCC-TAFs in 2D cultures were largely attributed to prolifera-
tion changes [6], our results support that altered SMAD2/3
expression elicits a proliferation advantage in terms of number
density selectively in high SMAD2 fibroblasts as in SCC-TAFs in
basal conditions that is not consistent with the larger TAF
accumulation observed in ADC. Collectively these findings
strongly support that the larger histologic TAF accumulation in
ADC is driven, at least in part, by the enhanced migration of ADC-
TAFs caused by their high SMAD3 expression at early stages, when
active TGF-β1 is expected to be low [18], as summarised in Fig. 6g.
DISCUSSION
SMAD2/3 are important transcription factors of the canonical TGF-
β1 pathway that are expressed in virtually all cell types [13,45]. In
fibroblasts, the TGF-β1/SMAD3 pathway has been extensively
documented as a positive regulator of fibrosis in the lung and
other organs [9,14,48]. In contrast, previous studies on the role of
SMAD2/3 in fibroblast migration are scarce and limited to 2D
cultures or transwells [15,16], which do not capture the
physiologic complexity of 3D microenvironments. To address
these limitations, we used a microfluidic device to examine for the
first time a panel of protrusions and migration descriptors of
single fibroblasts in dense 3D collagen gels. Our gel density was
>50% higher than that commonly used in other studies (1.5–2 mg/
ml) [17,29,31,49] to mimic the high collagen content reported in
lung cancer patients [7]. In addition, we used a technical
improvement in the protrusion analysis by checking different Z
planes instead of a single Zplane as commonly reported
[29,36,49].
We found that SMAD2/3 have a markedly distinct impact on
migration depending on the TGF-β1 context. Specifically, our
results revealed for the first time that high SMAD3 conditions (as
in shSMAD2 fibroblasts and ADC-TAFs) provide a migratory
advantage in terms of protrusions (fewer and longer) and
subsequent migration (faster and more directional) compared to
high SMAD2 conditions (as in shSMAD3 fibroblasts and SCC-TAFs)
selectively in the absence of exogenous TGF-β1, whereas TGF-β1
markedly abrogated this migratory advantage, promoting the
spatial confinement of fibroblasts. In contrast, high SMAD2
conditions were poorly responsive to TGF-β1 and consistently
exhibited the largest number of protrusions concomitantly with
the shortest protrusion length with and without TGF-β1, which
were subsequently associated with impaired migration through
less directional movement and a shorter net displacement.
Consistently, downregulating SMAD3 in ADC-TAFs was sufficient
to reduce basal migration in 3D and in Transwells. In agreement
with our findings, TGF-β1 was associated with reduced migration
in keratinocytes [50] and cardiac fibroblasts in Transwells [16].
Likewise, SMAD3 null cardiac fibroblasts exhibited a lower
migration than wild-type cells upon 1% FBS stimulation in
Transwells [16], in agreement with our high SMAD2 observations.
In contrast, unlike our 3D findings, no migratory effects were
observed in cardiac fibroblasts stimulated with 1% FBS upon
knocking down either SMAD2 or SMAD3 with siRNA using
Transwells [51], supporting that our multiparametric
microdevice-based 3D migration analysis may be more sensitive
in detecting migration changes in response to altered SMAD2/3
expression.
High SMAD3 fibroblasts were also the most sensitive to the
presence of exogenous TGF-β1, exhibiting more and shorter
protrusions that elicited a less directional movement and
subsequently a shorter net displacement, thereby increasing their
spatial confinement. Indeed, because TGF-β1 increased the
number of fibroblast protrusions in all SMAD2/3 settings and it is
known to increase traction forces in TAFs [52], it is likely that TGF-
β1 elicits more simultaneous traction in different (random)
directions, yielding an ineffective (non-persistent) movement
despite holding or increasing cell speed. In line with this
interpretation, ADC-TAFs exhibited lower basal migration than
control fibroblasts, which is consistent with the higher basal
secretion of TGF-β1 and expression of the contractility marker α-
SMA [9,53] in ADC-TAFs. In qualitative agreement with our
observations, TGF-β1 did not promote migration in cardiac
fibroblasts in Transwells [16], or even impaired it in keratinocytes
[50]. In contrast, our results are not consistent with the TGF-β1
stimulation of in vitro wound healing reported in human vocal
cord fibroblasts and keratinocytes or with its downregulation upon
SMAD3 inhibition using a scratch assay [15,54], since we observed
either no migratory changes in 3D or even a moderate increase in
Transwells in shSMAD3 fibroblasts in the presence of TGF-β1.
However, the marked differences between the scratch assay and
our 3D/Transwell assays may account for this discrepancy.
Fig. 4 Impact of altered SMAD2/3 in 3D collagen migration in lung fibroblasts and ADC-TAFs stimulated with TGF-β1. a Representative
phase contrast images of single control fibroblasts (#5) for each group (shControl, shSMAD2, shSMAD3) cultured in 3D collagen gels within
the microdevice with TGF-β1 during the experimental time-window (24–48 h). Yellow, green and red dots indicate the position of the
fibroblast centre at 24 h, 36 h and 48 h, respectively. Scale bar, 50 μm. bTrajectory maps corresponding to the tracking of the centre of
fibroblasts from each group in basal conditions throughout the experimental time-window. c–eAverage cell speed (c), net displacement (d)
and persistence ratio (e) with TGF-β1 of single control fibroblasts (#5) for each group (shControl, shSMAD2, shSMAD3) gathered from three
independent microdevices per condition (shown as bars). Each dot indicates the average of a single fibroblast. f–hFold average cell speed (f),
net displacement (g) and persistence ratio (h) assessed in the presence and absence of TGF-β1. Error bars were computed using error
propagation [67]. iRepresentative phase contrast images of single shControl or shSMAD3 ADC-TAFs (#37) cultured as in a. Yellow, green and
red dots indicate the position of the fibroblast centre at 24, 36 and 48 h, respectively. Scale bar, 50 μm. jTrajectory maps corresponding to the
tracking of the centre of ADC-TAFs from each group with TGF-β1 throughout the experimental time-window as in b.k–mFold average cell
speed (k), net displacement (l) and persistence ratio (m) of ADC-TAFs (#37) for each group assessed in the presence and absence of TGF-β1.
###
p< 0.005 comparing shSMAD2 and shSMAD3. **p< 0.01; ***p< 0.005 with respect to shControl or comparing shControl and shSMAD3
ADC-TAFs. Statistical comparisons were done using Student’sttest or Mann–Whitney (c–e). Mean values correspond to three independent
microdevices.
Y. Juste-Lanas et al.
10
British Journal of Cancer
0
20
40
60
0
20
40
60
Migration (% area)
0
20
40
60
Migration (% area)
0
20
40
60
Migration (% area)
0
20
40
60
Migration (% area)
###
Low glucose High glucose
dbac
Fold migration
(+TGf
β
/-TGF
β
)
#
0.0
0.5
1.0
1.5 #
– TGF-
β
1+ TGF-
β
1– TGF-
β
1– TGF-
β
1
f
Migration (% area)
0
20
40
60
Migration (% area)
Glucose
concentration:
1 g/L 4.5 g/L 1 g/L 4.5 g/L
shSMAD2
(ADC-like)
shSMAD3
(SCC-like)
ADC-TAFs (#13)
siControl
ADC-TAFs (#13)
siSmad3
0
20
40
60
2.0
1.5
1.0
0.5
0.0
+
1 g/L 4.5 g/L
#
#
+
+
+++
1 g/L 4.5 g/L
hg
klm
ADC-TAFs (#37)
shSmad3
ADC-TAFs (#37)
shControl
-+ -+
i
25
μ
m
25
μ
m25
μ
m
25
μ
m
+
ADC-TAFs (#13)
siSmad3
ADC-TAFs (#13)
siControl
-+ -+
25
μ
m
pErk1/2
shSMAD3
shSMAD2
shSMAD3
shSMAD2
Erk1/2
D-Tubulin
0.86 0.53
0.92 0.74
pErk1/2
DTubulin :
1.24 1.62
ratio shSMAD2
shSMAD3:
– TGF-
β
1 (basal)
e
25
μ
m
j
pErk1/2
Erk1/2
D-Tubulin
Low glucose High glucose
ADC-TAFs (#13)
1.98 1.47
0.37 0.59
0.62 1.35
pErk1/2
DTubulin :
ratio
siControl
siSMAD3:
pErk1/2 /
α
-tubulin
shSMAD2 / shSMAD3
Migration (% area)
0
20
40
60
Migration (% area)
0
20
40
60
Migration (% area)
shSMAD3
(SCC-like)
MEK inhibitor:
(Trametinib) -+ -+
shSMAD2
(ADC-like)
25
μ
m
Glucose
concentration: 1 g/L 4.5 g/L
siSMAD3
siControl
siSMAD3
siControl
shSMAD2
(ADC-like)
shSMAD3
(SCC-like)
shSMAD2
(ADC-like)
shSMAD3
(SCC-like)
shSMAD2
(ADC-like)
shSMAD3
(SCC-like)
ADC-TAFs (#37)
shSMAD3
ADC-TAFs (#37)
shControl
ADC-TAFs (#37)
shSMAD3
ADC-TAFs (#37)
shControl
25
μ
m25
μ
m
Fig. 5 Mechanistic insights on the intrinsic basal migration priming of shSMAD3 fibroblasts and ADC-TAFs: role of high
glucose, insulin–transferrin–selenium and Erk1/2. a–cAverage migration assessed using the Boyden Transwell assay of shSMAD2 and
shSMAD3 control fibroblasts (#5) in the absence (a) or presence of TGF-β1(b), and corresponding fold values (c). The same plots including
shControl fibroblasts are shown in Supplementary Fig. 5A, B. Bottom panels show representative images of the porous membrane of the
Transwell insert containing the migratory fibroblasts stained with crystal violet at the end of the experimental time-window (16 h) henceforth.
d,eAverage Transwell migration of shControl versus shSMAD3 ADC-TAFs (#37) (d) and siControl versus siSMAD3 ADC-TAFs (#13) (e) in the
absence of TGF-β1. SMAD3 mRNA levels of siControl and siSMAD3 ADC-TAFs are shown in Supplementary Fig. 5C. f,gAverage Transwell
migration with low or standard high glucose concentration in either shSMAD2 and shSMAD3 control fibroblasts (#5) (f) or siControl and
siSMAD3 ADC-TAFs (#13) (g) in the absence of TGF-β1 (i.e. basal conditions). Low glucose conditions started 3 days before seeding cells for the
migration experiment. Basal average Transwell migration with low or high glucose in the presence or absence of insulin-transferrin-selenium
(ITS) are shown in Supplementary Fig. 5F, G. h,iRepresentative Western blot analysis of phosphorylated Erk1/2 (pErk1/2), total Erk1/2 and
loading (α-tubulin) of shSMAD2 and shSMAD3 fibroblasts cultured as in fand examined 30 min after seeding. Corresponding densitometric
values of pErk1/2/α-tubulin are shown at the bottom (h) thereafter. Average densitometry ratio of pErk1/2 / α-tubulin in shSMAD2 with respect
to shSMAD3 are shown in i.jWestern blot of pErk1/2, total Erk1/2 and loading of siControl and siSMAD3 ADC-TAFs (#13) cultured as in h.
Densitometry ratio of pErk1/2/α-tubulin in siControl with respect to siSMAD3 is shown at the bottom. k–mAverage Transwell migration with
or without the MEK1/2 inhibitor Trametinib (100 nM) in either shSMAD2 or shSMAD3 control fibroblasts (#5) (k), shControl and shSMAD3 ADC-
TAFs (#37) (l) or siControl and siSMAD3 ADC-TAFs (#13) (m) in the absence of TGF-β1. Error bars represent mean ± s.e.m.
#
p< 0.05;
###
p< 0.005
comparing shSMAD2 and shSMAD3. *p< 0.05; ***p< 0.005 with respect to shControl.
+
p< 0.05;
+++
p< 0.005 comparing either low and high
glucose or the presence and absence of Trametinib. Statistical comparisons were done using Student’sttest. Mean values correspond to n≥2
experiments.
Y. Juste-Lanas et al.
11
British Journal of Cancer
Collectively, our results clarify the migration regulation of SMAD2/3
in fibroblasts in 3D and how it depends on the TGF-β1 context.
TGF-β1 inhibits growth through SMAD3 in numerous cell types
including epithelial cells and keratinocytes [55–57]. In contrast, we
observed cytostatic effects in high SMAD3 conditions in the
absence of TGF-β1 only, whereas TGF-β1 elicited a moderate
but consistent increase in number density particularly in high
SMAD3 fibroblasts and ADC-TAFs. Likewise, a TGF-β1-dependent
a
ADC
SCC
0.0
0.5
1.0
1.5
2.0 #
#
bH. Bellvitge cohort
Cell density
e
-TGF-
β
1
3D
p = 0.06 p = 0.09
p = 0.09
f
c
d
2D
360
45
30
15
0
60
45
30
15
0
2
1
0
3
2
1
0
+ TGF-
β
1
ADC-TAFs
SCC-TAFs
Cell density
g
α-SMA H&E
ADC
20 μm
SCC
20 μm
EARLY STAGE
( low active TGF-β1)
↑↑ Migration
Fast
recruitment
High SMAD3 fibroblasts
(as in ADC-TAFs)
Early cancer
cell
dissemination?
LATE STAGE
(high active TGF-β1)
Spatial
confinement
Cytostasis
Tumour evolution
↓↓ Migration ↑ Number
density
Mild TAF
expansion
Long-term TAF-cancer cell
crosstalk
MEK inh
Cell density Cell density
Fold fibroblasts density
shControl
shSMAD2
(ADC-like)
shSMAD3
(SCC-like)
ADC-TAFs
SCC-TAFs
shControl
shSMAD2
(ADC-like)
shSMAD3
(SCC-like)
Fig. 6 Differential TAF number density in ADC and SCC and potential contribution of SMAD2/3-regulated migration versus proliferation.
aIllustrative haematoxylin–eosin (H&E) (top) and α-SMA (bottom) staining of ADC and SCC patients. Arrow heads point to scattered and
spindle-shaped nuclei, which are histologic hallmarks of TAFs. bTAF number density assessed by morphometric analysis of haematoxylin
images from the Hospital de Bellvitge patient cohort (10 ADC, 9 SCC). Independent validation with a smaller cohort is shown in Supplementary
Fig. 6A. c,dFibroblast number density of TAFs randomly selected from our cohort (5 ADC, 5 SCC) cultured in 2D in the absence (c) or presence
(d) of TGF-β1. e,fFibroblast number density of the control fibroblasts (#5) from all groups (shControl, shSMAD2, shSMAD3) cultured within the
3D collagen gels used to analyse migration in the absence (e) or presence (f) of TGF-β1. Corresponding values of fibroblasts cultured in 2D as
in care shown in Supplementary Fig. 6B, C. gEmerging model for the impact of the interplay between SMAD3, migration and proliferation in
ADC-TAFs in early and late stages. Error bars represent mean ± s.e.m. Each dot corresponds to a different patient (b–d).
#
p< 0.05; comparing
either ADC-TAFs and SCC-TAFs or shSMAD2 and shSMAD3. Other p-values with respect to shControl. Statistical comparisons were done using
Student’sttest. Mean values correspond to n≥2 experiments.
Y. Juste-Lanas et al.
12
British Journal of Cancer
proliferation increase was reported in dermal fibroblasts [58],
further supporting that TGF-β1 is not an effective cytostatic
cytokine in fibroblasts. Although our mechanistic understanding
of these TGF-β1-independent pro-migratory and cytostatic func-
tions in high SMAD3 fibroblasts is limited, our results implicated
high glucose-dependent Erk1/2 hyperactivation in the migratory
advantage. Erk1/2 are important regulators of migration in
numerous cell types [42], and we found that standard high
glucose culture medium markedly increased both migration and
pErk1/2 selectively in high SMAD3 fibroblasts compared to low
glucose conditions. Conversely, the MEK inhibitor Trametinib
consistently attenuated basal migration in high SMAD3 but not
high SMAD2 fibroblasts. Our results are in agreement with
previous work reporting high glucose-dependent Erk1/2 hyper-
activation in kidney cells [41], and implicate for the first time
SMAD3 in this hyperactivation and a subsequent migration
enhancement in fibroblasts.
The molecular underpinnings of these new SMAD3-specific
fibroblast functions remain to be determined. Yet, it is unlikely
that they involve the C-terminal residues of SMAD3 that become
phosphorylated in response to TGF-β1 as part of the canonical
TGF-β1/SMAD3 pathway [59], since we previously showed that
these residues remain unphosphorylated in the basal conditions
(i.e. absence of TGF-β1) [9] in which the migration enhancement
was observed. Alternatively, the linker region of SMAD3 could be
implicated, since this region contains several phosphorylation
sites that can be regulated by Erk1/2 and other kinases in the
absence of TGF-β1[43,45], and there is growing evidence that the
linker regions may modulate different cell functions in SMAD2 and
SMAD3 independently of TGF-β1[43,60]. Moreover, it has been
shown that high glucose enhanced the sensitivity to exogenous
TGF-β1 in mouse embryonic fibroblasts and kidney epithelial cells
[41,61], and the SMAD3 linker was involved in this enhancement
[41], in agreement with our observed strongest response to TGF-
β1 selectively in high SMAD3 fibroblasts. On the other hand, it is
worth noting that a larger glucose uptake was recently reported in
ADC-TAFs compared to SCC-TAFs [62], which could also contribute
to our observed high glucose-dependent Erk1/2 activation in high
SMAD3 conditions as in ADC-TAFs. However, our current knowl-
edge of the Erk1/2-SMAD2/3 crosstalk in the absence of TGF-β1is
still very scarce and warrants further investigations. Likewise, it
would be interesting to assess how our observed high SMAD2-
driven poor response to MEK inhibition in fibroblasts may
contribute to the resistance to MEK inhibitors in cancer [63].
Our findings provide new insights on how TAFs may contribute
to tumour progression in lung cancer depending on the TGF-β1
context and the histologic subtype. Under low active TGF-β1
conditions as in early tumour stages [18], the enhanced migration
of high SMAD3 fibroblasts may be a major contributor to the
larger TAF recruitment observed in ADC compared to SCC. In
addition, because TAFs can promote cancer cell dissemination by
leading collective cancer cell migration [64], it is conceivable that
the basal migration priming of high SMAD3 fibroblasts may
contribute to the early dissemination of cancer cells in ADC, which
is a common clinical observation whose underlying mechanisms
remain unknown [65]. However, testing this hypothesis awaits
future investigations. Moreover, our results support that the
tumour-promoting effects associated with the enhanced migra-
tion of high SMAD3 fibroblasts could be prevented with
Trametinib or other MEK inhibitors in ADC but not SCC, since
high SMAD2 fibroblasts as in SCC-TAFs were largely non-
responsive to MEK inhibition. In contrast, at later stages where
active TGF-β1 is expected to be abundant at doses within the
same range used in our experimental settings [18,21], the
impaired migration observed in all SMAD2/3 conditions con-
comitantly with the expected α-SMA-dependent increased con-
tractility may collectively increase the spatial confinement of TAFs
and ultimately facilitate the long-term interactions with adjacent
cancer cells to promote their epigenetic reprogramming [52,66]
(Fig. 6g).
In summary, we unveil how altered SMAD2/3 expression
provides migration and proliferation advantages only in the
absence of TGF-β1, although in opposite directions, since
enhanced migration (but not proliferation) was observed selec-
tively in high SMAD3 conditions, strongly supporting that the
larger TAF accumulation in ADC occurs at early stages (under low
active TGF-β1). Moreover, our results provide a rationale for the
implication of ADC-TAFs in the early ADC cancer cell dissemination
observed in clinical settings and for the therapeutic use of MEK
inhibitors against the enhanced migratory phenotype of ADC-
TAFs (Fig. 6g).
DATA AVAILABILITY
The data sets generated and analysed in the study are available from the
corresponding author on reasonable request.
REFERENCES
1. Siegel RL, Miller KD, Jemal A. Cancer statistics, 2019. CA Cancer J Clin.
2019;69:7–34.
2. Chen Z, Fillmore CM, Hammerman PS, Kim CF, Wong K-K. Non-small-cell lung
cancers: a heterogeneous set of diseases. Nat Rev Cancer. 2014;14:535–46.
3. Kalluri R. The biology and function of fibroblasts in cancer. Nat Rev Cancer.
2016;16:582–98.
4. Ohlund D, Elyada E, Tuveson D. Fibroblast heterogeneity in the cancer wound. J
Exp Med. 2014;211:1503–23.
5. Sahai E, Astsaturov I, Cukierman E, DeNardo DG, Egeblad M, Evans RM, et al. A
framework for advancing our understanding of cancer-associated fibroblasts. Nat
Rev Cancer. 2020;20:174–86.
6. Puig M, Lugo R, Gabasa M, Gimenez A, Velasquez A, Galgoczy R, et al. Matrix
stiffening and beta(1) integrin drive subtype-specificfibroblast accumulation in
lung cancer. Mol Cancer Res. 2015;13:161–73.
7. Alcaraz J, Carrasco JL, Millares L, Luis I-C, Fernández-Porras FJ, Martínez-Romero
A, et al. Stromal markers of activated tumor associated fibroblasts predict poor
survival and are associated with necrosis in non-small cell lung cancer. Lung
Cancer. 2019;135:151–60.
8. Hasegawa Y, Takanashi S, Kanehira Y, Tsushima T, Imai T, Okumura K. Trans-
forming growth factor-beta 1 level correlates with angiogenesis, tumor pro-
gression, and prognosis in patients with nonsmall cell lung carcinoma. Cancer.
2001;91:964–71.
9. Ikemori R, Gabasa M, Duch P, Vizoso M, Bragado P, Arshakyan M, et al. Epigenetic
SMAD3 repression in tumor-associated fibroblasts impairs fibrosis and response
to the antifibrotic drug nintedanib in lung squamous cell carcinoma. Cancer Res.
2020;80:276–90.
10. Gabasa M, Ikemori R, Hilberg F, Reguart N, Alcaraz J. Nintedanib selectively
inhibits the activation and tumor-promoting effects of fibroblasts from lung
adenocarcinoma patients. Br J Cancer. 2017;117:1128–38.
11. Reck M, Kaiser R, Mellemgaard A, Douillard J-Y, Orlov S, Krzakowski M, et al.
Docetaxel plus nintedanib versus docetaxel plus placebo in patients with pre-
viously treated non-small-cell lung cancer (LUME-Lung 1): a phase 3, double-
blind, randomised controlled trial. Lancet Oncol. 2014;15:143–55.
12. Duch P, Díaz-Valdivia N, Ikemori R, Gabasa M, Radisky ES, Arshakyan M, et al.
Aberrant TIMP-1 overexpression in tumor-associated fibroblasts drives tumor
progression through CD63 in lung adenocarcinoma. Matrix Biol. 2022;111:207–25.
13. Brown KA, Pietenpol JA, Moses HL. A tale of two proteins: differential roles and
regulation of Smad2 and Smad3 in TGF-beta signaling. J Cell Biochem.
2007;101:9–33.
14. Huang S, Chen B, Humeres C, Alex L, Hanna A, Frangogiannis NG. The role of
Smad2 and Smad3 in regulating homeostatic functions of fibroblasts in vitro and
in adult mice. Biochimt Biophys Acta Mol Cell Res. 2020;1867:118703.
15. Branski RC, Bing R, Kraja I, Amin MR. The role of Smad3 in the fibrotic phenotype
in human vocal fold fibroblasts. Laryngoscope. 2016;126:1151–6.
16. Dobaczewski M, Bujak M, Li N, Gonzalez-Quesada C, Mendoza LH, Wang X-F, et al.
Smad3 signaling critically regulates fibroblast phenotype and function in healing
myocardial infarction. Circ Res. 2010;107:418–28.
17. Grinnell F, Ho CH, Tamariz E, Lee DJ, Skuta G. Dendritic fibroblasts in three-
dimensional collagen matrices. Mol Biol Cell. 2003;14:384–95.
18. Prud’homme GJ. Pathobiology of transforming growth factor beta in cancer,
fibrosis and immunologic disease, and therapeutic considerations. Lab Investig.
2007;87:1077–91.
Y. Juste-Lanas et al.
13
British Journal of Cancer
19. Giatromanolaki A, Koukourakis MI, Harris AL, Polychronidis A, Gatter KC, Sivridis E.
Prognostic relevance of light chain 3 (LC3A) autophagy patterns in colorectal
adenocarcinomas. J Clin Pathol. 2010;63:867–72.
20. Bankhead P, Loughrey MB, Fernández JA, Dombrowski Y, McArt DG, Dunne PD,
et al. QuPath: Open source software for digital pathology image analysis. Sci Rep.
2017;7:16878.
21. Domagała-Kulawik J, Hoser G, Safianowska A, Grubek-Jaworska H, Chazan R.
Elevated TGF-beta1 concentration in bronchoalveolar lavage fluid from patients
with primary lung cancer. Arch Immunol Ther Exp. 2006;54:143–7.
22. Gabasa M, Radisky ES, Ikemori R, Bertolini G, Arshakyan M, Hockla A, et al. MMP1
drives tumor progression in large cell carcinoma of the lung through fibroblast
senescence. Cancer Lett. 2021;507:1–12.
23. Shin Y, Han S, Jeon JS, Yamamoto K, Zervantonakis IK, Sudo R, et al. Microfluidic
assay for simultaneous culture of multiple cell types on surfaces or within
hydrogels. Nat Protoc. 2012;7:1247–59.
24. Plou J, Juste-Lanas Y, Olivares V, del Amo C, Borau C, García-Aznar JM. From
individual to collective 3D cancer dissemination: roles of collagen concentration
and TGF-β. Sci Rep. 2018;8:12723.
25. Movilla N, Borau C, Valero C, García-Aznar JM. Degradation of extracellular matrix
regulates osteoblast migration: a microfluidic-based study. Bone. 2018;107:10–17.
26. Farahat WA, Wood LB, Zervantonakis IK, Schor A, Ong S, Neal D, et al. Ensemble
analysis of angiogenic growth in three-dimensional microfluidic cell cultures.
PLoS ONE. 2012;7:e37333.
27. Movilla N, Valero C, Borau C, García-A znar JM. Matrix degradation regulates
osteoblast protrusion dynamics and individual migration. Integr Biol.
2019;11:404–13.
28. Merino-Casallo F, Gomez-Benito MJ, Juste-Lanas Y, Martinez-Cantin R, Garcia-
Aznar JM. Integration of in vitro and in silico models using Bayesian optimization
with an application to stochastic modeling of mesenchymal 3D cell migration.
Front Physiol. 2018;9:1246.
29. Giri A, Bajpai S, Trenton N, Jayatilaka H, Longmore GD, Wirtz D. The Arp2/3
complex mediates multigeneration dendritic protrusions for efficient
3-dimensional cancer cell migration. FASEB J. 2013;27:4089–99.
30. Rahman-Zaman A, Shan S, Reinhart-King CA. Cell migration in microfabricated 3D
collagen microtracks is mediated through the prometastatic protein Girdin. Cell
Mol Bioeng. 2018;11:1–10.
31. Fraley SI, Feng Y, Krishnamur thy R, Kim D-H, Celedon A, Longmore GD, et al. A
distinctive role for focal adhesion proteins in three-dimensional cell motility. Nat
Cell Biol. 2010;12:598–604.
32. Juste-Lanas Y, Guerrero PE, Camacho-Gómez D, Hervás-Raluy S, García-Aznar JM,
Gomez-Benito MJ. Confined cell migration and asymmetric hydraulic environ-
ments to evaluate the metastatic potential of cancer cells. J Biomech Eng.
2022;144:074502.
33. Moreno-Arotzena O, Borau C, Movilla N, Vicente-Manzanares M, García-Aznar JM.
Fibroblast migration in 3D is controlled by haptotaxis in a non-muscle myosin II-
dependent manner. Ann Biomed Eng. 2015;43:3025–39.
34. Del Amo C, Olivares V, Cóndor M, Blanco A, Santolaria J, Asín J, et al. Matrix
architecture plays a pivotal role in 3D osteoblast migration: the effect of inter-
stitial fluid flow. J Mech Behav Biomed Mater. 2018;83:52–62.
35. Gabasa M, Arshakyan M, Llorente A, Chuliá-Peris L, Pavelescu I, Xaubet A, et al.
Interleukin-1βmodulation of the mechanobiology of primary human pulmonary
fibroblasts: potential implications in lung repair. Int J Mol Sci. 2020;21:8417.
36. Carey SP, Goldblatt ZE, Martin KE, Romero B, Williams RM, Reinhart-King CA. Local
extracellular matrix alignment directs cellular protrusion dynamics and migration
through Rac1 and FAK. Integr Biol. 2016;8:821–35.
37. Galgoczy R, Pastor I, Colom A, Giménez A, Mas F, Alcaraz J. A spectrophotometer-
based diffusivity assay reveals that diffusion hindrance of small molecules in
extracellular matrix gels used in 3D cultures is dominated by viscous effects.
Colloid Surf B Biointerfaces. 2014;120:200–7.
38. Huang W, Zhang L, Yang M, Wu X, Wang X, Huang W, et al. Cancer-associated
fibroblasts promote the survival of irradiated nasopharyngeal carcinoma cells via
the NF-κB pathway. J Exp Clin Cancer Res. 2021;40:87.
39. Franco-Barraza J, Raghavan KS, Luong T, Cukierman E. Engineering clinically-
relevant human fibroblastic cell-derived extracellular matrices. Methods Cell Biol.
2020;156:109–60.
40. Cukierman E, Pankov R, Stevens DR, Yamada KM. Taking cell-matrix adhesions to
the third dimension. Science. 2001;294:1708–12.
41. Pan X, Phanish MK, Baines DL, Dockrell MEC. High glucose-induced Smad3 linker
phosphorylation and CCN2 expression are inhibited by dapagliflozin in a diabetic
tubule epithelial cell model. Biosci Rep. 2021;41:BSR20203947.
42. Huang C, Jacobson K, Schaller MD. MAP kinases and cell migration. J Cell Sci.
2004;117:4619–28.
43. Ooshima A, Park J, Kim SJ. Phosphorylation status at Smad3 linker region mod-
ulates transforming growth factor-β-induced epithelial-mesenchymal transition
and cancer progression. Cancer Sci. 2019;110:481–8.
44. Howard BV, Mott DM, Fields RM, Bennett PH. Insulin stimulation of glucose entry
in cultured human fibroblasts. J Cell Physiol. 1979;101:129–38.
45. Derynck R, Zhang YE. Smad-dependent and Smad-independent pathways in TGF-
beta family signalling. Nature. 2003;425:577–84.
46. Han J, Liu Y, Yang S, Wu X, Li H, Wang Q. MEK inhibitors for the treatment of non-
small cell lung cancer. J Hematol Oncol. 2021;14:1.
47. Alexander J, Cukierman E. Cancer associated fibroblast: mediators of tumor-
igenesis. Matrix Biol. 2020;91-92:19–34.
48. Meng XM, Huang XR, Chung ACK, Qin W, Shao X, Igarashi P, et al. Smad2 protects
against TGF-beta/Smad3-mediated renal fibrosis. J Am Soc Nephrol.
2010;21:1477–87.
49. Fraley SI, Feng Y, Giri A, Longmore GD, Wirtz D. Dimensional and temporal
controls of three-dimensional cell migration by zyxin and binding partners. Nat
Commun. 2012;3:719.
50. Hosokawa R, Urata MM, Ito Y, Bringas P Jr, Chai Y. Functional significance of
Smad2 in regulating basal keratinocyte migration during wound healing. J Invest
Dermatol. 2005;125:1302–9.
51. Huang S, Chen B, Su Y, Alex L, Humeres C, Shinde AV, et al. Distinct roles of
myofibroblast-specific Smad2 and Smad3 signaling in repair and remodeling of
the infarcted heart. J Mol Cell Cardiol. 2019;132:84–97.
52. Vizoso M, Puig M, Carmona FJ, Maqueda M, Velásquez A, Gómez A, et al. Aberrant
DNA methylation in non small cell lung cancer associated fibroblasts. Carcino-
genesis. 2015;36:1453–63.
53. Lugo R, Gabasa M, Andriani F, Puig M, Facchinetti F, Ramírez J, et al. Heterotypic
paracrine signaling drives fibroblast senescence and tumor progression of large
cell carcinoma of the lung. Oncotarget. 2016;7:82324–37.
54. Kocic J, Bugarski D, Santibanez JF. SMAD3 is essential for transforming growth
factor-β1-induced urokinase type plasminogen activator expression and migra-
tion in transformed keratinocytes. Eur J Cancer. 2012;48:1550–7.
55. Uemura M, Swenson ES, Gaça MDA, Giordano FJ, Reiss M, Wells RG. Smad2 and
smad3 play different roles in rat hepatic stellate cell function and alpha-smooth
muscle actin organization. Mol Biol Cell. 2005;16:4214–24.
56. Kim SG, Kim H-A, Jong H-S, Park J-H, Kim NK, Hong SH, et al. The endogenous
ratio of Smad2 and Smad3 influences the cytostatic function of Smad3. Mol Biol
Cell. 2005;16:4672–83.
57. Ashcroft GS, Yang X, Glick AB, Weinstein M, Letterio JL, Mizel DE, et al. Mice
lacking Smad3 show accelerated wound healing and an impaired local inflam-
matory response. Nat Cell Biol. 1999;1:260–6.
58. Sapudom J, Müller CD, Nguyen K-T, Martin S, Anderegg U, Pompe T. Matrix
remodeling and hyaluronan production by myofibroblasts and cancer-associated
fibroblasts in 3D collagen matrices. Gels. 2020;6:33.
59. Massagué J. How cells read TGF-beta signals. Nat Rev Mol Cell Biol.
2000;1:169–78.
60. Kamato D, Do BH, Osman N, Ross BP, Mohamed R, Xu S, et al. Smad linker region
phosphorylation is a signalling pathway in its own right and not only a modulator
of canonical TGF-βsignalling. Cell Mol Life Sci. 2020;77:243–51.
61. Wu L, Derynck R. Essential role of TGF-beta signaling in glucose-induced cell
hypertrophy. Dev Cell. 2009;17:35–48.
62. Zhang W, Bouchard G, Yu A, Shafiq M, Jamali M, Shrager JB, et al. GFPT2-
expressing cancer-associated fibroblasts mediate metabolic reprogramming in
human lung adenocarcinoma. Cancer Res. 2018;78:3445–57.
63. Kun E, Tsang YTM, Ng CW, Gershenson DM, Wong KK. MEK inhibitor resistance
mechanisms and recent developments in combination trials. Cancer Treat Rev.
2021;92:102137.
64. Labernadie A, Kato T, Brugués A, Serra-Picamal X, Derzsi S, Arwert E, et al. A
mechanically active heterotypic E-cadherin/N-cadherin adhesion enables fibro-
blasts to drive cancer cell invasion. Nat Cell Biol. 2017;19:224–38.
65. Hoffman PC, Mauer AM, Vokes EE. Lung cancer. Lancet. 2000;355:479–85.
66. Alcaraz J, Ikemori R, Llorente A, Díaz-Valdivia N, Reguart N, Vizoso M. Epigenetic
reprogramming of tumor-associated fibroblasts in lung cancer: therapeutic
opportunities. Cancers. 2021;13:3782.
67. Taylor JR. An introduction to error analysis. Sausalito: University Science Books;
1997.
ACKNOWLEDGEMENTS
We thank Raimon Sunyer, Pere Roca-Cusachs, Patricia Fernández, Kate Neal (UB),
Zanetta Zoi and Joan Montero (IBEC) for technical support, and Ramon Farré (UB) for
support.
AUTHOR CONTRIBUTIONS
Study concept, design and supervision: JMG-A, JA; acquisition of data: YJ-L, ND-V, AL,
RI, AB, MA, CB, JCR; analysis and interpretation of data: YJ-L, ND-V, AL, JA;
Y. Juste-Lanas et al.
14
British Journal of Cancer
development of the methodology: YJ-L, RI, CB, AL, JMG-A, JA; drafting of the
manuscript: YJ-L, ND-V, AL, JA; provided patient samples and clinical information: JCR,
EN, NR, JR; critical revision of the manuscript: AB, NR, JMG-A; obtained funding: JMG-
A, EN, JA; technical or material support: CB, JR.
FUNDING
This work was supported by grants from the Agencia Estatal de Investigación (AEI/
FEDER) (PI13/02368, SAF2016-79527-R and PID2019-110944RB-I00 to JA, PID2021-
122409OB-C21, RTI2018-094494-B-C21 to JMG-A), European Research Council (Adg-
101018587 to JMG-A), Instituto de Salud Carlos III (PI14/01109, PI18/00920 to EN) (co-
funded by European Regional Development Fund. ERDF, a way to build Europe),
Fundació Privada Cellex (to JA), Generalitat de Catalunya (AGAUR SGR 661 and CERCA
Programme to JA), Junta Provincial de Barcelona de l’Associació Espanyola Contra el
Càncer (AECC B16-917 to JA), Sociedad Española de Neumología y Cirugía Torácica –
SEPAR (SEPAR 951/2019 to JA), Spanish Society of Medical Oncology grant for
emerging research groups (to EN), and by fellowships from Spanish Ministry of
Science and Innovation (FPU17/03867 to YJ-L), Ciência sem Fronteiras CNPq (to RI),
CONICYT (to NDV) and IBEC (to ALL).
COMPETING INTERESTS
The authors declare no competing interests.
ETHICS APPROVAL AND CONSENT TO PARTICIPATE
This study was approved by the Institutional Review Board of the Hospital Clínic,
Hospital de Bellvitge and Universitat de Barcelona. Informed consent was obtained
from all participants involved in the study, and all experiments were conducted in
line with the principles of the Declaration of Helsinki.
CONSENT FOR PUBLICATION
There are no individual person’s data from all participants involved in the study.
ADDITIONAL INFORMATION
Supplementary information The online version contains supplementary material
available at https://doi.org/10.1038/s41416-022-02093-x.
Correspondence and requests for materials should be addressed to Jordi Alcaraz.
Reprints and permission information is available at http://www.nature.com/
reprints
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims
in published maps and institutional affiliations.
Open Access This article is licensed under a Creative Commons
Attribution 4.0 International License, which permits use, sharing,
adaptation, distribution and reproduction in any medium or format, as long as you give
appropriate credit to the original author(s) and the source, provide a link to the Creative
Commons license, and indicate if changes were made. The images or other third party
material in this article are included in the article’s Creative Commons license, unless
indicated otherwise in a credit line to the material. If material is not included in the
article’s Creative Commons license and your intended use is not permitted by statutory
regulation or exceeds the permitted use, you will need to obtain permission directly
from the copyright holder. To view a copy of this license, visit http://
creativecommons.org/licenses/by/4.0/.
© The Author(s) 2022
Y. Juste-Lanas et al.
15
British Journal of Cancer
