Fibroblast Senescence-Associated Extracellular Matrix Promotes

Heterogeneous Lung Niche

Andrew M. Howes1,2,3, Nova C. Dea1, Deepraj Ghosh1, Krishangi Krishna2, Yihong Wang3, Yanxi Li1
Braxton Morrison1, Kimani C. Toussaint2, and Michelle R. Dawson1,a)
1Department of Molecular Biology, Cell Biology, and Biochemistry, Brown University, Providence, RI 029012 U.S.
2School of Engineering, Brown University, Providence, RI 02912 U.S.
3Department of Pathology and Laboratory Medicine, Brown University, Providence, RI 02912 U.S.
a)Author to whom correspondence should be addressed: [email protected]

ABSTRACT
Senescent cell accumulation in the pulmonary niche is associated with heightened susceptibility to age-
related disease, tissue alterations, and ultimately a decline in lung function. Our current knowledge of
senescent cell-ECM dynamics is limited, and our understanding of how senescent cells influence spatial
ECM architecture changes over time is incomplete. Herein is the design of an in vitro model of SA-ECM
remodeling using senescent lung fibroblast-derived matrix that captures the spatiotemporal dynamics of an
evolving senescent ECM architecture. Multiphoton second-harmonic generation microscopy was utilized to
examine the spatial and temporal dynamics of fibroblast SA-ECM remodeling, which revealed a biphasic
process that established a disordered and heterogeneous architecture. Additionally, we observed that
inhibition of transforming growth factor-β (TGF-β) signaling during SA-ECM remodeling led to improved
local collagen fiber organization. Lastly, we examined patient samples diagnosed with pulmonary fibrosis
to further tie our results of the in vitro model to clinical outcomes. Moreover, we observed that the
senescence marker p16 is correlated with local collagen fiber disorder. By elucidating the temporal
dynamics of SA-ECM remodeling, we provide further insight on the role of senescent cells and their
contributions to pathological ECM remodeling.

INTRODUCTION instability, and loss of proteasome activity [10].

Aging of the lung is linked to the Upon increase in any of these factors, pre-
accumulation of senescent cells, increased senescent cells exit the cell-cycle and become
chronic inflammation and subsequent post-mitotic by the activation of the p53/p21CIP1
modifications to the extracellular matrix (ECM) and p16INK4a/pRB pathways. After the induction of
[1]. It has been established that this senescence, senescent cells then release SASP
accumulation of senescent cells generates a into the surrounding environment. contains
higher risk for age-related diseases such as multiple factors capable of remodeling the ECM
cancer, cardiovascular disease, environment, such as matrix metalloproteinases
neurodegenerative disease, and fibrotic (MMPs), collagen crosslinkers (LOX, TGM2), and
disease [2-5]. Chronic respiratory diseases are pro-inflammatory regulators known to influence
currently the 3rd leading cause of death in the US the matrix remodeling potential of surrounding
for those over 65 years of age [6]. By 2050, it is cells (TGF-β, IL-6, IL-8) [11-13]. The buildup of
estimated that 1 in 4 Americans will be above the senescent cells consequently dysregulates tissue
age of 65, leading to an increase in the homeostasis, resulting in a heightened
prevalence of age-related pulmonary disease [7]. inflammatory state through the overexpression of
Therefore, it is of pertinence to study the influence ECM-degrading enzymes, chemokines,
of senescent cells on the architecture of the lung. cytokines, and growth factors [14, 15].
Cellular senescence is a prevailing aspect Consequent chronic inflammation by apoptosis
of pulmonary fibrosis, whereby senescent cells resistant senescent cells disrupts ECM
directly remodel ECM or indirectly through remodeling homeostasis, resulting in a perpetual
senescence-associated secretory phenotype pro-fibrotic niche that leads to the loss of tissue
(SASP) signaling to neighboring cells [8, 9]. This function [16].
state of irreversible cell-cycle arrest is induced by Due to the pleiotropic nature of
pro-aging factors like mitochondrial dysfunction, senescence, our understanding of how senescent
epigenetic alterations, telomere attrition, genomic fibroblasts remodel the ECM architecture

1
organization over time is misunderstood. Muñoz- cells. Previously, it was not clear if senescent
Espín, et al. propose that transient cellular fibroblasts promote ECM disorder biophysically.
senescence promotes tissue remodeling in three Our findings indicate that senescent fibroblast
sequential events: senescence, clearance, and actin cytoskeleton disorganization and elevated
regeneration. However, persistent damage over integrin adhesion results in multilateral tension on
time compromises this model and the lack of the matrix that promotes disorganized collagen
senescent cell clearance and resolution results in fiber organization.
senescent lesions [17]. It is indicated that Based on this latter observation that
idiopathic pulmonary fibrosis (IPF), a chronic age- disordered senescent fibroblast polarization and
related pulmonary fibrotic disease, is influenced cytoskeleton tension is applied to the matrix in a
by cellular senescence [18]. Furthermore, IPF multilateral fashion, we separately inhibited
fibroblasts have been shown to display mechanoresponsive myocardin-related
senescence-like phenotypes [9, 19]. Herein, the transcription factor-A (MRTF-A) and signaling of
disruption of immunosurveillance due to the TGF-β pathway through TGFβR1. The MRTF-
senescent cells themselves, or in part an aging SRF pathway is a critical regulator of Cell-ECM
immune system, results in the accumulation of dynamics that links actin cytoskeleton
senescent cells within the tissue niche [20]. organization with transcriptional regulation,
Moreover, TGF-β, a profibrotic cytokine known to providing essential signals for adhesion, cell
orchestrate the development of IPF and other polarization and contractility [29]. Furthermore,
fibrotic diseases, is abundantly found in mechanical stress due to stiffening of the ECM
senescent cell SASP [21-24]. Additionally, it has can modulate MRTF-A signaling, contributing to
been shown to induce an immunosuppressive an α-smooth muscle actin (αSMA) stress-fiber
senescent cell state and induce cell homing expressing myofibroblast phenotype that is
defects [25, 26]. Although TGF-β is involved in a responsible for fibrotic remodeling [30-32].
myriad of other signaling processes, its ability to Moreover, Elevated levels of αv (CD51) and β1
stimulate profibrotic ECM remodeling is one of its (CD29) integrins have been shown to increase
hallmark traits. Excessive TGF-β-induced ECM the localization of MRTF-A to the nucleus,
remodeling ultimately leads to progressive fibrotic resulting in heightened activation and contractility
tissue-level alterations resulting in decreased [33]. By inhibiting the nuclear localization of
lung function in both mice and humans [27]. MRTF-A, we aimed to reduce SA-ECM
We expand upon the work done by Choi disorganization by limiting senescent fibroblast
et. al by examining senescent ECM structural multilateral tension on the matrix. Secondly, TGF-
alterations that influence cell and tissue behavior β is a well-known pro-fibrotic cytokine that
over time [28]. This study expands upon their influences ECM remodeling potential and is
work by quantifying the spatial and temporal expressed within SASP. By blocking TGF-β
dynamics of senescent pulmonary fibroblasts in signaling through TGFβR1, we intended to limit
vitro and their impact on the ECM architecture of the senescent fibroblast capability to remodel a
the lung niche. Previous senescence models disorganized SA-ECM architecture. Inhibition of
have not examined dynamic senescent cell the TGF-β pathway resulted in the improvement
changes of the senescent matrix and how of local SA-ECM architecture, generating more
senescent cells influence matrix architecture over organized anisotropic collagen fibers compared
time. Therefore, we designed an in vitro to MRTF-A inhibition. Lastly, we examined clinical
senescent lung fibroblast-derived matrix (FDM) patient samples to understand how collagen fiber
model to study SA-ECM architecture changes organization changes with the age and the
and spatial remodeling during the induction of percentage of p16 positive cells within patient
senescence. Using multiphoton second- tissue diagnosed with interstitial pulmonary
harmonic generation (SHG) microscopy, we fibrosis. Here we show a linear relationship
identified distinct structural biomarkers of an between the increasing age of pulmonary fibrosis
evolving pathological SA-ECM architecture patients and local collagen fiber disorganization.
undergoing phase-like changes to stabilize a Moreover, we also observed a correlation
heterogeneously disorganized senescent matrix between the percent of p16 positive cells and
over time. We hypothesized that senescent local collagen fiber disorder within these patient
fibroblasts contributed to the disorganization of samples.
the matrix mechanically by having disordered By spatiotemporally modeling the
actin cytoskeleton and an increased abundance stabilization of a senescent matrix, our results
of integrin binding compared to pre-senescent further demonstrate that an abundance of

2
senescent cells plays a potent role in the 1(g)]. DNA damage was observed via staining for
deregulation of tissue remodeling and exhibit DNA double-strand break marker γH2A.X. The
temporally different ECM remodeling phases to percentage of senescent fibroblasts positive for
establish a distinct heterogeneously disorganized γH2A.X was significantly higher at day 7 (31%
pathological architecture over time. Additionally, versus 0.4%) and remained increased compared
our findings report an innovative in vitro to pre-senescent conditions [Fig. 1(h), see
senescent fibroblast-derived matrix model that Supplemental Fig. 1(b)]. Lastly, qRT-PCR was
captures the spatial and temporal dynamics of used to measure expression levels of cyclin-
senescent fibroblast matrix remodeling. Utilizing dependent kinase inhibitors p16 and p21.
multiphoton SHG microscopy, we establish that Senescent fibroblast p16 expression levels were
SA-ECM remodeling is temporally biphasic and not significantly different from pre-senescent
establishes a heterogeneously disorganized fibroblasts [Fig. 1(i)]. However, p21 expression
ECM architecture. We further connect these levels were significantly higher than that of pre-
results by showing a correlation between age, senescent fibroblasts [Fig. 1(j)].
p16 positivity, and ECM disorganization within
patients diagnosed with pulmonary fibrosis. Generation of senescent primary lung
These findings build upon previous observations fibroblast-derived matrix model
that the senescent fibroblast phenotype promotes To grow senescent ECM and quantify
a pathological SA-ECM architecture and plays a SA-ECM remodeling by senescent pulmonary
pivotal role in promoting tissue dysfunction. fibroblasts, we designed a senescent FDM
model. Pre-senescent and senescent lung
RESULTS fibroblasts were seeded to gelatin-coated
Primary lung fibroblasts acquire senescent coverslips and fed matrix promoting growth
phenotype media to develop FDMs of 20-30 μm in thickness
To generate a population of primary [Fig. 2(a)]. Scanning electron microscopy (SEM)
senescent pulmonary fibroblasts, γ-irradiation of decellularized pre-senescent and senescent
was used to induce a stress-induced premature FDMs were used to qualitatively observe the
senescent (SIPS) phenotype. A concurrent ultrastructure of the pre-senescent and
analysis of the senescent cell phenotype was senescent extracellular matrix [Fig. 2(b)].
conducted, examining cell-cycle arrest, Multiphoton SHG microscopy was used to
morphological changes, DNA damage and optically section the FDM architectures [Fig.
senescence-associated β-galactosidase (SA-β- 2(c)]. Using an in-house 2D-Fast-Fourier
Gal) was used to validate fibroblast senescence Transform (2D FFT) MATLAB code, individual
[34]. Irradiated fibroblasts increased in cellular SHG optical sections were overlaid with a grid
and nuclear area over time compared to non- consisting of 16x16 pixel regions (1048 total
irradiated conditions (pre-senescent) [Fig. 1(a), regions) to quantify local individual collagen fiber
1(b), and 1(c)]. Cell-cycle arrest was measured organizational changes within each 16x16 pixel
using bromodeoxyuridine (BrDU) whereby day 14 region [36-39]. Circular variance was used as a
post-irradiation less than 4% of irradiated metric to quantify collagen fiber disorganization
fibroblasts were positive for BrDU [Fig. 1(d)]. The and ranged from 0 to 1. A circular variance of 0
presence of the senescence marker SA-β-Gal on a polar coordinate system would result in a
was measured post-irradiation and reached 90% single line and parallel collagen fiber
positivity by day 10 [Fig. 1(e)]. Senescence and organization. On the contrary, a circular variance
myofibroblast differentiation have previously of 1 would result in a circle on a polar coordinate
been linked, showing that senescent cells system and indicate complete collagen fiber
heterogeneously express α-smooth muscle actin disorganization. The circular variance of
(αSMA) stress-fibers [35]. Therefore, individual collagen fiber orientations identified
fluorescence microscopy was used to measure within each region was used to classify the
the percentage of αSMA positive fibroblasts. The regions as isotropic (disorganized-fibers),
percentage of irradiated fibroblasts positive for anisotropic (organized-fibers) or dark (no signal).
αSMA was higher than pre-senescent conditions Global circular variance was used as a bulk
over 14 days (5-15% versus 0.4 – 1% metric that depicts the overall circular variance of
accordingly) [Fig. 1(f), see Supplemental Fig. collagen fiber orientations across the entire
1(a)]. SIPS also resulted in an increase in image and consisted of both anisotropic and
multinucleation over time compared to pre- isotropic regions. Anisotropic circular variance
senescent conditions (25% versus ≤ 1%) [Fig. was used as a localized collagen fiber orientation

3
metric that compared only anisotropic circular variance after day 7 [Fig. 3(e) and 3(f)].
(organized) regions with a preferred orientation to Consistent with previous reports, senescent
one another. It is important to note that two fibroblasts had significantly higher proteolytic
anisotropic regions identified can have different activity at day 10, which coincided with the
orientations, as exampled in the cartoon depiction decrease in the percent of collagenous regions
[Fig. 2(d)]. These metrics in culmination were and normalized SHG intensity [see
used to quantitatively measure heterogeneous Supplemental Fig. 2(a)] [13]. Like previous
organizational changes of SA-ECM remodeling observations, qRT-PCR analysis of senescent
over time. fibroblasts revealed a significant decrease in
COL1A1 expression and increased expression of
Timeline of SA-ECM fiber density and COL3A1, TGM2, and LOX over 21 days [14, 40]
orientation alterations [see Supplemental Fig. 2(b), 2(c), 2(d), and
Quantitative analysis of multiphoton SHG 2(e)]. Interpreting the collective data, senescent
microscopy images was used to measure SA- cells establish a heterogeneously disorganized
ECM remodeling over time. We assessed SA-ECM through multi-step remodeling; first
collagen remodeling of pre-senescent and degrading the ECM and then accumulating
senescent FDMs on days 7,10, 14, and 21 [Fig. collagen that could be over-modified and unable
3(a)]. Of the regions identified through the grid to be degraded due to excessive crosslinking.
analysis, pre-senescent FDMs had consistent
percentages of anisotropic regions throughout 21 Senescent fibroblasts have disorganized
days (63%-74%). However, senescent conditions actin cytoskeleton and upregulated integrin
decreased in the percentage of anisotropic profile
regions from 56% to 40% from days 7 to 10 and Senescent cell-matrix adhesion has
then increased to 66% by day 21 [Fig. 3(b)]. The previously been characterized as being
percentage of isotropic regions identified stayed significantly elevated compared to non-
relatively consistent for both pre-senescent and senescent cells; with an increased force profile,
senescent conditions over the course of 21 days, focal adhesion size, and focal adhesion kinase
only fluctuating 5% [Fig. 3(b)]. SHG intensity was activity [41, 42]. ITGB3 (integrin beta 3 or β3) was
quantified and normalized to cell number per also shown to regulate senescence through TGF-
region imaged, which served as a proxy for β signaling [43]. Observing that senescent
collagen deposition per cell [Fig. 3(c)]. Here, we fibroblasts increase collagen fiber disorganization
observed that pre-senescent conditions had globally and locally during SA-ECM remodeling,
minimal changes in SHG intensity. Nevertheless, we hypothesized that senescent fibroblasts are
senescent FDMs showed a parabolic pattern in pulling in a multilateral fashion to promote SA-
normalized SHG intensity by first decreasing and ECM disorganization. Using fluorescent imaging
then significantly increasing [Fig. 3(c)]. To of actin cytoskeleton, we found that senescent
examine regional matrix density changes, the fibroblasts have higher actin cytoskeleton
percentage of regions containing matrix disorganization compared to pre-senescent
(anisotropic and isotropic regions) was quantified conditions [Fig. 4(a) and 4(b)]. Next, the
as regional collagen signal. We measured a abundance of CD29 (ITGB1), CD61 (ITGB3),
similar parabolic trend in the percent of regions CD51 (ITGAV), and CD49b (ITGA2) were
having collagen for senescent matrices. Between measured using flow cytometry. CD49b and
days 7 and 10, the percent of regions with CD61 were upregulated compared to pre-
collagen signal decreased from 79% to 68% for senescent conditions, however, were not
the senescent conditions (P-value < 0.0001). significantly different (P-value = 0.0636 and
Interestingly, from days 10 to 14 we saw a 0.1893 accordingly). CD51 was significantly
recovery in the percent of regions containing upregulated compared to that of pre-senescent
collagen signal to 78% (P-value < 0.0001) and a conditions and could be influencing matrix
further significant increase to similar levels as organization through increased cell-matrix
pre-senescent FDMs by day 21 (87% versus 91% adhesion (P-value = 0.0149) [Fig. 4(c)].
accordingly, P-value = 0.0930) [Fig. 3(d)]. Even Upregulation of TGM2 matrix crosslinking of the
though senescent matrices recovered to similar lung has also been associated with pulmonary
levels of collagenous regions as pre-senescent fibrosis and linked to increased fibroblast
conditions, circular variance analysis of collagen spreading and adhesion [40, 44]. We Propose
fibers revealed significant increases in matrix that SA-ECM disorganization is in part due to the
disorganization based on global and anisotropic increased senescent cell-matrix adhesion and

4
multilateral contraction of the matrix due to senescent cells were used to grow FDMs. FDMs
disorganized actin cytoskeleton [Fig. 4(d)]. were treated every 48 hours and isolated on day
14 to quantitatively measure SA-ECM remodeling
Inhibition of TGFβR1 and MRTF-A induces using SHG microscopy [Fig. 6(a)]. Using this
divergent senescent pulmonary fibroblast inhibition treatment plan to grow the matrices, we
phenotypes did not observe a difference in the average
To further probe SA-ECM remodeling, we number of cells per image for the inhibited
separately inhibited TGFβR1 and MRTF-A during conditions [see Supplemental Fig. 4(a)].
senescence induction. We then characterized the Normalized SHG intensity for SB505124 and
senescent phenotypes over 14 days post- CCG-1423 treated conditions were not
irradiation. TGF-β, being one of the most pro- significantly different, however, TGFβR1
fibrotic stimulators of ECM production and inhibition did significantly reduce normalized
fibroblast activation, has been demonstrated to SHG intensity compared to DMSO negative
mediate paracrine senescence [22, 45-47]. control and untreated senescent conditions. It
MRTF-A inhibition was chosen due to its role in was also not significantly different from that of
the mechanosensitive activation of pro-fibrotic pre-senescent conditions [Fig. 6(b)]. Similarly,
gene expression and downstream location of CCG-1423 and SB505124 regional collagen
TGF-β signaling processes [30, 32, 48]. Using a signals were not significantly different; however,
selective inhibitor of TGFβR1 (SB505124 – 5 μM) SB505124 did significantly reduce regional
and MRTF-A inhibitor (CCG-1423 – 5 μM) that collagen signal compared to pre-senescent and
inhibits the nuclear localization of MRTF-A, we DMSO conditions [Fig. 6(c)]. Neither inhibitor
pre-treated pre-senescent fibroblasts 24 hours treatments improved global circular variance of
prior to irradiation and continuously treated every collagen fibers [Fig. 6(d)]. Interestingly, inhibition
48 hours post-irradiation [49-51] [Fig. 5(a) and of TGFβR1 did improve local collagen fiber
see Supplemental Fig. 3(a)]. Senescent cell organization based on the significantly decrease
nuclear and cellular morphologies were in anisotropic circular variance compared to
characterized at 14 days post-irradiation using CCG-1423, DMSO and senescent conditions.
fluorescence microscopy [Fig. 5(b)]. Inhibition of Therefore, TGFβR1 inhibition decreased the
TGFβR1 resulted in a decrease in nuclear and heterogeneity in collagen fiber organization back
cellular area, whereas inhibition of MRTF-A to pre-senescent condition levels [Fig. 6(e)].
resulted in an increase in nuclear and cellular
area [Fig. 5(c) and 5(d)]. Interestingly, MRTF-A Human pulmonary interstitial fibrosis tissue
inhibition increased the percent of cells with samples increase in collagen fiber
multinucleation compared to negative control disorganization with age and p16 expression
DMSO (19.2% vs 26.9%) and TGFβR1 inhibition Having quantified that senescent lung
decreased the percent of multinucleated cells fibroblast SA-ECM remodeling establishes a
(5%) [Fig. 5(e)]. The percentage of cells disordered collagen architecture and prior studies
expressing the DNA double strand break marker indicating that senescent cells play an intricate
γH2A.X, was not significantly different across role in the development of pulmonary fibrosis, we
treatment conditions [Fig. 5(f)]. TGFβR1 and examined the clinical relation between senescent
MRTF-A both influence fibroblast activation to a cells and collagen fiber disorganization in human
myofibroblast state, therefore their inhibition pulmonary fibrosis patient samples. Human
reduced the percent of senescent fibroblasts interstitial pulmonary fibrosis patient tissue
expressing αSMA [Fig. 5(g)]. qRT-PCR analysis microarrays stained with Masson’s Trichrome
revealed TGFβR1 did result in lower expression ranged in age from 37-72 years (n = 2 samples
of SASP factors such as IL1A, IL1B, and IL6; per patient, N = 23 patients) [Fig. 7(a)]. Using
whereby inhibiting MRTF-A increased the SASP color deconvolution to extract stained collagen
factor expressions compared to TGFβR1 fiber regions, the same regional collagen fiber
inhibition [see Supplemental Fig. 3(b), 3(c), analysis used to analyze individual fiber
3(d)]. organization for the senescent FDMs was applied
to the stained patient tissue [see Supplemental
TGFβR1 inhibition reduces SA-ECM Fig. 5(a)]. Patient samples were binned by age
remodeling heterogeneity according to the Freedman-Diaconis rule to
To observe functional differences in examine if there were any changes in the percent
TGFβR1 and MRTF-A inhibited senescent of regions with no collagen (dark), anisotropic
fibroblast SA-ECM remodeling, inhibited (organized), and isotropic (disorganized)

5
collagen fibers. Patient samples did not vary activity increased [14]. We believe that healthy
significantly in the proportion of regions identified soluble collagen is initially degraded through
based on age [see Supplemental Fig. 5(b)]. A SASP, and an increase in matrix density during
simple linear regression was used to examine if the second phase of SA-ECM remodeling is due
there were linear trends between global circular to the accumulation of overly modified insoluble
variance and anisotropic circular variance versus matrix; rather than an increase in matrix
age. No trend was found between age and global deposition by senescent fibroblasts.
circular variance (P-value = 0.0874, r2 = 0.07690) Although at an early time-point
[Fig. 7(b)]. However, there was a significant trend senescent and pre-senescent matrix
observed between anisotropic circular variance architectures were not significantly different, the
and age, indicating a linear relationship between resulting stabilized senescent matrix was
age and local collagen fiber disorder (P-value = significantly disorganized compared to pre-
0.0311, r2 = 0.1110) [Fig. 7(c)]. Having observed senescent lung fibroblast ECM architectures. It
a trend between age and local collagen fiber has been shown that the senescent cell
disorganization, a second tissue microarray phenotype is hyper-adhesive; with increased
consisting of the same human interstitial mechanical tension that can regulate gene
pulmonary fibrosis patients was stained for the expression through integrin binding [41-43].
senescence marker p16 (n = 2 samples per However, it was previously unclear whether
patient, N = 23 patients) [Fig. 7(d)]. We identified senescent fibroblasts biophysically influenced the
an age-dependent increase in the percentage of surrounding local matrix organization during
p16-positive cells within the tissue (P-value = senescent matrix development [52]. Our study
0.0247, R2 = 0.1119) [Fig. 7(e)]. Combining the demonstrated that senescent fibroblasts do in
p16-positive staining and Masson’s Trichrome fact promote matrix disorganization and we
analysis per patient, we then used a simple linear propose that fibroblast SA-ECM disorder is in
regression to investigate the relationship part, the result of multilateral tension applied to
between the percentage of p16-positive cells and the matrix through disordered actin cytoskeleton
matrix organization parameters. Herein, we did and heightened matrix adhesion. Having
not observe a significant relationship between observed a biophysical difference between
global circular variance and the percentage of senescent and pre-senescent fibroblasts, we
p16-positive cells identified (P-value = 0.065, R2 sought to inhibit the biomechanical contributions
= 0.08063) [Fig. 7(f)]. Nonetheless, we did of SA-ECM remodeling through MRTF-A during
measure a significant positive correlation senescent induction. Profibrotic cytokine TGF-β
(Pearson r = 0.4705, P-value = 0.0235) when was also previously shown to mediate paracrine
comparing the percentage of p16-positive cells senescence and SASP, therefore, we also
and anisotropic circular variance. This resulted in inhibited TGF-β signaling through TGFβR1
a significant linear trend based on a simple linear during senescence induction [22]. Doing so, we
regression, demonstrating that as the percentage generated divergent senescent fibroblast
of p16-positive cells increased, local collagen phenotypes to grow senescent lung FDMs. We
fiber disorganization also did (P-value = 0.0029, found that inhibition of TGFβR1 improved SA-
R2 = 0.1966) [Fig. 7(g)]. ECM matrix organization.
Cellular senescence has previously been
CONCLUSIONS shown to mediate pulmonary fibrosis
In this study, we comprehensively development and progression [8, 18, 27]. Having
mapped the spatiotemporal dynamics of SA-ECM established that SA-ECM remodeling varies
remodeling by designing an in vitro senescent temporally to establish a stable disorganized
lung fibroblast-derived matrix (FDM) model to senescent architecture, we hypothesized that
further understand the previously ambiguous role ECM architecture of patients diagnosed with
of SA-ECM remodeling in age-related disease interstitial pulmonary fibrosis would vary
development. Applying multiphoton SHG depending on age. Using Masson’s Trichrome
microscopy, we observed a temporally staining and color deconvolution, the same ECM
heterogeneous biphasic SA-ECM remodeling regional analysis approach applied to the
process that first underwent matrix degradation senescent lung FDM model was utilized to
followed by an increase in matrix density. Like characterize patient sample matrix architecture.
prior findings, we found that ECM-related gene Doing so, we identified that as patient age
expression decreased over time, while MMP increases, local matrix disorganization does also.
Having observed this age and matrix

6
disorganization relationship, we stained tissue Irradiation
samples from the same patients for the LF1s were cultured to 80% confluence
senescence marker p16. Furthermore, increased and exposed to 15 Gy γ-irradiation using a Mark
percentages of p16-positive cells were positively I 68A137CS Irradiator at a rate of 5.04 Gy/min for
correlated with an increase in local collagen fiber 3 minutes. For the generation of FDMs, irradiated
disorder. These findings suggest that local loss in cells were immediately transferred to gelatin-
structural organization is related to age and the coated coverslips. This mode of senescence
senescence marker p16. Moreover, this loss in induction was selected to generate large
structure during disease development could populations of senescent lung fibroblasts for the
further contribute to the decline in tissue function. direct comparison of pre-senescent and
Likewise, the age-dependent decline in lung senescent conditions.
function has been shown to be directly related to
pulmonary ECM remodeling and the increase in Fibroblast-Derived Matrices
cellular senescence within tissue [53, 54]. Fibroblast-derived matrices were
Through the design, development, and prepared using a modified version of the following
spatiotemporal monitoring of in vitro senescent protocol [58]. Within a 12-well plate (Corning), 15
lung FDMs, our study further provided insights mm diameter, 1.5-μm-thick glass coverslips
into the understudied topic of SA-ECM (Electron Microscopy Sciences) were coated with
remodeling. The senescent fibroblast phenotype sterile 0.2% gelatin (Fisher Science) and
undergoes a parabolic-like remodeling process incubated at 37oC for 1 hour before fixation with
that establishes a heterogeneously disorganized 1% glutaraldehyde (Electron Microscopy
SA-ECM architecture. We also demonstrated that Sciences) for 30 minutes. Post-fixation,
local SA-ECM collagen fiber disorder can be coverslips were washed using phosphate
therapeutically targeted and improved. The buffered saline (PBS) (Corning). 1M
results of this study offer further untapped ethanolamine (Acros Organics) was added to
potential for the utilization of senescent fibroblast each coverslip for 30 minutes and subsequently
matrix as an aged biomaterial for the washed using PBS. Senescent and pre-
development of 3D age-related disease models senescent LF1 cells were then seeded at 80%
in the future. By collecting and reconstituting confluence to each prepared coverslip. LF1 cells
decellularized SA-ECM into a 3D hydrogel, 3D were maintained using a matrix media consisting
bioprinting could be used to design aging ECM of DMEM-F12, 10% fetal bovine serum (FBS), 15
architectures. Doing so would allow for further mM HEPES, 1% P/S, 50 μg/mL L-ascorbic acid
pioneering on the influence of senescent cell salt (Acros Organics), and 60 μg/mL L-Proline
matrix on age-related disease, senescent cell (Alfa Aesar). Media was replaced every 48 hours
biology, cancer, and therapeutic developments throughout experimentation.
[55]. This study unveils new insights into
senescent cell ECM remodeling by spatially and Scanning Electron Microscopy
temporally quantifying SA-ECM dynamics. In Pre-senescent and senescent fibroblast-
doing so, we demonstrate that senescent cells in derived matrices were cultured for 14 days.
vitro and ex-vivo play an intricate role in the Matrices were decellularized using 20 mM
disorganization of the matrix architecture. ammonium hydroxide and 0.5% Triton-X in DI
water for 2 hours. Matrices were then washed
METHODS with PBS 3 times for 30 minutes and a final wash
in PBS overnight. Matrices were fixed in
Cell Culture Karnovsky’s fixative (Electron Microscopy
The primary human lung fibroblast cell Sciences, #15720) overnight. Fixed matrices
line, LF1 (a gift from Dr. John Sedivy, Brown were then washed 5 times for 3 minutes in cold
University, Providence, RI) was previously 0.15 M sodium cacodylate buffer (Electron
described [56, 57]. Fibroblasts were cultured in Microscopy Sciences, #11654). Matrices were
Dulbecco’s Modified Eagle’s Medium F-12 sequentially dehydrated in 25%, 50%, 75%, and
(DMEM-F12) (Corning) supplemented with 10% 100% ethanol (Fresh) for 15 minutes each.
fetal bovine serum (FBS) (Atlanta Biologicals), 15 Samples were critical point dried with liquid CO2
mM HEPES (Fisher Bioreagents), 1% L- and sputter coated with 100 Å gold palladium and
Glutamine (Cytivia), and 1% penicillin- visualized with a Thermo Fisher Scientific Apreo
streptomycin (P/S) (Corning). All cells were VS scanning electron microscope.
maintained at 37oC, 5% CO2 and 2% oxygen.

7
Multiphoton Second-Harmonic Generation 3M HCL and blocked with 5% horse serum in
and Single Photon Imaging PBS. Cells were then incubated in biotinylated
To generate and detect second-harmonic anti-BrdU primary antibody (1:100, Biolegend,
signal, an Olympus FV-10000MPE multiphoton No. 339810) followed by secondary antibody
microscope (Olympus, Tokyo, Japan) fitted with a DyLightTM 488 Streptavidin (1:200, Biolegend,
25x water-immersion objective (NA 1.05, WD cat. No.405218) and propidium iodide (PI).
2mm) and second-harmonics bandpass filter Coverslip were mounted to glass slides using
(405/40) was used. The excitation wavelength Fluoromount-G (Southern Biotech, cat. No. 0100-
was set to 800 nm using a Mai Tai HP Tunable 01) and imaged using a Nikon Eclipse TI inverted
Laser. Samples were fixed with 4% microscope at 10x magnification. LF1
formaldehyde, permeabilized using 0.5% Triton proliferation was quantified based on the number
X-100, and stained with RedDot2TM far-red of BrdU-positive cells normalized to the total
nuclear stain (1:200, Biotium, #40061). Single- number of cells using ImageJ.
photon fluorescence imaging was done using the
same setup listed prior. Images were exported as Immunofluorescence Staining
OIB (Olympus Image Binary) and converted to Within a 24-well plate (Corning), cells
tiffs using ImageJ. Quantitative analysis was seeded on 12 mm-diameter, #1.5 glass coverslips
done through an in-house MATLAB code were fixed using 4% paraformaldehyde.
previously published on (23, 24, 26). Following this, cells were permeabilized using
0.5% Triton X-100. Next, coverslips were
2D Fourier Transform Second-Harmonic incubated overnight in anti-αSMA (1:200, Cell
Generation Microscopy Image Analysis Signaling, #19245) and anti-γH2A.X (1:100, Cell
Using an in-house MATLAB code Signaling, #80312) primary antibodies. Goat anti-
developed in the Toussaint Lab previously rabbit Alexa Fluor 488 secondary antibody
described in literature, 2D Fast-Fourier Transform (1:1000 dilution, Invitrogen, A32731) and goat
(2D-FFT) analysis was combined with SHG anti-mouse Alexa Fluor 647 secondary antibody
microscopy to quantify spatial collagen fiber (1:500, Southern Biotech, cat. No.1030-31) were
organization of lung fibroblast-derived matrices used in conjunction with Alexa Fluor 555
(24, 25). A 16-pixel x 16-pixel grid array was then Phalloidin (1:400, Thermo Fisher Scientific,
overlaid on SHG images and within each 16-pixel A34055), and Hoechst 33342 (1:2000, Thermo
x 16-pixel grid region the individual collagen fiber Fisher Scientific, #62249) for 1 hour. Coverslips
orientations were calculated and averaged. were washed with PBS then mounted to glass
These regions were classified as anisotropic slides using Fluoromount-G (Southern Biotech,
(organized; collagen fibers have a preferred cat. No. 0100-01) and imaged using a Nikon
orientation), isotropic (disorganized, collagen Eclipse TI inverted at 10x and 40x magnification.
fibers do not have a preferred orientation), and Nuclear areas and nuclear shape factors were
dark (no signal) based on circular variance. In processed through a custom pipeline in
each region of the image, the orientation of the CellProfiler (Broad Institute). Images were
collagen fibers was calculated based on the thresholded and segmented using the
gradient of pixel intensities. The orientation is IdentifyPrimaryObjects module. Cellular area and
then represented as a normal resultant vector. cell form factor measurements were manually
The normalized resultant vectors for each region done through tracing with ImageJ. Percent
were then averaged to get a mean resultant positivity of α-SMA and γH2A.X was quantified
vector (R). Circular variance (CV) is then based on the number of positive cells normalized
calculated as one minus the length of the mean to the total number of cells.
resultant vector.
CV = 1 − R Real-Time PCR
All reagents were purchased from Bio-
Proliferation Rad for the RNA isolation, cDNA synthesis, and
1 mg/mL Bromodeoxyuridine/5-bromo- qRT-PCR. Total mRNA was extracted from the
2’-deoxyuridine (BrdU) (Millipore Sigma) in cell senescent and pre-senescent fibroblasts using
culture media was used to treat LF1 cells seeded RibozolTM RNA extraction reagent (VWR Life
on 12-mm-diameter coverslips. After a 24-hour Science, Cat. No. N580-100ML) according to the
incubation, cells were fixed using 4% manufacturer’s instructions. RNA purity and
formaldehyde and permeabilized using 0.5% integrity was measured using a Nanodrop 1000
Triton X-100 in PBS. Cells were then treated with and iScript gDNA clear cDNA synthesis kit (Bio-

8
Rad, Cat. #1725035) was used for cDNA [37, 38]. The MATLAB code overlaid a grid
synthesis. 1 μg of RNA was used to generate system, consisting of 16-pixel x 16-pixel regions.
cDNA and conduct qRT-PCR using a CFX In each region of the image, the orientation of
Connect Real-Time System (Bio-Rad) using the actin fibers was calculated based on the gradient
following primers. of pixel intensities. This orientation was then
P16, (Sense) 5’-CACTCACGCCCTAAGC-3’ represented as a normal resultant vector. The
and (antisense) 5’- normalized resultant vectors for each region were
GCAGTGTGACTCAAGAGAA-3’; P21, (sense) then averaged to get a mean resultant vector (R).
5’-CGATGGAACTTCGACTTTGTCA-3’ and Circular variance (CV) of the actin fibers was then
(antisense) 5’-GCACAAGGGTACAAGACAGTG- calculated as one minus the length of the mean
3’; COL1A1, (sense) 5’- resultant vector. There were no changes to the
ATCAACCGGAGGAATTTCCGT-3’ and MATLAB analysis pipeline when analyzing the
(antisense) 5’-CACCAGGACGACCAGGTTTTC- FDMs or the cell actin alignment.
3’; COL3A1, (sense) 5’-
GGAGCTGGCTACTTCTCGC-3’ and (antisense) FDM DQ-Gelatin Proteolytic Activity
5’-GGGAACATCCTCCTTCAACAG-3’; TGM2, Pre-senescent and senescent FDMs
(sense) 5’-GAGGAGCTGGTCTTAGAGAGG-3’ were treated with 100 µg/mL of DQ-Gelatin in
and (antisense) 5’ media and incubated at 37oC for 2 hours.
CGGTCACGACACTGAAGGTG3’; LOX, (sense) Samples were fixed with 4% formaldehyde,
5’-CGGCGGAGGAAAACTGTCT-3’ and permeabilized with 0.5% Triton X-100, and
(antisense) 5’-TCGGCTGGGTAAGAAATCTGA- stained with RedDot2TM far-red nuclear stain
3’; IL1A, (sense) 5’- (1:200, Biotium, #40061). Images were exported
TGGTAGTAGCAACCAACGGGA-3’ and as OIB (Olympus Image Binary) and converted to
(antisense) 5’- tiffs using ImageJ. DQ-Gelatin fluorescence
ACTTTGATTGAGGGCGTCATTC-3’; IL1B, intensity was quantified using ImageJ and
(sense) 5’-TGCACGCTCCGGGACTCACA-3’ normalized using the cell number per image.
and (antisense) 5’-
CATGGAGAACACCACTTGTTGCTCC-3’; IL6, MTT Assay
(sense) 5’-ACTCACCTCTTCAGAACGAATTG-3’ Senescent pulmonary fibroblasts were
and (antisense) 5’- seeded into 96-well plates and treated with either
CCATCTTTGGAAGGTTCAGGTTG-3’; β-Actin, culture media, DMSO, 1, 5, 10 μM SB505124
(sense) 5’-AGAGCTACGAGCTGCCTGAC-3’ (Cayman Chemical, No. 11793) or CCG-1423
and (antisense) 5’- (Cayman Chemical, No. 10010350) every 48
AGCACTGTGTTGGCGTACAG. hours for 14 days. Samples were incubated in
MTT solution for 3 hours at 37oC before MTT
β-actin was used as the housekeeping gene and solvent was added. Samples were incubated in
the pre-senescent condition for each timepoint MTT solvent for 30 minutes at 37oC prior to
was used as the control group for normalization. absorbance reading at 590 nm with a reference
filter of 620 nm. Sample MTT absorbances were
Confocal Microscopy normalized to the culture media condition for
Pre-senescent and senescent LF1s were comparisons.
cultured for 14 days and fixed in 2%
paraformaldehyde. Cells were permeabilized Flow Cytometry
using 0.5% Triton X-100 and stained with Pre-senescent and senescent LF1s were
Hoechst 33342 (1:2000, Thermo Fisher cultured for 14 days. Cells were detached,
Scientific, #62249) and Alexa Fluor 555 pelleted, and suspended in FACS buffer (PBS,
Phalloidin (1:400, Thermo Fisher Scientific, 2% FBS, 1mM EDTA), and fixed in 2% PFA. Cells
A34055). Spinning disk confocal microscopy was were stained for the surface markers anti-human
used to image actin cytoskeleton alignment CD29 (Biolegend #303003), anti-human CD49b
(Olympus IX83 inverted microscope) (Silicone oil (Biolegend #359309), anti-human CD51
immersion obj. = 60x, N.A. = 1.3). Confocal (Biolegend #327907), or anti-human CD61
microscope images were uploaded to ImageJ (Biolengend #336405). 500,000 cells per
(FIJI) and channels were separated. Actin fiber condition were analyzed. Samples were run on a
circular variance was extracted from tiff images Cytek Aurora flow cytometer.
containing individual cells using an in-house 2D
fast Fourier Transform (2D FFT) MATLAB code

9
Tissue Microarray Histology Analysis ACKNOWLEDGEMENTS
Unstained human pulmonary interstitial This work was funded by the National
fibrosis tissue microarrays were purchased from Science Foundation Advanced Manufacturing
US Biomax, Inc. (LC561). The unstained tissue Program Award (2043243). Authors would like to
microarrays were stained with Masson’s thank Brown University Environmental Health
Trichrome and for p16 by trained pathologist Dr. and Safety, specifically Richard Shea and Mark
Yihong Wang. Images were taken using an Dirksen, for access to the irradiator. Authors
Olympus VS200 Slide Scanner (Obj. 40x, N.A. = would also like to thank Geoffry Williams and the
0.95). Masson’s Trichrome images were color Leduc Bioimaging Core for the use of the Thermo
deconvoluted using ImageJ to extract only the Apreo VS SEM, which was purchased with a
ECM component of the sample. Using the same high-end instrumentation grant from the Office of
in-house 2D FFT MATLAB pipeline previously the Director at National Institutes of Health
used, regional collagen fiber analysis was (S10OD023461).
conducted. Tissue microarrays stained for p16
were pre-processed prior to analysis using AUTHOR DECLARATIONS
ImageJ. Hue and saturation thresholding were Conflict of Interest
consistently applied across all images to remove The authors have no conflicts to disclose
carbon-laden regions. Next, a brightness
threshold was set using ImageJ to enable the Ethics of Approval
specific identification of the carbon-laden cells Ethics approval not required
[see Supplemental Fig. 5(c)]. A white mask was
superimposed over the carbon cell regions Author Contributions
leaving the p16 stained cells unaltered [see Andrew M. Howes: Conceptualization
Supplemental Fig. 5(d)]. CellProfiler was then (lead); Data Curation (lead); Formal Analysis
used to analyze the ImageJ pre-processed p16 (lead); Investigation (lead) Methodology (lead);
stained tissue samples. A simple linear Software (lead); Visualization (lead); Validation
regression analysis was conducted to determine (lead); Writing - original draft preparation (lead);
data trends. For the analysis of the histology, no Writing - review & editing (lead). Nova C. Dea:
adjustments were made to the MATLAB circular Formal Analysis (supporting); Investigation
variance analysis pipeline. (supporting); Validation (supporting);
Visualization (supporting). Deepraj Ghosh:
Statistical Analysis Conceptualization (supporting); Formal Analysis
Prism software was used for statistical (supporting); Methodology (supporting);
analysis. Results are reported as mean ± SEM or Supervision (supporting). Krishangi Krishna:
in violin plots. For the comparison of multiple Formal Analysis (supporting); Validation
groups, one-way ANOVA was used with Tukey’s (supporting). Yihong Wang: Resources
method for statistical hypothesis testing. For (supporting); Investigation (supporting). Yanxi Li:
comparisons between only two groups, Student’s Formal Analysis (supporting); Investigation
t-test was utilized. To examine clinical patient (supporting); Software (supporting); Validation
data, a simple linear regression was used. (supporting); Visualization (supporting). Braxton
*P<0.05 was considered significant. For Morrison: Investigation (supporting); Formal
consistency, significance values were as follows: Analysis (supporting). Kimani C. Toussaint:
*P<0.05, **P<0.01, ***P<0.001, ****P<0.0001. Conceptualization (supporting); Funding
Acquisition (equal); Resources (supporting);
SUPPLEMENTARY MATERIAL Supervision (supporting); Writing – review &
See the supplementary material for Fig. editing (supporting). Michelle R. Dawson:
1: αSMA and γH2A.X characterization; Fig 3: Pre- Conceptualization (supporting); Funding
senescent and senescent lung fibroblast cell Acquisition (equal); Resources (lead); Writing –
count, proteolytic activity, and ECM remodeling review & editing (supporting).
gene regulation; Fig. 5: Inhibitor dose-dependent
lung fibroblast viability and interleukin gene DATA AVAILABILITY
regulation changes; Fig. 6: Inhibitor treated The data that support the findings of this
average senescent cell number per image; Fig. 7: study are available from the corresponding author
Masson’s Trichrome collagen fiber extraction upon reasonable request.
analysis and p16 sample carbon-laden regions.

10
REFERENCES and modifiers. FEBS J, 2020.
287(13): p. 2636-2646.
1. Borghesan, M., et al., A Senescence- 12. Freitas-Rodriguez, S., A.R.
Centric View of Aging: Implications Folgueras, and C. Lopez-Otin, The
for Longevity and Disease. Trends role of matrix metalloproteinases in
Cell Biol, 2020. 30(10): p. 777-791. aging: Tissue remodeling and
2. Di Micco, R., et al., Cellular beyond. Biochim Biophys Acta Mol
senescence in ageing: from Cell Res, 2017. 1864(11 Pt A): p.
mechanisms to therapeutic 2015-2025.
opportunities. Nat Rev Mol Cell Biol, 13. Mavrogonatou, E., et al.,
2021. 22(2): p. 75-95. Senescence-associated alterations
3. Baker, D.J., et al., Naturally occurring in the extracellular matrix:
p16(Ink4a)-positive cells shorten deciphering their role in the
healthy lifespan. Nature, 2016. regulation of cellular function. Am J
530(7589): p. 184-9. Physiol Cell Physiol, 2023. 325(3): p.
4. Coppe, J.P., et al., The senescence- C633-C647.
associated secretory phenotype: the 14. Mavrogonatou, E., et al.,
dark side of tumor suppression. Annu Extracellular matrix alterations in
Rev Pathol, 2010. 5(1): p. 99-118. senescent cells and their significance
5. Bussian, T.J., et al., Clearance of in tissue homeostasis. Matrix Biol,
senescent glial cells prevents tau- 2019. 75-76: p. 27-42.
dependent pathology and cognitive 15. Wilson, M.S. and T.A. Wynn,
decline. Nature, 2018. 562(7728): p. Pulmonary fibrosis: pathogenesis,
578-582. etiology and regulation. Mucosal
6. Merkt, W., et al., Senotherapeutics: Immunol, 2009. 2(2): p. 103-21.
Targeting senescence in idiopathic 16. Lasry, A. and Y. Ben-Neriah,
pulmonary fibrosis. Semin Cell Dev Senescence-associated
Biol, 2020. 101: p. 104-110. inflammatory responses: aging and
7. Jonathan Vespa, L.M., and David M. cancer perspectives. Trends
Armstrong, Demographic Turning Immunol, 2015. 36(4): p. 217-28.
Points for the United States: 17. Munoz-Espin, D. and M. Serrano,
Population Projections for 2020 to Cellular senescence: from physiology
2060, in Current Population Reports. to pathology. Nat Rev Mol Cell Biol,
2020, U.S. Census Bureau. p. 25- 2014. 15(7): p. 482-96.
1144. 18. Selman, M. and A. Pardo, Revealing
8. Schafer, M.J., et al., Cellular the pathogenic and aging-related
senescence mediates fibrotic mechanisms of the enigmatic
pulmonary disease. Nat Commun, idiopathic pulmonary fibrosis. an
2017. 8(1): p. 14532. integral model. Am J Respir Crit Care
9. Alvarez, D., et al., IPF lung fibroblasts Med, 2014. 189(10): p. 1161-72.
have a senescent phenotype. Am J 19. Yanai, H., et al., Cellular senescence-
Physiol Lung Cell Mol Physiol, 2017. like features of lung fibroblasts
313(6): p. L1164-L1173. derived from idiopathic pulmonary
10. Campisi, J. and F. d'Adda di fibrosis patients. Aging (Albany NY),
Fagagna, Cellular senescence: when 2015. 7(9): p. 664-72.
bad things happen to good cells. Nat 20. Burton, D.G.A. and A. Stolzing,
Rev Mol Cell Biol, 2007. 8(9): p. 729- Cellular senescence:
40. Immunosurveillance and future
11. Levi, N., et al., The ECM path of immunotherapy. Ageing Res Rev,
senescence in aging: components 2018. 43: p. 17-25.

11
21. Ye, Z. and Y. Hu, TGF‑beta1: in response to myocardial infarction.
Gentlemanly orchestrator in Circ Res, 2010. 107(2): p. 294-304.
idiopathic pulmonary fibrosis 32. Sisson, T.H., et al., Inhibition of
(Review). Int J Mol Med, 2021. 48(1). myocardin-related transcription
22. Acosta, J.C., et al., A complex factor/serum response factor
secretory program orchestrated by signaling decreases lung fibrosis and
the inflammasome controls paracrine promotes mesenchymal cell
senescence. Nat Cell Biol, 2013. apoptosis. Am J Pathol, 2015. 185(4):
15(8): p. 978-90. p. 969-86.
23. Fernandez, I.E. and O. Eickelberg, 33. Hermann, M.R., et al., Integrins
The impact of TGF-beta on lung synergise to induce expression of the
fibrosis: from targeting to biomarkers. MRTF-A-SRF target gene ISG15 for
Proc Am Thorac Soc, 2012. 9(3): p. promoting cancer cell invasion. J Cell
111-6. Sci, 2016. 129(7): p. 1391-403.
24. Martinez, F.J., et al., Idiopathic 34. Gonzalez-Gualda, E., et al., A guide
pulmonary fibrosis. Nature Reviews to assessing cellular senescence in
Disease Primers, 2017. 3(1): p. vitro and in vivo. FEBS J, 2021.
17074. 288(1): p. 56-80.
25. Matsuda, S., et al., TGF-beta in the 35. Mellone, M., et al., Induction of
microenvironment induces a fibroblast senescence generates a
physiologically occurring immune- non-fibrogenic myofibroblast
suppressive senescent state. Cell phenotype that differentially impacts
Rep, 2023. 42(3): p. 112129. on cancer prognosis. Aging (Albany
26. Capron, C., et al., A major role of NY), 2016. 9(1): p. 114-132.
TGF-beta1 in the homing capacities 36. Ambekar, R., et al., Quantifying
of murine hematopoietic stem collagen structure in breast biopsies
cell/progenitors. Blood, 2010. 116(8): using second-harmonic generation
p. 1244-53. imaging. Biomed Opt Express, 2012.
27. Calhoun, C., et al., Senescent Cells 3(9): p. 2021-35.
Contribute to the Physiological 37. Sivaguru, M., et al., Quantitative
Remodeling of Aged Lungs. J analysis of collagen fiber
Gerontol A Biol Sci Med Sci, 2016. organization in injured tendons using
71(2): p. 153-60. Fourier transform-second harmonic
28. Choi, H.R., et al., Restoration of generation imaging. Optics Express,
senescent human diploid fibroblasts 2010. 18(24): p. 24983-24993.
by modulation of the extracellular 38. Rao, R.A., M.R. Mehta, and K.C.
matrix. Aging Cell, 2011. 10(1): p. Toussaint, Jr., Fourier transform-
148-57. second-harmonic generation imaging
29. Gau, D. and P. Roy, SRF'ing and of biological tissues. Opt Express,
SAP'ing - the role of MRTF proteins 2009. 17(17): p. 14534-42.
in cell migration. J Cell Sci, 2018. 39. Lau, T.Y., et al., Application of Fourier
131(19). transform-second-harmonic
30. Velasquez, L.S., et al., Activation of generation imaging to the rat cervix.
MRTF-A-dependent gene expression J Microsc, 2013. 251(1): p. 77-83.
with a small molecule promotes 40. Stephens, P., et al., Crosslinking and
myofibroblast differentiation and G-protein functions of
wound healing. Proc Natl Acad Sci U transglutaminase 2 contribute
S A, 2013. 110(42): p. 16850-5. differentially to fibroblast wound
31. Small, E.M., et al., Myocardin-related healing responses. J Cell Sci, 2004.
transcription factor-a controls 117(Pt 15): p. 3389-403.
myofibroblast activation and fibrosis

12
41. Brauer, E., et al., Dissecting the an inhibitor of Rho signaling. PLoS
influence of cellular senescence on One, 2014. 9(2): p. e89016.
cell mechanics and extracellular 51. DaCosta Byfield, S., et al., SB-
matrix formation in vitro. Aging Cell, 505124 is a selective inhibitor of
2023. 22(3): p. e13744. transforming growth factor-beta type
42. Shin, E.Y., et al., Integrin-mediated I receptors ALK4, ALK5, and ALK7.
adhesions in regulation of cellular Mol Pharmacol, 2004. 65(3): p. 744-
senescence. Sci Adv, 2020. 6(19): p. 52.
eaay3909. 52. Blokland, K.E.C., et al., Regulation of
43. Rapisarda, V., et al., Integrin Beta 3 cellular senescence by extracellular
Regulates Cellular Senescence by matrix during chronic fibrotic
Activating the TGF-beta Pathway. diseases. Clin Sci (Lond), 2020.
Cell Rep, 2017. 18(10): p. 2480- 134(20): p. 2681-2706.
2493. 53. Brandenberger, C. and C. Muhlfeld,
44. Olsen, K.C., et al., Transglutaminase Mechanisms of lung aging. Cell
2 and its role in pulmonary fibrosis. Tissue Res, 2017. 367(3): p. 469-
Am J Respir Crit Care Med, 2011. 480.
184(6): p. 699-707. 54. Miller, M.R., Structural and
45. Broekelmann, T.J., et al., physiological age-associated
Transforming growth factor beta 1 is changes in aging lungs. Semin
present at sites of extracellular matrix Respir Crit Care Med, 2010. 31(5): p.
gene expression in human 521-7.
pulmonary fibrosis. Proc Natl Acad 55. Schnellmann, R. and S. Gerecht,
Sci U S A, 1991. 88(15): p. 6642-6. Reconstructing the ageing
46. Frangogiannis, N., Transforming extracellular matrix. Nature Reviews
growth factor-beta in tissue fibrosis. J Bioengineering, 2023. 1(7): p. 458-
Exp Med, 2020. 217(3): p. 459.
e20190103. 56. Wei, S., W. Wei, and J.M. Sedivy,
47. Tomasek, J.J., et al., Myofibroblasts Expression of Catalytically Active
and mechano-regulation of Telomerase Does Not Prevent
connective tissue remodelling. Nat Premature Senescence Caused by
Rev Mol Cell Biol, 2002. 3(5): p. 349- Overexpression of Oncogenic Ha-
63. Ras in Normal Human Fibroblasts.
48. Zhou, Y., et al., Inhibition of Cancer Research, 1999. 59(7): p.
mechanosensitive signaling in 1539-1543.
myofibroblasts ameliorates 57. Brown, J.P., W. Wei, and J.M. Sedivy,
experimental pulmonary fibrosis. J Bypass of senescence after
Clin Invest, 2013. 123(3): p. 1096- disruption of p21CIP1/WAF1 gene in
108. normal diploid human fibroblasts.
49. Jin, W., et al., Increased SRF Science, 1997. 277(5327): p. 831-4.
transcriptional activity in human and 58. Franco-Barraza, J., et al.,
mouse skeletal muscle is a signature Preparation of Extracellular Matrices
of insulin resistance. J Clin Invest, Produced by Cultured and Primary
2011. 121(3): p. 918-29. Fibroblasts. Curr Protoc Cell Biol,
50. Hayashi, K., et al., RPEL proteins are 2016. 71: p. 10 9 1-10 9 34.
the molecular targets for CCG-1423,

13
FIG. 1. Primary lung fibroblasts acquire senescent phenotype. (A) Representative confocal microscopy images of pre-senescent
and senescent lung fibroblasts stained with Hoechst 33342 and phalloidin rhodamine (Obj. = 60x, N.A. = 1.3, scale bar = 50 μm).
(B) Nuclear area of pre-senescent and senescent lung fibroblasts over time (n ≥ 162, N = 3). (C) Cell area changes were tracked
over time for pre-senescent and senescent lung fibroblasts (n ≥ 81, N = 3). (D) Bromodeoxyuridine (BrDU) labeling confirmed
growth arrest of irradiated fibroblasts post-irradiation (n ≥ 151, N = 3). (E) The percentage of SA-β-Gal expressing lung fibroblasts
for pre-senescent and senescent fibroblasts (n ≥ 238, N = 3). (F) The percent of pre-senescent and senescent fibroblasts positive
for αSMA stress-fibers (n ≥ 151, N = 3). (G) The percentage of multinucleated pre-senescent and senescent fibroblasts over time
(n ≥ 151, N = 3). (H) The percentage of pre-senescent and senescent nuclei positive for γH2A.X over time (n ≥ 151, N = 3). (I)
qRT-PCR analysis of p16 for pre-senescent and senescent fibroblasts over time (N =3). (J) qRT-PCR analysis of p21 for pre-
senescent and senescent fibroblasts over time (N = 3). Mean ± SEM; ANOVA, *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001, ****P ≤ 0.0001.

14
FIG. 2. Generation of senescent primary lung fibroblast-derived matrix model. (A) pre-senescent and senescent lung fibroblasts
were used to generate fibroblast-derived matrices (FDM) over time. (B) SEM images of decellularized pre-senescent and
senescent lung fibroblast FDMs (Scale bar = 5 μm). (C) Representative second harmonic generation (SHG) microscopy images
of pre-senescent and senescent FDMs at day 14. Examples of organized (anisotropic) regions (blue) and disorganized (isotropic)
(red) are denoted (Obj. = 25x, N.A. = 1.05, scale bar = 100 μm.) (D) Schematic depicting the SHG image regional analysis used
to measure changing collagen fiber organizations and heterogeneity of pre-senescent and senescent FDMs. 16-pixel x 16-pixel
regions generated a grid array over SHG images and within each region the individual collagen fiber angles were measured.
Regions were then labeled isotropic (disorganized; fibers do not have a preferred orientation), anisotropic (organized; fibers have
a preferred orientation), and dark (no signal) (Illustrations designed in BioRender).

15
FIG. 3. Timeline of SA-ECM fiber density and orientation alterations. (A) Representative multiphoton second harmonic generation
(SHG) images of pre-senescent and senescent fibroblast-derived matrices (FDM) imaged at days 7, 10, 14 and 21 (Obj. = 25x,
N.A. = 1.05, scale bar = 100 μm. (B) Distribution of anisotropic, isotropic, and dark regions classified over 21 days for pre-
senescent and senescent FDMs (n ≥ 36, N = 3, N = 2 for day 21 senescent). (C) SHG intensity normalized to cell number per
imaged region (n ≥ 8, N = 3, N = 2 for day 21 senescent). (D) Percent regional collagen signal of senescent and pre-senescent
FDM (n ≥ 36, N = 3, N = 2 for Day 21 Senescent). (E) Global circular variance of individual collagen fibers (n ≥ 36, N = 3, N = 2
for Day 21 Senescent). (F) Anisotropic circular variance for pre-senescent and senescent FDMs (n ≥ 36, N = 3, N = 2 for Day 21
Senescent). ANOVA, *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001, ****P ≤ 0.0001.

16
FIG. 4. Senescent fibroblasts have disorganized actin cytoskeleton and upregulated integrin profile. (A) Representative confocal
microscopy images of pre-senescent and senescent fibroblasts stained with Hoechst 33342 and phalloidin rhodamine (Obj. = 60x,
N.A. = 1.3, scale bar = 20 μm). (B) Circular variance of actin cytoskeleton fibers for pre-senescent and senescent lung fibroblasts
(n ≥ 25, N = 3). (C) Analysis of CD29, CD61, CD51 and CD49b integrin profile for pre-senescent and senescent lung fibroblasts
(n ≥ 500,000 cells per condition, N = 3). (D) Schematic of proposed method of SA-ECM matrix disorganization by senescent lung
fibroblasts. Mean ± SEM; T-Test, *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001, ****P ≤ 0.0001.

17
FIG. 5. Inhibition of TGFβR1 and MRTF-A induces divergent senescent lung phenotype. (A) 24 hours prior to stress-induced
premature senescence is induced by γ-irradiation (15 Gy), lung fibroblasts were treated with either selective TGβFR1 inhibitor
(SB505124) or MRTF-A inhibitor (CCG-1423) and cultured over 14 days (Illustration designed in BioRender). (B) Representative
immunofluorescent images stained with Hoechst 33342 (Blue), phalloidin rhodamine (Red), anti-SMA (green) (Obj. = 40x, N.A. =
1.3, Scale bar = 50 μm). (C) Inhibited senescent lung fibroblast nuclear area measured at day 14 post-irradiation (n ≥ 149, N =
3). (D) Inhibited senescent lung fibroblast cell area measured at day 14 post-irradiation (n ≥ 100, N = 3). (E) Percent of inhibited
senescent fibroblasts with multinucleation (n ≥ 140, N = 3). (F) Percent of inhibited senescent fibroblasts positive for γH2A.X (n ≥
100, N = 3). (G) Percent of inhibited senescent lung fibroblasts positive for αSMA stress-fibers (n ≥ 100, N = 3). Mean ± SEM;
ANOVA. *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001, ****P ≤ 0.0001.

18
FIG. 6. TGFβR1 inhibition reduces SA-ECM remodeling heterogeneity. (A) Representative second harmonic generation (SHG)
images of collagen fiber organization of senescent lung fibroblast-derived matrices (FDM) treated with DMSO, SB505124, and
CCG-1423 at day 14 post-irradiation (Obj. = 25x, N.A. = 1.05, scale bar = 100 μm). (B) Inhibited senescent lung FDM SHG intensity
normalized to cell number per image (n ≥ 20, N = 3). (C) Inhibited senescent lung FDM regional collagen signal (n ≥ 34, N = 3).
(D) Global circular variance of inhibited senescent lung FDM individual collagen fibers (n ≥ 34, N = 3). (E) Anisotropic circular
variance of inhibited senescent lung FDM individual collagen fibers (n ≥ 34, N = 3). ANOVA; *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001,
**** P ≤ 0.0001.

19
FIG. 7. Regional matrix analysis of human interstitial pulmonary fibrosis tissue samples. (A) Representative image of Masson’s
Trichrome stained interstitial (Obj. = 40x, N.A. = 0.95, large image scale bar = 200 μm, zoom scale bar = 50 μm). (B) Simple linear
regression of global circular variance versus age of pulmonary fibrosis patient (n = 2 samples per patient, N = 23 patients). (C)
Simple linear regression of anisotropic circular variance versus age of pulmonary fibrosis patient (n =2 samples per patient, N =
23 patients). (D) Representative image of p16-stained pulmonary fibrosis patient tissue sample (Obj. = 40x, N.A. = 0.95, Large
image scale bar = 200 μm, zoom scale bar = 20 μm) (E) Simple linear regression of percentage of p16-positive cells versus age
of pulmonary fibrosis patient (n = 2 samples per patient, N = 23 patients). (F) Simple linear regression of percentage of p16-
positive cells versus global circular variance of pulmonary fibrosis patient (n = 2 samples per patient, N = 23 patients). (G) Simple
linear regression of percentage of p16-positive cells versus anisotropic circular variance of pulmonary fibrosis patient (n = 2
samples per patient, N = 23 patients). Simple Linear Regression. *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001, ****P ≤ 0.0001.

20