Abstract
Scar tissue formation is a hallmark of wound repair in adults and can
chronically affect tissue architecture and function. To understand the
general phenomena, we sought to explore scar-driven imbalance in tissue
homeostasis caused by a common, and standardized surgical procedure,
the uterine scar due to cesarean surgery. Deep uterine scar is
associated with a rapidly increasing condition in pregnant women,
placenta accreta spectrum (PAS), characterized by aggressive
trophoblast invasion into the uterus, frequently necessitating
hysterectomy at parturition. We created a model of uterine scar,
recapitulating PAS-like invasive phenotype, showing that scar matrix
activates mechanosensitive ion channel, Piezo1, through
glycolysis-fueled cellular contraction. Piezo1 activation increases
intracellular calcium activity and Protein kinase C activation, leading
to NF-κB nuclear translocation, and MafG stabilization. This
inflammatory transformation of decidua leads to production of IL-8 and
G-CSF, chemotactically recruiting invading trophoblasts towards scar,
initiating PAS. Our study demonstrates aberrant mechanics of scar
disturbs stroma-epithelia homeostasis in placentation, with
implications in cancer dissemination.
Subject terms: Extracellular signalling molecules, Urogenital
reproductive disorders, Disease model, Molecular medicine
__________________________________________________________________
The uterine scar that results from cesarean surgery is associated with
placenta accreta spectrum (PAS) during subsequent pregnancies. Here,
the authors created a model of uterine scar and show that scar matrix
activates mechanosensitive ion channel, Piezo1, which causes
inflammatory transformation of the decidua, leading to recruitment of
extravillous trophoblasts towards the scar, thereby initiating PAS.
Introduction
Scar tissue formation as a result of surgery, injury, trauma, or
infarction is a hallmark of adult wound repair, resulting in profound
changes in the tissue microenvironment^[50]1,[51]2. Pathological
scarring by surgery or trauma occurs through progressive remodeling of
the granulation tissue, and is characterized by high type I collagen
content, decreased cellularity, high mechanical rigidity, and aberrant
matrix ultrastructure^[52]3,[53]4. Although mechanisms of scar
formation have been extensively studied^[54]5–[55]8, relatively less is
understood about the effect of existing scar or fibrosis on tissue
homeostasis.
To understand the mechanisms driving scar matrix-induced dysregulation
in tissue homeostasis, we sought to investigate a disease attributable
directly to a standard and commonly occurring pre-existent scar:
placenta accreta spectrum (PAS). PAS is a general definition for
pathologies characterized by abnormally invasive placentation (accreta,
increta, and percreta)^[56]9–[57]11, and is primarily considered to be
an outcome of pre-existent uterine scar from cesarean
surgery^[58]12–[59]15 (Fig. [60]1A). Uterine scars due to cesarean
sections have increased rapidly, accounting for 21% of all childbirth
worldwide^[61]16,[62]17. PAS has also increased in parallel, rising
from about 1 in 30,000 in 1950s to about 1 in 300 pregnancies in the US
alone^[63]18–[64]20. Therefore, uterine scar-induced PAS could be
considered as a model to investigate scar-induced pathology. PAS
results in serious maternal and fetal complications, including
life-threatening hemorrhage, necessitating hysterectomy. PAS may also
result in damage to adjacent organs, maternal mortality, and preterm
birth, resulting in exorbitant clinical, psychological, and economic
costs^[65]21,[66]22. Although the most important predictor for PAS is
previous uterine scar due to cesarean delivery, surgeries or curettage,
or other damage to the uterine wall, little is understood about the
mechanisms driving scar-induced PAS onset and progression.
Fig. 1. In vitro model of PAS recapitulates the abnormally deep trophoblastic
invasion.
[67]Fig. 1
[68]Open in a new tab
A Schematic showing abnormally deep invasion of extravillous
trophoblasts (EVTs) in the decidual stroma proximal to scar. SCT:
syncytiotrophoblast, CT: cytotrophoblast, dF: decidual fibroblast,
Scar-dF: transformed dFs proximal to scar, Physio-dF: normal dFs distal
to scar, V: placental villi, M: myometrium. B H&E and
immunohistochemistry images of maternal-fetal interface tissue sections
from PAS patients showing HLA-G^+ EVT (green), and Vimentin labeled dFs
(red); nuclei marked with DAPI (blue). n = 6 biological replicates. C
Picrosirius red staining of tissue sections from regions proximal, and
distal to pre-existent scar, with orientation distribution of collagen
fibers from different PAS patients quantified in (D). n = 3 biological
replicates. E Surface topography of Scar matrix imaged by atomic force
microscopy (AFM) in PBS; F Photo and spatial rigidity characterization
of normal endometrial tissue; Graph (bottom) shows mean rigidity of
endometrial tissue, Physio, and Scar matrices. n = 8, 25, and 8
biologically independent experiments. p = 3 × 10^−42. G Schematic
showing workflow to establish in vitro Scar induced PAS model with
distinct invasion assays. H, I Phase contrast image showing in-situ
HTR8 spheroid invasion into ESFs decidualized on Physio; Graphs showing
line integral convolution representation of HTR8 invasion into ESFs
decidualized on Physio, or Scar matrices. I Normalized HTR8 invasion
area (S/S[0]) as a function of invasion time. n = 10 biological
replicates. J Fluorescent images of HTR8 (red) invasion into ESFs
pre-decidualized on Physio, or Scar at time 0, and 24 h; Graph showing
aerial invasion normalized to initial interface length; n = 8
interfaces; p = 0.005. K Apotome scanning of HTR8 spatial nuclear
locations relative to dESF monolayers 72 h after invasion; L
Quantification of individual HTR8 distance to dESF monolayer; n = 255
cells; p = 6 × 10^−79. M Volcano plot showing differentially expressed
genes in dESFs on Scar and Physio. N Ingenuity Pathway Analysis based
prediction of activated transcription factors in dESFs on Scar and
Physio; n = 3 biological replicates. Data in all bar graphs are showing
as mean ± s.d.; statistical significance is determined by unpaired
two-tailed t-test (**p < 0.01, ****p < 0.0001, and ns not significant).
Source data are provided as a Source Data file. A, G created with
BioRender.com released under a Creative Commons
Attribution-NonCommercial-NoDerivs 4.0 International license
([69]https://creativecommons.org/licenses/by-nc-nd/4.0/deed.en).
Placentation in humans involve deep invasion of fetal extravillous
trophoblasts (EVTs) into the maternal endometrial stroma^[70]23. In
anticipation of placentation, the maternal endometrial stroma undergoes
a profound transformation, termed decidualization. We, and others have
shown that decidualization is, in part, a mechanism to acquire
resistance against EVT invasion^[71]24,[72]25. The decidua has to both
support the deep maternal invasion by EVTs, as well as prevent
excessive invasion potentially leading to PAS-like pathology.
Physiological placentation is a tightly regulated phenomenon which
requires careful balancing between the pro- and anti-invasive
mechanisms, resulting in a negotiated homeostasis for optimal
invasion^[73]26,[74]27. This negotiation is partly arrived at by
molecular communication between fetal and maternal cells^[75]28.
Although very little is known about the mechanisms driving the
progression of PAS, the current prevailing hypothesis explaining the
PAS pathogenesis is that the trophoblasts may preferentially invade
into the acellular scar tissue left by a previous cesarean surgery or
other trauma to the uterine wall. We posit counter arguments to this
thesis. If cell permeable collagen tracks were present in acellular
collagenous scar, other cells within the decidua could enter and
re-cellularize the region. Instead, the absence of cells in the highly
collagenous acellular region of scar proper indicates a physical
barrier to cell invasion. Based on these considerations, we posit that
the scar may transform the decidua in its proximity, promoting
aggressive trophoblast invasion. Despite the frequent absence of
decidua at the invasion sites in the third trimester of PAS cases,
histological examination indicates endometrial stroma overlies the
cavity of scar after cesarean sectioning^[76]29–[77]34. We hypothesize
that uterine scarring leads to a failure of normal decidualization in
its proximity, which causes abnormal interaction between the decidua
and the EVTs. This aberrant communication between decidua and EVTs
leads to failure of maternal tissues to restrain the invading
trophoblasts.
Based on histological characterization of maternal-fetal interface from
PAS patients, we created a model of scar-decidua encapsulating the
essential microenvironmental features of a uterine scar. We found that
endometrial stromal fibroblasts (ESFs) obtained from normal patients
exhibited dysregulated decidualization on Scar matrix, and
recapitulated the phenotype of deep trophoblast invasion. We present
here the mechanisms by which mechanical signals from the scar matrix
modulate decidual fibroblasts to produce inflammatory cytokines
recruiting EVTs towards the scar. Specifically, we show that this
inflammatory transformation of dESF is achieved by Scar matrix-induced
activation of mechanosensitive ion channel, Piezo1, resulting in
increased intracellular calcium activity that leads to Protein kinase C
(PKC) activation. PKC activation results in NF-κB phosphorylation,
stabilization of a small Maf protein, MafG, and onset of
inflammation-related transcription. In addition, we also found that
scar transformed dESFs to a high contractile force-producing state,
fueled by glycolysis, further enhancing PAS-like manifestation of deep
placental invasion. The phenotypes were similar to the promotion of
cancer dissemination by cancer-associated fibroblasts
(CAFs)^[78]35–[79]37, suggesting that the mechanisms we identify are
implicated in other scar or fibrosis-related pathologies, including
cancer metastasis, and wound healing.
Results
Creating decidua-trophoblast interface on substrate mimicking uterine scar
Histological analysis of PAS tissues from multiple patients showed deep
invasion of placental villi into the myometrium, marked by disseminated
HLA-G^+ extravillous trophoblasts (EVTs) (Fig. [80]1A, B). Collagenous
regions, which were largely acellular, showed highly aligned collagen
fibers in the region proximal to scar (Fig. [81]1C, D and Supplementary
Fig. [82]1A). Anisotropy of collagen fibers has been observed in scars
of other tissues, including the ovary and endometrium by second
harmonics imaging^[83]38.
We asked if the aberrant mechanical signals present at the scar
transform decidual fibroblasts to be more invasable to EVT invasion. To
answer this question, we sought to create a scar-like model of decidua,
and ask if it manifests PAS-like phenotype characterized by enhanced
EVT invasion. To mimic the directional collagen bundles present in
scar, we patterned an anisotropic nanowrinkled pattern on the hydrogel,
with individual pitch matching those of collagen fibrils in fibrotic
tissue, confirmed by atomic force microscopy (Fig. [84]1E and
SupplementarySupplementary Fig. [85]1B). For scar-mimetic matrix
(Scar), we created polyacrylamide hydrogel substrate rigidity matched
to those reported in elastography data^[86]39 (Fig. [87]1F). For the
control physiological matrix (Physio), we mechanically profiled normal,
healthy endometrium, and created a flat substrate mimicking the tissue
rigidity (Fig. [88]1F and Supplementary Fig. [89]1C, D). Physio was
conjugated with 40 µg/ml type I collagen, while anisotropic high
rigidity Scar was conjugated with 100 µg/ml type I collagen. ESFs
isolated from normal patient endometrial biopsies were seeded on Physio
and Scar, and decidualized for 4 days before downstream experimentation
(Fig. [90]1G).
Scar matrix transforms decidual fibroblasts to chemotactically recruit
extravillous trophoblasts
We have previously shown that stromal permissibility to invasion, both
in decidua and in cancer stroma, is an evolved phenotype in mammals,
with stromal fibroblasts being active players in determining their own
invasibility^[91]40. Indeed, while decidualization is an evolved
response to resist trophoblast invasion, EVTs cooperate with decidual
fibroblasts via intercellular paracrine signaling to assist in their
own invasion^[92]41. We asked if EVTs invade more on Physio or Scar
decidua using multiple invasion assays designed to elicit different
plausible mechanisms regulating stromal invasion by EVTs (Fig. [93]1G).
First, we observed spheroids created from HTR8, a cell line derived
from EVTs, measuring their spreading and dissemination on a layer of
ESFs decidualized either on Physio or Scar substrate using live cell
phase contrast microscopy (Fig. [94]1H and Supplementary Movie [95]1).
Particle Image Velocimetry (PIV) analysis of this early placentation
model revealed that Scar decidua significantly promoted HTR8 spreading
and invasion (Fig. [96]1H, I, SupplementarySupplementary Fig. [97]1E,
and Supplementary Movie [98]2). We then asked if the Scar matrix itself
was causal in increased HTR8 invasion, or if it transforms dESFs to be
more invasable to HTR8 invasion. We therefore decidualized ESFs on
Physio or Scar, suspended them after trypsinization and patterned them
in a monolayer juxtaposed to a monolayer of H2B-mCherry labeled HTR8,
the interface being orthogonal to an underlying nanopatterned
substrate. We have used this platform, termed Accelerated Nanopatterned
Stromal Invasion Assay (ANSIA) to measure stromal invasibility with
high sensitivity^[99]24,[100]42. ANSIA eliminates the variation in
invasion by preexisting cell orientation in the monolayer, as well as
accelerates the invasion phenotype by unidirectional alignment of
cellular actomyosin assemblies, facilitating rapid quantitative
screening of stromal parameters. Time course analysis on ANSIA showed
that dESFs decidualized on scar exhibited reduced resistance to HTR8
invasion (Fig. [101]1J, Supplementary Fig. [102]1F-G, and Supplementary
Movie [103]3).
Stromal invasion is a complex process and is a composite outcome of
several potential sub-phenotypes, including migration of epithelial
cells^[104]43, stromal matrix degradation^[105]44, mechanical coupling
between epithelial and stromal cells^[106]37, paracrine recruitment of
epithelial cells, breach of cell-cell adhesion^[107]45, etc. As
decidualization on Scar resulted in more invasable dESFs, we asked if
paracrine signals from these Scar-altered dESFs contribute to increased
HTR8 invasion. We first spaced trophoblast spheroids away from the
dESFs monolayer with a layer of collagen gel before the initiation of
invasion (Fig. [108]1G and Supplementary Fig. [109]1H), and then
recorded the relative nuclear spatial position of trophoblasts to dESFs
monolayer after two days of invasion using structured illumination
(Apotome) based imaging. Our results showed that trophoblasts are
closer to dESFs on Scar than dESFs on Physio matrices (Fig. [110]1K,
L). These data suggested that ESFs decidualized on Scar produce
cytokines that recruit HTR8s towards scar (Fig. [111]1L). To gain
insight into the underlying mechanisms contributing to the aberrant EVT
invasion, we isolated the RNA from dESFs on Physio and Scar for
sequencing. We found large differences in gene expression
(Supplementary Fig. [112]1I-J and Supplementary Fig. [113]2), showing a
marked reduction in decidual marker genes like IGFBP1, as well as
increase in inflammatory cytokine CXCL8 (Fig. [114]1M). TransFac
predicted activation of transcription factors (TFs) highlighted
inflammatory TFs, including those belonging to the NF-κB pathway
(NFKB1, RELA, NFKBIA), HDAC1, which converts fibroblast into
cancer-associated fibroblasts (CAFs)^[115]46, RUNX1 which
prognosticates immune infiltrate in CAFs^[116]47, as well as NEF2L2
which encodes the key antioxidant TF Nrf2, also known to promote
metastasis in CAFs^[117]48 (Fig. [118]1N). Overall, our data suggest
that mechanical cues presented by the scar matrix alter dESF state to
chemotactically recruit EVTs preferentially towards the scar.
Scar promotes NF-κB-driven dESFs inflammation
Immunocytochemistry staining of tissue slides from PAS patients
revealed that many HLA-G^+ EVTs were present in the decidual region
proximal to the collagenous scar vs decidual location distal from the
scar, consistent with our PAS model (Fig. [119]2A and Supplementary
Fig. [120]3A). KEGG pathway analysis on differentially expressed genes
in Scar decidua confirmed increased activation of inflammatory pathways
including NF-κB, a master regulator of immune activation, and IL-17, as
well as mechanotransduction pathways associated with focal adhesion and
actin cytoskeletal regulation (Fig. [121]2B and Supplementary
Fig. [122]3B). A broader gene set analysis revealed activation of
several inflammation related ontologies on Scar, the top being those
related to tumor necrosis factor alpha (TNFA) and NF-κB signaling
(Fig. [123]2C). Gene set enrichment analysis, a non-parametric
statistical test, also showed high enrichment of inflammatory response
on Scar vs Physio, as well as TNFα signaling via NFKB (Supplementary
Fig. [124]3C). Activation of NF-κB associates with the accumulation of
RelA subunit in the nucleus, driving transcription of several
inflammatory genes. Immunofluorescence confirmed increased RelA
localization in the nuclei on ESFs decidualized on Scar vs Physio
(Fig. [125]2D and Supplementary Fig. [126]3D), also confirmed for
phosphorylated RelA abundance (Fig. [127]2E). We then sorted HLA-G^+
cells from decidua basalis (Supplementary Fig. [128]3E) and
functionally tested the effect of NF-κB activation in regulating
invasibility of dESFs using ANSIA. We have previously shown that NFKB1
increases invasibility of skin fibroblasts to cancer invasion^[129]49.
NFKB1 silencing significantly decreased dESF resistance on Scar-like
matrix, highlighting its role in regulating stromal response to
epithelial invasion (Fig. [130]2F and Supplementary Movie [131]4).
Histology of PAS tissue with residual decidua showed that decidual
fibroblasts in region proximal to acellular collagenous scar-like
location mostly exhibited nuclear localization of RelA, key indicator
of NF-κB activation. (Fig. [132]2G and Supplementary Fig. [133]3F–H).
In contrast, distal decidua had very few decidual fibroblasts with
nuclear RelA localization (Fig. [134]2G, H). We also found the nuclear
colocalization of p50 with RelA in scar-proximal decidual fibroblasts
(Supplementary Fig. [135]3G). Furthermore, we found a strong
correlation between EVT density in decidual regions with nuclear
localization of RelA in decidual fibroblasts (Fig. [136]2I),
highlighting that NF-κB activation in decidual fibroblasts promoted EVT
recruitment. Additionally, our pathway enrichment analysis of a
recently available PAS single-cell RNA sequencing (scRNAseq)
data^[137]50 showed that NF-κB pathway is indeed highly enriched in
adherent decidua of PAS patients (Supplementary Fig. [138]3I, J).
Overall, these data showed that scar matrix transformed the
decidualization of ESFs in its proximity to a more inflammatory state,
mediated by NF-κB, promoting more than optimal EVT invasion.
Fig. 2. Scar transforms decidual fibroblasts into NF-κB mediated inflammatory
state.
[139]Fig. 2
[140]Open in a new tab
A Representative immunohistochemistry images of maternal-fetal
interface (MFI) tissue sections from PAS patients showing EVTs present
in decidual regions proximal, and distal to collagenous acellular scar
regions; Quantification of EVT density in either region in right panel;
n = 5 and 3 for regions proximal and distal to scar respectively; EVTs
and dESFs are marked with HLA-G (red; arrow heads) and Vimentin
(green), respectively. p = 0.005. B KEGG pathway enrichment analysis
showing signaling pathways differentially enriched in dESFs on Scar and
Physio; C Gene ontologies related to inflammation enriched in dESFs on
Scar and Physio. D Representative immunofluorescence images showing
RelA (p65) location in dESFs on Physio and Scar; Quantification showing
percentage of dESF with nuclear RelA in lower panel. n = 4 and 5 fields
of view for Physio and Scar, respectively. p = 5 × 10^−8. E Immunoblot
showing abundance of phosphorylated RelA (p-RelA) in dESFs on Physio
and Scar. Experiments are repeated twice with similar results. F ANSIA
based analysis of stromal invasion of primary EVTs into dESF
compartment with scrambled, or gene silenced for NFKB1; n = 10 and 13
locations for scrambled and NFKB1^KD, respectively. p = 0.005. G
Representative immunohistochemistry images of MFI tissue sections from
PAS patient showing RelA intracellular localization in decidual regions
proximal, and distal to scar; H Quantification of percentage of
decidual fibroblasts with nuclear RelA; n = 3 and 6 locations for
distal and proximal, respectively. p = 0.0005. I Pearson correlation
test shows Pearson coefficient (r) of EVT number per field of view, and
ratio of decidual fibroblasts with nuclear and cytoplasmic RelA and
total decidual fibroblasts in PAS MFI tissue sections; a two-tailed
p-value for Pearson’s r is calculated; n = 8 field of views. Data in
figures A, D, F, H are showing as mean ± s.d.; statistical significance
is determined by unpaired two-tailed t-test (**p < 0.01, ***p < 0.001).
Source data are provided as a Source Data file.
dESFs on Scar matrix chemotactically recruit EVTs via secreted IL-8 and G-CSF
To identify potential paracrine signals, which induced dESFs on Scar to
recruit EVTs, we searched for ligand encoding genes differentially
regulated between Physio and Scar dESFs, finding several inflammatory
cytokines in the list (Fig. [141]3A). These genes included CSF3 which
encodes granulocyte colony-stimulating factor G-CSF, several immunocyte
recruiting C-X-C motif family ligands including CXCL1, CXCL2, and
CXCL8, the latter encoding for IL-8, other interleukins IL-1A, IL-1B
and IL-11, and several ligands of the tumor necrosis factor superfamily
(Fig. [142]3A). It has been reported that IL-8 and G-CSF can regulate
trophoblast migration in various contexts^[143]51,[144]52, and so we
tested if dESFs on Scar may recruit EVTs via IL-8 and G-CSF. We
confirmed increased abundance of IL-8 in dESFs both in 2D Scar model
using immunostaining (Fig. [145]3B and Supplementary Fig. [146]4A), and
in 3D Scar model with ELISA (Fig. [147]3C and Supplementary
Fig. [148]1H), compared to respective Physio models. We also confirmed
increased IL-8 and G-CSF secretion on 2D Scar using ELISA
(Fig. [149]3D). We then asked if NF-κB regulated IL-8 and G-CSF
production. TNFα treatment of dESFs significantly increased IL-8
secretion (Fig. [150]3E). To test if dESFs produced IL-8 and G-CSF
could influence EVT migration, we tracked H2B-mcherry labeled HTR8s
using live epifluorescence microscopy in the presence of conditioned
medium from dESFs on Scar, silenced for genes encoding IL-8, G-CSF, or
scrambled control. We verified that gene silencing of CXCL8 showed no
effect on the inflammatory phenotype of dESFs by assessing the
expression of α-SMA and vimentin (Supplementary Fig. [151]4B–E).
Conditioned medium from dESF^CXCL8-KD significantly reduced HTR8
displacement, which was reversed on addition of recombinant human IL-8
(Fig. [152]3F–H, Supplementary Fig. [153]4F, and Supplementary
Movie [154]5). A similar effect was observed for conditioned medium
from dESFs silenced for CSF3 (dESF^CSF3-KD), which significantly
reduced HTR8 velocity, while addition of recombinant G-CSF increased it
again (Fig. [155]3F–H and Supplementary Fig. [156]4F).
Fig. 3. IL-8/G-CSF secreted by Scar transformed decidual fibroblasts
chemotactically recruit EVTs.
[157]Fig. 3
[158]Open in a new tab
A Heatmap showing significant differential ligand encoding genes
expressed in dESFs on Scar and Physio matrices. B Representative IL-8
immunofluorescence images of dESFs treated with protein transport
inhibitor GolgiStop for 6 h on Physio and Scar matrices, quantification
shown in right panel; n = 49 and 53 cells for Physio and Scar,
respectively. p = 5 × 10^−11. C, D ELISA based analysis of IL-8 and
G-CSF concentration in supernatant of dESFs on Physio and Scar in 3D
and 2D; n = 3 samples. p = 0.0007 in (C) and p = 0.04 and 2 × 10^−5 in
(D). E ELISA based measurement of IL-8 concentration in supernatant of
dESFs treated overnight with DMSO, or 100 ng/ml TNFα; n = 3 samples.
p = 0.0097. F Experimental workflow to test migration of HTR8 in medium
conditioned from dESFs with gene silenced for IL-8 and G-CSF encoding
genes, CXCL8 and CSF3, respectively. G Migration trajectories (initial
location (x, y = 0,0)) of HTR8 conditioned with medium from dESFs
silenced for CXCL8 and CSF3 genes, without or with addition of
recombinant human (rh) IL-8 and G-CSF; Quantification of averaged
velocities over 24 h shown in (H); p = 1×10^−8, 8×10^−7, 8×10^−9, and 6
× 10^−5; n listed below each condition. I 3D chemotaxis of primary EVTs
in collagen gel towards IL-8 and G-CSF gradient; Shown is a
representative image of EVTs in collagen gel (left); Trajectories of
individually tracked EVTs from their initial location (0,0) (middle and
right); Cell trajectory with mean displacement towards cytokine end are
labeled red, and counted (n); p value showing Rayleigh test of cell
trajectories: p < 0.05 is considered chemotaxis. J ANSIA-based stromal
invasion analysis of HTR8 in monolayer of dESFs silenced for CXCL8 and
CSF3 genes; Control refers to scrambled sgRNA. p = 0.003 and 0.005.
Data in figures B–E, H and J are showing as mean ± s.d.; statistical
significance is determined by unpaired two-tailed t-test (*p < 0.05,
**p < 0.01, ***p < 0.001, and ****p < 0.0001). Source data are provided
as a Source Data file. Figure 3F created with BioRender.com released
under a Creative Commons Attribution-NonCommercial-NoDerivs 4.0
International license
([159]https://creativecommons.org/licenses/by-nc-nd/4.0/deed.en).
The presence of a scar is a spatially localized presentation of
mechano-chemical stimuli. We therefore asked if cytokines secreted by
dESFs at the scar can chemoattract EVTs through the endometrial stroma
towards its source, the Scar decidua. Using a microfluidically
generated gradient of IL-8 and G-CSF on primary EVTs and HTR8 embedded
in a 3D collagen gel, we tracked their displacement over time. We found
that both exhibited a strong displacement bias parallel to IL-8, as
well as G-CSF gradients (Fig. [160]3I, Supplementary Fig. [161]4G, and
Supplementary Movie [162]6). Finally, we used ANSIA to quantitatively
measure collective EVT invasion into dESFs. Our ANSIA results showed
that gene silencing of CXCL8 and CSF3 significantly reduced dESF
invasibility (Fig. [163]3J and Supplementary Fig. [164]4H).
Piezo1 activation on Scar decidua increases NF-κB mediated secretion of IL-8
and G-CSF
A scar results in a significant change in the mechanical milieu of
decidual stroma, to which decidual fibroblasts are likely to respond
differently than to the normal matrix. Indeed, immunocytochemistry
analysis of PAS patients’ tissue sections showed that dESFs with
nuclear RelA were located closer to the acellular areas compared to
dESFs with cytoplasmic RelA, indicating that scar matrix likely drove
nuclear translocation of RelA in decidual fibroblasts (Fig. [165]4A and
Supplementary Fig. [166]5A). Because cells sense extracellular
mechanical cues through a class of proteins on the plasma membrane
known as mechanosensitive ion channels (MSICs)^[167]53, we first asked
if dESFs express any MSICs. Gene expression analysis revealed that
PIEZO1, which encodes a key mechanosensitive ion channel, Piezo1, was
the only MSIC highly expressed and significantly upregulated on Scar vs
Physio dESFs (Fig. [168]4B). Our analysis of PAS scRNAseq data showed
that PIEZO1 expression is indeed upregulated in adherent decidua of PAS
patients (Supplementary Fig. [169]5B). ANSIA based quantitative
analysis of dESF invasibility showed that PIEZO1 knockdown
significantly increased dESF resistance to primary EVT spheroids and
HTR8 invasion (Fig. [170]4C, Supplementary Fig. [171]5C-F, and
Supplementary Movie [172]7). To further confirm the role of PIEZO1 in
regulating dESF invasibility, we performed EVT spheroids invasion in
Matrigel plugs embedded with wildtype and gene edited dESFs in mouse
(Fig. [173]4D, Supplementary Fig. [174]6). Similar to our in vitro
invasion assay, we found that PIEZO1 knockdown significantly increased
dESFs resistance to HTR8 spheroid invasion, suggesting Piezo1’s
causality in regulating PAS-like phenotype of increased decidual
invasibility (Fig. [175]4E, Supplementary Fig. [176]6). Piezo1 is a
membrane stretch gated Ca^2+ permeable channel^[177]54; therefore, we
tested if Scar matrix results in altered calcium dynamics within dESFs
using genetically encoded calcium indicator unless otherwise mentioned.
Scar matrix induced significantly more frequent, and higher amplitude
Ca^2+ oscillations (Fig. [178]4F–H, Supplementary Fig. [179]5G-H, and
Supplementary Movie [180]8). To confirm that these increased Ca^2+
levels and oscillations on Scar were Piezo1 mediated, we measured Ca^2+
dynamics upon treatment of Yoda1, a potent Piezo1 activator. Yoda1
resulted in significantly higher basal levels of Ca^2+ intensity
(Fig. [181]4I and Supplementary Movie [182]9). In contrast, gene
silencing for PIEZO1 in ESFs, and then decidualization on Scar resulted
in nearly complete cessation of calcium oscillations (Fig. [183]4J).
Piezo1 is classically described to be activated by changes in membrane
tension^[184]55. We therefore asked if directly perturbing cellular
membrane properties could regulate calcium signaling in dESFs. When
dESFs were treated with methyl-β-cyclodextrin (MβCD), which chelates
cholesterol from lipid rafts and increases membrane stiffness^[185]56,
we observed a dramatic increase in frequency of Ca^2+ oscillations
(Supplementary Fig. [186]5I).We then asked if Piezo1 activation in
dESFs on Scar could regulate NF-κB activation by checking RelA
phosphorylation, a key regulator in NF-κB activation by enhancing its
transactivation potential^[187]57,[188]58. Immunoblot showed that
phosphorylated RelA was reduced in dESFs treated with Piezo1 inhibitor,
GsMTx-4, while Piezo1 activator Yoda1 increased it (Fig. [189]4K).
Since GsMTx-4 also inhibits calcium activities gated by other TRP
channels, we further confirmed the causal link between Piezo1 and NF-κB
activation using CRISPR/Cas9 gene silencing in ESFs followed by
decidualization. As expected, Piezo1 gene silencing resulted in
decreased phosphorylated RelA abundance (Fig. [190]4K). Furthermore,
ELISA showed that Yoda1 increased IL-8 and G-CSF production, while
GsMTx4 significantly decreased production of both cytokines
(Fig. [191]4L). Gene silencing of PIEZO1 also markedly reduced IL-8 and
G-CSF production in dESFs on Scar (Fig. [192]4M). Remarkably, ELISA
also showed that direct perturbation of membrane stiffness by MβCD
resulted in significant increase in IL-8 production, as well as of
G-CSF (Supplementary Fig. [193]5J). These data show that increased
membrane tension of dESFs on Scar matrix can transform the fibroblasts
into an inflammatory state, producing chemotactic cytokines for EVTs,
and that this transformation is dependent on Piezo1 activation.
Fig. 4. Piezo1 dependent decidual mechanoregulation drives IL-8/G-CSF
production.
[194]Fig. 4
[195]Open in a new tab
A Immunohistochemistry images of RelA expression in decidual
fibroblasts from PAS patient; Quantification showing distance from scar
for dESFs classified according to RelA localization; n = 104 and 253
cells. p = 0.002. B Heatmap showing tpm values of genes encoding
mechanosensitive ion channels in dESFs on Physio or Scar; n = 3
biological replicates. C ANSIA analysis of primary EVTs spheroids
invasion into scrambled and PIEZO1^KD dESFs; n = 10 and 13 spheroids.
p = 0.0099. The scrambled control is shared with Fig. [196]2F since
these conditions are studied in the same round of experiment. D
Schematic showing HTR8 spheroids invasion into wildtype and gene edited
dESFs embedded in Matrigel plugs in mouse. E Invasion area of EVT
spheroids in Matrigel plugs containing scrambled and PIEZO1^KD dESFs.
n = 21 and 15 spheroids. p = 0.004. F Snapshot and calcium transients
of dESFs transduced with GCamP6f on Physio and Scar. G Ca^2+ peak/basal
ratio and transient events (H) in dESFs on Physio and Scar. and n = 44
and 92 cells (G); n = 9 and 25 cells (H) p = 0.03 (G) and 0.002 (H). I
Images of dESFs loaded with Fluo4-AM treated with DMSO or Yoda1; Graph
showing Ca^2+ peak/basal levels. n = 20 cells. J Ca^2+ dynamics in
scrambled and PIEZO1^KD dESFs. n = 48 and 32 cells. K Immunoblots
showing abundance of RelA, and phosphorylated RelA in dESFs treated
with GsMTx-4, or Yoda1, and in PIEZO1^KD dESFs. n = 2 biologically
independent experiments. L ELISA measurement of IL-8 and G-CSF
concentration in supernatant of dESFs treated with Yoda1 or GsMTx-4.
n = 3 biological replicates. p = 4 × 10^−6, 0.0005, 0.005, and 0.0009.
M Concentration of IL-8 and G-CSF in supernatant of scrambled and
PIEZO1^KD dESFs. n = 3 biological replicates. p = 0.003 and 0.004. N
Immunohistochemistry images and graph showing Piezo1 expression in
decidual fibroblasts with cytoplasmic or nuclear RelA localization in
PAS patients. n = 42 and 66. p = 2 × 10^−7. Data in all bar graphs are
showing as mean ± s.d.; statistical significance are determined by
unpaired two-tailed t-test (*p < 0.05, **p < 0.01, ***p < 0.001, and
****p < 0.0001). Source data are provided as a Source Data file. Figure
4D created with BioRender.com released under a Creative Commons
Attribution-NonCommercial-NoDerivs 4.0 International license
([197]https://creativecommons.org/licenses/by-nc-nd/4.0/deed.en).
Finally, we asked if the observed bimodal presentation of NF-κB
activation (Fig. [198]4A) is indeed related to mechanical signal from
the scar matrix in PAS patients. We co-analyzed Piezo1 and RelA
localization in decidual fibroblasts using immunohistochemistry after
verifying the specificity of Piezo1 antibody (Supplementary
Fig. [199]5C, D). Regions with mostly nuclear RelA localization also
showed higher Piezo1 expression (Fig. [200]4N). To note, in vitro Scar
matrix had also resulted in a similar twofold change in expression for
PIEZO1 mRNA (Fig. [201]4B), consistent with PAS in vivo scRNAseq data
(Supplementary Fig. [202]5B). These results are noteworthy as it shows
that chronic mechanical stimuli from scar can even change expression of
Piezo1. Utilizing scRNAseq data from normal pregnancy^[203]59,[204]60,
and PAS patients^[205]50, we also found coexpression of PIEZO1 with
CXCL8 (Supplementary Fig. [206]5K, L). Together with increased
expression, and higher activity of Piezo1, aberrant matrix on scar
results in activation of NF-κB signaling in decidual fibroblasts.
Scar matrix increases contractile force generation and membrane tension in
decidual fibroblasts driving Piezo1 activation
We asked how Scar matrix activates Piezo1. Uterine scar presents a
composite chemo-mechanical stimulus to the ESFs decidualizing in its
proximity. These include higher rigidity, high collagen content, as
well as reduced isotropy in the ultrastructural arrangement of collagen
fibrils. Indeed, Yap, a key mechanical transcriptional regulator active
in cells on rigid surfaces, was mostly cytoplasmic in decidual cells
distal to the acellular scar in PAS tissue. In contrast, Yap was mostly
nuclear in the fibroblasts proximal to scar (Fig. [207]5A and
Supplementary Fig. [208]7A). In vitro ESFs decidualized on Scar and
Physio matrix also showed a similar Yap localization (Fig. [209]5B and
Supplementary Fig. [210]7B). Moreover, we found elevated YAP1
expression in adherent decidua of PAS patients (Supplementary
Fig. [211]7C). Furthermore, elevation of nuclear Yap expression led to
increased Piezo1 expression in dESFs (Supplementary Fig. [212]7D).
Several gene ontologies related to cellular contractility, actomyosin
organization, and cell-matrix adhesion were significantly upregulated
in Scar vs Physio (Supplementary Fig. [213]2 and [214]7E). Focusing on
gene sets related to cellular contractile machinery, we found key
ontologies upregulated in Scar vs Physio, including focal adhesion,
stress fiber assembly, contractile actin fiber bundle etc.
(Fig. [215]5C and Supplementary Fig. [216]7F). It is now well
established, including in some of our previous works, that both
stiffness and anisotropic arrangement of the extracellular matrix can
promote the intracellular actomyosin assembly, driving its maturation,
and increasing cellular contractility^[217]61. Immunostaining indeed
revealed highly abundant, organized, parallel bundles of F-actin on
Scar vs Physio, with higher mean lengths indicating maturated stress
fibers (Fig. [218]5D).
Fig. 5. Scar activates Piezo1 through glycolysis fueled actomyosin
contraction.
[219]Fig. 5
[220]Open in a new tab
A Immunohistochemistry images of Yap localization in decidual sections
proximal, or distal to scar from PAS patients; n = 3 sections;
p = 0.005. B Immunofluorescence and graph showing Yap localization in
dESFs on Physio and Scar; n = 3 biological replicates. p = 2 × 10^−7. C
RNAseq-based enrichment analysis of contractility ontologies in dESFs.
D F-actin staining in dESFs on Physio, and Scar (left), n = 61 cells
for each condition; p = 6 × 10^−5. Graph showing length and intensity
of F-actin bundles (right), n = 21 and 17 cells. p = 9 × 10^−11. E
Traction force map of dESFs on Physio and Scar (left). Quantification
of strain energy and energy density for each cell (right); n = 50 and
82 cells; p = 0.005 and 0.01. F Time-lapse images showing Ca^2+
dynamics in three individual dESFs: C1, C2, and C3; heatmap showing
corresponding traction force of each cell; G Correlation analysis of
energy density with Ca^2+ events frequency. r: Pearson correlation
coefficient; n = 27 cells. H Ca^2+ events and abundance of
phosphorylated RelA (I) in dESFs treated with DMSO or Blebbistatin;
n = 60 and 59 cells; p = 2 × 10^−6. J Concentration of IL-8 and G-CSF
in supernatant of dESFs treated DMSO or Blebbistatin by ELISA; n = 3
biological samples; p = 0.002 and 0.01. K Heatmap showing gene
expression of PFKs and PFKFBs in dESFs. L Images of two dESF cells
showing 2-NBDG uptake (left) and their co-measured traction force maps
(right). n = 50 biological replicates. M Pearson correlation analysis
of cellular energy density and mean 2-NBDG intensity in dESFs; n = 50
cells. N Traction force maps and strain energy of dESFs maintained in
glucose and pyruvate withmatching C molarity; n = 164 and 94 cells;
p = 0.006. O Energetics profiling of oxygen consumption rate (OCR), and
extracellular acidification rate (ECAR) in dESFs at basal levels,
coupled (oligomycin sensitive), and uncoupled (FCCP sensitive)
respiration. n = 3 biological replicates. P dESFs on Scar show
significant increase in glycolysis. n = 3 biological replicates;
p = 0.03. Q Schematic showing Scar promoted cellular contractility is
fueled by increased glycolysis. Data in bar graphs are showing as mean
± s.d.; statistical significance is determined by unpaired two-tailed
t-test (*p < 0.05, **p < 0.01, and ****p < 0.0001; ns not significant).
Source data are provided as a Source Data file. Q created with
BioRender.com released under a Creative Commons
Attribution-NonCommercial-NoDerivs 4.0 International license
([221]https://creativecommons.org/licenses/by-nc-nd/4.0/deed.en).
To functionally confirm increased coupling of Scar matrix with the
intracellular actomyosin assembly, we used traction force microscopy
(TFM) on ESFs which were decidualized on Scar and Physio, and
thereafter replated on TFM substrates of 6 kPa rigidity matching normal
endometrium (Fig. [222]1E). A planar monolayer of fluorescent beads is
embedded in a pliable polyacrylamide hydrogel conjugated with Collagen
type I. As cells form focal adhesion with the matrix, intracellular
actomyosin-generated contractile forces are transmitted to the gel as
traction force, resulting in a measurable displacement of fluorescent
beads, used to back-calculate strain energy of the cell^[223]62. We
found that Scar resulted in significant increase in dESFs’
contractility both at per cell level (strain energy), as well as
spatially (energy density) (Fig. [224]5E).
ESFs decidualized on Scar showed higher Piezo1 activation, resulting in
increased NF-κB mediated inflammatory transformation. We therefore
asked if the intracellular mechanical changes within dESFs on Scar
could contribute to Piezo1 activation. We co-measured Ca^2+
oscillations in a population of dESFs along with their traction forces
using TFM at a single-cell level. Harnessing the population-wide
co-variance of either quantity, we found that frequency of Ca^2+
oscillations was correlated with cellular strain energy (Fig. [225]5F,
G, Supplementary Fig. [226]7G, and Supplementary Movie [227]10).
Addition of Blebbistatin, a myosin II inhibitor expected to reduce
contractile force generation, resulted in significant reduction in
Ca^2+ dynamics (Fig. [228]5H, Supplementary Fig. [229]7H, and
Supplementary Movie [230]11). Finally, we also found that Blebbstatin
treatment decreased phospho-RelA abundance, as well as IL-8 and G-CSF
production, measured by ELISA (Fig. [231]5I, J). Overall, these data
showed that increased contractile force on Scar activated
Piezo1-mediated Ca^2+ signaling, contributing to increased IL-8/G-CSF
production.
Scar induced enhancement in fibroblast traction force generation is
accompanied with increased reliance on glycolytic metabolism
Scar matrix resulted in increased contractile force generation in
dESFs, which requires maturation of the actomyosin assembly
(Fig. [232]5D). High cellular contractility necessitates increased
energy utilization. Differential gene expression analysis showed that
genes encoding isoforms of
6-phosphofructo-2-kinase/fructose-2,6-bisphosphatase (PFKFBs) were
upregulated on Scar (Fig. [233]5K). PFKFB3 is of particular interest
among the four isozymes due to its highest capability of promoting
glycolytic flux and keeping glycolysis high. It’s frequently
overexpressed in numerous human tumors, including ovarian, lung,
breast, colon, pancreatic, and thyroid tumors^[234]63,[235]64. To test
the relationship between glucose uptake and cellular contractility, we
harnessed the population-wide variation between either quantity in
single cells and measured their correlations. A strong correlation
existed between glucose uptake and strain energy, suggesting that dESFs
primarily require glycolysis for contractile force generation
(Fig. [236]5L, M and Supplementary Fig. [237]7I). To explore the link
between NF-κB pathway to glycolysis, we knocked down RelA expression in
dESFs and performed 2-NBDG uptake assay. We found that RelA knockdown
reduced 2-NBDG uptake in dESFs (Supplementary Fig. [238]7J, K),
indicating that NF-κB pathway and glycolysis are correlated. Culture in
equimolar (C content normalized) pyruvate significantly decreased
strain energy, suggesting that a shift to the citric acid cycle reduces
contractile force generation on Scar (Fig. [239]5N). Recent report
indicated that F-actin bundling sequesters E3 ligase TRIM21, which
modulates the degradation of rate-limiting metabolic enzyme
phosphofructokinase, thereby coupling actin polymerization with
glycolytic flux^[240]65. To document the energetic state of dESFs on
Scar, we tested the oxygen consumption rate (OCR) and extracellular
acid reflux rate (ECAR) using Seahorse XFe metabolic analyzer. Although
we did not find any significant difference in OCR rates between dESFs
previously cultured on Physio and Scar, glycolysis was significantly
increased on Scar matrix (Fig. [241]5O, P). These data suggested that
increased energy demand of the Scar transformed dESFs is primarily met
by increased glycolysis (Fig. [242]5Q).
Protein Kinase C (PKC) mediates increased Ca^2+ signaling related
inflammatory transformation of dESFs on Scar
We sought to identify the potential signaling intermediaries regulating
Ca2^+-mediated NF-κB activation in dESFs on Scar. There are multiple
cellular sensors of the cytosolic Ca^2+ levels, including calmodulin,
phosphatidylinositol 3-kinase (PI3K)/Protein kinase B pathway, and
Protein kinase Cs (PKC), all of which have been previously shown to
regulate NF-κB activation^[243]66. Kinase enrichment analysis
identified AKT1, PRKCA, EKR1/2 as top enriched kinases in dESFs from
Scar (Fig. [244]6A). We therefore tested if PKC was activated in dESFs
on Scar using immunoblot for PKC substrates. Indeed, we found increased
signal for PKC-mediated phosphorylation on Scar (Fig. [245]6B). Next,
we ask if the PKC activation is Piezo1 dependent by treating dESFs with
GsMTx-4 (Piezo1 inhibitor), Yoda1 (selective Piezo1 activator), or
Gö6983 (PKC inhibitor). We found that GsMTx-4 reduced phosphorylation
of PKC substrates, while Yoda1 increased it, confirming that PKC
activation in dESFs is Piezo1 dependent. (Fig. [246]6C and
Supplementary Fig. [247]8). We then asked if NF-κB phosphorylation on
Scar could be explained by PKC activation. Gö6983 reduced abundance of
phospho-RelA (Fig. [248]6D). Furthermore, Gö6983 reversed the increased
phospho-RelA abundance achieved by Yoda1 (Fig. [249]6E), together
suggesting that Piezo1 mediated increased Ca^2+ signaling activated
PKC, which phosphorylated NF-κB. Could PKC activation therefore
contribute to inflammatory cytokine production on Scar? To test this,
we perturbed PKC activity in dESFs on Scar, and used ELISA to measure
IL-8/G-CSF, cytokines responsible for recruiting EVTs towards the
Scar-decidua. For both IL-8/G-CSF, we observed that Gö6983
significantly reduced production, while PKC activator, PMA increased it
significantly more than control (Fig. [250]6F). Finally, we tested if
PKC is a key intermediate second messenger for Piezo1-mediated
IL-8/G-CSF production. ELISA confirmed that while Yoda1 increased
IL-8/G-CSF production in dESFs as previously noted, addition of Gö6983
along with Yoda1 reversed the increase (Fig. [251]6G). These data
showed that opening of Piezo1 mechanosensory channel on Scar resulted
in increased Ca^2+ activity which activated NF-κB via PKC signaling.
Fig. 6. Piezo1-mediated inflammatory transformation of dESFs depends on
Protein kinase C (PKC) activation.
[252]Fig. 6
[253]Open in a new tab
A Kinase enrichment analysis (KEA3) predicted top activated kinases in
Scar vs Physio dESFs using Fisher’s Exact Tests (pval) on RNAseq data.
B Substrate immunoblot of PKC activated targets in ESFs decidualized on
Physio and Scar matrices, as well as (C) in dESFs on Scar treated with
4 µM GsMTx-4 and 2 µM PKC inhibitor Gö6983 for 4 h. D Immunoblot
showing abundance of phosphorylated RelA (p65) in dESFs on Scar
without, and after overnight treatment with 2 µM Gö6983, as well as (E)
10 nM PKC activator Phorbol 12-myristate 13-acetate (PMA), 5 µM Piezo1
activator Yoda1, and 3 µM Yoda1 plus 5 µM Gö6983; GAPDH is loading
control in D, E. Experiments are one of the two biological replicates
with similar results. F ELISA based measurement of IL-8 and G-CSF
concentrations in supernatant of dESFs on Scar after overnight
treatment with 2 µM Gö6983 and 10 nM PMA; n = 3 replicates; p = 0.01,
0.0003, 0.001, and 0.003. G ELISA-based measurement of IL-8 and G-CSF
concentrations in supernatant of dESFs on Scar after treatment with
3 µM Yoda1, and 3 µM Yoda1 plus 5 µM Gö6983; n = 3 replicates;
p = 0.02, 0.002, 9 × 10^−6, and 8 × 10^−6. Data in F and G are showing
as mean ± s.d.; statistical significance is determined by unpaired
two-tailed t-test (*p < 0.05, **p < 0.01, ***p < 0.001, and
****p < 0.0001; ns not significant). Source data are provided as a
Source Data file.
MafG stabilization by Piezo1 regulates transcription of trophoblast
recruiting cytokines
Predicted transcription factor activation on Scar also showed several
zinc finger nucleases, PAX3, and MAFG (Fig. [254]7A). MAFG is enriched
in endometrium and placenta, but very little is known about its role in
pregnancy. Small Maf factors seem to play a role in regulating
transcription of inflammatory cytokines, with a report in the
myometrium^[255]67, and another in the central nervous system^[256]68.
MAFG transcripts level was significantly increased in Scar
(Fig. [257]7B), as well as MafG abundance (Fig. [258]7C). Consistently,
we also found increased MAFG expression in adherent decidua of PAS
patients^[259]50 (Fig. [260]7D). Immunoblot showed that MafG abundance
was directly regulated by Piezo1-mediated signaling. While addition of
Yoda1 dramatically increased MafG levels within an hour of addition,
indicating of post translational regulation, addition of Piezo1
inhibitor GsMTx-4 decreased MafG abundance (Fig. [261]7E).
Interestingly, presence of PKC inhibitor Gö6983 abrogated Yoda1
mediated increase in MafG levels (Fig. [262]7E). Moreover, inhibition
of ERK1/2, a known PKC target, also reduced MafG expression in a dose
dependent manner (Fig. [263]7F and Supplementary Fig. [264]9A).
Fig. 7. MafG stabilization by Piezo1 mediates transcription of trophoblast
recruiting cytokines.
[265]Fig. 7
[266]Open in a new tab
A Prediction of upstream transcription factors on Scar vs Physio dESFs.
B TPM values of MAFG on Physio and Scar; n = 3 replicates; p = 0.03. C
Immunoblot showing MafG abundance in dESFs on Physio and Scar matrices.
D MAFG scRNAseq expression levels in adherent and non-adherent PAS
decidua, and in normal decidua (GEO accession number:
[267]GSE212505)^[268]50. n = 1836, 1595, and 1246 cells. p = 0.04 and
0.001. E Immunoblot showing abundance of MafG and Nrf2 in dESFs on Scar
treated with GsMTx4, Yoda1, and Yoda1 plus Gö6983 for 1 h. n = 2
biological replicates. F Representative immunofluorescence image of
dESFs treated with DMSO, or SCH772984; Graph showing quantification of
MafG levels in dESFs treated with SCH772984 (SCH) (right). n = 599,
323, 384, and 331 cells for each condition; p = 4 × 10^−298, 0, and 4 ×
10^−140. G IL-8 levels in supernatants from scrambled and MAFG^KD dESFs
treated with Yoda1 by ELISA; n = 3 replicates; p = 0.03, 0.002, and
0.004. H G-CSF levels in supernatants from scrambled and MAFG^KD dESFs;
n = 3 replicates; p = 40.003. I Migration trajectories and mean
velocities of HTR8 conditioned with medium from scrambled and MAFG^KD
dESFs, without, or with addition of 300 ng/mL rh G-CSF. p = 1 × 10^−6
and 0.03. J ANSIA-based analysis of primary EVTs invasion into
scrambled and MAFG^KD dESFs, without, or with addition of 300 ng/mL rh
G-CSF; n = 10 spheroids for each condition; p = 0.03. The scrambled
control is shared with Fig. [269]2F and Fig. [270]4C since these
independent conditions are studied in the same round of experiment. K
Invasion area of EVT spheroids in Matrigel plugs containing scrambled
and MAFG^KD dESFs. n = 16 and 21 spheroids, respectively. p = 0.0096.
The scrambled control is shared with Fig. [271]4E since these
conditions are studied in the same round of mouse injection. L
Schematic showing the plausible mechanism driving inflammatory
transformation of decidual fibroblasts and EVT recruitment proximal to
existent uterine scar. Data in all bar graphs are shown as mean ± s.d.;
Statistical significance is determined by unpaired two-tailed t-test
(*p < 0.05, **p < 0.01, and ****p < 0.0001; ns not significant). Source
data are provided as a Source Data file. L created with BioRender.com
released under a Creative Commons Attribution-NonCommercial-NoDerivs
4.0 International license
([272]https://creativecommons.org/licenses/by-nc-nd/4.0/deed.en).
We therefore asked if increased MafG levels in Scar could contribute to
increased EVT recruitment by the transformed decidual fibroblasts. When
MAFG gene was silenced in dESFs, there was a significant reduction in
IL-8 and G-CSF production, which could not be rescued by Yoda1,
highlighting the essential role of MafG in regulating Piezo1-triggered
inflammation (Fig. [273]7G, H). In cell migration assay, HTR8
displacement decreased significantly in conditioned medium from dESFs
silenced for MAFG vs control dESFs, which was rescued by addition of
recombinant G-CSF (Fig. [274]7I, Supplementary Fig. [275]9B, and
Supplementary Movie [276]12). Indeed, ANSIA based quantification of
stromal resistance to primary EVT (Fig. [277]7J) and HTR8
(Supplementary Fig. [278]9C, D) invasion showed a similar trend, with
dESFs silenced for MAFG gene displaying reduced invasibility.
Similarly, our Matrigel plug based EVT spheroids invasion of MAFG^KD
dESFs in mouse showed the same trend (Fig. [279]7K, Supplementary
Figure [280]9E). Overall, these data suggest key role of calcium driven
PKC signaling, as well as MafG-mediated transcription in regulating the
inflammatory transformation of decidual fibroblasts on Scar, promoting
chemotactically driven EVT invasion.
Discussion
The pathogenesis of scarring has long been a research interest since
early studies in development and tissue morphogenesis. Therefore, the
mechanistic details of the molecular drivers of scarring have been
identified. However, the long-term effect of scar on tissue homeostasis
is less well documented. Existing scars can profoundly and chronically
influence cellular functions in various tissues. Pre-existent scars due
to myocardial infarction can increase the chance of heart
failure^[281]69,[282]70, and fibrotic response to a malignant neoplasm
can regulate cancer dissemination^[283]71,[284]72. Although there are
many case reports of scar-associated invasive pathologies,
epidemiological studies are few, and the mechanistic understanding of
scar induced pathogenesis is severely lacking. In a small number of
patients burn scars can result in development of squamous cell
carcinoma, called Marjolin’s ulcers^[285]73,[286]74. Malignant
degeneration of scars have been noted in other cases^[287]75. Keloids,
benign dermal fibro-proliferative tumors with a genetic basis, are
positively correlated with several skin cancers and even pancreatic
cancers^[288]76. Mechanistic studies are unavailable because
association of scar with invasive processes is small, and takes many
years to emerge. Potential emergence of neoplasm is also a relatively
smaller concern with scars which typically emerge from significant
trauma.
Here, we sought to understand how scar may dysregulate tissue
homeostasis by studying a common, surgically induced deep scar, which
has a causative pathology: the uterine scar resulting from cesarean
procedures which is positively associated with PAS. Using an in vitro
model of uterine scar with patient-derived decidua, we showed that
scar-like matrix promotes stromal transformation through
mechanosensation-mediated inflammation. Scar impairs the physiological
endometrial decidualization process, leading to production of
chemoattractive cytokines which recruit the EVTs towards the scar.
Specifically, the altered mechanical milieu of the scar drives the
surrounding stromal fibroblasts to produce IL-8, and G-CSF, which
depends on glycolysis-fueled contraction and Piezo1-mediated calcium
signaling resulted NF-κB p65 phosphorylation and MafG
stabilization (Fig. [289]7L).
Our findings differ from the currently held hypothesis explaining PAS
pathogenesis, which posits that the acellular scar presents an empty
pavement for the EVTs to rapidly invade into, resulting in
manifestation of PAS. A counterargument to the hypothesis is that any
acellular scar region which is penetrable for EVTs is also by
definition, penetrable to numerous other maternal cell types, including
the immune cells. Therefore, if scar matrix is merely a spatial void
which gets filled in post implantation, it should have been
re-celluarized before implantation. Indeed, in patient derived term
tissues with PAS, we frequently found acellular, highly collagenous
regions, which were not invaded by maternal or placental cells.
Remarkably, residual decidua proximal to these acellular scar regions
contained more HLA-G^+ EVTs compared to distal decidua. Confirming our
hypothesis even more was the observation that decidual regions proximal
to scar contained maternal decidual cells with nuclear RelA, and also
expressed higher Piezo1 levels than distal regions. We therefore posit
that the altered mechanical cues presented by the scar matrix transform
the fibroblasts into an inflammatory state, which then promote EVT
recruitment via chemotactic cues. It is possible that once EVTs are
recruited more than optimally proximal to the scar, other mechanisms
may promote deeper invasion, some of which our study also hints at,
including increased contractile force generation by the fibroblasts at
the scar.
In recent years, molecular analysis of PAS samples has revealed markers
characterizing, or priming aggressive trophoblast
invasion^[290]77,[291]78. Our study details a mechanistic link from the
non-physiological mechanical cues present at the scar to the altered
intracellular signaling, resulting in PAS. In particular is the role of
a key stretch-gated Ca^2+ permeable channel, Piezo1, which we found to
be both activated as well as increased in expression on Scar. Increased
Ca^2+ signaling activated PKC mediated NF-κB phosphorylation, and
transcription of inflammation related genes. Among these inflammatory
cytokines, we showed that IL-8, and G-CSF, which significantly increase
in decidual Scar, are potent recruiters of EVTs towards the Scar,
presenting attractive targets to modulate placental invasion. Piezo1
activation on decidual scar is particularly notable, as it has been
described to be important in mechanosensation in cells with large
ranges of stretch stimuli, e.g. lung alveolar cells^[292]79. Further
downstream, PKC activation has been previously described to interfere
with decidualization^[293]80. We show that PKC-driven activation of
NF-κB, which is well documented in other tissues^[294]81, can play a
critical role in transformation of the decidual cells into inflammatory
fibroblasts.
We also found a hitherto less studied transcription co-regulator, MafG,
as a key contributor to the PAS-like phenotype. Small Maf proteins act
as bZIP (basic region leucine zipper) type transcription factors which
can bind to DNA, and are well known to form heterodimers with other CNC
(cap and collar) transcription factors, which include the antioxidant
Nrf (1,2,3) factors^[295]82. Recently there have been reports of role
of MafG in the inflammation of the central nerve system, as well as in
cancer^[296]68,[297]83. We found that scar-like matrix resulted in
increased expression and abundance of MafG, driven by Piezo1
activation, suggesting mechanoregulation of MafG activity. The link
from cellular mechanosensation to MafG mediated inflammatory response
may be more general in other mechanically active tissues.
PAS is a rapidly growing concern during pregnancy, with devastating
effects on maternal health and future reproductive plans. We used this
model of scar owing to its standardized presentation in many women,
which has also resulted in a clearly identifiable pathology caused by
the scarring. However, deep scars resulting from trauma to other
tissues are also common, and similar scar-mediated inflammatory
transformation of the resident fibroblasts can contribute to pathology.
Our work presents a mechanistic understanding of the long-term
scar-induced effect on fibroblast phenotype, as well as avenues to
prevent fibroblast inflammation, a harbinger of various chronic
pathologies.
Methods
Tissue collection, cell isolation and culture
Human endometrium tissues and FFPE tissue samples from medically
necessary biopsies or hysterectomies were obtained from the
biorepository at University of Connecticut Health Center after
de-identification in accordance with the guidelines and Accio Biobank
Online with IRB exemption (Supplementary Table [298]1). Endometrial
tissue was minced on the day of collection, and then either plated on
Physio or Scar matrices, or on polystyrene culture plates for isolation
of endometrial stromal fibroblasts (ESFs) by selective
attachment^[299]84. Isolated ESFs were maintained in phenol red free
DMEM/F12 50:50 containing 25 mM glucose, and supplemented with 10%
charcoal-stripped fetal bovine serum (Thermo Fisher), 1%
antibiotic/antimycotic, and ITS (insulin, transferrin, and selenium).
To induce decidualization, ESFs were treated with 0.5 mM 8-bromo-cyclic
3′,5′-(hydrogen phosphate)-adenosine (Cayman Chemicals) and 1 µM
Medroxyprogesterone acetate (Cayman Chemicals) in DMEM/F12 medium with
2% FBS for 4 days. Chemical perturbations are implemented during the
last day of decidualization unless otherwise stated. For 3D
decidualization, cells seeded on Physio and Scar matrices were embedded
in 1 mg/ml type I collagen from rat tail (Thermo Fisher) and incubated
at 37 °C for 1 h to allow 1 mm thick collagen gel formation.
Decidualization medium was then added without disturbing the matrices.
To isolate HLA-G^+ cells, decidual basalis was separated from the
maternal side of the termed placenta tissue and then minced. Enzyme
cocktail containing dispase, collagenase I, and DNase I was added to
equal volume of minced tissue. The mixture was fixed on a shaker with
500 rpm at 37 °C for 1 h. Equal volume of culture medium containing 20%
FBS was added to the mixture to block the digestion. The mixture was
then centrifuged at 340 × g for 5 s to remove the undigested tissue.
The supernatants containing isolated cells were collected and directed
plated into flasks coated with 20 µg/ml fibronectin. After 48 h of
culture, non-attached cells were gently washed away. The isolated
placental cells were then detached and blocked with 5% goat serum and
1% BSA at room temperature for 20 min, before stained with Alex Flour
488 conjugated human anit-HLA-G antibody (Biolegend, 1:50) at 4 °C for
30 min. HLA-G+ cells were sorted using flow cytometry (BD FACSAria II
sorter) and maintained in DMEM/F12 with 10% FBS, ITS, and antibiotics.
Human extravillous trophoblast cell line-HTR8/SVneo derived from first
trimester of pregnancy were obtained from ATCC (CRL-3271). Cells were
cultured in RPMI medium supplemented with 10% FBS and 1%
antibiotic/antimycotic (Gibco). HTR8 were stably transduced with
plasmid expressing H2B-bound mCherry driven by CMV promoter unless
otherwise stated.
Mechanical indentation
Endometrial tissue samples from patients were cut into 1 cm × 1 cm ×
0.4 cm pieces and half-embedded in low-melting agarose and then
immersed in RPMI medium. Mechanical indentation was implemented using
Mach-1 micromechanical testing system (Biomomentum) mounted with an
indentation probe with a 2 mm bead at the tip. The indentation speed is
50 µm/s and the indentation depth is 500 µm. At least 20 locations were
measured per sample.
Matrix fabrication
Scar matrix was fabricated by sandwiching polyacrylamide precursors
between nanotextured poly(urethane acrylate) (PUA) molds and
saline-activated coverslips. PUA molds are either
pre-fabricated^[300]61, or fabricated from a silicone mold fabricated
using electron-beam lithography followed by deep-reactive ion etching.
Saline-activated coverslips for gel attachment were cleaned with
ethanol and sonication, treated with air plasma, and activated with
0.5% glutaraldehyde and 0.5% (3-Aminopropyl)triethoxysilane (Sigma
Aldrich). Polyacrylamide precursor solution containing 10% acrylamide
and 0.3% bis-acrylamide (Bio-Rad) was degassed for 30 min and mixed
with 0.1% tetramethylethylenediamine and 0.1% ammonium persulfate
(Sigma Aldrich) before sandwiched between silane-activated coverslips
and PUA molds for 20 min. The physiological matrices were fabricated
following similar procedures but with precursor containing 5%
acrylamide and 0.12% bis-acrylamide sandwiched between saline-activated
and Rain-X coated coverslips. The crosslinked gels were peeled off and
coated with 40 μg/ml (Physio) and 100 μg/ml (Scar) collagen type I
using sulfo-SANPAH (Thermo Fisher) overnight at 4 °C. Gels were
sterilized under UV for at least 2 h before cell seeding.
Collagen orientation analysis
Tissue slides from PAS patients were stained with Picrosirius red.
Collagen signals in the regions of interest were obtained after color
deconvolution using ImageJ. Collagen orientation was quantified using
OrientationJ plugin in ImageJ.
AFM imaging
AFM imaging of the surface topography of the Scar matrix was performed
using Asylum Research Cypher AFM in PBS. A 0.08 N/m triangle PNP-TR
probe (NanoWorld) was used. The scanning speed was set to 0.3 Hz.
In vitro invasion assays
For in-situ invasion, HTR8 spheroids prepared using ultra-low
attachment round bottom microplate (Corning) were plated on top of
dESFs on Physio and Scar substrates. Time-lapse images were taken every
2 h for at least 24 h. Invasion area was normalized by the initial
projection area of the spheroid. For spheroid invasion, invading
primary EVT or HTR8 spheroids were seeded on dESFs monolayers on
nanopatterned scar-mimicking substrates. Time-lapse images were taken,
and invasion area was analyzed as aforementioned. For ANSIA invasion,
invading primary EVTs or HTR8s and dESFs monolayers were patterned
juxtaposed to each other as previously described^[301]24,[302]42.
Briefly, a custom-made polydimethylsiloxane stencil was placed on the
nanogrooves-patterned substrate. The device was kept in a vacuum to
remove air bubbles under the stencil. Then, HTR8-mcherry were seeded at
a density of 5 × 10^5 cells and allowed to attach to the substrate
overnight. The stencil was removed carefully using blunt-end tweezers.
The unlabeled stromal cells were seeded at a density of 5 × 10^5 to
fill and attach that area covered by the stencil before. The unattached
cells were washed off after 5 h of incubation. Invasion were recorded
for 24 h. Invading cells were imaged by using time-lapse microcopy
every 1 h for 24 h. The area occupied by the invading cells was traced
manually using Region of Interest (ROI) panel in the Fiji software. The
normalized extent of invasion was calculated by dividing total
[MATH:
δArea(t) :MATH]
(of invading cells) by the length of the initial trophoblast–stroma
interface.
[MATH: δAreat=Areat−Areat0 :MATH]
1
[MATH: δAreat=δAreatLInterf<
/mi>ace
:MATH]
2
Automated peak identification was performed on the converted ROIs to a
one-dimensional mask. The masks’ profiles were smoothed by moving the
average over 20 pixels. The mean signal to either side of every point
with 40 pixels each, was calculated by using smoothed profiles. A peak
also was identified when the boundary of smoothed profile was larger
than both side average. Finally, the number of invasive forks as well
as the distribution of the depth of invasion in stromal monolayer were
measured.
For 3D invasion, ESFs were seeded on Physio and Scar. After cell
attachment, 1 mg/ml collagen solution were casted on ESFs and incubated
at 37 °C for 1 h to form a gel layer with thickness of 1 mm. After 4
days of decidualization, HTR8 spheroids prepared as aforementioned were
suspended in 1 mg/ml collagen solution and plated on the 3D
decidualization ESFs and incubated at 37 °C for 1 h to allow spheroid
settle down and gel formation. dESF monolayer and HTR8 nuclear
locations were recorded by Zeiss Apotome 3D scanning. Distance between
each individual HTR8 nucleus and each dESF layer were calculated after
deconvolution and nuclear segmentation.
In vivo invasion
Matrigel plug based EVT spheroid in vivo invasion was performed in
mouse. Briefly, target genes in primary ESFs were knockdown using Neon
NxT electroporation system (Invitrogen) following the manufacturer’s
guideline. ESFs were then seeded on scar-like substrate and
decidualized for 8 days. One day before injection, growth factor
reduced high concentration Matrigel (Corning) was thawed in ice at 4 °C
overnight for cell encapsulation. On the same day, HTR8 cells were
labeled with 4 µM CM-DiI (Invitrogen) for 5 min at 37 °C and 15 min at
4 °C. Afterwards, HTR8 spheroids were prepared by centrifuging 1000
cells in each well of the 384-well spheroid microplate (Corning) at 350
× g for 5 min. After 24 h, HTR8 spheroids were collected in 1.5 mL
tubes (25 spheroids/tube). Meanwhile, dESFs were collected and
resuspended in 50 µL culture medium in the same tubes containing HTR8
spheroids (1 million cells/tube). The tubes were then placed in ice.
Before injection, 1 mL syringes and 20 G needles were also cooled in
ice. Then, 150 µL Matrigel was draw into each syringe without needle.
After mounting the capped needle onto syringe, the whole syringe was
embedded in ice immediately to prevent Matrigel gelation. Before
injection, the back fur of male mice was removed by shaving and cream.
After anesthetization, HTR8 spheroids and dESFs in each 1.5 mL tube
were mixed with Matrigel pre-loaded in each syringe and injected
subcutaneously into the back of the mice. No more than 4 injections
were performed for each mouse, with control and experimental conditions
in the same mouse. Adult (three months old) SCID/beige male mice
(Inotiv) were used in this study to reduce the variations caused by
female progesterone on dESFs. 10 mg MPA suspension were also injected
subcutaneously into each mouse to help dESFs maintain their
differentiation status. At least two injections on two mice were
performed for each condition. Injection sites were marked with marker
pen to help Matrigel plug excision 3 days after injection. To maintain
the integrity of Matrigel plugs, skin tissue was excised together with
the plugs. The whole tissue was then fixed in 0.5% Glutaraldehyde (EMS)
for at least 2 h to crosslink Matrigel, and then in 4% PFA overnight.
Skin tissue was then carefully removed to release Matrigel plugs for
imaging. To ensure that the integrity of EVT spheroids is not affected
by the forces during mixing and injection, the same procedures were
also performed in vitro in 96-well plate instead of in mouse. We
confirmed that the integrity of spheroids was not affected, and its
capability of invasion remained (Supplementary Fig. [303]6). All animal
protocols were approved by the Institutional Animal Care and Use
Committee (IACUC) at the University of Connecticut Health Center before
study initiation. All experiments were performed in accordance with
IACUC guidelines, and abides by the ARRIVE guidelines for reporting
animal experiments. Mice were kept in a 12 h light–dark cycle,
temperature-controlled (22 ± 2 °C) and humidity-controlled (55 ± 5%)
environment and fed a standard chow diet. Mice were anaesthetized with
isoflurane and cells were subcutaneously injected into the mice. After
3 days injection, Mice were euthanized with carbon dioxide and Matrigel
plugs, skin tissue were harvested for analysis.
RNA sequencing and transcriptomic analysis
Cells were lysed and RNA was isolated with RNeasy Mini Kit (Qiagen)
following manufacturer’s instructions. RNA integrity was evaluated with
Bioanalyzer 2100 (Agilent) and RIN ~ 8 was used for library
preparation. Library prep and RNA sequencing were performed by Novogene
Inc. HISAT2 pipeline was used to align reads to NCBI GRCh38 genome
assembly. HTSeq was used for reads were counting, and DESeq2 was used
for statistical significance (p-values) and fold-changes for
differential expression. Enrichment of gene sets in the differentially
expressed (DE) genes were calculated using Fisher exact test to
calculate the overrepresentative of terms (Gene Ontology, Hallmark,
Wikipathways) using hypergeometric test followed by correction for
multiple testing^[304]85. Ingenuity Pathway Analysis (Qiagen Inc) was
used to calculate predicted scores for transcription factors activation
and canonical pathways analysis. Gene set enrichment analysis (GSEA)
was performed on the genes^[305]86. Hierarchical clustering was
performed using UPGMA method with Euclidian distance on z-scores as
mentioned earlier^[306]61. KEGG pathway analysis was performed using
ShinyGO8^[307]87,[308]88. Kinase enrichment analysis was performed
using the webserver application Kinase Enrichment Analysis 3^[309]89.
For single-cell RNA sequencing analysis, decidua cells were extracted
from the data following the same annotation as described by the study
(GEO accession number: [310]GSE212505)^[311]50. Decidua cells from PAS
patients at the adherent and non-adherent sites, and from normal
pregnancy were analyzed. Genes of interest were visualized using violin
plots with t-test. Correlation analysis of gene expression was
performed using the Spearman method. Pathway enrichment analysis for
differential expression genes in decidua cells was performed using GSEA
on hallmark pathways.
Immunoblotting
Cells were harvested and lysed in cell lysis buffer containing RIPA
buffer (Bio-Rad), protease and phosphatase inhibitor cocktail (Sigma
Aldrich). BCA kit (Pierce) was used for protein concentration
measurement and normalization. Denatured samples (70 °C for 10 min in
SDS) were loaded on 4–12% NuPAGE Bis-Tris Gel (Thermo Fisher) along
with Lamelli loading buffer. Proteins were then transferred to PVDF
membranes, and blocked with 3% BSA for 1–2 h at room temperature and
incubated with primary antibodies (1:1000) overnight at 4 °C. Membranes
were washed with TBST for 5 times and re-blocked with 3% BSA.
HRP-linked anti-rabbit or mouse IgG secondary antibodies (GE
healthcare; 1:10,000) were cross-linked at room temperature for 1 h,
and protein bands visualized using enhanced chemiluminescence reagent
(Thermo Fisher) using an Imager (Molecular Biosciences). Antibody
details are listed in Supplementary Table [312]2.
Immunofluorescence
Cells were fixed in 4% paraformaldehyde for 15 min, permeabilized in
0.1% Triton-X100 for 10 min, and blocked in 1% BSA and 5% goat serum
for 1 h at room temperature. Cells were then incubated with primary
antibodies at 4 °C overnight and with secondary antibodies for 1 h at
room temperature. FFPE sections were rehydrated by immersing the slides
in xylenes twice for 3 min, 1:1 xylenes:ethanol for 3 min, 100% ethanol
twice for 3 min, followed by 90%, 75%, and 50% ethanol for 3 min each.
Antigen retrieval was performed after rehydration by immersing the
slides in citrate buffer at 95 °C for 30 min. Slides were then
incubated with primary antibodies (1:100) at 4 °C overnight and with
secondary antibodies (1:400) for 1 h in the dark at room temperature.
Antibody details are listed in Supplementary Table [313]2.
Gene silencing
Gene silencing was achieved by using pre-prepared synthetic sgRNA
(IDT). ESFs were transfected with sgRNA and TrueCut Cas9 protein
(Invitrogen) using Lipofectamine CRISPRMAX transfection reagent
(Invitrogen) or Neon NxT Electroporation System (Invitrogen).
Specifically, a cocktail was created by mixing (i) solution1: 24 µl
Opti-MEM and 1 µL CRISPRMAX, and (ii) solution 2: 10 nmol sgRNA, 15
nmol Cas9, 1.5 µL CRISPRMAX Plus reagent and remaining Opti-MEM to make
a 30 µL solution. Solution 1 and 2 were mixed and incubated for 10 min
before being drop dispensed into one well of a 24-well plate containing
cells at 50% confluency in culture medium. Cells were used after 48 h
of transfection, and observation completed within 48 h thereafter.
Electroporation based CRISPR/Cas9 gene editing was performed following
the manufacturer’s guideline for fibroblast gene editing
(1650 V/20 ms/1 pulse).
Cell migration and 3D chemotaxis
For cell migration, HTR8-mCherry cells were cultured in conditioned
dESF medium mixed with fresh medium (1:1) and monitored using
microscope for at least 9 h and time-lapse images were taken every
20 min. Single cell tracking was then performed using Fiji/ImageJ
TrackMate plugin. Cell migration velocity and distance were quantified
for individual cell. Cell trajectories were plotted using ibidi
Chemotaxis and Migration Tool V2.0. 3D Chemotaxis of HTR8-mCherry was
performed using μ-Slide Chemotaxis (ibidi) following the manufacturer’s
instruction. Briefly, 3 × 10^6 cells/ml HTR8 were resuspended in
1 mg/ml collagen type I. Six microliters of cell suspension was loaded
into the microchannel with a width of 1 mm and a height of 70 μm. For
the two chambers separated by the microchannel, one chamber was loaded
with 100 ng/ml recombinant human IL-8 or G-CSF (Proteintech), and the
other chamber was loaded with vehicle. Cell migration trajectories were
quantified as aforementioned. Rayleigh test was performed to determine
the statistical significance of chemotaxis. Rayleigh p < 0.05 is
considered chemotaxis.
Actin filament analysis
F-actin images were enhanced using the Contrast Limited Adaptive
Histogram Equalization (CLAHE) filter, then masks of actin filaments of
individual cells were generated by the threshold function in ImageJ.
Mean intensity value of actin was then quantified. To obtain actin
length, the BoneJ plugin were used.
Traction force microscopy
Traction force gels were fabricated using protocols previously
described^[314]90. Briefly, coverslips for gel attachment were cleaned
with ethanol and sonication, treated with air plasma, and activated
with 0.5% glutaraldehyde and 0.5% (3-Aminopropyl)triethoxysilane (Sigma
Aldrich). Coverslips for beads coating were treated with air plasma and
coated with 0.01% poly-L-lysine (Sigma Aldrich) before coated with
carboxylate-modified microspheres with diameters of 0.2 µm (Thermo
Fisher). Gel precursor solution containing 7.5% acrylamide and 0.15%
bis-acrylamide was degassed for 30 min and mixed with 0.1%
tetramethylethylenediamine and 0.1% ammonium persulfate before
sandwiched between silane-activated coverslips and bead-coated
coverslips for 20 min. After peeling off the bead-coated coverslips,
the resulting traction force gels were coated with 30 µg/ml collagen
type I using sulfo-SANPAH (Thermo Fisher) overnight at 4 °C. Gels were
sterilized under UV for at least 2 h before cell seeding. Images
containing microbeads location before and after cells trypsinization
were recorded using Zeiss Observer A1 microscope. Traction forces were
calculated following protocols previously described^[315]91.
Glucose uptake
Pre-warmed glucose-free DMEM (Thermo Fisher) with 10% FBS and 400 µM of
2-NBDG (Invitrogen) were added to cells on traction force gel at 37 °C
and 5% CO[2] for 30 min. Cells were then washed with DMEM twice, added
with culture medium, and mounted on microscope for live cell imaging.
Calcium imaging
To image cellular calcium dynamics, we generated 3rd gen. lentivirus
for GCamP6f. Briefly, GCamP6f probe was obtained from Addgene and
transferred to pENTR™/SD/D-TOPO® vector (ThermoFischer Scientific).
Subsequently, LR reaction was used to transfer it to pLEX_307 vector
(Addgene plasmid # 41392). The vector was sequence verified. The virus
was generated in HEK293-FT cells using packaging mix of three plasmids
(pLP1, pLP2, and pLP/VSVG) from ThermoFisher Scientific. The virus was
concentrated using PEG-it™ Virus Precipitation Solution (SBI
biosciences) and MOI was calculated using qRT-PCR. Cells were
transduced using Polybrene (Millipore Sigma) using MOI 5 (multiplicity
of infection). Calcium dyes Fluo4-AM (2 µM, Invitrogen) or Calbryte 590
(5 µM, AAT Bioquest) were premixed with 0.04% Pluronic-F127 (Sigma
Aldrich) in serum-free medium and loaded to cells for 20 min. Calcium
activities were imaged 20 min after dye washing out. To prevent
photobleaching, time-lapse images were taken using light with low
intensity and short exposure time at 1 Hz for 2 or 3 min. Photo
bleaching of FLuo4-AM was corrected using the Bleaching Correction
plugin in ImageJ. To image calcium activities for cells treated with
blebbistatin, Calbryte 590-AM (AAT Bioquest) was used as a calcium
indicator to avoid autofluorescence of Blebbistatin (Sigma Aldrich).
Measurement of oxidative phosphorylation and glycolysis
Seahorse XF analyzer (Agilent Technologies) was used to monitor
cellular energetics^[316]61. Cells were dissociated and cultured in a
96 wells XF plate, and 6 replicates were used for each condition.
Oxygen consumption rate (OCR) was used to estimate oxidative
phosphorylation while change in extracellular acidification rate (ECAR)
was used to estimate glycolysis. Basal rates show changes in O[2] or pH
in the absence of any added compounds or metabolic inhibitors. ATP
synthase was inhibited using Oligomycin (4 µM), and rotenone (2 µM) and
antimycin A (2 µM) was used to inhibit complex 1/3 respectively.
Uncoupling was achieved using FCCP (500 nM) to estimate maximum
respiratory capacity, and iodoacetate (100 µM) was used to inhibit
glycolysis (glyceraldehyde-3-phosphate dehydrogenase). Respiratory
rates were normalized to DNA content using Picogreen DNA assay (Thermo
Fisher Scientific) following manufacturer’s instructions.
Enzyme-linked immunoabsorbent assay (ELISA)
Human IL-8 (Biolegend) and G-CSF (SinoBiological) kits were used to
test measure secreted IL-8 and G-CSF levels from dESFs culture
supernatants according to the manufacturer’ instructions. Briefly, a
96-well plate was coated with capture antibody overnight at 4 °C,
blocked for 1 h at room temperature, followed by sample incubation for
2 h and detection antibody incubation for 1 h. Then the plate was
incubated with Avidin-HRP for 30 min, 10 mg/ml Tetramethylbenzidine for
15 min before 2 N H[2]SO[4] was added to the plate. The absorbance
signals were read using SpectraMax i3x multi-mode microplate reader
(Molecular Devices) at 450 nm and background at 570 nm. IL-8 and G-CSF
concentrations were quantified using SoftMax Pro.
Plasmids
pLenti6-H2B-mCherry was obtained from Addgene (#89766). Lentiviral
particles prepared with 3rd generation packaging plasmids. HTR8 cells
were transduced, and selected using blasticidin resistance, as well as
sorted with FACS for reduced heterogeneity of mCherry expression in the
cell cultures.
Microscopy
All cell invasion and migration assays and traction force microscopy
were performed using Zeiss Axio Observer Z1 microscope with PECON
Incubation System S for live cell imaging. 3D imaging and deconvolution
was performed using ZEISS Apotome.2 and Zen Blue 2.6 Pro software.
Whole slide scanning was performed using ZEISS Axioscan 7.
Statistical analyses
Statistical analyses were performed using GraphPad Prism. Two-tailed
unpaired Student’s t-test was implemented for statistical significance
of the differences between two groups. One-way ANOVA was performed
followed by Tukey test if multiple groups were presented in a graph.
For graphs showing correlation analysis, Pearson correlation
coefficients were provided together with two-tailed statistical
significance values. All bar graphs show mean values with standard
deviation as error bars. Statistical significance levels are defined as
*p < 0.05, **p < 0.01, ***p < 0.001, and ****p < 0.0001. P values
higher than 0.05 are considered non-significant.
Reporting summary
Further information on research design is available in the [317]Nature
Portfolio Reporting Summary linked to this article.
Supplementary information
[318]Supplementary Information^ (3.2MB, pdf)
[319]Peer Review File^ (3.8MB, pdf)
[320]41467_2024_52351_MOESM3_ESM.pdf^ (145.6KB, pdf)
Description Of Additional Supplementary File
[321]Supplementary Movie 1^ (10.8MB, avi)
[322]Supplementary Movie 2^ (12.9MB, avi)
[323]Supplementary Movie 3^ (695.9KB, avi)
[324]Supplementary Movie 4^ (464.8KB, avi)
[325]Supplementary Movie 5^ (6.2MB, avi)
[326]Supplementary Movie 6^ (9.8MB, avi)
[327]Supplementary Movie 7^ (404.9KB, avi)
[328]Supplementary Movie 8^ (7.7MB, avi)
[329]Supplementary Movie 9^ (16.5MB, avi)
[330]Supplementary Movie 10^ (6.2MB, avi)
[331]Supplementary Movie 11^ (5MB, avi)
[332]Supplementary Movie 12^ (430.6KB, avi)
[333]Reporting summary^ (3.9MB, pdf)
Source data
[334]Source Data^ (7.6MB, xlsx)
Acknowledgements