Abstract
Endothelial cells respond to changes in subendothelial stiffness by
altering their migration and mechanics, but whether those responses are
due to transcriptional reprogramming remains largely unknown. We
measured traction force generation and also performed gene expression
profiling for two endothelial cell types grown in monolayers on soft or
stiff matrices: primary human umbilical vein endothelial cells (HUVEC)
and immortalized human microvascular endothelial cells (HMEC-1). Both
cell types respond to changes in subendothelial stiffness by increasing
the traction stresses they exert on stiffer as compared to softer
matrices, and exhibit a range of altered protein phosphorylation or
protein conformational changes previously implicated in
mechanotransduction. However, the transcriptome has only a minimal role
in this conserved biomechanical response. Only few genes were
differentially expressed in each cell type in a stiffness-dependent
manner, and none were shared between them. In contrast, thousands of
genes were differentially regulated in HUVEC as compared to HMEC-1.
HUVEC (but not HMEC-1) upregulate expression of TGF-β2 on stiffer
matrices, and also respond to application of exogenous TGF-β2 by
enhancing their endogenous TGF-β2 expression and their cell-matrix
traction stresses. Altogether, these findings provide insights into the
relationship between subendothelial stiffness, endothelial mechanics
and variation of the endothelial cell transcriptome, and reveal that
subendothelial stiffness, while critically altering endothelial cells’
mechanical behavior, minimally affects their transcriptome.
Subject terms: Mechanotransduction, Actin
Introduction
Subendothelial stiffening often occurs with aging and in multiple
pathologies such as atherosclerosis and hypertension constituting a
risk factor for development of cardiovascular disease^[28]1,[29]2.
Endothelial cells (ECs) that form a single monolayer in vivo to line
the inner lumen of blood vessels, respond to changes in the mechanics
of their extracellular matrix (ECM), such as its stiffness, by changing
their migration, proliferation and barrier integrity, thus contributing
to the emergence of these pathologies^[30]3–[31]5. Understanding the
interplay between the micro-environmental mechanical determinants and
EC behavior is therefore pertinent to understanding vascular biology
and might have important therapeutic implications.
ECs exhibit remarkable phenotypic heterogeneity, and the basis of these
morphological, molecular and functional differences is still not
completely characterized^[32]6,[33]7. It has been previously proposed
that the spatiotemporal differences in chemical and also mechanical
cues relayed to ECs by their environment theoretically could be
sufficient to explain their structural and functional
differences^[34]8. Examples of mechanical signals relayed to ECs
include subendothelial stiffness, fluid shear flow and mechanical
strains. However, even when ECs from different anatomical locations are
placed in vitro in the same biomechanical environment, they can still
display a unique behavior intrinsic to the ECs themselves and not
determined by differential culture or microenvironmental
conditions^[35]9–[36]11. For instance, the response of human umbilical
cord endothelial cells (HUVEC) to changes in curvature or shear stress
applied in tissue culture is completely distinct from that of brain
microvascular ECs^[37]9.
Transcriptomic profiling has advanced our understanding of how
differential gene expression is linked to altered cell behavior.
Specifically, it has provided insight into the complex biological
pathways and molecular mechanisms that regulate changes in cellular
behavior in response to mechanical cues for certain cells types, such
as mesenchymal stem cells, vascular smooth muscle cells and certain
endothelial cell types, all of which were found to be extremely
sensitive to substrate stiffness^[38]12–[39]17. However, in most of
these studies cell confluency was either low or not explicitly stated.
Cell density plays a crucial role in the response of ECs to mechanical
cues and in the forces transduced by ECs on their ECM and on each
other^[40]18,[41]19 and increased cell density can even override the
effect of ECM stiffness in certain cell types^[42]20. Inspired by these
studies, we sought to answer two important previously unexplored
questions: (1) Are the biomechanical changes in response to
subendothelial stiffness observed for ECs in monolayers due to
transcriptional regulation of key stiffness-sensitive genes? and (2) Is
the transcriptomic profile of ECs in monolayers dominated by the
specific EC type or by the mechanical microenvironment, in particular
subendothelial stiffness?
In this study, we compared the responses of two different types of ECs
to growth on stiff versus soft hydrogel substrates, primary human
umbilical vein endothelial cells (HUVEC) cultured from normal human
tissue and immortalized human microvascular endothelial cells (HMEC-1)
that were transformed using SV40 large T antigen^[43]21. Both cell
types in confluent monolayers changed their mechanical behavior in
response to increasing subendothelial stiffness similarly, by elevating
their cell-matrix traction stresses on stiffer as compared to softer
matrices, and altering protein phosphorylation profiles associated with
mechanotransduction. However only very modest stiffness-dependent
alterations in gene expression were observed using RNA sequencing.
Results
ECs in monolayers exert increased cell-matrix traction stresses when residing
on stiff as compared to soft hydrogels
To assess how subendothelial stiffness affects EC mechanics and how
that is related to changes in the endothelial transcriptome, we
cultured ECs as confluent monolayers on substrates of varying
stiffness. Trying to mimic (patho)physiologic subendothelial stiffness,
we built collagen-I coated soft hydrogels of 3 kPa (lower range of EC
basement membrane reported stiffness where cells can still in vitro
form monolayers) and stiffer hydrogels of either 35 kPa or 70 kPa
(higher ranges of basement membrane reported stiffness often associated
with aging or pathologies)^[44]22–[45]24. The resulting stiffness of
the gels was confirmed through atomic force microscopy (Supplementary
Fig. [46]S1a). Hydrogels were embedded with fluorescent beads such that
when ECs establish their focal adhesions and start pulling on the
hydrogels and on each other, we can use time-lapse microscopy to
monitor the cell and bead movement. Using the fluorescence images of
the beads we then infer the deformations the cells impart and the
stresses they exert on their matrix through traction force microscopy
(TFM)^[47]25. 3 kPa is the lowest stiffness we examined since as
previously reported, we find that ECs seeded on <3 kPa stiffness gels
do not consistently form monolayers but rather vessel-like patterns
with intermittent gaps^[48]26,[49]27. 70 kPa gels are considered as our
“STIFF” condition given previous reported values of pathological
subendothelial stiffness measured in clinical studies^[50]22–[51]24.
However, for TFM only we used as “STIFF” 35 kPa gels since this was
found to be the highest stiffness on which the cells can still deform
the gels.
By performing TFM, we found that both HUVEC and HMEC-1 followed via
time-lapse microscopy between 24 h to 32 h post-seeding generate
smaller deformations on stiffer 35 kPa gels as opposed to the softer 3
kPa, and overall HMEC-1 generate larger deformations than HUVEC under
similar conditions (Fig. [52]1). Interestingly, for both cell types the
traction stresses they exert and the total traction force magnitude
over the field of view is significantly higher for ECs residing on 35
kPa gels as compared to 3 kPa gels (Fig. [53]1a,b,d,e). This finding
suggests that the internal cytoskeleton and contractility of these two
EC types changes depending on the mechanical properties of their
subendothelium, in a way that might be similar in both EC types. Note
that the strain energy, which is the mechanical work spent by the cells
to deform their matrix, for both cell types is similar irrespective of
substrate stiffness (Fig. [54]1c,f).
Figure 1.
[55]Figure 1
[56]Open in a new tab
Endothelial cells in monolayers exert higher cell-matrix traction
forces onto stiff as compared to soft hydrogels. (a) Representative
phase contrast images (phase, first column) and cell-matrix deformation
maps (second column, color indicates deformation magnitude in μm) and
traction stresses (third column, color indicates stress magnitude in
Pa) exerted by confluent HUVEC adherent onto soft 3 kPa or stiff 35 kPa
hydrogels. (b,c) Time evolution of the integral of the traction force
magnitude over the whole field of view to its area (nN/μm^2) (b) and of
the total strain energy imparted by the cells per area of field of view
(nN/μm) (c) calculated for two different regions within confluent HUVEC
monolayers for cells residing on soft 3 kPa (blue) or stiff 35 kPa
(red) matrices. (d–f) Same as in panels a-c but corresponding to HMEC-1
monolayers.
Given that previous studies have shown that EC protein levels can be
sensitive to subendothelial stiffness^[57]28,[58]29, we sought to
evaluate whether in our system protein expression or activity would be
modulated by matrix stiffness under our experimental conditions. We
thus performed Western blot analysis for a number of EC proteins whose
activity or conformation has been previously shown to depend on
subendothelial stiffness^[59]28–[60]32. To that end we measured protein
levels for the active form of integrin β1 and integrin β1^[61]30,
p397FAK and FAK^[62]32, p-Vav2 and Vav2^[63]28, p-p70S6K and
p70S6K^[64]31, p-ERK and ERK^[65]5. We found that the phosphorylation
level or conformation of all these proteins changes for one or both
cell types (Fig. [66]2; Supplementary Fig. [67]S2). For both cell types
the protein level of the active form of integrin β1, but not total
integrin β1 expression, is increased for cells residing on stiff as
compared to soft substrates (Fig. [68]2a–c). Similarly, pVav2 and
p-p70S6K, but not total protein levels of Vav2 and p70S6K, are
increased for both cell types on stiff as opposed to soft matrices.
ERK2 protein levels remain constant while phosphorylation increases
with stiffness for HUVEC (but not HMEC-1). Finally, total FAK levels
remain constant but FAK activity increases significantly for HMEC-1
only (but not HUVEC). These results demonstrate that known
mechanosensitive proteins in these two EC types are responding as
expected to changes in matrix stiffness, and that this response is
largely mediated by changes in protein phosphorylation or conformation
(Fig. [69]2b,c).
Figure 2.
[70]Figure 2
[71]Open in a new tab
Post-transcriptional changes on ECs in monolayers grown on soft versus
stiff hydrogels. (a) Western blots from whole HUVEC or HMEC-1 lysates
of cells previously residing on soft gels (3 kPa) or stiff gels (70
kPa). Representative cropped blots are displayed and full-length blots
can be found in the supplementary material (Supplementary Fig. [72]S2).
Each row shows a different protein whose expression, phosphorylation or
conformation was probed, namely: active form of integrin β1, total
integrin β1, p397FAK, total FAK, p-Vav2, total Vav2, p-p70S6K, total
p70S6K, p-ERK, total ERK and GAPDH (used as loading control).
Experiments were performed N = 3 times. (b) Bar plots show relative
expression of the proteins probed in panel a for HUVEC cells residing
on soft (blue) or stiff (red) gels. All measurements were normalized to
GAPDH expression for each condition, and expressed as fold-change
relative to the median expression level on soft substrates. One or two
asterisks denote statistically significant differences between the
medians of two distributions (<0.05 or <0.01 respectively; unpaired
t-test) and non-significant differences are denoted as ns. (c) Same as
panel b but for HMEC-1.
Differentially expressed genes (DEGs) arise due to EC type rather than
subendothelial stiffness
We sought to understand whether the subendothelial stiffness-dependent
elevation of the EC traction stresses is additionally regulated by
changes in the EC transcriptome. To this end, we implemented RNA
sequencing to comprehensively define gene expression in HUVEC and
HMEC-1 grown in monolayers and in similar densities, in response to
physiologically soft (3 kPa, on SOFT) and pathologically stiff (70 kPa,
on STIFF) matrices (Supplementary Fig. [73]S1a,b). Due to inherent
variability of biological samples and to increase our confidence in
identification of DEGs, we decided to sequence 6 replicates per
condition. Using GENCODE annotations (grch38)^[74]33, we identified a
total of 15,797 genes expressed in HUVEC and 16,027 genes expressed in
HMEC-1, after disregarding genes with base mean normalized count lower
than 10^[75]34.
To our surprise, differential expression (DE) analysis of HUVEC on
STIFF versus HUVEC on SOFT led us identify just 24 DEGs using typical
thresholds (see Methods; Fig. [76]3a,b and Supplementary Fig. [77]S3
and Table [78]S1)^[79]34. When we then compared HMEC-1 on STIFF versus
SOFT, we found just 8 DEGs (Fig. [80]3c,d and Supplementary Fig. [81]S4
and Table [82]S2). In contrast, comparison of HUVEC versus HMEC-1 on
STIFF revealed a total of 10020 DEGs (Fig. [83]3e,f and Supplementary
Fig. [84]S5 and Table [85]S4) while comparison of HUVEC versus HMEC-1
on SOFT revealed 10162 DEGs (Fig. [86]2g,h and Supplementary
Fig. [87]S6 and Table [88]S4). Of those, 44 genes were solely expressed
in HMEC-1 and 147 solely in HUVEC (Supplementary Table [89]S5). Note
that for the above we used a low threshold of at least 1.23-fold
change. If a threshold of 2-fold change was applied, there would be no
stiffness-dependent DEGs identified for either EC type, while HUVEC
versus HMEC-1 on STIFF would yield 4449 DEGs and HUVEC versus HMEC-1 on
SOFT 4558 DEGs. These results suggest that the specific EC type
profoundly defines the transcriptome of ECs, and that matrix stiffness
has only a minimal effect, at least for these two specific EC types
studied and under the conditions where we examined them (i.e. confluent
monolayers, with cells seeded for 24 h on collagen I-coated gels). In
addition, we discovered that HUVEC show more stiffness-sensitive DEGs
as compared to HMEC-1, and that none of these few identified DEGs are
shared in common between the two EC types.
Figure 3.
[90]Figure 3
[91]Open in a new tab
Endothelial origin but not matrix stiffness strongly determines the
transcriptome of endothelial cells. (a,c,e,g) Scatter plots of
expressed genes showing normalized counts of gene expression in the x
and y axes for the indicated groups: (a) HUVEC on stiff 70 kPa (N = 6)
versus soft 3 kPa matrices (N = 6), (c) HMEC-1 on stiff 70 kPa (N = 6)
versus soft 3 kPa matrices (N = 6), and (d) HUVEC (N = 6) versus HMEC-1
on stiff 70 kPa matrices (N = 6) and (g) HUVEC (N = 6) versus HMEC-1 on
soft 3 kPa matrices (N = 6). Light gray dots represent genes that are
not differentially expressed while differentially expressed genes
(DEGs) are shown as dots color-coded by their -log[10] p-values.
(b,d,f,h) Volcano plots showing DEGs between the same groups compared
as above. The -log[10] p-values (y-axis) are plotted against the
average log[2] fold changes in expression (x-axis). Non DEGs are
plotted in light gray. DEGs are color-coded depending on the log[10] of
their mean normalized counts.
Principal component analysis of the top 500 most variable genes across
all samples confirmed that the EC type but not ECM stiffness is the
major determinant of expression differences and accounts for
approximately 99% of the variation between transcriptomes
(Fig. [92]4a,c). ECM stiffness appears to be a minor contributor
accounting for <1% of the expression variance for HUVEC only, while
HMEC-1 on STIFF fall exactly in the same PCA space as on SOFT,
appearing indistinguishable (Fig. [93]4b). Consistent with this
analysis, when we performed hierarchical clustering on the rlog
transformed counts and created dendrograms of the Euclidean distance
between pairs of samples, we found that the separation emerges based on
cell type but not on subendothelial stiffness (Supplementary
Fig. [94]S7a). Similarly, hierarchical clustering on the 200 most
variant genes across all samples clearly separates the two EC types,
while subendothelial stiffness does not separate HMEC-1, but separates
HUVEC with low confidence (Supplementary Fig. [95]S7b and
Table [96]S6).
Figure 4.
[97]Figure 4
[98]Open in a new tab
Principal component analysis (PCA) confirms that endothelial origin but
not matrix stiffness is a major contributor of expression differences.
(a,b) PCA on top 500 DEGs for HUVEC on soft 3 kPa (turquoise circles)
and on stiff 70 kPa (purple circles) matrices and for HMEC-1 on soft 3
kPa (orange circles) and on stiff 70 kPa (green circles) matrices. PC1
versus PC2 is shown in panel A and PC3 versus PC4 is shown in panel B.
The 95% confidence ellipse is shown for each group with the
corresponding colors. (c) Scree plot showing in decreasing order the
proportion of variance explained by each PCA mode up to PC10.
Next, given that the differences between the two EC types are so
dramatic, we sought to understand what major biological pathways
differ, and to that end we performed Kyoto Encyclopedia of Genes and
Genomes (KEGG) pathway analysis and Gene Ontology (GO) functional
enrichment for the DEGs of HUVEC versus HMEC-1^[99]35. Pathway
enrichment analysis reveals 9 KEGG pathways significantly perturbed
between the two EC types with 4 of those being clearly pertinent to
endothelial biomechanics (Supplementary Table [100]S7). Cell adhesion
molecules and particularly tight junction genes essential for proper
intercellular stress transmission and EC barrier integrity are
upregulated for HUVEC as compared to HMEC-1 (hsa04514; Supplementary
Fig. [101]S8). Most genes related to arachidonic acid metabolism,
important for the regulation of vascular tone in response to shear
stresses, are also upregulated for HUVEC^[102]36 (hsa00590;
Supplementary Fig. [103]S8). Additionally, genes related to
ECM-receptor interactions are significantly different between the two
EC types with HMEC-1 expressing significantly more collagen (and
exclusively COL1A1), tenascin and vitronectin as compared to HUVEC
which express more laminin, perlecan and fibronectin among other ECM
proteins (hsa04512; Supplementary Fig. [104]S9). Finally, genes related
to calcium signaling, involved in shear stress- and matrix
stiffness-sensing of ECs^[105]37, are upregulated for HUVEC as compared
to HMEC-1 (hsa04020; Supplementary Fig. [106]S9). Among the other genes
upregulated by HUVEC is myosin light chain kinase (MLCK) central in
cellular contractile behavior (Supplementary Fig. [107]S9). Altogether
these results suggest that HUVEC as compared to HMEC-1 might be more
sensitive and responsive to mechanical cues and perturbations.
To confirm the KEGG results we also mapped the DEGs between HUVEC and
HMEC-1 to GO Biological Process (BP) terms. The resulting functional
analysis revealed that the majority of significantly perturbed GO terms
in the BP category are upregulated for HUVEC as opposed to HMEC-1
(Supplementary Table [108]S8). Interestingly, all the 11 GO terms that
are upregulated for HMEC-1 correspond to cell cycle or cell division.
In contrast, the 47 upregulated terms found for HUVEC are largely
related to endothelial-specific functions, many of which are central in
EC biomechanical behavior, in accordance with the KEGG pathways
analysis (e.g. EC migration, calcium-dependent cell-cell adhesion).
TGF-β2 and TGF-β2-related genes are among the stiffness-sensitive DEGs
identified in HUVEC
When we looked more closely on the 24 stiffness-sensitive DEGs in
HUVEC, we found that 8 out of the 24 DEGs correspond to pseudogenes,
uncharacterized genes, mitochondrial RNA or microRNA and were therefore
not examined further (Supplementary Table [109]S1). Out of the
remaining 16 genes, 14 of them are upregulated on stiff and only 2 on
soft matrices. Of the 14 genes that are upregulated on stiff matrices,
5 genes encode proteins that are secreted to or remodel the ECM of ECs:
TGF-β2, ADAMTSL1, LAMB3, SMOC1, STC1^[110]38,[111]39. TGF-β2 is a
cytokine that is secreted to the ECM in a latent form. Once it switches
into an active conformation, TGF-β2 can bind to TGFβ receptors and
modulate the synthesis and accumulation of ECM components and the
expression of cell surface receptors for ECM
components^[112]40–[113]42. ADAMTSL1 is a latent TGF-β binding protein
that is key in ECM remodeling^[114]43, LAMB3 encodes the β3 subunit of
laminin which is a protein of EC basement membranes and is expressed
more in response to TGF-β2^[115]44, SMOC1 is a matricellular protein
(i.e. does not have a structural but rather a regulatory role when
secreted to the ECM) found in basement membranes and is also a TGF-β
regulator^[116]39 and STC1 is a secreted glycoprotein found downstream
of TGF-β2 in osteoclasts^[117]38. From the 11 remaining upregulated
genes 4 affect gene expression or transcription, namely: ZNF703, a
transcriptional co-repressor related to TGF-β^[118]45; ZSCAN31, a
DNA-binding transcription factor; ID1, a transcriptional
regulator^[119]46; and KLF10, a zinc finger DNA-binding protein that
regulates gene expression and is also known as transforming growth
factor-β (TGFβ) inducible early gene-1 (TIEG1)^[120]47. In addition, we
found the following genes to be upregulated: CXCL12, a secreted
chemokine that contributes to cancer^[121]48; GADD45B an effector of
TGF-β signaling^[122]49,[123]50; KRT7 a cytoskeletal protein expressed
in blood vessels and upregulated in response to TGF-β^[124]51; INSR a
transmembrane receptor that is activated by insulin^[125]52; and EPHA5
a receptor belonging to the protein-tyrosine kinase family^[126]53.
Finally, the only two upregulated genes discovered for HUVEC grown on
soft matrices were XAF1 an antagonist of XIAP (suppresses caspase-3
activation and cell death) activities and is suppressed in response to
TGF- β stimulation^[127]54 and OFML3 a secreted scaffold protein
important in endothelial to mesenchymal transition (EndMT)^[128]55.
DE analysis of HMEC-1 monolayers residing on soft versus stiff gels led
to the identification of only 7 DEGs (Supplementary Table [129]S2). 6
of them were upregulated on stiff gels and consisted of histone-related
genes, along with a pseudogene and a mitochondrially-encoded
dehydrogenase. The one gene that was upregulated on soft matrices
corresponded to SLC6A6, which encodes a multi-pass membrane protein, a
member of a family of sodium and chloride-ion dependent
transporters^[130]56. Overall, HUVEC express more ECM
stiffness-sensitive genes as compared to HMEC-1, and seem to retain
more EC specific functions and mechanosensitivity according to the
pathway enrichment analysis. Given these findings, we decided to focus
our attention on this specific cell type to further explore if and how
the stiffness-sensitive DEGs we identified are related to each other
and to HUVEC mechanobiology.
RT-qPCR on HUVEC treated with recombinant TGF-β2 confirms changes in
expression of several genes identified to be stiffness-sensitive through RNA
sequencing
One of the ECM stiffness-sensitive genes that we identified to be
upregulated in HUVEC on stiff as opposed to soft gels is TGF-β2.
Interestingly, for certain cell types there is a correlation between
expression of TGF-β2 and enhanced cellular
contractility^[131]57,[132]58 together with deposition of ECM
components^[133]59 that might even contribute to the development of
fibrotic diseases^[134]60. Triggered by our finding and these past
studies, we hypothesized that the differential regulation of many of
the genes expressed by HUVEC on stiff but not on soft matrices could be
primarily due to TGF-β2 upregulation. To assess the plausibility of
this hypothesis, we seeded cells on soft or stiff matrices in
monolayers and treated them for 24 h with either vehicle control, or
1 ng/mL or 10 ng/mL of recombinant TGF-β2. We then performed RT-PCR to
assess which genes are differentially expressed in the presence or
absence of TGF-β2 as compared to controls and whether that depends on
subendothelial stiffness (Supplementary Table [135]S9).
We found that addition for 24 h of recombinant TGF-β2 onto HUVEC
residing on soft matrices led to a 4-fold or to a 5-fold increase in
endogenous TGF-β2 expression for cells residing on soft or stiff
matrices respectively (Fig. [136]5a). Consistent with the idea that
TGF-β2 is upstream of several of the DEGs identified, we additionally
found a 7-fold increase in expression of KRT7, 2-fold increase of
CXCL12, 1.5-fold increase of GADD45B, 1.3-fold increase of INSR and
1.3-fold increase of KFL10 (Fig. [137]5b–f). Also consistent with the
DE analysis, addition of TGF-β2 led to 30% decrease in OFML3 and in 20%
decrease of XAF1 for cells residing on stiff matrices only
(Fig. [138]5k,l). To our surprise STC1, LAMB3, SMOC and ADAMTLS (the
DEGs expressing proteins involved in ECM remodeling) showed decreased
expression compared to cells treated with vehicle control irrespective
of substrate stiffness (Fig. [139]5g–j). STC1 showed 80% decrease in
expression, LAMB3 40% decrease, SMOC1 50% decrease and ADAMTLS a 20%
decrease (Fig. [140]5g–j) as compared to control cells. EPHA5, ZSCAN31,
ZNF703 and ID1 did not show any appreciable changes in gene expression
irrespective of the treatments (Fig. [141]5m–p). These results suggest
that at least 7 of the DEGs identified when comparing HUVEC on stiff
versus soft matrices are regulated downstream of TGF-β2.
Figure 5.
[142]Figure 5
[143]Open in a new tab
TGF-β2 modulates expression of some stiffness-sensitive genes in HUVEC.
(a–p) Relative with respect to GAPDH expression levels of the indicated
stiffness-sensitive DEGs obtained by RT-qPCR. For each boxplot N = 3
replicates are shown for each group treated with either vehicle control
(#1, #4), or treated for 24 h with 1 ng/mL (#2, #5) or 10 ng/mL TGF-β2
(#3, #6). HUVEC either were residing on soft 3 kPa (blue) or stiff 70
kPa (red) matrices and the relative levels of expression in each
treated sample (#2-#4 or #6-#8) are expressed relative to the vehicle
control samples of cells residing on soft (#1) or stiff (#4) matrices
respectively.
HUVEC but not HMEC-1 treated with recombinant TGF-β2 show a significant
elevation in their traction stresses on both stiff and soft matrices
We next asked whether TGF-β2 and its downstream effectors might
contribute to the elevation in the traction stresses that HUVEC display
on stiff as opposed to soft matrices. For that we grew HUVEC in
monolayers on soft 3 kPa or stiff 35 kPa hydrogels and added 1 ng/mL
TGF-β2 or vehicle control for 24 h. For comparison, we also examined if
addition of recombinant TGF-β2 would influence HMEC-1 mechanics by
performing in parallel TFM on this cell type. Consistent with our
previous findings, we found that both HMEC-1 and HUVEC when treated
with vehicle control exert higher traction stresses on stiffer as
opposed to softer matrices (Supplementary Figs. [144]S10 and [145]S11).
When HUVEC were treated with TGF-β2 they displayed an elevation of
their cell-matrix deformations, traction stresses and strain energy as
compared to controls both on soft and stiff matrices (Supplementary
Fig. [146]S10a–c; Supplementary Movies [147]S1 and [148]S2). Addition
of TGF-β2 led to a higher increase in strain energy for HUVEC residing
on stiff (4-fold increase) as opposed to soft (3-fold increase)
matrices (Fig. [149]6). We also found that cells are less dynamic upon
addition of TGF-β2 since the correlation of the deformation maps
between subsequent frames decreased much faster for cells treated with
vehicle control as compared to cells treated with TGF-β2 (Supplementary
Fig. [150]S12). Unlike HUVEC, HMEC-1 did not increase either their
traction stresses or strain energy upon addition of TGF-β2 (Fig. [151]6
and Supplementary Fig. [152]S10a–c). Indeed, when we plotted the mean
strain energy per field of view for multiple recordings and normalized
it for each condition (cell type or stiffness) with the strain energy
of vehicle control, we found an approximately 2 or 3-fold increase in
the strain energy for HUVEC residing on soft or stiff matrices
respectively but no change for HMEC-1 irrespective of stiffness
(Fig. [153]6).
Figure 6.
Figure 6
[154]Open in a new tab
HUVEC but not HMEC-1 treated with recombinant TGF-β2 show a significant
elevation in their traction stresses on both stiff and soft matrices.
Boxplots of the strain energy imparted by confluent EC monolayers
calculated during different instants of time and for multiple fields of
view (N = 200). Boxplots refer to HUVEC or HMEC-1 residing on soft 3
kPa (blue) or stiff 35 kPa matrices (red) treated with vehicle control
or 1 ng/mL TGF-β2 for 24 h prior to imaging. Each boxplot is normalized
with respect to the mean value of the vehicle control case. Two
asterisks denote statistically significant differences between the
medians of two distributions (p < 0.01; Wilcoxon rank sum test).
HUVEC exposed to TGF-β2 show increased amount of F-actin and pMLC
The increase in cell-matrix traction stresses that HUVEC exert upon
addition of TGF-β2 suggest major actin cytoskeletal rearrangements
occurring within the cells or changes in their actomyosin contractility
or both. To explore this possibility, we seeded HUVEC on soft (3 kPa)
and stiff (70 kPa) hydrogels and added recombinant TGF-β2 or vehicle
control for 24 h. Samples were then fixed and stained with phalloidin
to visualize F-actin and DAPI to stain the cells’ nuclei, to ensure
that confluency is similar between different conditions (Supplementary
Fig. [155]S13a,b). Since actomyosin contractility is mediated through
the phosphorylation of myosin regulatory light chain (MLC)^[156]61, we
also used an antibody against pMLC (Thr18/Ser19) to assess potential
changes in myosin activity that could explain the increase in traction
stresses upon addition of TGF-β2. As shown by the representative images
in Fig. [157]7, HUVEC on stiff as opposed to soft gels showed slightly
increased F-actin fluorescence intensity and increased actin stress
fiber formation, reinforcing the idea that actin organization rather
than gene expression is what different in these two distinct mechanical
regimes (Fig. [158]7a,c). In addition, samples of wells treated with
TGF-β2 showed a significant increase both in F-actin fluorescence
intensity and in actin stress fiber formation, particularly on soft but
also on stiff substrates (Fig. [159]7b,d,e). In parallel, we
investigated the effect of TGF-β2 on the phosphorylation of MLC and
found an increase in pMLC fluorescence intensity for HUVEC treated with
TGF-β2 residing on both soft and stiff substrates, similar to the
increase in F-actin (Fig. [160]7b,d,f). In addition, pMLC appeared in
part to co-localize with F-actin stress fibers, suggesting that F-actin
reorganization and actomyosin contractility in response to TGF-β2 for
both substrate stiffnesses might in part explain the increase in
traction stresses for HUVEC. We also immunostained HMEC-1 under the
same conditions but were unable to see any changes in F-actin or pMLC
intensity and localization, upon addition of TGF-β2, consistent with
the absence of changes in the magnitude of traction stresses these
cells exert upon addition of TGF-β2 (Fig. [161]7g,h).
Figure 7.
[162]Figure 7
[163]Open in a new tab
F-actin and pMLC increase for HUVEC monolayers exposed to TGF-β2. (a–d)
Representative images depicting the phase image of cells (first
column), F-actin fluorescence (second column), anti-pMLC antibody
fluorescence (third column) and the image of the nuclei (fourth column)
for HUVEC residing on soft 3 kPa matrices and treated with vehicle
control (a) or 1 ng/mL TGF-β2 for 24 h (b) or HUVEC residing on stiff
70 kPa matrices and treated with vehicle control (c) or 1 ng/mL TGF-β2
for 24 h (d). Scale bar is 50 μm. (e,f) Boxplots showing the integral
of the F-actin fluorescence intensity (e) or pMLC (f) over different
fields of view (N = 10) for HUVEC residing on soft 3 kPa matrices
(blue) and treated with vehicle control or 1 ng/mL TGF-β2 for 24 h and
for HUVEC residing on stiff (red) 70 kPa matrices and treated with
vehicle control or 1 ng/mL TGF-β2 for 24 h. N = 400–500 cells were
analyzed for each condition. One or two asterisks denote statistically
significant differences between the medians of two distributions (<0.05
or <0.01 respectively; Wilcoxon rank sum test). (g,h) Same as panels
e-f but for HMEC-1 cells.
Discussion
In this study we showed for the first time that subendothelial
stiffness has a minimal effect on the transcriptome of HUVEC and HMEC-1
cells when those are seeded in confluence for 24 h on collagen I-coated
hydrogels. The effect of ECM stiffness on the transcriptome of cells
has been examined before for certain cell types and has yielded novel
insight into the importance of stiffness in dictating the cellular
transcriptome^[164]12–[165]14. In hepatocytes 4000 genes were
differentially-regulated depending on ECM stiffness although cell
confluency in this study was not explicitly stated^[166]14. Similarly,
the mesenchymal stem cell transcriptome was found extremely
stiffness-sensitive when these cells are sparsely encapsulated within
alginate gels^[167]12. When vascular smooth muscle cells were placed on
soft or stiff matrices, 2842 stiffness-sensitive genes were identified
common in both cell types corresponding to both protein coding and
lncRNAs^[168]13, although no fold change cutoff was considered for DEG
identification in this study. Interestingly, in most of these studies
the characterized cells were not plated under conditions where they
formed monolayers or they were plated as single cells. Single or
subconfluent ECs placed on matrices of varying stiffness show much more
intense differences in their morphology, cytoskeletal architecture and
mechanics as compared to cells placed in confluency and studies have
shown that cell confluency can even override the effect of matrix
stiffness^[169]5,[170]20,[171]62. In this study we chose to examine ECs
in confluency, as this scenario is more physiologically relevant (i.e.
ECs are not found in vivo as single cells, but in a healthy endothelium
cells are confluent and non-proliferating).
We discovered that both confluent HUVEC and HMEC-1 exert higher
traction forces when residing on stiff as compared to soft substrates
but this change in mechanical behavior does not originate from
differential regulation of gene expression. It is likely that
post-transcriptional regulation might control ECM stiffness-sensitive
responses of ECs^[172]63. In accordance with this argument and past
studies, we found that the phosphorylation level or the conformation of
key mechanosensitive proteins is differentially regulated depending on
subendothelial stiffness^[173]28–[174]30. It could also be that the
changes in traction forces arise due to changes not in the amount of
genes expressed but rather on the localization of the corresponding
proteins. In the past we discovered increased amounts of surface
vimentin for HMEC-1 on stiffer matrices, while the total amount of
vimentin remained the same between conditions^[175]32. Similarly,
herein we found increased actin stress fiber formation for HUVEC on
stiff versus soft matrices although no changes in the expression of the
ACTB gene were detected (Fig. [176]6a,c). Finally, miRNAs, that
typically do not affect gene expression but rather the translation and
stability of mRNAs^[177]64,[178]65, rather than mRNA or lncRNA can play
a critical role in EC mechanosensing and
mechanotransduction^[179]16,[180]17. These mechanosensitive miRNAs can
even contribute in arterial stiffening, fibrosis and
hypertension^[181]16,[182]66, however we did not assess miRNA
expression in this study.
An important conclusion of our study is that, unlike primary HUVEC that
exhibit 24 stiffness-sensitive DEGs when grown on stiff versus soft
matrices, HMEC-1 cells express fewer DEGs, appearing
stiffness-insensitive in a transcriptomic level. However, when
comparing HUVEC to HMEC-1 irrespective of substrate stiffness gene
expression differences are dramatic (Fig. [183]2). Through functional
analysis of DEGs, we found that multiple pathways related to
proliferation and cell cycle regulation are upregulated for HMEC-1,
whereas HUVEC show upregulation of many more pathways related to
endothelial functions^[184]67. This behavior of HMEC-1 might arise from
their immortalized nature since immortalized human EC lines demonstrate
significant differences in their ability to respond to cytokines
compared to primary ECs^[185]21,[186]68. Whether the differences of
primary HUVEC versus HMEC-1 are caused by immortalization procedures of
the latter or just reflect heterogeneity of ECs arising from different
vascular beds could be the focus of future studies.
From the 24 stiffness-sensitive DEGs identified for HUVEC, we decided
to focus our attention on TGF-β2 cytokine for several reasons: (1) It
has been shown to be mechanosensitive. It is secreted in a latent form
on the ECM and can switch into an active conformation by mechanical
force applied by the cells onto their matrix on which it is embedded.
That allows active TGF-β2 to bind to TGF-β receptors eliciting a
cascade of events crucial for EC homeostasis^[187]40–[188]42 but which
also when unregulated can contribute to the development of fibrosis,
cardiovascular disease, or may be usurped during tumor growth^[189]59;
(2) TGF-β2 is also implicated in ECM protein production that is thought
to lead to tissue remodeling and fibrosis characterized by ECM
stiffening^[190]69. Indeed a lot of the DEGs we identified to be
upregulated on stiff matrices are ECM proteins^[191]38,[192]39; (3)
Previous literature suggests direct or indirect relationship with
TGF-β2 for most of the identified
DEGs^[193]38,[194]44–[195]52,[196]54,[197]55; (4) Recently TGF-β2 was
shown to be downregulated in lymphatic ECs seeded on soft as opposed to
stiff matrices^[198]70.
Indeed, we showed that addition of recombinant TGF-β2 increases
endogenous TGF-β2 more for cells residing on stiff and to a lesser
degree for cells residing on soft matrices. Together with the
observation that traction stresses for HUVEC on stiff matrices are
higher than on soft, this result might be suggestive of the cells being
able to mechanically activate more easily the latent TGF-β2 on stiff
matrices which then leads to more endogenous TGF-β2 expression compared
to soft matrices. We also discovered that upon addition of recombinant
TGF-β2 most of the DEGs identified to be up- or down- regulated on
stiff as compared to soft matrices followed similar trends, suggestive
of these genes being downstream of TGF-β2. The only genes that followed
the opposite trend were STC1, LAMB3, SMOC and ADAMTLS which showed
decreased expression compared to cells treated with vehicle control
irrespective of substrate stiffness. We speculate that this result
might be due to the high potentially non-physiological amount of added
TGF-β2 or to the fact that these genes are just not regulated via
TGF-β2. For instance, increased STC1 levels have shown to decrease
TGF-β2 expression and reduce fibrotic effects resulting from
it^[199]71.
We hypothesized that if the increase in traction stresses that HUVEC
exert on stiff matrices is partially due to enhanced TGF-β2 expression,
then addition of recombinant TGF-β2 that enhances endogenous TGF-β2
expression should lead to an increase in the traction stresses that
HUVEC exert. Indeed, upon addition of recombinant TGF-β2 we found a
2-fold and 3-fold increase in the strain energy that HUVEC exert on
soft and on stiff matrices respectively. Activation of extracellular
TGF-β from its latent state to an active state can happen via cells
mechanical pulling the latent complex via their integrin based focal
adhesions^[200]59. Matrix elasticity might also play a role in TGF-β
activation, as shown for myofibroblasts^[201]72. However, it is not
conclusive whether increased matrix stiffness allows cells to grab
stronger thereby mechanically activating TGF-β2 and leading to
subsequent enhancement in production of TGF-β2. Our findings are
suggestive of such a mechanism because addition of recombinant TGF-β2
leads to further increase in endogenous TGF-β2 for cells residing on
stiff as opposed to soft matrices. Consistent with this idea, when we
measure the traction force increase that HUVEC produce on their matrix
upon addition of recombinant TGF-β2 we measure a 3-fold increase in
strain energy, as opposed to a 2-fold increase on softer matrices.
TGF-β2 has been previously implicated in modulating cellular
contractility for certain cell types^[202]73,[203]74. In these studies,
active TGF-β2 by binding to TGF-β receptors elicits a cascade of
downstream signaling either through the canonical Smad-dependent
pathway or through non-canonical pathways in which MAPK and ROCK/RhoA
are involved inducing increased actin polymerization and/or actomyosin
contractility^[204]58,[205]74. Our immunofluorescence imaging of HUVEC
treated with recombinant TGF-β2 are consistent with the idea that
activation of TGF-β2 in HUVEC, signals most probably through the
non-canonical RhoA/ROCK pathway leading to increased actomyosin
contractility and elevated cell-ECM traction stresses. In turn that
appears to lead to further activation of TGF-β2, synthesis of
additional ECM molecules and expression of more TGF-β2. Unlike HUVEC,
HMEC-1 do not respond mechanically to addition of TGF-β2, a discrepancy
which is consistent with TGF-β2 not being a stiffness-sensitive DEG for
HMEC-1. Different EC types are well known to exhibit local
morphological and functional specializations and distinct gene
expression profiles depending on the tissue they originate from^[206]75
and can therefore respond differently to cytokines including
TGF-β2^[207]76. Which EC types are sensitive to TGF-β2 and how much
that depends on the anatomical location they come from or their local
microenvironment, including its mechanics, could be the focus of future
studies.
Materials and Methods
Fabrication of polyacrylamide hydrogels on multi-well plates
Polyacrylamide hydrogel fabrication was done as previously
described^[208]77. Glass-bottom plates with 24 wells (MatTek;
P24G-1.5-13-F) were incubated for 1 h with 500 μL of 1 M NaOH, then
rinsed with distilled water, and incubated with 500 μL of 2%
3-aminopropyltriethoxysilane (Sigma; 919-30-2) in 95% ethanol for
5 min. Following rinsing with water 500 μL of 0.5% glutaraldehyde were
added to each well for 30 min. Wells were rinsed with water again and
dried at 60 °C. To prepare polyacrylamide hydrogels of varying
stiffness, mixtures containing 3–10% acrylamide (Sigma; A4058) and
0.06–0.6% bis-acrylamide (Fisher; BP1404–250) were prepared.
Specifically, 3 kPa hydrogels contained 5% acrylamide and 0.1%
bis-acrylamide, 35 kPa hydrogels contained 8% acrylamide and 0.26%
bis-acrylamide, and 70-kPa hydrogels contained 10% acrylamide and 0.6%
bis-acrylamide^[209]77. For all experiments described herein we chose 3
kPa stiffness for our “SOFT” ECM condition, since it is the lowest
stiffness where we can still attain confluent monolayers. Below 3 kPa
neither cell type (HUVEC or HMEC-1) consistently forms monolayers and
cells often cluster or form vessel like structures with gaps as
previously shown^[210]26,[211]27. For all experiments other than TFM,
we used 70 kPa for our “STIFF” ECM condition given that this stiffness
lies on the upper range of what can be encountered in
vivo^[212]22–[213]24. However, when we performed TFM on cells at 70 kPa
the gels were too stiff to enable cells to deform them so that we could
consistently capture displacements of the beads (embedded into the
hydrogels on which cells are seeded). We therefore used only for TFM 35
kPa hydrogels as “STIFF” matrices since this is the highest stiffness
on which cells are still able to consistently deform the hydrogels. For
each stiffness, two mixtures were prepared, the second of which
contained 0.03% 0.1 μm–diameter fluorescent beads (Invitrogen, F8803)
for TFM experiments.
0.06% APS and 0.43% TEMED were added to the solutions to initiate
polymerization. First, 3.6 µL of the first mixture without the beads
was added at the center of each well, capped with 12-mm untreated
circular glass coverslips, and allowed to polymerize for 20 min. Then
2.4 µL of the mixture containing tracer beads was added, sandwiched
again with a 12-mm untreated circular glass coverslip and allowed to
polymerize for 20 min. Next, 50 mM HEPES at pH 7.5 was added to the
wells, and coverslips were removed. Hydrogels were UV-sterilized for
1 h and then activated by adding 200 µL of 0.5% weight/volume
heterobifunctional cross-linker Sulfo-SANPAH (ProteoChem; c1111) in 1%
dimethyl sulfoxide (DMSO) and 50 mM HEPES, pH 7.5, on the upper surface
of the hydrogels and exposing them to UV light for 10 min. Hydrogels
were washed with 50 mM HEPES at pH 7.5 and were coated with 200 µL of
0.25 mg/ml rat tail collagen I (Sigma-Aldrich; C3867) in 50 mM HEPES at
pH 7.5 overnight at room temperature. Note that for all assays other
than TFM, stiff gels were 70 kPa. For TFM only stiff gels were 35 kPa
because that was determined to be empirically the limit of the ECs
being able to still deform their matrix. On 70 kPa hydrogels we were
unable to detect any matrix deformations (data not shown).
Atomic force microscopy (AFM) for determination of hydrogel stiffness
AFM force-distance measurements were performed on polyacrylamide
hydrogel samples immersed in 50 mM HEPES pH 7.5 buffer with a Park
NX-10 AFM (Park Systems, Santa Clara, CA) using silicon nitride
cantilevers CP-PNP-SiO with a sphere tip (sQube, 0.08 N/m stiffness,
sphere radius ~1 μm) and gold coating on the reflective side.
Temperature was 37 °C throughout the experiment. Tip calibration curves
were performed on a glass surface considered to be infinitely hard for
the soft tips used and two approach-withdraw cycles were performed. The
XEI (Park Systems, Santa Clara, CA) and SPIP softwares (Image
Metrology, Hørsholm, Denmark) were used for data analysis of the FD
curves and for calculation of the gel stiffness.
Cell culture conditions
HMEC-1 (generous gift from the Welch lab, University of California,
Berkeley) were cultured in MCDB 131 medium (Fisher Scientific;
10372–019) supplemented with 10% FBS (GemBio; 900-108), 10 ng/mL
epidermal growth factor (Sigma; E9644), 1 μg/mL hydrocortisone (Sigma;
H0888), and 2 mM l-glutamine (Sigma; 56-85-9). HUVEC (Lonza C2517A)
were cultured according to the manufacturer’s instructions (EGM Bullet
Kit-2, Lonza CC-3162). Passages used were between P4–P8.
Traction force microscopy
TFM assays were performed as previously
described^[214]25,[215]78,[216]79. Two layered polyacrylamide hydrogels
were manufactured as described above. After hydrogel equilibration with
cell media for 30 min at 37 °C, cells were seeded to a concentration of
2 × 10^5 cells per well directly onto the hydrogels 24 h prior to
imaging. Multi-channel time-lapse sequences of fluorescence (to image
the beads within the upper portion of the hydrogels) and phase contrast
images (to image the cells) were acquired using an inverted Nikon
Diaphot 200 with a CCD camera (Andor Technologies) using a 40X Plan
Fluor NA 0.60 objective and the MicroManager software package (Open
Imaging). The microscope was surrounded by a cage incubator (Haison)
maintained at 37 °C and 5% CO[2]. Images were acquired every 10 min for
approximately 8 h. Subsequently, at each time interval we measured the
2D deformation of the substrate at each point using an image
correlation technique similar to particle image velocimetry. We
calculated the local deformation vector by performing image correlation
between each image and an undeformed reference image which we acquired
by adding 10% SDS at the end of each recording to detach the cells from
the hydrogels. We used interrogation windows of 32 × 8 pixels (window
size x window overlap). We calculated the two-dimensional traction
stresses that cell monolayers exert to the hydrogel as described
elsewhere^[217]78,[218]80. We calculated the strain energy (U[s]) as
the mechanical work imparted by the cell to deform its hydrogel:
[MATH:
Us=12∫sτ→(z=h)⋅u→(z=h)ds :MATH]
, where
[MATH: u→ :MATH]
is the measured displacement vector field on the free surface of the
hydrogel and
[MATH:
∫sds
:MATH]
represents a surface integral.
Antibodies and reagents
Hoechst (Thermofisher; D1306) was dissolved at 1 mg/ml in DMSO and used
at 1:1000 to stain nuclei. Recombinant TGF-β2 (Sigma; T2815) was
dissolved in water containing 0.1% BSA at stock concentration 50 μg/mL,
stored at −80 °C and was added to cells for 24 h at either 1 ng/mL or
10 ng/mL. Primary antibody used for staining of pMLC (Cell Signaling;
3674S) was rabbit polyclonal phospho-Myosin Light Chain 2 (Thr18/Ser19)
antibody. For actin staining 0.2 µM AlexaFluor488 phalloidin (Thermo
Fisher; A12379) was used. For western blot analysis, active form of
integrin β1 (Millipore; MAB2079), integrin β1 (Millipore; MAB2053),
phospho-Tyr-397 FAK (Invitrogen; PA5-17084), FAK (Santa Cruz; sc158),
phospho-Vav2 (Santa Cruz; sc16409R), Vav2 (Abcam; AB86699), phospho-ERK
(Cell Signaling; 9109S), ERK2 (Santa Cruz; sc154-G), phospho-p70S6K
(Cell Signaling; 9205S), p70S6K (Cell Signaling; 9202) and GAPDH (Santa
Cruz; sc20538) were used.
Western blotting for HUVEC or HMEC-1 lysates coming from cells in monolayers
residing on different ECM stiffness substrates
To assess phosphorylation and expression levels of different
mechanosensitive proteins^[219]28, cells were seeded at a concentration
of 2 × 10^5 cells/well (24-well plates) on soft 3 kPa or stiff 70 kPa
hydrogels for 24 h, and then lysed with a buffer containing 1% Nonidet
P-40, 0.5% sodium deoxycholate, and a protease inhibitor mixture
(phenylmethylsulfonyl fluoride [PMSF], leupeptin, aprotinin, and sodium
orthovanadate). The total cell lysate was separated by SDS–PAGE (10%
running, 4% stacking) and transferred onto a nitrocellulose membrane
(Immobilon P, 0.45-μm pore size). The membrane was then incubated with
the designated antibodies. Immunodetection was performed using the
Western-Light chemiluminescent detection system (Applied Biosystems).
RNA isolation and RNA sequencing
Sample preparation
4^th passage HUVEC and HMEC-1 cells were placed at a concentration of
2 × 10^5 cells/well on soft 3 kPa hydrogels or stiff 70 kPa hydrogels
(N = 6 replicates for each condition) built on wells of a 24-multi well
plate. Cells residing for 24 h on these gels were then harvested and
lyzed using the QIAshredder Kit (Qiagen; 79656). mRNA was harvested
using the RNeasy Plus Micro Kit (Qiagen; 74004) and eluted in 30 μL
RNAase free water. RNA concentrations were measured using the nanodrop
machine and were comparable between conditions. RNA quality and
quantity were confirmed via bioanalyzer analysis performed by the
Stanford Protein and Nucleic Acid (PAN) Facility.
Library preparation and RNA sequencing
RNA libraries were prepared by the Stanford Functional Genomics
Facility using the KAPA stranded RNA-seq kit with RiboErase (Kit code
KK8483, Roche Cat. # 07962282001), for a fragment length of 200–300 bp.
Sequencing was run on the Illumina NextSeq. 500 System using the
High-Output Kit with 2 × 75 read length. We had an average of 30
million reads per sample. We used 2 lanes with 12 samples each ensuring
that from each condition 3 replicates go to lane A and the other 3 to
lane B (total number of replicates N = 6).
Transcriptome assembly and differentially expressed gene identification
Sequencing reads were obtained in Fastq format and evaluated using
FastQC v0.11.5 according to the directions on the following website:
[220]http://www.bioinformatics.babraham.ac.uk/projects/fastqc/. For
sequence alignment we used HISAT2
([221]https://ccb.jhu.edu/software/hisat2/index.shtml) and reads were
mapped to reference genome grch38 to generate.bam files. Python script
HTSEQ was then used to generate counts per read
([222]https://htseq.readthedocs.io/en/release_0.11.1/count.html). Both
HISAT2 and HTSEQ were run from the command line. For differential gene
expression analysis, we used Deseq2R package (Bioconductor version:
Release 3.8,
[223]https://bioconductor.org/packages/release/bioc/html/DESeq2.html).
The thresholds that we used for identifying DEGS were: (1) DESeq. 2
mean normalized counts >10; (2) padj-value < 0.05; and (3)
lof2foldchange >0^[224]34.
Hierarchical clustering, PCA and gene set analysis
The R package Bioconductor 3.8 and imported the libraries genefilter
and RcolorBrewer to perform hierarchical clustering dendrograms based
on the Euclidean distance of the sample and on the 200 top variance
genes. Principal Component Analysis (PCA) was performed using the
plotPCA function in R. Kyoto Encyclopedia of Genes and Genomes (KEGG)
pathway and Gene Ontology (GO) term analyses of the whole data set of
DEG were performed using the R package GAGE “Generally Acceptable Gene
set Enrichment” (GAGE v.2.22.0) package implemented in R. Briefly,
default parameter settings were used for comparisons of log-scaled gene
set expression (i.e. enrichment) data between different time points
(q-value < 0.01). Gene sets were defined using annotations obtained
from GAGE v2.22.0, go.db v3.2.2, and kegg.db v3.2.2. The R package
“Pathview” v.1.12.0 and KEGGGraph v1.30.0 were used to visualize gene
set expression data in the context of functional pathways.
RT-qPCR
4^th passage HUVEC were placed at a concentration of 2 × 10^5
cells/well on soft 3 kPa hydrogels or stiff 70 kPa polyacrylamide
hydrogels (N = 3 replicates for each condition) built on wells of a
24-multi-well plate. Cells residing for 24 h on these gels were treated
either with vehicle control or recombinant TGF-β2 for 24 h, without the
cells being previously starved. Cells were then harvested and lyzed
using the QIAshredder Kit (Qiagen; 79656). mRNA was harvested using the
RNeasy Plus Micro Kit (Qiagen; 74004) and eluted in 30 μL RNAase free
water. RNA concentrations were measured using the nanodrop machine and
were comparable between conditions. cDNA was prepared using the
Superscript III First-strand Synthesis SuperMix (Invitrogen; 18080085).
RT-qPCR was performed using the SYBR qPCR Master mix by Arraystar Inc.
Genes of interest were amplified using primers indicated in
Supplementary Table [225]S9. Briefly the steps followed were: (1)
Performance of RT-qPCR for each target gene and the housekeeping gene
GAPDH; (2) According to the standard curve, the gene concentration of
each sample is generated directly by Rotor-Gene Real-Time Analysis
Software 6.0; (3) For each sample, the relative amount of the target
gene is determined by calculating the ratio between the concentration
of the target gene and that of GAPDH.
Immunostaining
HMEC-1 and HUVEC cells residing on soft 3 kPa hydrogels or stiff 70 kPa
polyacrylamide hydrogels were incubated for 24 h with 1 ng/mL TGF-β2
without the cells being previously starved. Prior to fixation 1 μg/mL
Hoechst (Thermofisher; D1306) was added in each well to stain the
cells’ nuclei for 10 min. Cells were washed once with PBS and fixed
with 4% formaldehyde of EM grade in PBS for 10 min. Following a wash
with PBS, samples were permeabilized for 5 min in 0.2% Triton X-100 in
PBS and then washed again with PBS. Samples were then blocked for
30 min with 5% BSA in PBS and then incubated with anti-pMLC primary
antibody (Cell Signaling; 3674S) diluted 1:100 in PBS containing 2% BSA
for 1 h. Samples were washed in PBS three times and then incubated with
Alexa Fluor 546 goat anti-rabbit IgG secondary antibody (Invitrogen
A-11035) diluted 1:250 in PBS containing 2% BSA for 1 h and 0.2 µM
AlexaFluor488 phalloidin. Samples were washed three times in PBS and
stored in 1 mL PBS for imaging. N > 500 cells were analyzed per
condition. For imaging, we used an inverted Nikon Diaphot 200 with a
charge-coupled device (CCD) camera (Andor Technologies) and a 60× air
Plan Fluor NA 0.60 or a 100× oil objective. The microscope was
controlled by the MicroManager software package.
Supplementary information
[226]Supplementary Figures and Captions^ (13.9MB, pdf)
[227]Supplementary Movie 1^ (5.3MB, avi)
[228]Supplementary Movie 2^ (5.3MB, avi)
[229]Supplementary Table 1^ (135.6KB, xlsx)
[230]Supplementary Table 2^ (131.1KB, xlsx)
[231]Supplementary Table 3^ (3.7MB, xlsx)
[232]Supplementary Table 4^ (3.7MB, xlsx)
[233]Supplementary Table 5^ (134.5KB, xlsx)
[234]Supplementary Table 6^ (197.1KB, xlsx)
[235]Supplementary Table 7^ (129.7KB, xlsx)
[236]Supplementary Table 8^ (133.6KB, xlsx)
[237]Supplementary Table 9^ (10.4KB, xlsx)
Acknowledgements