Culture of Mesenchymal Stem Cells in a Hydrogel Model of Vocal Fold Lamina Propria

Abstract

Stem cell injection has been proposed as an alternative approach for the restoration of vocal fold (VF) function in patients with VF scarring. To assess the therapeutic efficacy of this treatment strategy, we evaluated the behaviors of human mesenchymal stem cells (hMSCs) in hydrogels derived from thiolated hyaluronic acid (HA-SH) and poly(ethylene glycol) diacrylate (PEG-DA) entrapping assembled collagen fibrils (abbreviated as HPC gels). Three hydrogel formulations with varying amounts of collagen (0, 1 and 2 mg/mL) but a fixed HA-SH (5 mg/mL) and PEG-DA (2 mg/mL) concentration, designated as HPC0, HPC1 and HPC2, were investigated. The HPC gels exhibit similar pore sizes (35–50 nm) and AFM indentation moduli (~175 Pa), although the elastic shear modulus for HPC1 (~32 Pa) is lower than HPC0 and HPC2 (~55 Pa). Although HPC1 and HPC2 gels both promoted the development of an elongated cell morphology, greater cell spreading was observed in HPC2 than in HPC1 by day 7. At the transcript level, cells cultured in HPC1 and HPC2 gels had an increased expression of fibronectin and integrin β1, but a decreased expression of tissue inhibitor of metalloproteinase-1, collagen types I/III and HA synthase-1 when compared to cells cultured in HPC0 gels. Cellular expression of connective tissue growth factor was also elevated in HPC1 and HPC2 cultures. Importantly, the HPC2 hydrogels promoted a signficant up-regulation of matrix metalloproteinase 1, transforming growth factor β1, and epithelial growth factor receptor, indicating an increased tissue turnover. Overall, hMSCs cultured in HPC2 gels adopt a phenotype reminiscent of cells involved in the wound healing process, providing a platform to study the effectiveness of therapeutic stem cell treatments for VF scarring.

1. INTRODUCTION

The vocal fold (VF), comprised of a lamina propria (LP) situated between a stratified squamous epithelium and the vocalis muscle of the larynx, is one of the most dynamically oscillated tissues in the human body. During phonation, VFs experience airflow-induced vibrations at frequencies up to several kHz and reach axial strains of up to 30% [1]. The ability of the VF to undergo large deformations and to sustain repetitive collisions is a result of its unique tissue structure and composition. The LP of the adult human vocal fold can be divided into three layers; the superficial, the intermediate, and the deep layer, with a gradual increase in matrix stiffness from the luminal surface to the muscle. Mechanically, the human VF can be modeled as a body-cover system, with a body that is tension bearing and a top cover that is soft and pliable, supporting propagation of the mucosal wave [2].

Collagen and hyaluronic acid (HA) are the most abundant extracellular matrix (ECM) components of the VF. Mature type I and III collagen fibers, most abundantly distributed in the deep LP and generally oriented along the anterior-posterior direction, provide structural integrity to the tissue and contribute to the tensile properties of the VF [3]. Collagen fibers within the superficial layer of the LP are thin and sparsely distributed. HA, a non-sulfated glycosaminoglycan most enriched in the intermediate layer, serves as filler or lubricant. The highly negatively charged macromolecules recruit water to the tissue, contributing to the maintenance of optimal tissue viscosity to facilitate phonation and an optimal tissue stiffness that may be important for fundamental frequency control [4]. HA has been proposed to act as a shock absorber for the tissue during normal phonation [4] and a promoter of scarless wound healing in fetal VFs [5].

Vocal fold scarring is a significant clinical issue and remains a therapeutic challenge in the field of laryngology. VFs can be damaged by chemical insults, such as pollutant particles and refluxed stomach acids. Mechanical factors such as trauma introduced during intubation or stress induced by vocal abuse and overuse can result in VF scarring [6]. Following injury, resident cells produce excessive collagen that is structurally disorganized. As a result, tissue viscoelasticity is altered and mucosal wave propagation is hampered. Stem cell-based [7] regenerative therapy has been implicated in mediating VF wound healing and fibrosis. In vivo studies using a rabbit model show that treatment of scarred VFs with bone marrow or adipose-derived mesenchymal stem cells enhances VF regeneration, reduces the thickness of the LP, decreases collagen type I content, and restores tissue viscoelasticity. Stem cell transplantation has also resulted in the upregulation of fibroblast growth factor 2 (FGF2), hepatocyte growth factor (HGF) and hyaluronan synthases (HAS1, 2 and 3) [8].

Mechanistically, it is not clear how injected stem cells interact locally with the ECM of the lamina propria and how such interactions affect stem cell behaviors to improve VF function. Considering the abundance of HA and collagen in the VF tissue and their roles in maintaining VF function, a semi-synthetic matrix composed of HA and collagen would be an optimal tissue-mimetic scaffold for evaluating the potential of stem cell therapy in the treatment of VF scarring. Because the emergence of a trilayer LP during development is demarcated mainly by varying collagen content, comparing the behaviors of stem cells cultured in hydrogels with varying collagen content will shed light on the applicability of stem cell therapy toward individuals with varying tissue composition.

Herein, hydrogels were prepared under physiological conditions using thiolated HA (HA-SH, 5 mg/mL), poly(ethylene glycol) diacrylate (PEG-DA, 2 mg/mL) and type I collagen (0, 1 and 2 mg/mL). Hydrogel fibril distribution, swelling, pore size, and mechanical properties were evaluated. Human bone marrow derived MSCs were encapsulated in the HPC gels for up to 13 days. Matrix composition-dependent cell metabolic activity and morphology were analyzed. Relative expression of VF-relevant ECM proteins and wound healing-related signaling molecules was analyzed using quantitative polymerase chain reaction (qPCR). Overall, the HPC hydrogel platform containing VF ECM-relevant constituents allows for the evaluation of how changes in local collagen concentration during the wound healing process may affect injected therapeutic cells.

2. MATERIALS AND METHODS

2.1. Chemicals and Reagents.

Hyaluronic acid (HA, sodium salt, 500 kDa) was generously donated by Sanofi Genzyme (Cambridge, MA). 3,3’-Dithiobispropanoic acid (DTPA), hydrazine hydrate, 1-ethyl-3-(3-(dimethylamino)propyl) carbodiimide hydrochloride (EDC) and dithiothreitol (DTT) were purchased from Sigma Aldrich (Milwaukee, WI). Poly(ethylene glycol) diacrylate (PEG-DA, 3,500 Da) was obtained from JenKem Technologies (Plano, TX). Rat tail collagen type I was purchased as a 0.92 wt% solution in acetic acid from Thermo Fisher Scientific (Waltham, MA). TRIzol, calcein AM, ethidium homodimer, 4’,6-diamidino-2-phenylindole, dihydrochloride (DAPI), Alexa Fluor 568-conjugated phalloidin, and PrestoBlue cell viability reagent were purchased from Life Technologies (Grand Island, NY). hMSCs and hMSC growth media (MSCGM) were obtained from Lonza (Walkersville, MD).

2.2. Synthesis of Thiolated HA (HA-SH).

HA-SH was synthesized as described previously [9–11]. Briefly, 500 kDa HA was reacted with 3,3’-dithiobis-propanoic dihydrazide, synthesized using DTPA, ethanol and hydrazine, in water with pH adjusted to 4.75 and catalyzed by EDC, followed by reduction with DTT at pH 8.0. The solid product, HA-SH, was obtained at 80% yield with a 40% thiol incorporation based on ¹H NMR (Figure S1). For cell culture purposes, the dialyzed solutions were sterilized using a 0.22 μm filter prior to lyophilization. Following lyophilization, solid HA-SH was stored at −20 °C until use.

2.3. Preparation of HPC gels.

A 2 wt% solution of HA-SH and a 0.8 wt% solution of PEG-DA were prepared separately in phosphate-buffered saline (PBS). The stock collagen solution was diluted to 0.46 wt% in 0.2% acetic acid before further dilution with PBS to 0.40 and 0.20 wt%, depending on the final collagen concentration in the gels. All gel components were kept ice cold until crosslinking. After adjusting pH to ~8, HA-SH was mixed with collagen and PEG-DA at a volume ratio of 1:2:1. A control gel was created by replacing the collagen volume with PBS. In all gel formulations, HA-SH and PEG-DA concentrations were maintained at 5 and 2 mg/mL, respectively. Gels containing 0, 1, and 2 mg/mL of collagen were referred to as HPC0, HPC1, and HPC2, respectively. Subsequently, 100 μL of the mixture was immediately pipetted into glass bottom petri dishes (Mattek, Ashland, MA) and was allowed to crosslink for 20 min.

2.4. Confocal Reflectance Microscopy (CRM).

Gels were inspected using a Zeiss LSM 710 confocal microscope with a C-Apochromat 40x water objective (numerical aperture = 1.2). Collagen fibrils were visualized by detecting reflected light at 405 nm. Collagen gels, prepared following the procedure above with the exclusion of HA-SH and PEGDA at a collagen concentration of 0.5, 1 and 2 mg/mL, were included for comparison purposes. Reflectance images were then imported into Fiji [12] (version 1.50e) for analysis. The Auto Threshold plugin was used to threshold each image such that only collagen fibrils were visible then the area measurement function was selected to record fibril coverage.

2.5. Nanoindentation.

The indentation elastic modulus of HPC gels was determined via AFM nanoindentation using a Bruker Bioscope Catalyst in PeakForce quantitative nanoscale mechanical (QNM) characterization mode. Silicon nitride cantilever tips with a 25-μm polystyrene (PS) spherical probe were purchased from Novascan (Boone, IA). The tip used had a deflection sensitivity of 74.7 nm/V and a spring constant of 0.06 N/m. Gels were created in 60 mm petri dishes with 1 mm spacers on which a Teflon-coated coverslip rested to create a flat surface for probing. The AFM tip was lowered to the surface of each gel and indented until a trigger threshold of 10 nN was reached then retracted while the force was determined by tip deflection. Areas of 1 mm were surveyed at 10 μm intervals, producing 100 data points per region. Three regions were measured from two samples of each gel type. Force-separation curves were imported into the NanoScope Analysis (Bruker) software for curve-fitting. AFM indentation elastic modulus was determined using the Derjaguin-Muller-Toporov (DMT) model of contact mechanics and is shown as a histogram. Prior to the measurement of HPC gels, a standard agarose gel of known modulus (15 kPa) was probed to verify accuracy of the indentation method.

2.6. Oscillatory Rheology.

The viscoelastic properties of the HPC gels were measured using a stress-controlled rheometer (AR-G2, TA Instruments, New Castle, DE) with a 20 mm parallel plate geometry. Hydrogel mixture was added to the bottom plate and the top plate was lowered to a gap size of 100 μm before applying mineral oil around the sample circumference to minimize water evaporation. A strain sweep was carried out at a frequency of 6 rad/s at strains of 0.1% to 100% and a frequency sweep was conducted at 1% strain from 0.1–10 Hz. All measurements were performed at 37 °C in triplicate and the average storage (G’) and loss (G”) moduli are reported.

2.7. Cell Culture.

Primary hMSCs were cultured in Celltreat (Shirley, MA) 182 cm² cell culture flasks at an initial seeding density of 4,000–5,000 cells/cm². Cells were maintained in MSCGM and were kept in a humidified incubator at 37 °C and 5% CO₂, with media changes every three days. After 7–8 days (80–90% confluence) cells were lifted via accutase and counted, then pelleted for resuspension in HA-SH. HA-SH containing suspended hMSCs was mixed with collagen and PEG-DA as described above to produce cell-laden constructs with a final cell density of 1×10⁶ cells/mL. The cell-gel mixture was transferred to a glass dish and MSCGM (2 mL) was added 20 min after mixing. Constructs were incubated at 37 °C for 13 days and the medium was changed every 2 days.

2.8. Cell Viability.

On days 1, 3, 5, 7, 9, 11, and 13, the cell culture medium was aspirated and gels were washed briefly with PBS. Two milliliters of the PrestoBlue viability reagent (diluted in MSCGM at a 1:9 volume ratio) were added to each gel. Samples were incubated at 37 °C for 2 h. After 300 μL of supernatant from each sample were transferred to a 96-well plate. Samples were excited at 570 nm and fluorescence was measured at 610 nm. Fluorescence readings over subsequent days were normalized to each sample’s mean day 1 value. Separately, cell viability was analyzed by LIVE/Dead staining. After 7 days of culture, media was replaced with a buffer containing the LIVE/DEAD dye (Calcein AM and ethidium homodimer, 4 μM each in PBS). Following 15-min incubation at 37 °C, constructs were removed, the LIVE/DEAD dye was aspirated, and gels were washed twice with PBS before imaging using a confocal laser scanning microscope (Zeiss LSM 710) with an EC Plan-Neofluar 10x water objective (0.3 N.A.) and excitation wavelengths of 488 nm and 561 nm.

2.9. Cell Morphology.

After 0, 5, and 7 days of culture, media was aspirated and cells were fixed with 4% paraformaldehyde in PBS for 1 h at 4 °C, then the constructs were rinsed with PBST (0.05% Tween-20 in PBS). Next, gels were blocked and permeabilized in a 3% BSA/0.1% Triton X-100 mixture in PBS for 45 min followed by three PBST rinses. Gels were then incubated with phalloidin conjugated to Alexa Fluor 568 (diluted 1:400 in PBS) for 1 h before washing three times with PBST. Finally, samples were treated with DAPI (diluted 1:500) for 5 min. After additional PBS wash (3 times), samples were transferred to a confocal laser scanning microscope for imaging using a C-Apochromat 40x water objective with a numerical aperture of 1.2. Excitation wavelengths for DAPI and Alexa Fluor 568-phalloidin were 405 nm and 560 nm, respectively. To quantify the cell morphology z-stacks were obtained for each sample over an 850 × 850 μm² area which was 200–400 μm thick. The focal plane was adjusted such that there was a 4 μm vertical increment between each image. Z-stacks were then compiled in a maximum intensity projection. Morphological analysis was performed by importing images to Fiji, thresholding, and creating a region of interest (ROI) for each cell in the z-stack. These ROIs were measured to determine the area of each cell.

2.10. Gene Expression.

After 3 and 7 days of culture, samples were snap-frozen in a dry ice/isopropanol slurry and stored at −80 °C for analysis by real-time quantitative polymerase chain reaction (qPCR). RNA extracted via TRIzol digestion from snap-frozen constructs was reverse-transcribed to cDNA using a QuantiTect reverse transcription kit (Qiagen, Valencia, CA). Templates (4 ng) and primers (400 nM) were then combined with 10 μL Power SYBR green master mix (Invitrogen, Carlsbad, CA) before qPCR was carried out using an ABI 7300 real-time sequence detection system. GAPDH was used as a reference target. Primers (Table 1) were purchased from Integrated DNA Technologies (Coralville, IA) and resulting qPCR data were normalized and processed using commercially available qbase+ software (Biogazelle, Zwijnaarde, Belgium).

Table 1.

Summary of Quantitative Polymerase Chain Reaction Primer Information.

Gene	Forward Primer(5′−3′)	Reverse Primer (5′−3′)	GeneBank#	Efficiency	Product Size (bp)
GAPDH	GAAATCCCATCACCATCTTCCAGG	GAGCCCCAGCCTTCTCCATG	NM_001289746	2.08	120
FN	ACCTACGGATGACTCGTGCTTTGA	CAAAGCCTAAGCACTGGCACAACA	NM_001306132	2.10	116
ITGB1	CTACCAACACGCCCTTCATT	ATGTGAATGCCAAAGCGAAG	XM_005252448	2.03	105
MMP1	GGGAGATCATCGGGACAACTC	GGGCCTGGTTGAAAAGCAT	NM_001145938	2.03	72
TIMP1	TTTCTTGGTTCCCCAGAATG	CAGAGCTGCAGAGCAACAAG	NG_012533	1.90	99
COL3A1	TGGTGCCCCTGGTCCTTGCT	TACGGGGCAAAACCGCCAGC	NM_000090	2.03	87
COL1A1	AATGGTGCTCCTGGTATTGCTGGT	ACCAGTGTCTCCTTTGCTGCCA	XM_005257058	2.10	141
HAS1	GTGAGTGGCTGTACAACGCG	AGAGGGACGTAGTTAGCGGC	NM_001523	1.91	355
CTGF	AGGAGTGGGTGTGTGACGA	CCAGGCAGTTGGCTCTAATC	NM_001901.2	1.90	117
HGF	TCCAGAGGTACGCTACGAAGTCT	CCCATTGCAGGTCATGCAT	MM_001010932.2	1.95	70
EGFR	GGCAGGAGTCATGGGAGAA	GCGATGGACGGGATCTTAG	NM_005228.3	1.81	153
VEGFA	GACAAGAAAATCCCTGTGGGC	AACGCGAGTCTGTGTTTTTGC	NM_001287044.1	1.91	102
TGFB1	GCAGAAGTTGGCATGGTAGC	CCCTGGACACCAACTATTGC	XM_011527242.1	1.99	131

Gene abbreviations: GAPDH, glyceraldehyde-3-phosphate dehydrogenase; MMP1, matrix metalloproteinase 1; ITGBL integrin beta 1; COL1A1, procollagen I alpha 1 chain; COL3AL collagen III alpha 1 chain; TIMP1, tissue inhibitor of metalloproteinase 1; HAS1, hyaluronic acid synthase 1; CTGF, connective tissue growth factor; HGF, hepatocyte growth factor; EGFR, epithelial growth factor receptor; VEGFA, vascular endothelial growth factor A; TGFB1, transforming growth factor beta 1.

2.11. Statistical Analysis.

All quantitative analyses were conducted on data sets with n≥3. Data sets were compared using one-way analysis of variance (ANOVA) with Tukey-Kramer post-hoc (p≤0.05 considered significant) unless where indicated. Results are presented as mean ± standard error.

3. RESULTS

3.1. Characterization of HPC gels.

HPC gels were created under physiological conditions by Michael-type addition reaction between HA-SH and PEG-DA at a thiol to acrylate ratio of 4:1, while monomeric collagen underwent fibrilization (Figure 1A). Confocal reflectance microscopy (Figure 1B) enabled direct visualization of the collagen fibrils. Collagen gels assembled in the absence of the covalent network, designated as C0.5, C1 and C2, were relatively heterogeneous and exhibited substantial fibril aggregation (Figure 1B, i-iii). In contrast, collagen fibrils in the HPC gels were thinner and more evenly dispersed throughout the field of view (Figure 1B, iv-vi). Collagen fibrils were randomly oriented with no demonstrated anisotropy. While increasing the concentration of collagen in each gel type significantly increased the total fibril area (Figure S2), there was no noticeable effect on overall homogeneity. Due to fibril aggregation, the area in C0.5 is significantly lower than in HPC0.5 (p=0.003).

Figure 1: Synthesis of HPC gels using HA-SH, PEG-DA and Collagen monomer (A) and visualization of the collagen network (B).

(A) Collagen fibrillation occurs in a gelling medium consisting of thiolated HA (HA-SH) and poly(ethylene glycol) diacrylate (PEG-DA). The thiol:acrylate molar ratio was 4:1 and the collagen concentration varied from 0, 1, to 2 mg/mL to produce HPC0, HPC1, and HPC2 gels respectively. Not drawn to scale. (B) Confocal reflectance images of HPC gels as compared with pure collagen gels at the same collagen concentrations. (i-iii): Pure collagen gels at a collagen concentration of 0.5 (C0.5), 1 (C1), and 2 (C2) mg/mL. (iv-vi): HPC gels containing 0.5 (HPC0.5), 1 (HPC1), and 2 (HPC2) mg/mL collagen. Scale bar: 50 μm.

A nanoparticle retention experiment was carried out to estimate the average pore size of HPC gels (Figure S3A). Near 100% retention was observed for PEG-AuNPs with diameters greater than 35 nm. On the other hand, less than 10% of 35-nm particles were retained in the gels following 48 h incubation. As the particle size approached the average pore size of the hydrogel network, particle diffusion through the matrix and into the supernatant became significantly restricted. Because the particles are pegylated, minimal particle-matrix interaction is anticipated. Thus, the average pore size of HPC gels was less than 50 nm, irrespective of collagen content. The addition of collagen to the HA/PEG network led to a progressive reduction in gel swelling (Figure S3B). The average swelling ratio for HPC2 (71 ± 5) is half of the value measured for HPC0 (142 ± 39).

The mechanical properties of HPC gels were characterized by AFM nanoindentation (Figure 2A, Figure S4) and oscillatory shear rheometry (Figure 2B-D). For nanoindentation studies, a cantilever tip with 25 μm PS bead attached was used to probe the local microenvironment. The overall indentation depth was less than 5 μm, ensuring that indentation was no more than 10% of total thickness to remain within the linear elastic range, at the same time avoiding the effects of sample pressurization against the glass substrate. The force-separation curves (Figure S4) obtained by AFM nanoindentation of three different areas on the surface of two gels of each collagen concentration (HPC0, HPC1, and HPC2) were fitted. Change in voltage was used to determine tip deflection based upon cantilever sensitivity. Each 1 mm x 1 mm area was probed 100 times at 10 μm intervals and from the measured slope of the force-separation curves, F, indentation elastic modulus E was obtained from the following equation (1) [13, 14]:

$$F=\frac{4}{3}\frac{E}{\left(1-v^2\right)}\sqrt{R}{\delta }^{\raisebox{1ex}{$3$}\!\left/ \!\raisebox{-1ex}{$2$}\right.}$$ (1)

where R represents the probe radius, v is the Poisson’s ratio of the gels, assumed to be 0.5, [15] and is the indentation depth. The mean indentation elastic moduli of HPC0, HPC1, and HPC2 gels were 163 ± 8, 184 ± 7, and 177 ± 7 Pa, respectively. There was no significant change in modulus with changing collagen concentration. Data were plotted in a histogram to study the distribution of moduli in each sample. A majority (>90%) of the measured values were below 500 Pa. The measured modulus of the 2% agarose standard was 15 kPa, in good agreement with previous results from indentation performed on similar gels [16].

Figure 2. Characterization of hydrogel mechanical properties by AFM (A) and oscillatory shear rheometry (B-D).

(A) Histogram plot showing the frequency of each modulus as determined by curve-fitting analysis using AFM nanoindentation with a 25 μm PS bead attached to the AFM tip. (B-D): Rheological properties analyzed using AR G2 rheometer with a parallel plate geometry. Dotted lines represent a Gaussian fit. (B): Strain sweep; (C) Frequency sweep; (D) Dynamic viscosity (η’) as a function of frequency. *Significantly different from HPC0 and HPC2, p<0.05.

The viscoelastic properties of the HPC gels were analyzed by oscillatory rheometry. Strain sweep experiments (Figure 2B) on fully crosslinked HPC gels show a constant storage modulus (G’) independent of shear strain from 0.1% to 100%. G’ values recorded from the frequency sweep (0.2% strain) at frequencies of 0.1–10 Hz were 56.6 ± 2.4 Pa, 32.1 ± 1.8 Pa, and 50.6 ± 4.2 Pa for HPC0, HPC1, and HPC2 gels, respectively (Figure 2C). The elastic modulus of the HPC1 gels was statistically lower than that for HPC0 and HPC2 (p=0.0009). In all cases, G’ is independent of frequency and was significantly higher than the loss modulus (0.30 ± 0.03 Pa, 0.69 ± 0.30 Pa, and 2.66 ± 0.77 Pa for HPC0, 1, and 2, respectively), indicative of an elastic solid. The calculated damping ratio, tanδ = G”/G’, was 0.0054, 0.0208, and 0.0493 for HPC0, 1 and 2, respectively. The dynamic viscosity (η = G”/ϖ, Figure 2D) decreases linearly with frequency in log-log scale. The relationship between dynamic viscosity and frequency can be modeled by a power law: log η = log k + n log f, where f is frequency in Hz, and k and n are constants. The data-fitting results are summarized in Table 2 along with those obtained for human VF mucosa [17].

Table 2:

Summary of Dynamic Viscosity Log-Log Curve Fitting Analysis.

Sample	k(Pa s2)	n	R2
HPC0	0.3030	−0.8337	0.898
HPC1	0.8430	−0.8332	0.981
HPC2	2.8060	−0.9162	0.996
Male (n = 10) [17]	3.1440	−0.8755	0.999
Female (n = 5) [17]	0.7038	−0.8345	0.996

3.2. Cell Viability and Cell Morphology

A PrestoBlue assay was conducted to monitor the metabolic activity of hMSCs cultured in HPC gels over 13 days (Figure 3A). A moderate decrease in cell metabolism was observed for all three types of gels during the initial 5 days of culture when compared to the respective day 1 controls. From day 7 onwards, cells cultured in collagen containing gels (HPC1 and 2) started to metabolize substantially whereas those in HPC0 gels remained metabolically inactive. By day 9 and day 13, cellular metabolism detected from HPC0 constructs was 57% and 74% of the day 1 levels, respectively. Contrarily, hMSCs in HPC1 and HPC2 gels continued to metabolize up to day 13, with a 1.4-fold increase over day 1 when the culture was terminated. No discernable difference was observed between HPC1 and HPC2. Live/dead staining (Figure 3B-D) at day 7 revealed a similar percentage of dead cells for HPC0 (26.8 ± 0.2%), HPC1 (23.8 ± 0.8%), and HPC2 (22.7 ± 3.3%) gels. A similar cell density was seen at all collagen concentrations. Moreover, cells in HPC1 and HPC2 gels had developed elongated processes, adopting a spindle-shaped morphology, whereas those in HPC0 remained rounded.

Figure 3: Characterization of cell metabolism and cell viability.

(A): Overall metabolic activity of hMSCs in HPC gels measured by PrestoBlue assay as a function of culture time. *: statistically significant, p<0.05. (B-D): Cell viability as assessed by LIVE/DEAD staining in HPC0 (B), HPC1 (C) and HPC2 (D) gels after 7 days of culture. Live cells were stained green with calcein AM and dead cells were stained red by ethidium homodimer. Scale bar: 100 μm.

Next, cytoskeletal structure and cell morphology were characterized by confocal imaging after cells were stained with phalloidin (Figure 4A-I). Combined fluorescence and reflectance microscopy (Figure 4J-L) was used to visualize the interaction between the entrapped hMSCs and the surrounding collagen network. Cell morphology was further quantified by computing the area of each cell (Figure 5) based on maximum intensity projections. Day 0 samples contained rounded cells with areas below 500 μm² regardless of collagen concentration. Although some thin hair-like projections (white arrows, Figure 4C) were seen in HPC0 constructs on days 5 and 7, cells showed no bilateral extension. In HPC1 gels, cells took on a stellate morphology, extending projections radially on day 5. Cell area reached an average of 881 μm² on day 5 then extended further by day 7 with sparse, thin, and peripherally located stress fibers. Cell area increased significantly by day 7 to 958 μm². Within HPC2 cultures, cells had extended long, thin, branching projections (white arrow heads, Figure 4H) by day 5 and reached an average area of 725 μm². These projections continued to broaden on day 7, showing clearly defined stress fibers in a single direction and increasing overall cell area to 1426 μm². In some areas, cells form a continuous network with their processes. Cells in HPC1 and HPC2 gels interact with the assembled collagen fibers intimately, accumulating collagen fibers around the cell periphery particularly in HPC2 gels (Figure 4L).

Figure 4. Characterization of cell morphology and cell matrix interaction.

(A-I) Confocal fluorescence images of hMSCs stained with Phalloidin (red) and DAPI (blue) after 0, 5 and 7 days of culture in HPC0 (A-C), HPC1 (D-F) and HPC2 (G-I) gels. Scale bar: 50 μm. (J-L): Combined confocal fluorescence and reflectance images of hMSCs cultured in HPC0 (J), HPC1 (K) and HPC2 (L) gels for 7 days. Cells were stained with Phalloidin (red) and DAPI (blue). Collagen fibrils are shown in greyscale. Scale bar: 50 μm.

Figure 5: Characterization of hMSC spreading in HPC gels.

Cells were cultured in HPC0 (A), HPC1 (B) and HPC2 (C) gels for 0, 5 and 7 days before being stained with phalloidin and DAPI. Morphological analysis was performed by importing confocal images to Fiji (version 1.50e) thresholding and creating a region of interest (ROI) for each cell in the z-stack. These ROIs were measured to compute the area of each cell. Dotted lines represent Gaussian fits. *: statistically significant compared to day 0 (p<0.05); ^#: significantly different compared to day 5.

3.3. Gene Expression of ECM Proteins and Cytokines.

Finally, qPCR analyses were performed to identify the effect of matrix composition on hMSC functions. Gene expression was normalized to HPC0 levels and was represented as the log₂ transform so that downregulation was scaled equivalently. Figure 6A shows the changes in expression of various essential VF ECM components following 7 days of culture. First, the expression of crucial ECM proteins involved with matrix breakdown; matrix metalloproteinase 1 (MMP1) and tissue inhibitor of metalloproteinase 1 (TIMP1), were assessed. Compared to the HPC0 gels, cells in HPC2 gels significantly increased expression of MMP1 by 14.0 ± 4.5-fold (p=0.003). Conversely, the expression of TIMP1 decreased by 9.8 ± 3.0 and 4.7 ± 3.7-fold in HPC1 and HPC2 gels, respectively (p=0.006 and 0.025). In addition, proteins involved in VF ECM production and interaction were analyzed. These included collagen type I alpha 1 chain (COL1A1), collagen type III alpha 1 chain (COL3A1), hyaluronic acid synthase 1 (HAS1), fibronectin (FN), and integrin beta 1 (ITGβ1). For COL1A1, the expression was significantly downregulated in HPC2 gels. Day 7 expression of COL1A1 in HPC1 and HPC2 gels was 3.5 ± 0.6 and 2.8 ± 1.0-fold lower, respectively (p=0.016 and 0.026). For COL3A1 a similar trend was uncovered, where HPC1 and HPC2 gels downregulated expression by 3.5 ± 1.0 and 3.3 ± 0.9-fold (p=0.020 and 0.035). Similarly, in HPC1 and HPC2 gels HAS1 expression was downregulated significantly 5.7 ± 0.8 and 5.9 ± 0.5-fold (p=0.010). Although not significant, the expression of fibronectin in HPC1 and HPC2 gels was 2.6 ± 0.9 and 4.0 ± 1.5-fold greater than the HPC0 control. Finally, the expression of ITGβ1 was moderately upregulated by cells in HPC2 gels by 1.6 ± 0.1-fold (p=0.015). Markers of differentiation towards different tissue fates were also analyzed. Chosen genes included aggrecan (AGN), alkaline phosphatase (ALP), runt-related transcription factor 2 (RUNX2), fatty acid-binding protein 4 (FABP4), and adipocyte protein 2 (aP2). However, the concentration of these markers was below the detectable limit of qPCR, and was not altered by changing collagen concentration (data not shown).

Figure 6: Effects of hydrogel composition on cellular expression of ECM proteins (A) and wound healing related cytokines (B).

Changes in gene expression by cells cultured in HPC1 and HPC2 gels were normalized to cells encapsulated in HPC0 gels. Cells were cultured for 7 (A) and 3 (B) days before the mRNA was isolated for qPCR analysis. FN: fibronectin; ITGβ1: integrin β1; MMP1: matrix metalloproteinase-1; TIMP1: tissue inhibitor of metalloproteinase-1; COL3: collagen type III; COL1A1: procollagen type 1; CTGF: connective tissue growth factor; HGF: hepatocyte growth factor; EGFR: epithelial growth factor receptor; VEGFA: vascular endothelial cell growth factor A; TGFβ1: transforming growth factor β1. *: significantly different from HPC0 (p<0.05); ^#: significantly different compared to HPC1 (p<0.05).

qPCR analysis after 3 days of culture showed that connective tissue growth factor (CTGF) was upregulated on day 3 in both HPC1 and HPC2 gels by 1.3 ± 0.1 and 1.5 ± 0.1 fold respectively (p=0.045 and 0.001). Epidermal growth factor receptor (EGFR) was only significantly upregulated in the HPC2 gels, by a factor of 1.38 ± 0.05 (p=0.005). A similar trend was seen for transforming growth factor beta-1 (TGFβ1), where HPC2 cultured cells had 1.8 ± 0.1 times the expression of target sequence compared to HPC0 (p=0.0002). Cellular expression of hepatocyte growth factor (HGF) was low and was not significantly different between HPC cultures. The expression of vascular endothelial cell growth factor A (VEGFA) was moderate, but the difference was not significant.

4. DISCUSSION

Stem cell injection has been proposed as a promising approach for the treatment of VF scarring. Reported studies on the usage of hMSCs in VF regeneration involve the direct injection of hMSCs to unilaterally damaged VFs, followed by morphological and mechanical characterization of the treated tissue. In some cases, gene analysis was conducted on tissue samples. The damaged VF undergoes substantial restructuring, with resident fibroblasts and myofibroblasts depositing excessive amount of disorganized collagen [18, 19]. At various stages of tissue repair, the local matrix composition can vary greatly. Moreover, the composition and viscoelastic properties of the ECM vary across the layered structure of the lamina propria and are gender and age-dependent [20]. Thus, hMSCs introduced to different patients or different LP layers are expected to have a varied response to the local microenvironment [7]. Using HPC gels of varying collagen concentration, we aim to understand how matrix properties affect cell function.

HA and collagen were chosen as the major components of the HPC gels because of their abundance in the native tissue and their contribution to the maintenance of VF mechanical properties. While HA is known to promote scarless wound healing, excessive collagen deposition is a hallmark of scarred VFs. Additionally, the breakdown of collagen fibers during catabolic events has been shown to provide signals which can promote cell expansion and proliferation [21]. Rat tail collagen was used because it is readily available in bulk quantities with minimal batch-to-batch variation. It is structurally and compositionally similar to human collagen, and has proven utility in generating human skin equivalents [22, 23]. Mixing of HA-SH, collagen monomer and PEG-DA at pH 8 and ambient temperature initiated the formation of a covalent network through Michael-type addition reaction between thiol- and acrylate-moieties[24, 25] as well as the assembly and entanglement of collagen fibrils. CRM revealed a homogenous distribution of collagen fibrils within both the HPC1 and HPC2 gels, likely due to decreased kinetics of collagen fibrilization in the gelling liquid, thereby preventing the formation of larger and more heterogeneous fiber bundles commonly seen in pure collagen gels. Collagen fibrils had random orientation, which is indicative of a scar morphology [26]. In HPC2 gels, the network showed signs of interconnectivity while still not being completely percolated.

The HPC gels are composed of interstitial, amorphous gels derived from HA-SH and PEG-DA interpenetrated with relatively immature collagen fibrils. The hydrogel pore size, estimated by particle retention to be in the range of 35–50 nm, is dominated by the covalent network. On the other hand, the addition of collagen to the covalent network at 2 mg/mL decreased the gel swelling ratio by roughly half. Since swelling ratio is partially a measure of fluid flow into the matrix, it is dependent on permeability. Lower permeability associated with collagen deposition in scar tissue is implicated in poor wound healing outcomes, [27] thus the addition of collagen to HPC gels is capable of more closely approximating a scar phenotype.

During phonation, the VF LP experiences tensile, compressive and shear stresses [28]. Thus, the mechanical properties of the HPC gels were evaluated by AFM nanoindentation and oscillatory shear rheometry. Because the PS indenter is significantly larger than the size of individual collagen fibrils in the HPC gels (estimated at ~50 μm long, 1–2 μm wide), the measurement reflects the composite nature of the HPC gels [29]. All three types of HPC gels are soft and the elastic modulus is independent of collagen content. Since collagen fibrils were not chemically modified in any way so as to covalently conjugate them to the HA/PEG matrix, they are able to move freely when indented, sliding past the HA/PEG component with little resistance and adding little to no compressive resistance. Additionally, measurements were made while the gels were hydrated, and the fact that the HPC0 gels swelled more than the gels containing collagen could have caused the measured modulus to increase.

When probed macroscopically under oscillatory shear, the mechanical properties of HPC gels reflect the contribution of a fibrillar network. Collagen-free, HA/PEG-based covalent networks had an average G’ value of 57.6 ± 0.7 Pa. The decrease in G’ from HPC0 to HPC1 is likely a result of collagen fibrils inhibiting the Michael-type addition reaction between HA and PEG, reducing overall crosslinking efficiency and weakening the network. At a concentration of 1 mg/mL, the collagen fibrils have not reached a significant entanglement to contribute to the mechanical integrity of the composite network. When the concentration of collagen was increased to 2 mg/mL, the collagen network contributed a greater portion of shear strength, causing the overall modulus to increase back to HPC0 levels. The measured G’ of HPC gels falls within previous shear measurements done on the human VF cover, which generally fall between 0.01–1 kPa [17]. The low tanδ value measured for HPC0 reflects its covalent crosslinking. The HPC1 and HPC2 gels are significantly more damping than HPC0, owing to the presence of physically entrapped collagen fibrils. The measured damping ratio for porcine and human VFs is ~0.4. The dynamic viscosity of HPC gels decreased with frequency at approximately the same rate as that of human VF mucosa, with an average n value of −0.8610. The average k value is 0.3030, 0.8430, and 2.8060 Pa∙s² for HPC0, 1 and 2, respectively. Reported k values for female and male VF mucosa are 0.7038 and 3.1440, respectively.

Cells such as hMSCs cultured in HPC gels can potentially interact with the matrix through CD44/RHAMM or integrin [30]. These interactions, however, are not sufficient to promote cell elongation, [31] as evidenced by the rounded cell morphology seen in HPC0 gels. On the other hand, integrin receptors directly link collagen in the ECM to the cytoskeleton through β1’s intercellular tail, making this subunit a potent regulator of cell traction forces and mechanosensation [32]. In our case, the incorporation of collagen fibrils in the HA/PEG network increased cell metabolism and promoted cell spreading in 3D. Although HPC1 gels are significantly softer than HPC2 as determined by oscillatory rheometry, this difference was not sufficient to promote changes in cell metabolic activity, viability, or gene expression of ECM components.

The combined CRM and LSM images confirm the ability of cells in HPC1 and HPC2 gels to interact with the collagen fibrils and to drive reorganization of the local fibrillar structure [33]. Although cells can spread in both HPC1 and HPC2 gels, a longer culture time is needed for cells to fully elongate in HPC1 gels than HPC2 gels. By day 5, cells in HPC1 were still radially probing their environment, displaying a stellate-like morphology, whereas those in HPC2 were spindle-shaped and had already extended along a single axis. The higher ligand density, combined with the higher expression of ITGβ1 in HPC2 gels, promotes a faster cell/matrix interaction leading to more rapid elongation. The greater interconnectivity of the collagen fibril network in HPC2 gels enables hMSCs to extend their processes over a longer distance. Cell/matrix interactions through integrin receptors have been shown to activate the RhoA pathway, contributing to increased proliferation and metabolic activity [34]. Moreover, the breakdown of collagen fibrils provides signals to promote cell expansion and proliferation [21]. The integrin upregulation observed in our study suggests that cell response to changing collagen concentrations is integrin mediated and that the shift towards a more catabolic phenotype is likely initiated via integrin binding.

With the introduction of type I collagen the expression of VF relevant ECM proteins, in particular COL1, COL3, and HAS1, were downregulated. Comparable changes in protein production have been seen in several animal models of VF scarring. Tissue collagen content was found to decrease following injection of hMSCs into scarred rabbit VFs [35]. A decrease in HA synthesis and expression after introducing hMSCs into scarred rabbit VFs was also observed in multiple studies [35], in agreement with our qPCR results. A common collagenase, MMP1, was found to be was upregulated in HPC2 gels. Matrix metalloproteinases are commonly found to be upregulated during the wound healing response in MSCs, and have been implicated in turnover of fibrotic tissue in VF fibroblasts [36]. Upregulation of MMP1 along with downregulation of its inhibitor TIMP1, as well as COL1, COL3, and HAS1, indicate that cells are moving towards a net catabolic state in the HPC2 gels [37].

Changes in expression of common growth factors involved in the wound healing response were determined through qPCR. In HPC2 gels, both EGFR and TGFβ1 were upregulated significantly. During the initial VF wound healing process in vivo, EGF, EGFR, and TGFβ1 are all upregulated [38]. Both EGF and EGFR are known regulators of cell proliferation as well as ECM production and increased EGFR expression has been shown to increase cell responsiveness to environmental stimuli [39]. Interestingly, EGFR interaction with extracellular ligands and subsequent activation by mechanical stretch has been shown to induce tissue damage and inflammation in cells of fibroblastic lineage [40]. The observed TGFβ1 upregulation is also indicative of a scarring phenotype. This growth factor has long been known to become upregulated in the presence of inflammatory cytokines and its presence is associated with scar formation [41]. Studies of the effects of MSCs on hypertrophic scar formation have shown anti-scarring effects including down-regulation of TGFβ1 post-injection [42]. Other growth factors assayed, VEGFA and HGF, showed no significant change. This could mean that either those factors are not affected by collagen binding or that expression was time-dependent.

Our investigation showed that hMSC functions are dependent on collagen concentration, and at the highest collagen concentration (2 mg/mL) investigated herein, HPC2 gels elicit a wound healing cell response most similar to hMSCs injected into a scarred VF. The local ECM composition must be taken into consideration when delivering therapeutic stem cells such as hMSCs. Considering the fact that HA and collagen concentration is gender, age, layer, and disease dependent, hMSCs injected into the VF will respond differently from patient to patient. For example, since collagen deposition is increased 3–5 days following injury [43], an injection of hMSCs at that time could correspond with a more pronounced wound healing response. Currently, cells injected into the scarred VF environment have low survivability and the potential to become myofibroblastic. Observed therapeutic efficacy has been attributed to paracrine signaling from the surviving hMSC population. If high cell survival can be achieved, hMSC injection is therapeutically attractive. The availability of a VF LP-mimetic matrix in vitro will allow for the identification of conditions which are beneficial to therapeutic strategies.

In the current investigation, hMSCs were cultured in HPC gels under static conditions in the absence of high frequency vibrations. This is clinically relevant as vocal rest is highly recommended after certain VF injuries or surgeries to promote proper healing. Typically, patients are advised to have 4–8 days of voice rest after treatment [44]. The addition of vibratory stimulations at phonation frequencies will likely further alter hMSC behaviors. It is recognized that fibroblasts and macrophages in the VF play significant roles in maintaining tissue homeostasis and in modulating tissue response to injuries [45, 46]. Because our goal was to assess the effect of matrix composition on hMSC function, tissue-resident cells were not included in our model. Our on-going effort is dedicated to the development of a hydrogel-based cellular model of the human vocal fold, consisting of a LP-mimetic layer with dispersed fibroblasts and macrophages and a stratified squamous epithelium. The engineered tissue model can be used as a reliable platform for studying vocal fold physiology and pathology, as well as for the screening and testing of new treatment options.

5. CONCLUSION

Composite HA/collagen gels were developed using HA-SH, PEG-DA and collagen to approximate the ECM of the vocal fold lamina propria. The covalent HA/PEG network imparted mechanical strength to the gels while the assembled collagen fibrils provide anisotropic features and integrin binding sites to promote cell attachment and mechanosensing. The addition of collagen enhanced metabolic activity, downregulated the expression of COL1A1, COL3A1, HAS1 and TIMP1, but upregulated ITGβ1, MMP1, CTGP, EGFR and TGFβ1. At the highest collagen concentration, a shift towards a catabolic phenotype was observed. These studies show that incorporation of collagen fibrils in an amorphous HA matrix at a concentration of 2 mg/mL promoted the commitment of hMSCs towards a wound healing phenotype. We are currently studying the response of these cell-gel constructs to vibration at physiologically-relevant frequencies.