Single-cell analysis reveals IGF-1 potentiation of inhibition of the TGF-β/Smad pathway of fibrosis in human keratocytes in vitro

Corneal wound healing is often affected by TGF-β–mediated fibrosis and scar formation. Guided fibrosis with IGF-1 and antifibrotic substances might maintain corneal transparency. Primary human corneal keratocytes under serum-free conditions were used as a model of corneal stromal wounding, with markers of corneal fibrosis and opacity studied under TGF-β2 stimulation. Single-cell imaging flow cytometry was used to determine nuclearization of Smad3, and intracellular fluorescence intensity of Smad7 and the corneal crystallin aldehyde dehydrogenase 3A1. Extracellular matrix proteoglycans keratocan and biglycan were quantified using ELISAs. On the TGF-β2 background, the keratocytes were treated with IGF-1, and suberoylanilidehydroxamic acid (SAHA) or halofuginone ± IGF-1. IGF-1 alone decreased Smad3 nuclearization and increased aldehyde dehydrogenase 3A1 expression, with favorable extracellular matrix proteoglycan composition. SAHA induced higher Smad7 levels and inhibited translocation of Smad3 to the nucleus, also when combined with IGF-1. Immunofluorescence showed that myofibroblast transdifferentiation is attenuated and appearance of fibroblasts is favored by IGF-1 alone and in combination with the antifibrotic substances. The TGF-β/Smad pathway of fibrosis and opacity was inhibited by IGF-1, and further with SAHA in particular, and with halofuginone. IGF-1 is thus a valid aid to antifibrotic treatment, with potential for effective and transparent corneal wound healing.


IGF-1 alone and in combination with SAHA or halofuginone inhibits nuclearization of Smad3 and modulates Smad7 levels.
For the imaging flow cytometry analysis of Smad3 translocation, we ran an algorithm for co-localization of nuclear and Smad3 probes in the individual naïve human keratocytes treated and analyzed here. As shown in Fig. 1a,b, 70.8% of these keratocytes that were treated with 10 ng/ml TGF-β 2 alone (i.e., as positive control; Fig. 1a, CON+ ) promoted translocation of Smad3 to the nucleus, in contrast to the 17.5% translocation seen for the naïve keratocytes (i.e., as negative control; Fig. 1a, CON− ). On this background of 10 ng/ml TGF-β 2, treatment of the keratocytes with only 10 ng/ml IGF-1 led to a 30.0% loss in the Smad3 nuclearization compared to the TGF-β 2 positive control (40.8% vs. 70.8%, respectively), indicating the anti-TGF-β 2 activity of IGF-1 alone (Fig. 1a, IGF-1). The TGF-β 2-treated keratocytes were also treated with 10 nM SAHA alone and in combination with 10 ng/ml IGF-1 (Fig. 1a, SAHA, combo SAHA), which showed the lowest levels of Smad3 nuclearization, at 6.6% and 9.7%, respectively; i.e., significantly lower than both the TGF-β 2 positive control and the negative control (p < 0.001). The TGF-β 2-treated keratocytes were also treated with 5 ng/ml halofuginone alone and in combination with 10 ng/ml IGF-1 (Fig. 1a, HAL, combo HAL), which showed similar levels of Smad3 nuclearization to the negative control (22.6%, 22.9%, respectively), and were significantly lower than the TGF-β 2 positive control (p < 0.001).
This imaging flow cytometry analysis of Smad3 nuclearization thus demonstrated high activity for the TGF-β /Smad pathway in the TGF-β 2 positive control, and clear inhibition in these treatment groups. Immunoblotting studies of human and canine corneal fibroblasts 40,41 have recently shown SAHA-mediated inhibition of Smad2/3 phosphorylation, which might be the main mechanism for the lower nuclearization of Smad3 observed with these treated keratocytes. Indeed, Smad3 nuclearization was notably decreased by halofuginone, in agreement with a report from Nelson et al. 45 .
Halofuginone, on the other hand, did not show any significant increase in the intracellular levels of Smad7 compared to the negative control, which is not in agreement with the report of Saika et al. 47 .

IGF-1 and SAHA, alone and in combination, regulate the secretion of keratocan and biglycan.
Compared to the TGF-β 2-treated keratocytes of the positive control (Fig, 4a, CON+ ), the keratocan levels in the culture supernatants were significantly higher for all of the further additions ( Fig. 4a; p < 0.001), except for the treatments that included halofuginone ( Fig. 4a, HAL, combo HAL). On the other hand, the biglycan levels were significantly reduced compared to the TGF-β 2-treated keratocytes under all of these conditions (Fig. 4b, p < 0.001), except for the IGF-1 plus halofuginone combination (Fig. 4b, combo HAL). Biglycan is normally not present in the corneal stroma, and it is seen only for a disorganized ECM; i.e., in a corneal scar. However, along with other proteoglycans, keratocan is responsible for precise collagen fibril alignment in the normal cornea 7 . For the TGF-β 2-treated keratocytes there was a 38.0% ± 4.3% decrease in keratocan secretion and a 45.5% ± 7.2% increase in biglycan secretion compared to the negative control. The profiles of these proteoglycans were mainly influenced by SAHA addition to these TGF-β 2-treated keratocytes. Here, 10 nM and 5 nM SAHA were used alone and in the combination with 10 ng/ml IGF-1 ( Fig. 4a,b, SAHA, combo SAHA), which resulted in increases for keratocan secretion of 83.7% ± 11.8%, 48.5% ± 4.7%, and 48.3% ± 1.9%, respectively, and in decreases for biglycan secretion of 79.4% ± 6.6%, 86.6% ± 1.8%, and 63.4% ± 10.6%, respectively, compared to the TGF-β 2 positive control. The TGF-β 2-treated keratocytes treated with SAHA in combination with IFG-1 had similar expression profiles of these proteoglycans to the negative-control keratocytes ( Fig. 4a,b, combo SAHA, CON− ).

Inhibition of keratocyte transdifferentiation into myofibroblasts. These human keratocyte cultures
contained dendritic cells that have little cytoplasm, and that stain positive for keratocan and negative for α -SMA fibers. Under the stimulation with TGF-β 2, these keratocytes transdifferentiated into myofibroblasts, with a gain in cell volume and the formation of α -SMA stress fibers (Fig. 5a,b and Supplementary Fig. S1).
We used confocal microscopy to simultaneously analyze the expression of Thy-1 and the formation of α -SMA fiber bundles, which allowed us to roughly differentiate between keratocytes, wound-type corneal fibroblasts, and corneal myofibroblasts. The TGF-β 2-treated keratocytes of the positive control showed Thy-1-positive cells with α -SMA stress fibers, namely myofibroblasts (Fig. 5, CON+ ), while in the negative-control naïve human keratocytes there was very little Thy-1 labeling and only the diffuse presence of α -SMA in the cells (Fig. 5, CON− ). This diffuse α -SMA corresponded to the observations of unstructured actin in keratocytes reported previously 52 . When SAHA or halofuginone were added to the TGF-β 2-treated keratocytes in the wound model, the myofibroblastic phenotype was less common, although the cells were more spindle shaped and larger, compared to the negative control (Fig. 5, SAHA, HAL). Thy-1 was very prevalent in the TGF-β 2-treated keratocytes with SAHA, and less so with the halofuginone treatments, where the α -SMA stress fibers were less prominent than in the TGF-β 2-treated keratocytes of the positive control. In the presence of Thy-1, this indicates fibroblastic transdifferentiation, as opposed to myofibroblastic transdifferentiation 12 .
The combination of IGF-1 plus halofuginone, but not SAHA, increases keratocyte proliferation.
The proliferative activity of the TGF-β 2-treated keratocytes was determined after 48 h of treatment, using the crystal violet assay (Fig. 6, CON+ ). This positive control was not significantly diminished in the keratocytes treated with SAHA and halofuginone, as also for the negative control (Fig. 6); indeed, the TGF-β 2-treated keratocytes that were also treated with 5 ng/ml halofuginone alone and in combination with 10 ng/ml IGF-1 showed enhanced proliferation (Fig. 6, HAL, combo HAL). The combination of 10 nM SAHA with 10 ng/ml IGF-1 showed proliferation rates that were not significantly different from that of the TGF-β 2 positive control (Fig. 6, combo SAHA, CON+ ). Both concentrations of IGF-1 used alone (Fig. 6, 5 ng/ml, 10 ng/ml) with the TGF-β 2treated keratocytes demonstrated significantly increased proliferation compared to both the TGF-β 2 positive control and the negative control (p < 0.001), with no significant difference seen between these IGF-1 concentrations.

Discussion
Keratocytes have a unique role in vision, as they sustain the corneal stroma and maintain corneal transparency, and they have can provide injury repair, and return the corneal stroma to its non-opaque normalcy. The loss of stromal transparency during wound healing has been linked to myofibroblast transdifferentiation, and thereby to errant ECM production and increased intracellular light scattering. The in-vitro data from the present study indicate that these reference antifibrotic compounds that are approved for human use (i.e., SAHA and halofuginone) can provide beneficial influence over the corneal stromal healing process, especially when combined with IGF-1. They thus favor a non-scarring healing pathway, while propagating elements of transparency.
The canonical Smad signaling pathway of TGF-β 2 was successfully inhibited by SAHA and halofuginone in the present study. IGF-1 alone decreased Smad3 nuclearization, which is a novel finding for human keratocytes. The combination of IGF-1 with these well-established antifibrotics, SAHA and halofuginone, yielded more pronounced block of the TGF-β 2-promoted translocation. Inhibition of Smad3 phosphorylation might be the main mechanism behind this effect 40 . Smad7 is an inhibitory feedback transducer, and as such, it also inhibits phosphorylation of Smad3, as has been shown in corneal wounding models 31,40,47 . According to the present study, SAHA appears to have effects through a separate mechanism, by which it causes higher levels of Smad7 even though it decreases the transcriptional activity of TGF-β 2. It has been reported that proteasome-mediated degradation of Smad7 depends upon a balance between acetylation, deacetylation, and ubiquitination 53 . HDACs tend to deacetylase Smad7 and to decrease its stability, thus channeling it toward ubiquitination and degradation 54 . We suggest that as an HDAC inhibitor, SAHA disrupts the activity of specific HDACs, thereby preventing the degradation of Smad7 in the cytosol, and apparently promoting greater inhibition of fibrosis through the TGF-β /Smad pathway (see schematic representation in Fig. 7).
This study of Smad3 and Smad7 has provided relevant insight into TGF-β /Smad signaling, and it has shown that the antifibrotic compounds investigated (i.e., SAHA and halofuginone) inhibit this pathway to a significant extent. Furthermore, it has demonstrated that addition of IGF-1 inhibits translocation of Smad3 to the nucleus in its own right, with these effects potentiated when combined with SAHA (Fig. 7). This is the first study to show that IGF-1 can inhibit Smad3 translocation to the nucleus in human keratocytes. However, we did not investigate directly other intracellular pathways here, which can also propagate some of the TGF-β signals, such as mitogen-activated protein kinase 55 and the Rho family proteins, which might take part in this system. The imaging flow cytometry analysis performed here was applied to large numbers of cells (n = 3000), where each individual cell was taken into account. Thus, rather than using a generalization based upon population means for gene expression or proteome investigation, the events of Smad3 nuclearization in each individual cell were viewed and analyzed; i.e., using a single-cell approach. According to the literature, the present study is the first to use imaging flow cytometry for the characterization of intracellular events in keratocytes.
The IDEAS analysis software was used here to measure the fluorescence of the specific antibodies that were bound to the target proteins, with the calculation of the median fluorescence intensities, and thus the relative levels of these target proteins (see Methods). All of the keratocytes were prepared under the same conditions and were measured with identical machine calibration in a single day. The weakness of quantitative analysis of immunofluorescence using flow cytometry is the wide dispersion of the data obtained, which here were carefully weighted with the calculation of the median fluorescence intensities and the use of box and whisker plots. This approach for the examination of cell cultures at the subcellular level using imaging flow cytometry analysis as the first experimental step for the validation of a therapy was recently verified 56,57 .
While with flow cytometry the native shape of cells is lost, confocal fluorescence microscopy provided the proof-of-concept here, as an additional insight into the differentiation of these keratocytes. Hassell and Birk 11 reported that wound-type fibroblasts have the most important role in regenerative wound healing in vivo, while myofibroblasts lead to erroneous ECM formation and scarring. In the present study, IGF-1 and the selected antifibrotics prevented this transdifferentiation to myofibroblasts, with higher proportions of Thy-1-positive wound-type fibroblasts in this cell-culture model of corneal stromal wounding, which would favor the regenerative pathway of wound healing 11,12 .
A crystallin (i.e., ALDH3A1) and two proteoglycans (i.e., keratocan, biglycan) were studied here, which have been shown to be representative of intracellular and ECM transparency, respectively 8,11 . Differences in the crystallin levels were shown under the different stimulation conditions, where the lowest levels were seen for the TGF-β 2-treated keratocytes of the positive control, with a prevalence of myofibroblasts in the immunofluorescent microscopy. All of the treatments inhibited this TGF-β 2-promoted depletion of ALDH3A1 in the cytosol to some extent, although significant inhibition was only achieved by IGF-1 alone and in combination with SAHA. The levels of the crystallin in the other treated keratocytes were significantly lower compared to the negative control, which confirms previous findings that the expression of ALDH3A1 is lower in transdifferentiated fibroblasts than in native keratocytes 11,58 . Furthermore, the keratocytes in the ELISA for the two proteoglycans saw the secretion of the highest levels of keratocan and the lowest levels of biglycan under the treatments with SAHA and IGF-1 and their combination, which indicates that under these conditions the ECM should appear more transparent in vivo.
The secretion of keratocan was significantly increased by IGF-1 alone, which has also been reported for lumican 13 , both of which are significant structural proteoglycans in terms of a transparent collagenous ECM. Furthermore, IGF-1 promoted a beneficial increase in cytosolic ALDH3A1, which to the best of our knowledge has not been reported in previous studies. The decreased TGF-β 2-promoted Smad3 nuclearization that was seen here for IGF-1 was previously suggested in studies of muscle fibrosis 59,60 and prostate cancer 61 , but to date this has not been shown for human keratocytes. Similar to previous in-vitro studies in corneal wound healing 51 , the confocal microscopy in the present study indicated that TGF-β 2-promoted transdifferentiation to myofibroblasts is less likely to occur with the addition of IGF-1. As demonstrated by the proliferation assays here, the TGF-β 2-promoted proliferation rates under IGF-1 treatment of the keratocytes was significantly increased, which corresponds to a report of exogenous IGF use on human corneal fibroblasts 51 . We can deduce here that IGF-1 alone can decrease the probability of corneal stromal scar formation, while exerting its well-known positive effects on  53 , blocking Smad3 phosphorylation and nuclearization. SAHA can also inhibit HDACs, which leads to acetylation and subsequent ubiquitination of Smad7 54 , resulting in higher intracellular Smad7 levels, as observed in the present study. P, phosphate group; UB, ubiquitin; HDAC, histone deacetylase.
Scientific RepoRts | 6:34373 | DOI: 10.1038/srep34373 keratocyte proliferation and wound repair. Hence, IGF-1 is not only a valid aid to inhibitors of the TGF-β /Smad pathway in keratocytes, but it can act alone as a potent modulator of fibrosis.
Although our study indicates the need for further investigations and in-vivo validation to confirm these data, it appears that IGF-1 supplemented with SAHA and halofuginone can be advocated for anti-fibrotic therapy of corneal wounding.

Methods
Isolation and cultivation of primary human corneal keratocytes. The human keratocytes used in this study were isolated and cultured in serum-free medium according to a modified procedure described by Jester et al. 62 and Pei et al. 10 . Briefly, corneal tissue was harvested from donor corneas that are normally discarded during the standard surgical procedure for perforative keratoplasty. Written informed consent for perforative keratoplasty was obtained. The Slovenian National Ethics Committee approved the collection of these tissue samples (No. 132/11/13, 20.12.2013), furthermore, our study was done in accordance with the Declaration of Helsinki and the Oviedo Convention. The corneal epithelium and Descemet's membrane with the endothelium were scraped off using a hockey knife, while the remaining stroma was stored in sterile advanced Dulbecco's modified Eagle's medium (DMEM; Life Technologies Ltd, Paisley, UK). The tissue was cut into blocks and incubated at 37 °C overnight in 2.0 mg/ml collagenase type I (Sigma-Aldrich, Grand Island, USA) diluted in DMEM. The digested material was collected and centrifuged in a 50-ml tube at 200 × g for 5 min, to obtain the pellet of primary human keratocytes. Some of these cell pellets were frozen in liquid nitrogen and stored at − 80 °C until analysis. These isolated cells were washed in DMEM and cultured under serum-free conditions with the addition of 100 U/ml penicillin, 1 mg/ml streptomycin, and 2 mM L-glutamine. The cells were resuspended and transferred into collagen-coated tissue culture flasks for further cultivation at 37 °C and in an atmosphere of 5% CO 2 . After 48 h, the non-adherent cells were removed and the adherent cells were cultured further. At the time of the experiments, these human keratocytes had reached 3-6 passages.
TGF-β2 activation and treatment with IGF-1 and the antifibrotics. The human keratocytes were cultured in serum-free medium as described above (with the negative control as the untreated human keratocytes). First, the upper concentrations of the reference bioactives were set according to the 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) cell-viability assay (see Supplementary Fig. 2). A reductionist corneal stromal wounding model was initiated by addition of 10 ng/ml TGF-β 2 15 (R&D systems, Minneapolis, USA) into the DMEM for 48 h (as the positive control). These 'activated keratocytes' were also treated with the different bioactives, which were added at the time of the 'wounding' , with parallel incubation of the wound model for 48 h, due to the fast transient nature of these keratocytes. The bioactives used were: 10 ng/ml, 5 ng/ml recombinant IGF-1 (R&D systems, Minneapolis, USA), 10 ng/ml, 5 ng/ml halofuginone (Santa Cruz Biotechnology, Heidelberg, Germany), and 10 nM, 5 nM SAHA (Tocris Bioscience, Bristol, UK). The 10 ng/ml IGF-1 was also tested in combination with one concentration of each of the antifibrotics: 10 ng/ml halofuginone, and 10 nM SAHA. After completion of the first proliferative assays, the concentrations of halofuginone and SAHA used for the further experiments were chosen as 5 ng/ml and 10 nM, respectively. These human keratocytes were also used for further testing, as described below.
Analysis of the TGF-β/Smad signaling pathway using imaging flow cytometry. To quantitatively determine Smad7 expression and the rate of Smad3 nuclear translocation, these human keratocytes were analyzed using multispectral imaging flow cytometry (ImageStreamX; Amnis Coorporation, Seattle, USA). The keratocytes were harvested, followed by their permeabilization, with the following antibodies used for the staining, according to manufacturer protocols: mouse anti-Smad7 or rabbit anti-Smad3 antibodies, with Cy5 goat anti-mouse or goat AlexaFluor ® 488 anti-rabbit secondary antibodies (Abcam, Cambridge, UK), and 7AAD (BD Pharminogen) to visualize the nucleus. Cell images were acquired for 3000 events per sample at 40x magnification, using 488 nm and 658 nm lasers lines, with the fluorescence collected using two spectral detection channels. For double-stained keratocytes, two single-stained controls were used to compensate for the fluorescence between the channel images. The cell images were analyzed using the IDEAS image-analysis software (Amnis). First, cells within the focal plane were selected using a plot of the image contrast versus the root-mean-squared gradient. Then the aspect ratio versus the cell area was gated to isolate populations of single cells on a bivariate plot 57 . The Smad7 probe median fluorescence intensities were determined for the relative intracellular quantification. The Smad3 nuclear translocation was calculated with a similarity algorithm (similarity dilate), where the software compared the location of the Smad3 versus the location of the nucleus. The subpopulation of keratocytes in which Smad3 translocation occurred were calculated and expressed as percentages 56 .
Measurement of intracellular ALDH3A1 using imaging flow cytometry. The human keratocytes were harvested, permeabilized, and treated with a rabbit anti-ALDH3A1 antibody labeled with a AlexaFluor ® 488 goat anti-rabbit antibody (Abcam), according to the manufacturer protocols. 7AAD (BD Pharminogen, Heidelberg, Germany) was used for nuclear staining. The keratocyte images were obtained using imaging flow cytometry and the IDEAS software, as described above. For the relative quantification, the median fluorescence intensities of ALDH3A1 were determined.
Evaluation of keratocan and biglycan secretion using enzyme-linked immunosorbent assays. The human keratocytes were seeded at 10,000 cells/well in 96-well plates. They were treated as described above, and after 48 h the levels of keratocan and biglycan in the supernatants were determined using ELISA kits (Invitrogen Co., Camarillo, USA, and USC Life Science Inc., Wuhan, China), based on the manufacturer protocols. All of these assays were performed in triplicate.
Scientific RepoRts | 6:34373 | DOI: 10.1038/srep34373 Confocal fluorescence microscopy. The human keratocytes were immunostained with antibodies against Thy-1 and α -SMA, as previously described 12 . Briefly, the keratocytes were grown on collagen-coated 8-well glass chamber slides for 24 h until the respective treatments, as described above. After 48 h, they were fixed in 2% paraformaldehyde in phosphate-buffered saline (pH 7.4). They were then permeabilized and reacted with a PE/ Cy5-conjugated anti-human Thy-1 antibody (Abcam) and a FITC-conjugated anti-human α -SMA antibody (Abcam), according to manufacturer protocols. The samples were counter-stained for the nucleus, and mounted with Fluoroshield TM with DAPI (Sigma-Aldrich). These keratocytes were then evaluated using laser-scanning confocal microscopy (Leica TCS SP5 II; Leica Microsystems, Mannheim, Germany).

Cell proliferation.
To perform the cell proliferation assays, the human keratocytes were seeded at 10,000 cells/well in 96-well plates, at a concentration of 30 viable cells per well. After adding TGF-β 2 and the various treatments, as described above, the cells were stained with crystal violet and the absorbance was measured at 595 nm. All of these cell proliferation assays were carried out in triplicate.
Statistical analysis. The data were collected and analyzed using Statistical Package for Social Sciences (SPSS), version 16. The data are expressed as the means ± SD for the parametric tests, and median ± interquartile range for non-parametric tests. Data normality was analyzed using Shapiro-Wilk tests. Comparison of groups were performed using one-way ANOVA for multiple comparisons, followed by Holm-Sidak post-hoc analysis or Mann-Whitney rank sum tests for comparisons of medians. Chi 2 tests were used to analyze the significance of the Smad3 nuclear translocation. The differences were considered statistically significant for p < 0.001, with a 95% confidence interval.