Recombinant lysyl oxidase effects on embryonic tendon cell phenotype and behavior

Lysyl oxidase (LOX) plays an important role in the elaboration of tendon mechanical properties during embryonic development by mediating enzymatic collagen crosslinking. We previously showed recombinant LOX (rLOX) treatment of developing tendon significantly increased LOX‐mediated collagen crosslink density to enhance tendon mechanical properties at different stages of tissue formation. Working toward the future development of rLOX‐based therapeutic strategies to enhance mechanical properties of tendons that are compromised, such as after injury or due to abnormal development, this study characterized the direct effects of rLOX treatment on embryonic tendon cells from different stages of tissue formation. Tendon cell morphology, proliferation rate, proliferative capacity, and metabolic activity were not affected by rLOX treatment. Tenogenic phenotype was stable with rLOX treatment, reflected by no change in cell morphology or tendon marker messenger RNA (mRNA) levels assessed by reverse‐transcription polymerase chain reaction. Collagen mRNA levels also remained constant. Matrix metalloproteinase‐9 expression levels were downregulated in later stage tendon cells, but not in earlier stage cells, whereas enzyme activity levels were undetected. Bone morphogenetic protein‐1 (BMP‐1) expression was upregulated in earlier stage tendon cells, but not in later stage cells. Furthermore, BMP‐1 activity was unchanged when intracellular LOX enzyme activity levels were upregulated in both stage cells, suggesting exogenous rLOX may have entered the cells. Based on our data, rLOX treatment had minimal effects on tendon cell phenotype and behaviors. These findings will inform future development of LOX‐focused treatments to enhance tendon mechanical properties without adverse effects on tendon cell phenotype and behaviors.


| INTRODUCTION
Tendon is an important connective tissue that transmits forces from muscle to bone to enable musculoskeletal movements.
[4] Tendon mechanical properties are also significantly compromised in tendon-related heritable connective tissue disorders and birth abnormalities, 5,6 with tendon-related birth defects affecting more than 1 out of 100 live births. 7Considering the frequency at which these debilitating conditions occur, there is a dire need for approaches to recover normal tendon mechanical properties for abnormal fetal and adult tendons.
We recently discovered that treating embryonic tendons with recombinant lysyl oxidase (rLOX) during different stages of development increases elastic modulus and tensile strength by increasing LOXmediated collagen crosslinking. 8Furthermore, higher concentrations of rLOX led to a plateau in elastic modulus that correlated with a saturation in LOX-mediated collagen crosslinking, with elastic modulus and crosslink density exhibiting a linear correlation (r 2 = 0.92, p = 0.041). 8This work was inspired by previous findings that LOX naturally regulates embryonic tendon mechanical properties via LOXmediated crosslinking of collagen. 9,10Inhibition of LOX activity during tendon formation prevented new LOX-mediated collagen crosslinks from forming, which in turn abrogated increases in elastic modulus. 9,10rthermore, elastic modulus exhibited high correlations with LOXmediated collagen crosslink density (r 2 = 0.8, p < 0.0001) under normal and perturbed conditions. 10We also found that elastic modulus also correlates highly with LOX activity levels (r 2 = 0.97, p = 0.016). 11][10] Taken together, the apparent ability for embryonic tendon to intrinsically prevent overstiffening of the tissue via saturation of LOX-mediated collagen crosslinking highlights a distinct advantage to developing a therapy inspired by how tendons naturally use LOX to regulate mechanical properties during tissue formation.
Based on our previous studies, an approach to use rLOX treatment to improve tendon mechanical properties seems feasible and is attractive.3][14][15][16][17] For example, osteoblast marker gene expression levels and mineral module formation were significantly inhibited in mouse calvaria cells when LOX activity was inhibited. 15,16While not previously examined in tendon, rLOX effects on tendon markers should be assessed to determine its effects on phenotype stability, and because tendon markers could influence tendon mechanical properties.In particular, knockout of scleraxis led to significant disruption of ECM organization and inferior tendon mechanical properties, 18,19 and knockout of mohawk led to changes in tendon thickness, collagen fibril diameter, and ultimate tensile strength. 202][23][24][25] There are reported effects of rLOX on these molecules in other cell types.As examples, adult skin fibroblasts and lung fibroblasts responded to rLOX treatment by significantly upregulating collagen type I gene expression levels, which coincided with increased collagen content. 17In adult mouse aorta, inhibition of LOX activity significantly upregulated MMP2 expression levels, but significantly downregulated MMP9 levels. 26][10] Specifically, tendons exhibited no changes in cell viability, density, proliferation, and/or metabolic activity when treated with exogenous rLOX or inhibitor of LOX activity, despite changes to tissue material properties. 8,9Furthermore, these findings were consistent across a range of embryonic stages.Collectively, these prior findings suggest rLOX does not induce cell behaviors that would negatively affect tendon.To test this hypothesis, the direct effect that rLOX treatment has on tendon cells should be assessed.Such a study is necessary in further exploring the potential to use rLOX to enhance tendon mechanical properties.
This study aimed to understand the effects of rLOX on chick embryo tendon cells harvested from stages Hamburger-Hamilton (HH) 40 and HH42 of development.Active ECM deposition and elaboration of tendon mechanical properties are actively occurring during HH40 and HH42, with significantly higher mechanical properties and collagen crosslink density in HH42 than HH40 tendons. 9,10As described above, rLOX treatment of HH40 and older embryonic tendons improved mechanical properties without affecting collagen content or cell behaviors in previous work. 8However, the specific effects of rLOX on tendon cell expression of phenotype and ECM remodeling molecules were not examined.Here, we hypothesized that rLOX has no effect on tendon cell behaviors, including expression of tendon markers, collagens, MMPs, TIMPs, or LOX regulators.HH40 and HH42 cells were treated with rLOX and assessed for changes in tendon phenotype, cell morphology, proliferation rate, proliferative capacity, metabolic activity, and enzyme activity levels, as compared to vehicle controls.The results of these experiments support the potential to develop a LOX-based treatment strategy that does not induce cell behaviors with adverse effects on tendon.

| Experimental overview
Chick embryonic tendon cells were isolated from HH40 and HH42 (incubation days 14 and 16, respectively) calcaneal tendons.Six tendons from three embryos per stage were pooled together to have enough cells for one biological replicate (N).For all experiments, cells were seeded at 20,000 cells/cm 2 in six-well plates and treated at 70% confluency with either 1.5 µg/mL rLOX (Origene) or saline vehicle.Supplemented medium was changed every 12 h.Cells were harvested to assess gene expression, protein, enzyme activity, and metabolic activity levels, as well as proliferation rate.Each assay required N = 3 biological replicates.

| Embryonic tendon cell isolation
Fertilized white leghorn chick embryos (University of Connecticut Poultry Farm) were incubated in a humidified rocking incubator at 37.5°C and killed and staged at HH40 and HH42.Cells were isolated as previously described 27,28 and used for experiments within the first two passages based on our previous study. 27Briefly, calcaneal tendons from three chick embryos were dissected and pooled for each biological replicate (N = 1).Tendons were minced and then digested in 1% collagenase type II while shaking at 200 RPM at 37°C for 25 min.

| Cell morphology analysis
HH40 and HH42 cells were each cultured with rLOX or vehicle for 24 h and then imaged under brightfield microscopy (Primovert, Zeiss).Images were analyzed for cell surface area and aspect ratio using the cell tracing function in ImageJ (NIH) by three independent reviewers in a blinded fashion, as previously described. 8,27Three biological replicates (N = 3) were analyzed.For each biological replicate, three different wells (n = 3) were evaluated.Three different images were acquired for each well (n), and between 50 and 100 cells were analyzed in each image.

| Cell counting
HH40 and HH42 cells were each treated with rLOX or vehicle, and then trypsinized and counted using a hemocytometer at 0, 24, and 48 h, as previously described. 27Three biological replicates (N = 3) were analyzed.For each biological replicate, three technical replicates (n = 3) were evaluated.

| EdU (5-ethynyl-2-deoxyuridine) staining
HH40 and HH42 cells were each treated with rLOX or vehicle for 24 h.After the first 12 h, cells were incubated with EdU to allow incorporation of EdU into DNA during active DNA synthesis.After a total of 24 h, cells were washed with HBSS, fixed in 3.7% formaldehyde, and permeabilized with 0.5% Triton ® X-100.Incorporated EdU was detected using Alexa Fluor ® 488 according to manufacturer's protocol (Invitrogen).Hoechst 33342 (Thermo Fisher) was used to stain nuclei.Stained cells were imaged on a FLUOVIEW FV3000 confocal system (Olympus).Three biological replicates (N = 3) were analyzed, with one technical replicate per N.
2.6 | MTT (3-[4,5-dimethylthiazol-2-yl]-2, 5-diphenyltetrazolium bromide) metabolic assay HH40 and HH42 cells were each treated with rLOX or vehicle for 24 h.Then, cells were incubated in fresh media containing 10 µM of MTT at 37°C for an additional 3 h, according to manufacturer's protocol (Abcam).Absorbance readings at 590 nm measured the conversion of MTT to insoluble formazan product using a Synergy neo2 multimode plate reader (Biotek).Raw values were offset by the absorbance value of the blank well (medium alone).Three biological replicates (N = 3) were analyzed, with one technical replicate per N.

| Reverse-transcription polymerase chain reaction (RT-PCR)
HH40 and HH42 cells were each treated with rLOX or vehicle for 12 h and then harvested for RNA isolation and RT-PCR, as previously described. 27Cells were homogenized in TRIzol (Life Technologies), total RNA isolated and reverse-transcribed using the Superscript III First Strand Synthesis kit (Invitrogen), and PCR performed using PlatinumTM Taq DNA Polymerase High Fidelity (Invitrogen) on a Mastercycler ® (Eppendorf).Chick-specific primers were used to characterize gene expression levels of 18S as well as tendon phenotype markers, collagens, MMPs, TIMPS, and LOX and its regulators (BMP-1, periostin, fibronectin) (Table 1).

| Enzyme activity assays
HH40 and HH42 cells were each treated with rLOX or vehicle for 24 h, and then washed with cold HBSS twice.Cells were scraped using Nunc™ Cell Scrapers (Thermo Scientific) in 1X protease inhibitor RIPA buffer.Cell lysates were collected, vortexed, and rested on ice for 5 min.Cell lysates were then centrifuged for 5 min at 4°C and 13,000g.Protein isolates were collected from the supernatant and transferred to new tubes.LOX enzyme activity levels were measured using the LOX activity assay (Abcam), as previously described. 8,11BMP-1 enzyme activity levels were measured using BMP-1 activity assay (R&D Systems).MMP enzyme activity levels were measured using zymograms.Briefly, protein isolates from each sample were loaded into Novex TM Zymogram Plus Gels (Invitrogen) and separated using electrophoresis at 100 V for 130 min under nonreducing conditions.After electrophoresis, gels were incubated in Renaturing Buffer (Invitrogen) for 30 min at room temperature and then incubated in Developing Buffer (Invitrogen) at 37°C overnight.
Finally, gels were stained with Colloidal Blue Staining Kit (Invitrogen) to visualize areas of protease activity.Three biological replicates (N = 3) were analyzed, with one technical replicate per N.

| Statistical analyses
Power analysis, performed as previously described, 8,27,29 determined N = 3 biological replicates was needed per embryonic stage and treatment for each assay.Cell surface area, cell aspect ratio, cell number, percentage of EdU-positive cells, metabolic activity levels, gene expression levels, and enzyme activity levels were analyzed for statistically significant differences between rLOX and vehicle treatments using two-tailed unpaired Student t-test.All statistical analyses were performed using Prism v8 (GraphPad Software).Metabolic activity levels were not affected by rLOX treatment for both HH40 and HH42 cells (Figure 3).There were no differences in gene expression levels of scleraxis, tenomodulin, mohawk, tenascin-C, collagen types I and III, MMP2, TIMP2, TIMP3, TIMP4, LOX, periostin (PSTN), and fibronectin between rLOX and vehicle treatments for either HH40 or HH42 cells.
(Figures 4 and 5).In HH40 cells, rLOX treatment significantly upregulated gene expression levels of BMP-1 (Figure 4).In HH42 cells, rLOX treatment significantly downregulated gene expression levels of MMP9 (Figure 5).There were no statistical differences in pro and mature MMP2 enzyme activity levels between rLOX and vehicle treatments in either HH40 or HH42 cells (Figure 6A-D).LOX enzyme activity levels were significantly higher with rLOX-treatment than vehicle-treatment in both HH40 and HH42 cells (Figure 6E,F).There were no statistical difference in BMP-1 enzyme activity levels between rLOX and vehicle treatments in either HH40 or HH42 cells (Figure 6G,H).

| DISCUSSION
This study revealed rLOX treatment has no effect on HH40 and HH42 cell morphology, proliferation rate, proliferative state, metabolic activity, and tendon phenotype marker expression levels.Gene expression levels of collagens, LOX, TIMPs, and most MMPs also T A B L E 1 Forward and reverse primer sequences.F I G U R E 4 Gene expression levels in HH40 cells after 12 h of rLOX treatment, data normalized to vehicle controls.BMP-1 expression levels were significant upregulated, whereas other genes were not affected by rLOX treatment.Statistically significant differences between rLOX and vehicle treatments were determined by Student t-test with *p < 0.05.N = 3 per treatment group.
were not affected.LOX enzyme activity levels were significantly upregulated by rLOX treatment despite no changes in LOX gene expression levels or BMP-1 and MMP enzyme activity levels, suggesting intracellular uptake of exogenous rLOX.These data are significant because they suggest that rLOX treatment minimally affects tendon cell behaviors, supporting the hypothesis that previously reported effects of rLOX on tendon mechanical properties during tissue development 8 are due to rLOX-mediated crosslinking of collagen rather than downstream effects of rLOX on cells.
HH40 and HH42 tendon phenotypes were not affected by rLOX treatment based on constant gene expression levels of scleraxis, tenomodulin, mohawk, and tenascin-C (Figures 4 and 5), as well as maintenance of fibroblastic morphology with constant surface area and aspect ratio (Figure 1).In contrast, loss of tenogenic phenotype is associated with increases in tendon cell surface area and decreases in tendon phenotype marker expression levels, which we previously observed in embryonic tendon cultures at high passage number. 27e absence of rLOX effects on phenotype in our current study are interesting because LOX has been hypothesized to alter phenotype of smooth muscle cells and 3T3 fibroblasts by inducing crosslinking of lysine residues on chromatin within nuclei, thereby altering chromatin packing. 30,31Changes in scleraxis or mohawk expression can lead to tendon mechanical property changes, 19,20 Gene expression levels in HH42 cells after 12 h of rLOX treatment, data normalized to vehicle controls.MMP9 expression levels were significant downregulated, whereas other genes were not affected by rLOX treatment.Statistically significant differences between rLOX and vehicle treatments were determined by Student t-test with *p < 0.05.N = 3 per treatment group.
exogenous rLOX entered the cells, based on reports of intracellular uptake of exogenous LOX by smooth muscle cells and 3T3 fibroblasts. 30,31It is also possible that intracellularly stored proLOX was cleaved to produce active LOX, since proLOX is first produced intracellularly before export to the extracellular space.To explore this possibility, we examined BMP-1, periostin, and fibronectin expression.
4][35] Periostin and fibronectin expression levels were not affected by rLOX treatment in either HH40 or HH42 cells, whereas BMP-1 was significantly upregulated by rLOX treatment in HH40 cells but not in HH42 cells (Figures 4 and 5).However, BMP-1 intracellular enzyme activity levels did not change with rLOX treatment in either HH40 or HH42 cells (Figure 6).Based on the lack of increase in intracellular BMP-1 enzyme activity levels, we propose that the increase in intracellular LOX activity levels reflected intracellular uptake of rLOX.
The significant increase in intracellular LOX enzyme levels warrants further investigation, as other studies suggest LOX mediates covalent crosslinking of other substrates within cells, such as chromatins, to induce downstream effects on cell phenotype and behaviors.Future studies should evaluate potential effects of elevated intracellular LOX activity levels at later timepoints.

LOX reportedly promotes collagen expression in other tissues
such as adult skin fibroblasts, 17 whereas inhibition of LOX leads to increases in collagen synthesis in adult osteoblasts and chondrocytes in culture. 15,16Here, collagen types I and III mRNA levels were not affected by rLOX treatment in either HH40 or HH42 tendon cells (Figures 4 and 5), which agrees with our previous study in which treatment of HH40 and older embryonic tendons with exogenous rLOX did not affect collagen content, density, or maturity. 8In addition, it is notable that cell morphology and metabolic activity did not change with rLOX treatment (Figures 1 and 3), because previous studies have established a strong correlation between nuclear shape index and collagen synthesis. 36These data collectively suggest exogenous rLOX does not promote collagen synthesis during new tendon formation.However, there is the possibility that rLOX could influence post-transcriptional events to enhance collagen deposition.
For example, increasing LOX-mediated collagen crosslinking could enhance retention of collagen, thereby increasing collagen content without increasing collagen synthesis.For this reason, we did not perform soluble collagen assays, but future experiments could perform pulse-chase labeling experiments to evaluate procollagen synthesis, which was beyond the scope of the current study.
4][25] For example, inhibition of MMP2 and MMP9 activity with doxycycline significantly compromised mechanical properties of injured adult rat Achilles tendon, suggesting ECM remodeling by MMPs is important for tendon healing. 25In other tissues, LOX appears to regulate MMP2 and MMP9.For example, MMP2 and MMP9 expression levels in rat aorta cells increased and decreased significantly, respectively, when LOX activity was inhibited. 26While the roles of MMP2 and MMP9 in embryonic tendon development are unknown, both have been detected in embryonic tendons. 37,38Here, MMP2 expression and proMMP2 and MMP2 enzyme activity were detected but not affected by rLOX treatment in HH40 and HH42 cells (Figures 4, 5, and 6).In contrast, rLOX treatment downregulated MMP9 expression levels in HH42 cells, but not in HH40 cells (Figures 4 and 5).We did not detect proMMP9 and MMP9 enzyme activity in either HH40 or HH42 cells, perhaps because they were below the threshold of detection.Interestingly, proMMP2 and MMP2 activity were previously detected in chick embryo metatarsal tendons, whereas proMMP9 and MMP9 activity were not. 38On the other hand, we detected both MMP2 and MMP9 protein via immunohistochemical staining in chick embryo craniofacial tendons. 37TIMP2, TIMP3, and TIMP4 expression levels also did not change with rLOX treatment for either HH40 or HH42 cells (Figures 4 and 5).We did not examine TIMP1 expression because the chicken gene sequence is not available.Our data suggest rLOX does not regulate MMP2, MMP9, and TIMPs.However, future studies should also evaluate levels of MMPs and TIMPs that are secreted extracellularly.
Proliferation rates and the percentage of EdU-positive cells remained the same between rLOX and vehicle treatments for both HH40 and HH42 cell cultures (Figure 2).These in vitro results support our previous findings that rLOX treatment does not affect cell density of embryonic tendons in vivo. 8This is interesting because LOX appears to play a role in proliferation of other cell types.
Overexpression of LOX inhibited proliferation of vascular smooth muscle cells (VSMCs), but these inhibitory effects were abrogated by chemical inhibition of LOX activity, suggesting LOX is important for VSMC proliferation. 13Different effects of rLOX treatment on proliferation rates between our study and others suggest rLOX effects are dependent on cell type.While the underlying mechanisms of LOX regulation of cell proliferation are unknown, it has been suggested that LOX oxidizes lysine residues on different growth factors such as basic fibroblast growth factor (bFGF) and transforming growth factor (TGF)-β1 to either inhibit or promote downstream signaling pathways, thereby regulating cell behaviors including cell proliferation. 12,14For example, LOX was reported to bind TGF-β1 in bone and suppress TGF-β1-induced Smad3 phosphorylation and was hypothesized to inhibit proliferation of osteoblast progenitors. 14ile LOX binding to growth factors implicated in embryonic tendon development, such as FGF-4 and TGF-β2, has not been reported, we previously showed that neither recombinant FGF-4 nor TGF-β2 affected proliferation rates of mouse embryonic tendon cells. 39ken together, it appears rLOX does not regulate embryonic tendon cell proliferation, but future studies should examine rLOX effects on adult tendon cell proliferation.
We measured metabolic activity levels to gain insight into whether rLOX may have been inducing metabolically demanding cell functions, such as ECM protein synthesis.HH40 and HH42 metabolic activity levels did not change with rLOX treatment (Figure 3).Our findings support that the previously reported effects of rLOX treatment on enhancing collagen crosslinking and tendon mechanical properties 8  In addition to the suggested future studies mentioned earlier, we have additional recommendations.We examined gene expression levels after 12 h and protein levels after 48 h of rLOX treatment because gene expression changes occur earlier than protein changes.
Later timepoints could be interesting but cells could become overly confluent in vitro, and thus longer term effects of rLOX should be studied with in vivo studies.Future studies would also benefit from approaches to label and sort tenocytes from other cell populations in the tendon, as we could not in this study.We examined rLOX effects on intracellular enzyme activity levels, but future studies should also assess activity levels of enzymes secreted extracellularly that could directly affect the ECM during tissue formation.We tested one rLOX concentration based on our previous studies that showed this concentration significantly enhanced embryonic tendon mechanical properties, 8 but future studies could examine additional concentrations.Here, we used an in vitro culture model to study the effects of rLOX treatment under a more controlled environment.Future studies should examine the combined effects of rLOX and increased elastic modulus on tendon cells, as embryonic tendon cells are sensitive to their mechanical microenvironment. 28,39,43,44 summary, rLOX treatment did not directly affect tendon cell phenotype, morphology, proliferation rate, proliferative state, or metabolic activity at different stages of development.Our data here support the possibility that the previously reported ability for rLOX to enhance tendon mechanical properties in vivo 8 is due to direct mediation of collagen crosslinking and not also due to rLOX effects on tendon cell phenotype and behaviors.These findings contribute fundamental information about the effects of LOX on embryonic tendon cell functions, and also motivate future development of Standard deviations are represented by error bars.
3 | RESULTSHH40 and HH42 cells both exhibited elongated, thin morphologies with rLOX and vehicle treatments (Figure1A,B).Cell surface area and aspect ratio were not affected by rLOX treatment for both HH40 and HH42 cells (Figure1C−F).Similar growth curves between rLOX and vehicle-treated cells revealed proliferation rates were not affected by rLOX treatment in either HH40 or HH42 cells (Figure 2A,B).EdU staining revealed no significant differences in proliferative state between rLOX and vehicle treatments for both HH40 and HH42 cells (Figure 2C−E).

F I G U R E 1
Morphologies of HH40 and HH42 cells were not affected by 24 h of rLOX treatment.(A, B) Representative brightfield images of HH40 (A) and HH42 (B), with select cells outlined to highlight example tracings.(C, D) Average surface areas of HH40 (C) and HH42 cells (D) were not affected by rLOX treatment.(E, F) Average aspect ratios of HH40 (E) and HH42 cells (F) were not affected by rLOX treatment.Statistically significant differences were determined by Student t-test with p < 0.05.N = 3 per stage and treatment group.Scale bar = 50 µm.F I G U R E 2 Proliferation rates and proliferative states of HH40 and HH42 cells were not affected by rLOX treatment.(A, B) Growth curves revealed proliferation rates of HH40 (A) and HH42 cells (B) were not affected by rLOX treatment.(C) Representative images of EdU staining of HH40 and HH42 cells.(D, E) The percentages of EdU-positive cells in HH40 (D) and HH42 (E) cell cultures were not affected by rLOX treatment.Statistically significant differences between rLOX and vehicle treatments at the same timepoint were determined by Student t-test with p < 0.05.N = 3 per stage and treatment group.Scale bar = 100 µm.F I G U R E 3 Metabolic activity levels of HH40 and HH42 cells were not affected by 24 h of rLOX treatment.Statistically significant differences between rLOX and vehicle treatments at the same timepoint were determined by Student t-test with p < 0.05.N = 3 per stage and treatment group.

F I G U R E 6
Enzyme activity levels in HH40 and HH42 cells after 48 h of rLOX treatment.(A, B) ProMMP2 activity levels were not affected by rLOX treatment in both HH40 (A) and HH42 (B) cells.(C, D) MMP2 activity levels were not affected by rLOX treatment in both HH40 (C) and HH42 (D) cells.(E, F) LOX activity levels were significantly upregulated by rLOX treatment in both HH40 (E) and HH42 (F) cells.(G, H) BMP-1 activity levels were not affected by rLOX treatment in both HH40 (G) and HH42 (H) cells.Statistically significant differences between rLOX and vehicle treatments at each stage were determined by Student t-test with *p < 0.05.N = 3 per stage and treatment group.
therapeutics to enhance tendon mechanical properties during abnormal tissue formation, such as adult tendon healing or development of tendon-related birth defects.AUTHOR CONTRIBUTIONS Phong K. Nguyen designed and performed experiments and assays.Kaitlyn Hall and Iverson Holt helped with image analyses that contributed to Figure 1.Phong K. Nguyen and Catherine K. Kuo designed experiments, performed data analysis, interpreted the data, and wrote the manuscript.All authors have read and approved the submitted manuscript.