Close this search box.

Microinterfaces in biopolymer-based bicontinuous hydrogels guide rapid 3D cell migration – Nature Communications

Material synthesis and characterization

Macromer synthesis

All reagents were obtained from Sigma-Aldrich and Fisher Scientific unless otherwise specified. Hydrogel macromers were prepared as described previously64. Briefly, for all polymers, sodium hyaluronate (HA, MW = 60 kDa, Lifecore Biomedical) was modified with tetrabutylammonium salt (HA-TBA). Ad-HA modification was performed through the reaction of 1-adamantane acetic acid (Ad) and (dimethylamino)pyridine (DMAP) and ditert-butyl decarbonate (Boc2O) for 20 h at 45°C in anhydrous dimethyl sulfoxide (DMSO). CD-HA modification was performed through the reaction of Mono-(6-(1,6-hexamethylenediamine)-6-deoxy)-β-Cyclodextrin (Crysdot) and benzotriazol-1-yloxytris (dimethylamino) phosphonium hexafluorophosphate (BOP) for 3 h at RT in DMSO. AHA modification was performed through the reaction of acrylic anhydride at a pH 9–10 for 3 h at RT in DI water. All polymers were purified via dialysis, lyophilized and modification was confirmed using 1H NMR (Bruker Neo 400). AD-HA, CD-HA, and AHA of 12%, 20%, and 92% modification by 1H NMR, respectively, were used for all experiments (Supplementary Fig. 1, analysis performed in TopSpin and MestReNova).

Bicontinuous hydrogel fabrication

The GH concentration (0, 1, 3%) signifies the combined polymer weight percent, with a consistent ratio of 1:1 between adamantane and beta-cyclodextrin. Hydrogels were formed by separately preparing solutions of gelatin (0 or 5 wt%, gel strength 300, Type A from porcine skin) with desired AD-HA concentration and a separate solution that combined CD-HA with transglutaminase (enzymatic crosslinker). The two solutions were manually mixed at 25 °C and placed at 37 °C for at least 150 min to allow for full crosslinking prior to hydration. Pure GH hydrogels were similarly formed.

Uniform hydrogel fabrication

Stock solutions of 7.17 mg/mL fluorescent collagen solution was mixed by combining fluorescent collagen and 10 mg/mL collagen (Advanced BioMatrix) as previously65. Collagen was then diluted with PBS and neutralized with 1 NaOH and placed at 37 °C for at least 60 min for gelation prior to hydration. 5 wt% AHA hydrogels were fabricated by combining AHA with 10 mM MMP-degradable crosslinker (GCNSVPMSMRGGSNCG, Genscript) and 1 mM thiolated RGD (GCGYGRGDSPG).

Collagen-GH hydrogel fabrication

Collagen stock was mixed with AD-HA, and CD-HA was mixed with 1 N NaOH and PBS to obtain a neutral pH. Both solutions were mixed together and placed at 37 °C for at least 60 min for gelation prior to hydration.

Structural characterization

To assess structural heterogeneity, fluorescent-conjugated gelatin was added to bicontinuous hydrogels (0.5 wt%, from pig skin) and Z-stacks were acquired (Leica TCS SP5, 10x air objective – 1.44 μm/pixel x 5 μm/voxel, 25x water objective – 0.566 μm/pixel x 5 μm/voxel, or 63x oil objective – 0.115 μm/pixel x 1 μm/voxel). Fluorescent intensity profiles and quantification of coefficient of variation was performed using a custom MATLAB script. 3D rendering of bicontinuous hydrogel structure was conducted in Imaris, in which GR domains were fluorescently labeled, and GP domains were obtained from inversion of GR domain fluorescence. To label individual polymers, fluorescent peptides (GenScript) were covalently conjugated to methacrylated AD-HA and CD-HA as described previously66. Briefly, corresponding methacrylated polymer was dissolved in 0.2 M (pH ~8) triethanolamine buffer (TEOA) to a final polymer concentration of 2 wt%, with peptides achieving a final concentration of 5 mg of thiolated peptide/100 mg of methacrylated polymer. Reaction was allowed to occur overnight at 37 C. Polymers were then purified via dialysis, and lyophilized. Fluorescent polymers were doped into the hydrogel prior to gelation such that 0.4 wt% of the solution was fluorescently labeled.

3D object counter was performed on both GR and GP domains to obtain the volume, surface area, and number of discrete objects within a given 3D space. Volume fraction calculations were performed by dividing the total volume of either GR or GP domains by the theoretical total volume of the ROI. Connectivity was defined as the volume of the largest GR or GP discrete domain divided by the total volume of each respective domain x 100%. Total internal interfacial surface area was normalized by volume within each ROI. Normalized interfacial surface area was obtained by dividing by the volume of the ROI.

Mechanical characterization (Rheology)

Hydrogels were formed as described and deposited on the bottom plate of a rheometer immediately after mixing. An AR2000 stress-controlled rheometer (TA Instruments) was fitted with a 20 mm diameter cone geometry and placed at a 26 μm gap. Hydrogels were allowed to crosslink by applying an oscillating torque at 1 Hz and 1% strain for 150 min at 37 °C unless otherwise specified. After polymerization, the stress relaxation of the gel was tested at 10% strain for 10 min and the creep-recovery of the gel was tested at 100 Pa for 10 min. The frequency dependence at 1% strain was also characterized from 0.01 to 100 Hz and strain sweeps at 1 Hz were characterized from 0.001% to 1000% strain.

Mechanical characterization (atomic force microscopy – AFM-nanoindentation)

AFM-based nanoindentation (Bruker’s Dimension Icon AFM) was utilized to quantify microscale mechanical properties of the formed hydrogels. Cantilevers (HQ:CSC38/Cr-Au MikroMasch) had a nominal spring constant of 0.03 N/m and a 10 μm diameter spherical bead was attached to the tip extremity. Bicontinuous hydrogels were fabricated in between two glass coverslips with ~100 μm spacing. To query the mechanical properties of the hydrogel interior, the bicontinuous hydrogel was formed in a 1 mL blunt syringe and allowed to undergo complete gelation. The hydrogel was cut in half with a razor blade and AFM-nanoindentation was conducted on the newly exposed surface. Bicontinuous hydrogels were probed with an approaching and retraction rate of 10 μm/s in PBS, following established procedures67. Deflection error versus height sensor curves were collected in an 8 × 8 grid of 105 μm by 105 μm. The spring constant of the cantilevers were calibrated through thermal resonance frequency prior to the experiment. With an assumption of Poisson’s ratio of 0.49 for highly swollen hydrogels, effective indentation modulus, Eind, was obtained by fitting a Hertz model to the entire loading portion of each indentation force-depth curve68. Force map visualization and quantification of coefficient of variation was conducted with a custom MATLAB (MathWorks) script.

To probe mechanical properties within distinct GR and GP domains, 50 μm fluorescent beads (CD Bioparticles) were embedded at a 0.5 wt% concentration as fiducial markers. 3 × 8 grids of 30 μm by 105 μm were obtained. After indentation, the same hydrogels were imaged using confocal microscopy. The location of fiducial markers (beads) that appeared on the bright field AFM microscope were correlated to the location of beads within the confocal images to determine which fluorescent regions were probed (Supplementary Fig. 7b, c). Regions were manually denoted as GP or GR (or inconclusive if they were at or close to an interface; these regions were excluded from analysis). Outliers in AFM force maps were removed with the IQR method.

Mechanical characterization (dynamic mechanical analysis)

Hydrogel prepolymer solution (30 uL) was placed into a cylindrical mold (4.78 mm diameter) and allowed to gel (gelatin at 37 °C for greater than 150 min, agarose at 4 °C for 1 h and, then transferred to 37 °C for 1 h). Uniaxial compression testing was conducted (TA Instruments Q800 DMA), with hydrogels exposed to a preload force of 0.001 N and compressed at a rate of 0.05 N min-1. Compressive moduli were calculated as slopes between 10 and 20% strain on the stress-strain curve.

Degradation studies

3% bicontinuous hydrogels were fabricated and added to PDMS molds (6 mm diameter, 6 mm height). Every two days, hydrogel volumes were measured and supernatant was acquired for uronic acid release assays. Uronic acid content was characterized via the uronic acid carbazole reaction according to the previous literature69. Briefly, a standard curve was prepared using sodium hyaluronate dissolved in water at 1 mg mL-1 and subsequently serially diluted to produce a 7-point curve (1 mg mL-1 to 0.03125 mg mL-1, plus 0 mg mL-1). To initiate the reaction, the sample (50 μL) was heated to 100 °C for 10 min immediately after the addition of 25 mM sodium tetraborate (200 μL, dissolved in sulfuric acid) and allowed to cool to room temperature for 15 mins. Next, 0.125% (w/v%) carbazole (50 μL, dissolved in absolute ethanol) was added to each well, heated to 100 °C for 10 min, and cooled to room temperature for 15 min. Sample absorbance was read immediately at 550 nm on a Tecan M200 Pro plate reader. A linear regression curve fit was used to determine uronic acid concentration. Data is presented as cumulative uronic acid release %.

In vitro experiments

Spheroid encapsulation in 3D

Meniscal fibrochondrocytes (MFCs) were harvested from the medial meniscus of dissected juvenile bovine joints (Research 87) as described previously70. Briefly, menisci were minced into ~1 mm3 cubes and placed in tissue culture plates, where MFCs gradually emerged from tissue over two weeks. Mesenchymal stromal cells (MSCs) were isolated from juvenile bovine bone marrow71. All cells were cultured until 80% confluency of initial colonies and then stored in liquid nitrogen (90% FBS, 10% dimethylsulfoxide). For in vitro studies, MFCs and MSCs were used between passage 2-6. Spheroids (1000 cells/spheroid) were formed through 48-h culture on agarose (standard gelling temperature) microwell pyramid arrays (molds formed from AggreWellTM 400, Stem Cell Technologies). Spheroids were removed from microwells via pipette aspiration and mixing, allowed to settle, and then were transferred to hydrogels for assessment of outgrowth over time. Spheroids were added at a concentration of 1000 spheroids/mL of hydrogel solution, and hydrogels were completely gelled (for 150 min with transglutaminase) prior to adding media. For studies that consisted of networks with 5 wt% gelatin, 1 U/mL of transglutaminase and 3% soluble HA, an additional thin layer of 0% GH hydrogel was added atop these hydrogels to form a barrier to soluble HA dissolution from hydrogel. Unless otherwise noted, cells and spheroids were cultured in high glucose DMEM, supplemented with 10% Fetal Bovine Serum (FBS) and 1% penicillin and streptomycin at 37 °C and 5% CO2.

In vitro imaging and analysis

Unless otherwise specified, Z-stack images were acquired on an upright confocal microscope (Leica TCS SP5).

All viability studies were conducted by treating cells with calcein AM (2 μM) and ethidium homodimer-1 (4 μM) for 3 h at 4 °C. For actin staining, all constructs were fixed in 10% neutral buffered formalin for 30 min at RT, permeabilized with 0.1% Triton X for 30 min at RT, and then stained at 1:1000 with phalloidin-647 to visualize the actin cytoskeleton in 1% bovine serum albumin (BSA) overnight at 4 °C. Actin staining was performed after 1, 2, and 3 days of culture. Analysis was carried out in ImageJ (NIH), where outgrowth was determined by manually denoting original spheroid body and extrapolating spheroid radius, with live stain used for spheroids in guest-host only or gelatin and guest-host hydrogels, and actin stain used for all other studies (Supplementary Figs. 10, 11b). Eight lines were drawn from the center of spheroid towards the periphery of the outgrowth. These values were then averaged and spheroid radius was subtracted to yield average cell outgrowth from a single spheroid.

For MMP inhibition studies, spheroids were embedded in 0% GH hydrogels (5 wt% gelatin) and allowed to migrate for 3 days to determine a concentration of Marimastat that was sufficient to reduce cell outgrowth. Constructs were treated with 0, 1, 10, 100, and 1000 μM Marimastat. After identifying 100 μM as the desired concentration, all experimental groups were treated with Marimastat and a control of DMSO for 3 days and stained for actin. Outgrowth was analyzed as above. For RGD binding perturbation, soluble RGD (5 mM RGD peptide, GCGYGRGDSPG, Genscript) was added to media in the 3% experimental group and outgrowth after 3 days was compared to PBS controls, with outgrowth analyzed as above.

For proliferation, the constructs were fixed and soaked in sucrose overnight and then embedded in OCT and flash frozen in liquid nitrogen-cooled 2-methylbutane. Using a cryostat microtome, constructs were sectioned to 30 μm thickness. Slides were then rehydrated for 5 min with Phosphate Buffer Saline (PBS), permeabilized with 0.1% Triton X for 5 min and then blocked with 3% BSA for 1 h at RT. Ki67 antibodies (2 ug/mL, Abcam ab15580) were added in 3% BSA to the sections and incubated at 4 °C overnight. Sections were rinsed with PBS for 5 min three times. The secondary antibody Alexa Fluor 647 IgG H&L (1:1000, Invitrogen A32733TR) was added in 3% BSA for 1 h. Slides were then rinsed three times again for 5 min each and mounted with Prolong Gold Antifade reagent with DAPI overnight at RT prior to imaging. In ImageJ, the number of Ki67+ cells was divided by the total number of cells (Moments threshold) to obtain the fraction of proliferating cells that had migrated from the spheroids. Antibody validation was performed by seeding MFCs onto glass substrates for 3 days and imaging with an upright epifluorescent microscope (Supplementary Fig. 22).

To assess nascent protein deposition, studies were performed as previously described51,72. Briefly, the media consists of glutamine, methionine, and cysteine-free high glucose DMEM, supplemented with 0.201 mM cysteine, 50 ng/mL of Vitamin C, 100 μM L-methionine, 100 μg/mL sodium pyruvate, 8 mM Glutamax, 1% P/S, 10% FBS and 0.1 mM azidohomoalanine (AHA). At the desired time points, hydrogels were washed, and stained for 30 min with CellTrackerRed in 3% BSA at 5 μg/mL at 37 °C and 5% CO2. The gels were then washed several times with 3% BSA and stained with 30 μM DBCO-488 in 3% BSA for 30 min at 4 °C and washed 3 times again prior to fixation in 10% formalin for 30 min at RT. Constructs were imaged as described previously.

Spheroid outgrowth along 2D interfaces

Gelatin hydrogels (5 wt%, 1 U/mL transglutaminase) were prepared in 8 mm biopsy-punched PDMS wells and allowed to fully gel. MFC spheroids were placed atop gelatin hydrogels and allowed to adhere for 2 h. Afterwards, gelatin (5 wt%, 1 U/mL transglutaminase), GH (3 wt%) or molten agarose (0.25%, 1, 3, and 6 wt%, low gelling temperature) were placed atop gels. After gelation, media was added, and constructs were fixed, stained for actin, and imaged after 1 day. Outgrowth parallel to the 2D interface was analyzed as above. Outgrowth perpendicular to 2D interface was quantified in Imaris. Briefly, cells (spheroid body and outgrown cells) were united into a single surface and fit into an object-oriented bounding box. The box dimension was obtained to determine the extent of perpendicular outgrowth, where the perpendicular outgrowth diameter was matched to the shortest principal axis. As a result, outgrowth perpendicular to interface data includes spheroid radii.

Live cell outgrowth

Live-imaging studies were performed with a laser scanning confocal microscope (Nikon Ax 1) with a live-cell chamber (37 °C, 5% CO2). For outgrowth studies, a 20X water immersion objective (0.38 μm/pixel, 5 um/voxel) was used. Formed spheroids were stained with deep red Cell Mask Plasma membrane stain (5 μg/mL) and embedded along with 0.2 μm-diameter fluorescently labeled beads at 0.7% (v/v) in the Ad-HA/gelatin prepolymer solution and 0.3% (v/v) in the CD-HA/transglutaminase prepolymer solution. After spheroid encapsulation and gelation, gels were incubated in live imaging media (FluoroBrite DMEM, 10% FBS, 1% P/S). Z-stacks (100 μm) were acquired every hour for 3 days. Imaging parameters were adjusted to minimize photobleaching and maximize cell viability.

Confocal time-lapse z-stacks were analyzed with Imaris 10.0. Migration track quantification was obtained through spots analysis of cells. Cells were estimated to be 25 μm spots and tracked via autoregression motion algorithms, in which the maximum path length was set to 100 μm with a 5-spot gap max to predict future spot positions. Cell tracks were filtered and removed ( < 30 μm) to reduce artifact motion. Tracks that appeared from outside field of view were removed manually to ensure that only cells that migrated from spheroid were evaluated. Migration tracks were rendered with a custom MATLAB script, with each cell track beginning from a common origin point. For real-time outgrowth renderings, the spheroid bodies were rendered as surfaces and cells were rendered as surfaces and overlayed with quantified tracks in Imaris.

Anisotropic persistent random walk model

The statistical characteristics of cell migration in our hydrogel system were studied using the anisotropic persistent random walk (APRW) model37,73. According to this model, we first projected our 3D cell trajectories and obtained their corresponding x and y coordinates. Then, we verified that the cell motility is time-invariant after 30 h in our system (Supplementary Fig. 9a) and thus APRW model can be applied. We next calculated the MSD from the cell trajectories at different time lags as follows:

$${{{{{rm{MSD}}}}}}left(tau right)= < {, [xleft(t+tau right)-xleft(tright)]}^{2}+{[yleft(t+tau right)-yleft(tright)]}^{2} , > $$


where τ represents the time lag and ( < cdots > ) shows time averaging. The angular displacements (dtheta) between two consequent steps of movement with time lag τ were then calculated using the definition of inner product of the two velocity vectors as follows:

$$dtheta left(t,tau right)=arccos left(frac{{v}_{t}.{v}_{t+tau }}{left|{v}_{t}right|left|{v}_{t+tau }right|}right)$$


where (left|ldots right|) shows the vector magnitude. After that, we obtained the primary and non-primary axes of migration for each individual cell through singular value decomposition (SVD) analysis of its velocity matrix. We subsequently calculated average magnitudes of the cell velocities at different orientations relative to the primary axis and obtained the corresponding velocity magnitude profiles. According to the APRW model, cell motility is assumed to display different persistence random walks along the two axes of migration. As a result, we calculated the MSDs of cell movements along each of these orthogonal migration axes and then fitted them to the following equations:

$${{{{{{rm{MSD}}}}}}}_{p}left(tau right)={S}_{p}^{2}{P}_{p}^{2}left({e}^{-frac{tau }{{P}_{p}}}+frac{tau }{{P}_{p}}-1right)+2{sigma }^{2}$$


$${{{{{{rm{MSD}}}}}}}_{{np}}left(tau right)={S}_{{np}}^{2}{P}_{{np}}^{2}left({e}^{-frac{tau }{{P}_{{np}}}}+frac{tau }{{P}_{{np}}}-1right)+2{sigma }^{2}$$


where ({sigma }^{2}) is the variance of observation noise in cell position and ({S}_{p}({S}_{{np}})) and ({P}_{p}({P}_{{np}})) show the cell speed and its persistent time along (vec{{{{{{bf{p}}}}}}},) (({{{{{bf{n}}}}}}vec{{{{{{bf{p}}}}}}})) direction. The obtained values for the cell speeds and persistent times can be further used to calculate cell diffusivities in different migration directions using the following formula:





The ratio of these cell diffusivities gives the anisotropy index of the cell migration:



and the total cell diffusivity can be calculated as ({D}_{t}={D}_{p}+{D}_{{np}}). Finally, based on the obtained cell speeds and persistent times, cell trajectories were also simulated via the governing stochastic differential equations of the APRW model and the corresponding MSDs and velocity magnitude profiles were extracted to verify the model accuracy. The details of the governing differential equations and the simulation process are described further in Wu et al.37.

Bead displacement studies

Cells were allowed to migrate from spheroids and were imaged at 10 min intervals for 3 h on day 3 with a 40x water immersion objective (0.22 μm/pixel, 1 μm/voxel). Cells that migrated along one axis throughout the imaging time span were selected for further analysis. 3D stacks were drift corrected with ImageJ Plugin Correct 3D Drift. A stack of 5 μm was isolated based on a cell of interest, max projected, and oriented such that cells were migrating in the negative direction of the axis of migration. HyperStackReg was used as a second drift correction and bead displacements were quantified with the ImageJ PIV plugin (1st interrogation window size: 128 pixels, 1st search window size: 256 pixels; 2nd interrogation window size: 64 pixels, 2nd search window size: 128 pixels). A region of interest (40 μm x 150 μm – which corresponds to 2x the average cell volume) was manually identified in the forward path of cell migration. Vertical components of bead displacement vectors in this region of interest were averaged and then plotted based on time from the beginning of imaging session on day 3.

Visualization of 3D bead displacement

To remove noise, bead images were background subtracted in ImageJ with a 4-pixel radius. To extract the bead positions, we used a maximum likelihood estimator-based Lucy-Richardson deconvolution method. Bead displacements were then computed using a topology-based particle tracking (TPT) algorithm implemented in MATLAB74. To extract the cell body for 3D reconstruction, we first background subtracted the images in ImageJ with a 50-pixel radius and segmented the images using the Otsu’s method. Then, 3D objects were detected using the bwmorph3 function in MATLAB, followed by a size-exclusion procedure to remove cell debris. The displacement field was enlarged 6 times and overlayed with the cell body for visualization purposes.

Gelatin-agarose particle composite hydrogels

Extrusion fragmented agarose particles were prepared as described previously56. Briefly, 3 wt% agarose was cooled at 4 °C for 1 h, and then sequentially extruded through 18, 21, 23, 25, 27, 30, and 34 G needles. Particles were resuspended in gelatin solution (gelatin – 5 wt%, transglutaminase – 1 U/mL) and then centrifuged at 100, 500, and 9800 rcf (low, medium, and high density) for 30 seconds. Supernatant was aspirated and gelatin-microgel slurry was combined with a spheroid pellet, manually mixed, and then added to 4.78 mm molds. The construct was allowed to gel for 150 min prior to addition of media. After 3 days, constructs were fixed and analyzed for cell outgrowth as above. Structural analysis of composite hydrogels was conducted as above (where agarose particles were obtained by inverting gelatin fluorescence).

Ex vivo and in vivo experiments

Ex vivo studies

Using biopsy punches, cylindrical tissue explants (5 mm in diameter and 1 mm in height) were excised from the middle zone of the meniscus such that circumferential fibers were oriented normal to the surface of the hydrogel. Explants were subsequently incubated in basal media for 2 weeks to support cell occupation of the explant periphery3. Afterwards, hydrogels were fabricated within 8 mm diameter PDMS molds and explants were placed on top of hydrogels and cells were allowed to egress from the tissue onto gels for 6 days, at which point explants were removed, and hydrogels (and their infiltrated cells) were fixed as above. Constructs were then stained with Hoechst 33342 (20 μM) to identify nuclei and actin (as above). Cell infiltration quantification and 3D construct rendering were carried out in Imaris.

In vivo studies

To evaluate the effect of scaffolds on meniscal repair in an in vivo setting, a nude rat xenotransplant model was employed. All procedures were approved by the Animal Care and Use Committee of the University of Pennsylvania Office of Animal Welfare (Protocol 806580). Adult bovine meniscal explant cylinders (4 mm outer diameter, 1 mm inner diameter, n = 10 donors) were biopsy punched as previously. The defect was filled with 0, 1, or 3% GH hydrogel via syringe injection. Explants were allowed to incubate for at least 150 min before subcutaneous implantation into male nude rats (N  =  8; NTac: NIH-Foxn1rnu, 9-12 weeks old, ~300 g, Taconic). Rats were anesthetized with isoflurane and the dorsal area shaved and scrubbed with chlorhexidine and betadine. Four subcutaneous pockets were made with 1 on the cranial and caudal aspects on each side of the midline via blunt dissection. One construct per group was inserted into each pocket (with a repeat of one of the groups in each rat). The incision was closed with wound clips. At 2 weeks, rats were euthanized by CO2 asphyxiation and the constructs were removed from the subcutaneous tissue. Constructs were cut into 1 mm sections using a custom meniscus slicer and the slices at the end of the constructs were removed to minimize contributions from host cell infiltration. Constructs were fixed for 15 min at 37 °C and stained with Hoechst for 30 min. To assess cellularity, cell signal was Otsu thresholded in ImageJ and % positive cell area was quantified within each quartile of the manually segmented total defect area. To assess fluorescence loss, fluorescence % area was thresholded and subtracted to obtain % defect without fluorescence. Any constructs in which gel in defect did not remain intact (through lack of fluorescence and cells) were excluded from analysis. To assess CD68 labeling, constructs were flash-frozen and cryosectioned to 10 μm thickness as previously, and stained with CD68 (1:500, MCA341GA clone ED1; Bio-Rad) and secondary antibody Alexa Fluor 546 IgG H&L (1:500, Invitrogen) prior to mounting with DAPI. Imaging of CD68 was conducted with a Zeiss Axioscan Z1 slide scanner using a 40X objective and a Colibri 7 LED illumination source. In ImageJ, the number of cells positive for CD68 were divided by the total number of cells to obtain the fraction of CD68+ infiltrating cells.

Statistics and data presentation

All statistical analyzes were performed using Microsoft Excel and GraphPad. Unless otherwise specified, the robust regression and outlier removal method was used to remove outliers prior to performing statistical tests. Significance was assessed with two-way student’s t-test (for 2 experimental groups) or by one or two-way ANOVA with Tukey’s HSD post hoc testing, where a p value  <  0.05 was considered significant. When no experimental groups were significant, n.s. was used to denote no statistical significance. In ANOVAs, when some groups were significant, a straight line between significant groups was denoted, and unlabeled groups in these graphs have no statistical significance. N values and biological replicates are described in figure captions. All data are reported as mean ± standard deviation as denoted in the figure caption. All statistical test p, q, t, and DF values are reported in Source Data. Schematics were designed with Adobe Illustrator.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.