The collection of animal tissues was approved by the Royal Veterinary College Ethics and Welfare Committee (URN-2016-1627b). All tissues were sourced from horses euthanised for reasons unrelated to this study, and other than tendon injury at a commercial equine abattoir.

Superficial digital flexor tendons (SDFT) were harvested from forelimbs taken from young, skeletally mature horses (age = 3–8 years, n = 5, exercise history unknown). Prior to isolation, the forelimbs were clipped to remove hair and the skin sterilised by several applications of 4% chlorhexidine (HiBiScrub®; Mölnlycke Health Care). Portions of mid-metacarpal SDFT (6–10 cm) were dissected free of the limb and stored immediately in standard growth medium consisting of pyruvate and low glucose Dulbecco’s modified eagle medium (DMEM) supplemented with 1% (v/v) penicillin/streptomycin and 10% (v/v) qualified, heat-inactivated foetal bovine serum (FBS) until tissue processing (all from Gibco™). Excised tendons presenting with previously reported definitions of macroscopic evidence of injury were excluded from all experiments (Webbon, 1977; Dakin et al., 2012), and all tendons had a normal histological appearance. Dissections and subsequent cell processing were completed within 24 h of euthanasia.

SDFT frozen sections were prepared as previously described (Godinho et al., 2017). Tissues were briefly washed in Dulbecco’s phosphate-buffered saline (calcium and magnesium free, embedded with optimal cutting temperature compound (OCT; Cell Path, Newtown, United Kingdom) embedding matrix and snap-frozen in pre-cooled hexane on dry ice. Serial longitudinal (6–20 µm thickness) and transverse (30 µm thickness) sections were prepared using a cryostat microtome (OTF5000, Bright Instruments) equipped with MX35 Premier Disposable Low-Profile Microtome Blades (3052835, Fisher Scientific). Tissue sections were mounted on SuperFrost™ Plus Slides (10149870, Fisher Scientific), air-dried at room temperature (RT) for a maximum of 2 h and stored at −80 °C.

Periodic acid-Schiff (PAS) staining was used to detect mucins and basement membrane proteins. Staining was performed using an Alcian Blue (pH 2.5)/PAS staining kit according to manufacturer guidelines (Atomic Scientific). SDFT cryosections (20 µm) were thawed and fixed with 4% PFA/10% NBF for 10 min at RT. Slides were rinsed thoroughly with distilled water, stained with 1% Alcian blue in 3% acetic acid (pH 2.5) for 10 min, and washed thoroughly in distilled water. Slides were treated with 1% periodic acid solution for 10 min at RT, washed with distilled water, then treated with Schiff reagent (Feulgen) for 10 min at RT. Sections were then washed under running tap water until sections presented a magenta colour macroscopically. Sections were then counter-stained with haematoxylin, dehydrated and cleared using an automated slide stainer (Varistain^™ Gemini ES), and mounted with glass coverslips using DPX mountant. Slides were cured at RT overnight and imaged using brightfield microscopy (DM4000B upright microscope) in Leica Application Suite software version 2.6 (Leica Microsystems).

Proteins of interest for immunolabelling were selected based on their expression by tendon progenitor cell subpopulations in previous reports (Yin et al., 2016) and their predicted interactions in Equus caballus (NCBI taxid: 9796) using STRING (version 10.5) network-based predictions for CD146 and LAMA4 (Szklarczyk et al., 2017; Gumucio et al., 2020).

SDFT cryosections were thawed and fixed with acetone (pre-cooled at −20°C) for 10 min, washed three times for 5 min at RT with tris-buffered saline (TBS), incubated in “blocking” buffer (TBS supplemented with 1% (w/v) bovine serum albumin (Scientific Laboratory Supplies), 5% (v/v) goat serum (Sigma), and 5% (v/v) horse serum (Sigma) for 2 h. Horse serum was used to saturate Fc receptors on the surface of cells within the tissue. Sections were incubated with primary antibodies overnight at 4°C (details regarding primary and secondary antibodies are provided in Supplementary Table S1). For negative controls, sections were treated with blocking buffer only. For isotype controls, sections were treated with mouse and rabbit IgG isotype-matched controls diluted in blocking buffer at identical concentration to primary antibodies used. For fluorescent detection (10 µm sections), secondary antibodies diluted in blocking buffer were applied to sections and incubated for 1 h at RT under dark conditions. Sections were washed three times with TBS for 5 min, and mounted with glass coverslips using ProLong^™ Diamond antifade mountant with 4′,6-diamidino-2-phenylindole (DAPI) as a nuclei counterstain. Slides were cured for 24 h at RT under dark conditions, prior to imaging. Negative and isotype matched control images for fluorescent labelling are provided in Supplementary Figure S1. Immunohistochemical labelling (6 µm sections) was performed in a similar manner to fluorescent detection, using an EnVision®⁺ Dual Link System-HRP DAB⁺ system (Dako), with the inclusion of an of EnVision dual endogenous enzyme block for 15 min at RT under dark conditions prior to treatment with blocking buffer, and wash steps were performed using .05% (v/v) TBS-TWEEN20. For immunohistochemical detection, sections were incubated in EnVision peroxidase labelled polymer (conjugated to goat anti-mouse and goat anti-rabbit immunoglobulins) for 30 min at RT. Sections were then washed three times and incubated with EnVision DAB⁺ substrate buffer-3,3′-diaminobenzidine (DAB) chromogen solution for 3 min, rinsed three times with deionised water (diH₂O), counter-stained using haematoxylin according to Delafield, dehydrated and cleared using standard procedures on a Varistain^™ Gemini ES automated slide stainer, then finally mounted with glass coverslips using DPX mountant. Slides were cured at RT overnight and imaged using brightfield microscopy (DM4000B upright microscope) in Leica Application Suite software version 2.6 (Leica Microsystems). Regions clearly showing IFM vascular morphology and positively labelled structures were chosen to demonstrate protein localisation. Negative control images for immunohistochemistry are provided in Supplementary Figure S2.

To distinguish between regions of IFM and fascicular matrix (FM), boundaries between both phases were determined by light refraction in phase contrast images, as well as gross identification by nuclei number and cell morphology (Supplementary Figure S3). For quantification, all settings remained constant between samples including exposure, pixel size, z-step size, and laser settings with all images taken in one single session. For each sample, two distinct areas were imaged in two separate serial tissue sections (2 × sections per horse donor, n = 5). Confocal images are presented as maximum intensity projections from z-stacks containing image slices at a resolution of 512 × 512 × 40 pixels (227.9 × 227.9 × 13.09 µm; .34 µm z-step size) to fully capture tissue depths. Image processing and analysis was performed using Fiji/ImageJ software (Schindelin et al., 2012). For IFM measurements, an area fraction (%) of positively stained pixels were recorded in 8-bit binary images (black = negative, white = positive) to measure expression of markers of interest. To generate binary images for each marker, a background correction was performed to remove noise, followed by a median filter and threshold (Triangle for CD146/MKX = 555 nm, Huang for CD44/CD90 = 633 nm). The lookup table (LUT) of colour channels within images was changed for visualisation purposes.

3D immunolabelling of SDFT segments was performed as previously described (Marr et al., 2020). All steps were performed with orbital agitation. SDFT segments (5 mm × 5 mm  ×  2 mm, n = 2) were washed twice for 12 h with TBS at RT, and permeabilised sequentially in 50% (v/v) methanol:TBS, 80% (v/v) methanol:diH₂O, and 100% methanol for 2 h at 4°C. Samples were washed sequentially for 40 min at 4°C with 20% (v/v) DMSO:methanol, 80% (v/v) methanol:diH₂O, 50% (v/v) methanol:TBS, TBS, and TBS supplemented with .2% (v/v) Triton X-100. Prior to blocking, samples were incubated with a pre-blocking penetration buffer containing .2% TBS-TX100, .3 M glycine, and 20% DMSO for 6 h at 37°C. Equine SDFT segments were blocked for 80 h at 37°C in .2% TBS-TX100 supplemented with 6% (v/v) goat and 6% (v/v) donkey serum and 10% (v/v) DMSO. Primary antibody incubations for CD146 (1:100) were performed at 37°C for 80 h in wash buffer (TBS supplemented with .2% (v/v) TWEEN20), 3% (v/v) goat serum, 3% (v/v) donkey serum, and 5% (v/v) DMSO. Segments were washed 3 × 2 h with wash buffer, incubated with secondary antibodies (1:250, goat anti-rabbit Alexa Fluor® 594) for 36 h at 37°C, washed 5 × 5 min with wash buffer, and counterstained overnight with DAPI (1:2000) diluted in wash buffer. Segments were dehydrated with increasing concentration of methanol, and tissue cleared with immersion in Visikol® HISTO™-1 for 36 h, followed by immersion in HISTO™-2 for at least 36 h at RT. Samples were stored in HISTO™-2 at 4°C prior to confocal imaging. Confocal imaging of regions (approx. 1 mm × 1 mm  × .2 mm) within each sample was performed using a Leica TCS SP8 laser scanning confocal microscope with a motorised stage. Images were captured using lasers emitting light at 405 (blue channel; DAPI) and 561 (red channel; Alexa Fluor 594) nm with laser power <10% and scanning speed = 600 Hz with a HC PL FLUOTAR 10x/.32 dry objective lens, resolution = 1,024 × 1,024 px, pinhole size = 1 Airy unit, frame average = 1, line average = 8, and electronic zoom = .75. 3D renderings were captured in Leica LAS X software (version 3.5.5) within the 3D module.

SDFTs collected under sterile conditions were placed in Petri dishes containing Gibco™ Dulbecco’s PBS (without phenol red, calcium and magnesium) supplemented with 1% (v/v) antibiotic-antimycotic solution. Surrounding peritenon was removed to isolate the tendon core (6 g), which was diced into approximately 4 mm³ pieces, rinsed with DPBS, and digested with 1 mg/mL pronase E (39052, VWR) per 1 g tissue for 6–8 h at 37°C and 5% CO₂ under constant agitation. Following pronase digestion, tissue was digested for a further 24 h with .5 mg/ml collagenase type IV (CLS-4, Lorne Laboratories) and 1 mg/mL dispase II (17105041, Invitrogen) at 37°C and 5% CO₂ with constant agitation (Garvican et al., 2017).

Previous studies have shown that >50% expression of cell membrane proteins can be restored post-digestion by 24 h in vitro culture (Autengruber et al., 2012). Hence, to enhance antigen recovery, freshly digested tendon-derived cells (TDCs) were cultured overnight to maximise CD146 cell isolations. Following this recovery phase, adherent cells were dissociated at 37°C for 10 min using Accutase® solution according to manufacturer’s guidelines. Cells remaining in suspension (i.e. non-adherent populations) were also collected alongside dissociated cells (adherent populations). Cell isolates were washed by resuspension in fresh growth medium and centrifuged at 300 × g for 10–20 min depending on pellet formation. Cell pellets (passage 1; p1) were resuspended in growth medium and separated into single-cell suspensions (SCSs) by passing through a 70 μm cell strainer. SCSs were resuspended in freshly prepared, ice-cold MACS buffer containing sterile-filtered FACSFlow™ (342003, BD Biosciences) supplemented with 1% (w/v) BSA. SCSs were centrifuged for 10 min at 300 × g, resuspended in MACS buffer, and both cell viability and numbers determined by trypan blue (T8154, Sigma-Aldrich) and a haemocytometer. Suspensions with <90% viability were discarded. SCSs were incubated with anti-CD146 antibodies (ab75769, Abcam, Cambridge, United Kingdom) at a concentration of 1 μg/mL for 30 min at 4°C on ice. Following primary antibody incubation, SCSs were washed three times by centrifugation at 300 × g, resuspended in MACS buffer, and incubated with anti-rabbit IgG micro-beads (130-048-602, Miltenyi biotec) diluted in MACS buffer for 15 min at 4°C. SCSs were washed three times by centrifugation at 300 × g and resuspended in MACS buffer. MidiMACS™ LS columns (130-042-401, Miltenyi biotec) were mounted to a MidiMACS™ Separator and multistand (130-042-301, Miltenyi biotec) and washed with MACS buffer according to manufacturer guidelines. MACS-ready SCSs were passed through MidiMACS™ columns and washed with MACS buffer twice. All wash elutions containing negatively selected cells (i.e. CD146^- TDCs) were collected on ice until processing of sub-cultures. Following negative cell depletion, CD146⁺ cells were collected by removing the MACS column from the MACS magnet and eluting the column with MACS buffer and a plunger. All sub-cultures were maintained until a maximum of three passages (p3) to limit phenotypic drift. For downstream assays, cells were dissociated using Accutase® solution (A6964, Sigma-Aldrich). n = 9 (biological replicates).

For direct flow cytometry, .1–.2 × 10⁶ cells were resuspended in DPBS. For CD146⁺ cells, lower concentrations were used according to yields following MACS isolation. All tubes were stored on ice immediately prior to and during flow cytometry. Cell suspensions (50 µl) were incubated with a phycoerythrin (PE)-conjugated variant of the EPR3208 anti-CD146 antibody (1:100, ab209298, Abcam, Cambridge, United Kingdom) on ice for 30 min, washed with DPBS and spun at 400 × g. Supernatant was removed, and pellets resuspended in 500 µl DPBS for immediate flow cytometry analyses.

All flow cytometry acquisition was performed using an air-cooled 3-laser BD FACSCanto II™ flow cytometer (BD Biosciences) equipped with BD FACSDiva (version 8.0.1, BD Biosciences). Acquisition equipment and software were calibrated daily or immediately prior to acquisition using BD FACSDiva™ CS&T Research Beads (BD Biosciences). Data analyses was performed in FlowJo software (version 10, FlowJo LLC). Unstained controls (fluorescence minus one control) were used to gate and discriminate positively and negatively labelled populations (see Supplementaty Figure S4). The percentage of positive cells gated in unstained samples (i.e., autofluorescent cells) was subtracted from stained samples (i.e. experimental cells) to give an overall percentage of immunoreactivity. All experiments recorded a minimum of 10,000 total events (i.e. cells). n = 2 (biological replicates) per cell fraction.

For detection of CD146 in unsorted TDCs, .1–.2 × 10⁶ cells were seeded on sterile 16 mm borosilicate glass circle coverslips coated with poly-L-lysine solution (.01%, sterile-filtered, P4832, Sigma-Aldrich) until 70%–80% confluence. To detect CD146 within MACS-enriched CD146⁺ cells, immunocytochemistry was performed directly on cells (.1–.2 × 10⁶ seeding density) in non-coated culture vessels at 70%–80% confluence. Cells were washed 3 times with DPBS, fixed with pre-chilled (−20°C) acetone:methanol (1:1) for 20 min on ice, then washed three times with DPBS. Cells were blocked for 1 h with blocking buffer as described above. Cells were incubated overnight with primary antibodies overnight at 4°C as described above, washed three times with DPBS, incubated for 1 h with secondary antibodies (1:500, goat anti-rabbit Alexa Fluor® 488 and goat anti-mouse Alexa Fluor® 594).

For direct CD146 labelling in MACS-sorted populations, cells were incubated overnight at 4 °C with phycoerythrin (PE)-conjugated anti-CD146 antibodies (1:100, ab209298, Abcam, Cambridge, United Kingdom). Cells were washed three times with DPBS, labelled with DAPI (1 μg/mL) for 2 min, washed three times with DPBS and mounted using Prolong™ Diamond, cured at RT under dark conditions for 24 h before storing at 4°C until imaging. Fluorescent imaging of TDCs was performed using a Leica SP5 (40 × HCX PL FLUOTAR PH2 NA = .75 objective). For CD146⁺ cells, imaging was performed on a DMIRB inverted microscope (Leica Microsystems, Wetzlar, Germany; 40 × N PLAN L corr PH2 NA = .55 objective). n = 3 (biological replicates).

Bone marrow-derived mesenchymal stromal cells (MSCs) isolated as described previously were kindly provided by Dr Giulia Sivelli (Godwin et al., 2012). MSCs, unsorted TDCs, sorted CD146⁻ cells and CD146⁺ cells were seeded in 6-well plates at a density of 100 cells cm⁻³ (approx. 900 cells) and cultured for 7 d. At termination of cultures, cells were washed 3× with DPBS, fixed with 2.5% glutaraldehyde for 10 min, then washed 3× DBPS (all steps at RT). Cells were stained with .1% (v/v) crystal violet for 30 min at RT. Cells were washed 3x with DBPS and left to air dry at RT. Images were acquired using a flat-bed scanner (Epson Perfection 4990, Epson) at a resolution of 800 dpi. n = 3 for each cell type (biological replicates). n = 2–3 wells for each condition (technical replicates).

MSCs, unsorted TDCs, sorted CD146⁻ cells and CD146⁺ cells were seeded into 12-well plates at a density of .4 × 10⁵ cells per well and cultured for 48 h until adherence in standard growth medium. To induce adipogenesis, standard growth media was removed, and cells were cultured with StemPro® Adipogenesis differentiation media for a further 14 d. Cells were fed induction media every 72 h. Upon termination of culture, monolayers were washed once with DPBS before fixation with 4% PFA/10% NBF for 30 min at RT. To assess intracellular lipid vesicles produced by adipogenic conditions, cells were stained with Oil Red O. Fixed monolayers were rinsed once with distilled water then washed with 60% isopropanol for 5 min at RT. Monolayers were stained for 15 min at RT with a 3:2 working solution of 3-parts .3% (w/v) Oil Red O diluted in isopropanol and 1-part distilled water. Cells were washed repeatedly with distilled water until rinsed clear of precipitating Oil Red O, then counterstained with Harris haematoxylin for 1 min at RT. Imaging was performed on an Axiovert 135TV inverted microscope (Zeiss) using Image Pro Insight version 9.1.4 (Media Cybernetics). n = 3 for each cell type (biological replicates). n = 2-3 wells for each condition (technical replicates).

Following dissociation, MSCs, unsorted TDCs, sorted CD146⁺ and CD146^- cells were seeded into 12-well plates a density of .1 × 10⁶ cells per well with osteogenic media containing 2 mM sodium phosphate dibasic (DiP) or standard growth medium as a control with each condition supplemented with 50 μg/ml ascorbic acid to promote collagen synthesis (Barnes, 1975; Patel et al., 2019). DiP (free phosphate donor) is essential for bone/mineralised extracellular matrix metabolism during osteogenesis (Robey and Termine, 1985). Monolayers were fed with fresh half-media changes corresponding to each condition every 72 h. Cell cultures were terminated after 21 days to assess mineralisation with Alizarin Red S staining (Taylor et al., 2014). Monolayers were rinsed once with DPBS then fixed for 10 min at RT with 2.5% (v/v) glutaraldehyde. Fixed cells were rinsed once with DBPS then three times with 70% ethanol and air-dried at RT overnight. Dried monolayers were subsequently stained with 1% (w/v) Alizarin Red S in diH₂O for 5 min at RT, then washed three times with 50% ethanol and left to air-dry overnight. Imaging was performed as described above. n = 3 for each cell type (biological replicates). n = 2-3 wells for each condition (technical replicates).

Statistical analyses and graphs were produced using GraphPad Prism (version 9.1). Normality tests were performed according to Shapiro-Wilk tests (α = .05). All datasets passed normality tests and were analysed using unpaired two-tailed t-test (significance set to p < .05). Graphs were plotted as mean (µ) ± standard deviation (SD).

PAS staining demonstrated that the IFM contains mucin-rich basement membrane (Figure 1A). Using both CD146 and the IFM basement membrane marker LAMA4 in STRING predictions identified several potential interfascicular cell surface markers including CD44, CD90 (THY1) and CD133 (PROM1), as well as a broader network of interfascicular niche and basement membrane components, including dystroglycan 1 (DAG1), integrin subunit β1 (ITGB1) and fibronectin 1 (FN1) (Figure 1B). To validate these proposed interfascicular cell markers, fluorescent labelling of CD44, CD90, CD146 and MKX was quantified in both fascicular and interfascicular regions. All markers were enriched within IFM (72%–94% positive expression) and had significantly less expression within fascicles; fascicular CD146 expression was less than 1% whereas CD44, CD90 and MKX expression was between 4% and 15% (Figures 1C–J). Using 3D imaging of SDFT labelled with CD146, we identified an interfascicular network of vascular structures within which CD146 cells were localised (Figure 2A). The colocalisation of Desmin (mural cell marker) with CD31 (endothelial marker) and CD146 (pan-vascular marker), alongside LAMA4 (basement membrane; Figures 2B–D) observed in transverse sections confirmed that the structures were vascular, and often found in regions of IFM connecting 3 adjacent fascicles. Within IFM, we observed distinct CD31 endothelium surrounded by Desmin-rich layers, whilst CD146 and Desmin colocalised in smooth muscle and pericyte layers. Labelling of CD31, CD146 and Desmin was also confirmed in vascular layers of larger blood vessels within epitendinous regions (Supplementary Figure S5).

To characterise the distribution of the major components of interfascicular basement membrane, we performed immunolabelling of basement membrane proteins in longitudinal tendon sections, including full-length laminin (pan-laminin), type IV collagen, and Perlecan (Figures 3A–C), all of which localised to the vasculature within IFM. Further labelling with endothelial markers endomucin (EMCN) and von Willebrand factor (VWF) demonstrated abundant expression within the IFM (Figures 3D, E). STRING predictions in Equus caballus identified canonical basement membrane components integrin β1 (ITGB1) and dystroglycan 1 (DAG1), as part of the CD146-LAMA4 interaction network. Hence, we performed labelling of α-dystroglycan (IIH6) and ITGB1 (Figures 3F, H), in addition to network-predicted cell surface marker CD133 (Figure 3G) with all three labelled abundantly within the IFM. As LAMA4/LAMA5 ratios are critical for basement membrane integrity (Galatenko et al., 2018), we also demonstrated labelling for laminin α5 (LAMA5) within the IFM (Figure 3I). We also examined other reported angiogenic mediators, Netrin-1 (NTN1); a reported ligand of CD146, and Neuropilin-1 (NRP1), both of which also localised to the IFM (Figures 3J, K). In addition to demonstrating the presence of these vascular and basement membrane markers in the IFM, our results highlight the variability in IFM vasculature and morphology in the mid-metarcarpal region of the SDFT, with a complex vascular network distinguishable in transverse and longitudinal sections, as well as 3D imaging (Figure 2).

Upon isolation from the SDFT, in vitro labelling of cell surface markers demonstrated that the majority of TDCs exhibited abundant CD44 and CD90 labelling and limited CD146 expression (Figures 4A–C). To study CD146 cells in vitro, we therefore developed a MACS procedure for the enrichment of CD146 cells. Immunocytochemistry of positively sorted CD146 cells confirmed enrichment for cells expressing CD146 (Figure 4D). A single application of MACS was able to yield CD146 cells with enrichment of approximately 64% as determined by flow cytometry (Figures 4E, F). Comparison of cell numbers pre and post MACS showed that approximately 2% of unsorted cells were CD146 positive (Figure 4G), providing further emphasis on the rarity of CD146 cell subpopulations and requirements for optimal enrichment procedures. However, some CD146 positive cells were detected in negative fractions (approx. 14%; Figure 4H).

To assess their clonogenicity and multi-lineage potential, unsorted TDCs, CD146⁺, CD146⁻ cells were subjected to clonogenic, osteogenic and adipogenic assays using MSCs as a positive control. CD146⁺ cells showed no enhanced clonogenicity compared to CD146-negative cells or heterogenous TDCs (Figures 5A–D). For adipogenesis, TDCs, CD146⁺ and CD146⁻ cells all showed limited adipogenic potential when stimulated (Figures 5E–H). Under osteogenic conditions, unsorted TDCs displayed extensive calcium deposition with some mineralised nodules present, however virtually no calcium deposition nor mineralisation was detected in either CD146⁺ and CD146⁻ sorted cell populations (Figures 5M–P).