Postnatal Osterix but not DMP1 lineage cells significantly contribute to intramembranous ossification in three preclinical models of bone injury

Murine models of long-bone fracture, stress fracture, and cortical defect are used to discern the cellular and molecular mediators of intramembranous and endochondral bone healing. Previous work has shown that Osterix (Osx+) and Dentin Matrix Protein-1 (DMP1+) lineage cells and their progeny contribute to injury-induced woven bone formation during femoral fracture, ulnar stress fracture, and tibial cortical defect repair. However, the contribution of pre-existing versus newly-derived Osx+ and DMP1+ lineage cells in these murine models of bone injury is unclear. We addressed this knowledge gap by using male and female 12-week-old, tamoxifen-inducible Osx Cre_ERT2 and DMP1 Cre_ERT2 mice harboring the Ai9 TdTomato reporter allele. To trace pre-existing Osx+ and DMP1+ lineage cells, tamoxifen (TMX: 100 mg/kg gavage) was given in a pulse manner (three doses, 4 weeks before injury), while to label pre-existing and newly-derived lineage Osx+ and DMP1+ cells, TMX was first given 2 weeks before injury and continuously (twice weekly) throughout healing. TdTomato positive (TdT+) cell area and cell fraction were quantified from frozen histological sections of injured and uninjured contralateral samples at times corresponding with active woven bone formation in each model. We found that in uninjured cortical bone tissue, Osx Cre_ERT2 was more efficient than DMP1 Cre_ERT2 at labeling the periosteal and endosteal surfaces, as well as intracortical osteocytes. Pulse-labeling revealed that pre-existing Osx+ lineage and their progeny, but not pre-existing DMP1+ lineage cells and their progeny, significantly contributed to woven bone formation in all three injury models. In particular, these pre-existing Osx+ lineage cells mainly lined new woven bone surfaces and became embedded as osteocytes. In contrast, with continuous dosing, both Osx+ and DMP1+ lineage cells and their progeny contributed to intramembranous woven bone formation, with higher TdT+ tissue area and cell fraction in Osx+ lineage versus DMP1+ lineage calluses (femoral fracture and ulnar stress fracture). Similarly, Osx+ and DMP1+ lineage cells and their progeny significantly contributed to endochondral callus regions with continuous dosing only, with higher TdT+ chondrocyte fraction in Osx+ versus DMP1+ cell lineages. In summary, pre-existing Osx+ but not DMP1+ lineage cells and their progeny make up a significant amount of woven bone cells (particularly osteocytes) across three preclinical models of bone injury. Therefore, Osx+ cell lineage modulation may prove to be an effective therapy to enhance bone regeneration.


Introduction
Bone is one of the only tissues in the body that can heal with scarless tissue regeneration. This remarkable capacity for selfrepair requires a complex, multi-faceted process that involves growth factors, mechanical cues, and unique populations of cells. Based on these environmental factors, bone healing occurs either via endochondral or intramembranous ossification. In endochondral ossification, progenitor cells first differentiate and form a cartilage callus that is later replaced by bone. In contrast, intramembranous ossification results in direct bone formation from progenitor cells, bypassing the cartilage intermediate. Although still unclear, studies indicate that endochondral processes are favored in environments with low oxygen tension, vascular disruption, and some micromotion (non-rigid fixation) (Toyosawa et al., 2004;Marsell and Einhorn, 2011;Bahney et al., 2015;Miller et al., 2015). With nearly 5-10% of fractures progressing to delayed healing or nonunion (Woolf and Pfleger, 2003;Einhorn and Gerstenfeld, 2015) and resulting in increased medical cost and loss of productivity (Bonafede et al., 2013), understanding the cellular and molecular mediators of both and endochondral and intramembranous ossification following bone injury is paramount.
Preclinical models of bone injury are critical for dissecting the cellular and molecular processes controlling endochondral and intramembranous ossification. The most common injury model used is the transverse, full fracture ("Einhorn model") first developed by Bonnarens and Einhorn (Bonnarens and Einhorn, 1984). This model has been adapted for use in both the tibia and femur of rats and mice (Bonnarens and Einhorn, 1984;An et al., 1994;Zondervan et al., 2018;Buettmann et al., 2019), and utilizes blunt trauma to induce a mid-diaphyseal fracture that is stabilized with an intramedullary rod. Due to the semi-stable nature of fixation, this model heals by periosteal intramembranous woven bone formation near the callus periphery and endochondral ossification near the fracture site, with both woven bone tissue and cartilage visible by day 14 postinjury (Colnot, 2009;Buettmann et al., 2019). Tissue transplantation studies have determined that cells from the periosteum are the primary contributors to callus formation in this model, with smaller contributions from the adjacent skeletal muscle and marrow (Colnot, 2009;Julien et al., 2022). In contrast, the rodent stress fracture model, developed and characterized in our lab, utilizes forelimb cyclic fatigue loading to create a non-displaced ulnar fracture that heals predominantly by periosteal intramembranous woven bone formation 10-14 days post-injury (Hsieh and Silva, 2002;Uthgenannt et al., 2007;Wohl et al., 2009;Martinez et al., 2010). Bulk RNAseq analysis comparing the transverse, full fracture model versus stress fracture model in mice indicates that the stress fracture model has a shorter, less pronounced inflammatory phase and a more enriched osteogenic signature (Coates et al., 2019). Another widely used bone repair model is the monocortical defect injury. In this model, a small monocortical defect (0.4-0.8 mm in diameter) is drilled in the mid-diaphysis of the long-bone (Liu et al., 2018;Buettmann et al., 2019;Li and Helms, 2021). Healing progresses after injury with inflammation followed by small amounts of periosteal cartilage callus formation between days 3 and 7 (Hu and Olsen, 2016;Liu et al., 2018). By days 5-10 after injury, intramedullary intramembranous hard callus formation occurs, followed by resolution at days 14-21. Due to the differing healing modalities among these three bone injury models, their simultaneous utilization can provide insights into the unique cellular and molecular mediators of bone healing (Supplementary Figure S1).
Tracking the cellular mediators of bone healing has been aided by the recent development of many tamoxifen-inducible Cre constructs (Cre_ERT2) that can be crossed with fluorescent transgenic reporters (Ai9, Ai14, mTmG, YFP, etc.), allowing for longitudinal tracking of targeted cell populations that contribute to fracture healing in vivo (Feil et al., 2009;Madisen et al., 2010;Abe and Fujimori, 2013;Seime et al., 2015). The emerging role of different skeletal stem cells in fracture repair has been reported by numerous groups and was reviewed recently (Serowoky et al., 2020). We have focused on cells at the later stage of the osteoblast lineage (Osx and later), and used continuous tamoxifen dosing to demonstrate that Osx + lineage cells and their progeny (labeled in Osterix Cre_ERT2 (Maes et al., 2010) Ai9 (Madisen et al., 2010) mice) contributed greater cell numbers than DMP1 + lineage cells (labeled in Dentin Matrix-Protein 1 Cre_ERT2 (Powell et al., 2011) Ai9 (Madisen et al., 2010) mice) and their progeny to woven bone formation in femoral transverse, ulnar stress fracture, and tibial cortical defects (Buettmann et al., 2019). However, because cells were labeled before and during healing by continuous tamoxifen, we could not determine the contribution of pre-existing versus newly differentiated Osx + and DMP1 + lineage cells and their progeny (herein labeled Osx + or DMP1 + lineage cells) to fracture callus tissues. More recent work by our lab group used pulse-chase labeling strategies and demonstrated that pre-existing Osx + and DMP + lineage cells and their progeny contribute significantly to early lamellar bone formation following anabolic (non-damaging) skeletal loading (Zannit and Silva, 2019;Harris and Silva, 2022). Interestingly, we observed that these lineage-labeled cells, especially DMP1 + lineage cells, are rapidly depleted from the periosteal bone surface when a higher loading stimulus induces woven bone formation (Zannit and Silva, 2019;Harris and Silva, 2022). Together these data indicate that Osx + and DMP1 + lineage cells play a role in load-induced bone formation and bone healing, however the relative contributions of pre-existing versus newlyderived Osx + and DMP1 + lineage cells and their progeny across various bone injury types remains poorly defined.
Using both continuous and pulse-chase tamoxifen dosing strategies, we sought to determine the role of pre-existing and/or newly-differentiated Osx + and DMP1 + lineage cells and their progeny in three pre-clinical models of bone repair: transverse femoral fracture, ulnar stress fracture and tibial cortical defect (Supplementary Figure S1). Due to the wider resident bone cell population reported to be targeted with Osx Cre_ERT2 construct (Maes et al., 2010), we hypothesized, that pre-existing Osx + lineage cells target a greater portion of woven bone regions versus pre-existing DMP1 + lineage across all three injury models. Furthermore, we hypothesized that lineage-labeled cells in woven bone callus would be significantly increased with continuous dosing compared to pulse dosing in both Osx Cre_ERT2 Ai9 and DMP1 Cre_ERT2 Ai9 mice.

Mouse lines
All mouse breeding and experimental protocols were approved by Washington University in St. Louis IACUC. Mouse lines including Osx Cre_ERT2 (Maes et al., 2007), DMP1 Cre_ERT2 (Powell et al., 2011), and Ai9 (RCL-tdTomato) (Madisen et al., 2010) were previously generated and described. Osx Cre_ERT2 and DMP1 Cre_ ERT2 breeders were shared from the laboratories of Drs. Henry Kronenberg and Paola Pajevic, respectively. Ai9 (RCL-tdT; Catalog #007909) breeders were purchased from Jackson Laboratories. All mice were obtained from a previously backcrossed C57BL/6J line. To generate inducible Cre reporter mice, male mice hemizygous for Cre were crossed to female mice containing homozygous Ai9 alleles ( Figure 1).

Experimental overview and tamoxifen dosing timeline
Tamoxifen (TMX) was mainly administered by oral gavage dissolved in corn oil (Sigma-Aldrich, CAS #10540-29-1; 100 mg/kg). In an initial cohort of mice (~10% of study), TMX was given by chow diet (ENVIGO TD. 130859; 40 mg/kg daily) for continuous dosing strategies but later discontinued in favor of gavage dosing. We did not observe differences in TdTomato expression during bone healing between tamoxifen administration methods when used continuously (data not shown). Experimental mice harboring Osx Cre_ERT2 +/− ; Ai9 +/− ( + = presence of transgene; − = absence of transgene) or DMP1 Cre_ERT2 +/− ; Ai9 +/− and given TMX served as Cre reporter mice and are labeled as Osx TMX and DMP1 TMX , respectively (Table 1). Mice harboring Cre and Ai9 alleles and only given the vehicle corn oil or chow without tamoxifen were used to assess Cre non-inducible recombination (i.e. "leakiness") and are labeled as Osx VEH or DMP1 VEH , respectively. To label preexisting Osx + and DMP1 + lineage cells (and their progeny) as well as newly-derived Osx + and DMP1 + lineage cells and their progeny after bone injury, mice were given TMX continuously (2x weekly; 100 mg/kg) starting at 2 weeks before injury and throughout healing (Figure 2). These mice are referred to as Osx TMX ;Continuous and DMP1 TMX ;Continuous groups, respectively. To label only pre-existing Osx + and DMP1 + lineage cells (and their progeny), mice were given three TMX doses 4 weeks before bone injury at 8 weeks of age ( Figure 2). These mice are referred to as Osx TMX ;Pulse and DMP1 TMX ;Pulse groups, respectively. We have previously reported residual tamoxifen effects on bone formation are negligible following a 4-week clearance time (Zannit and Silva, 2019). Male and female mice were used as available and in approximately equal numbers among experimental groups. We utilized both males and females

FIGURE 1
Breeding Strategy for Osx + and DMP1 + Osteoblast Cell Lineage Analysis. For experimental animal generation, male mice hemizygous for inducible Cre (Cre_ERT2 +/− ) were crossed to female mice containing homozygous Ai9 alleles (Ai9 +/+ ). Male and female mice hemizygous for Osx or DMP1 Cre_ERT2 and containing the Ai9 allele were utilized for experiments. + = presence of transgene; -= absence of transgene. Figure created in Biorender.
Frontiers in Physiology frontiersin.org 03 in this study since both mouse sexes have been readily utilized in these inducible Cre lines in previous literature (Buettmann et al., 2019;McKenzie et al., 2019;Harris and Silva, 2022). Mice were group housed under a standard 12-h light/dark cycle and given access to food and water ad libitum.

Models of bone repair
For each bone repair model, the right limb was injured whereas the left contralateral limb served as the uninjured control. Mice were given buprenorphine SR LAB (1 mg/kg, s. c.) one hour before injury, and anesthetized during all procedures with isoflurane (1-3% v/v). The right limb was shaved and sterilized with betadine and alcohol (70%) before surgery. Following all bone injury procedures, mice were returned to their cage and placed on electronic heating pads (BeanFarm; Ultratherm) until awake and sternal. Mice were monitored daily for signs of pain and distress and open wounds were quickly resutured and treated with topical triple antibiotic ointment.

Femoral semi-stabilized transverse fracture
Right femurs were prepared for fracture as previously described (McBride-Gagyi et al., 2015;McKenzie et al., 2018). Briefly, a complete (full) transverse bone fracture was made in the femoral mid-diaphysis via three-point bending using a custom designed fixture on a materials testing machine (Instron, DynaMight 8841). The fracture was stabilized with a 24-gauge stainless steel pin (Microgroup, #304 H24RW) and the wound sutured with 3-0 nylon sutures in a simple interrupted pattern (Ethicon, #1669H). Immediately after fracture, lateral radiographs at ×3 magnification (Faxitron, Ultrafocus 100) were taken to verify proper fixation of the fracture site. Mice were allowed to heal for 14 days post-injury (PID 14), when the intramembranous woven bone on the callus periphery and cartilage undergoing endochondral ossification near the Timeline of Pulse Versus Continuous Dosing During Bone Repair. Mice were either given tamoxifen by oral gavage (syringe icon in Figure 2) (TMX; 100 mg/kg) continuously (2 days/wk) starting at 10 weeks old, or pulsed with TMX for three consecutive days starting at 8 weeks old. Then mice underwent bone injury via semi-stabilized femoral fracture, ulnar stress fracture or tibial defect injury at 12 weeks. Mice were sacrifice at predefined timepoints corresponding with robust woven bone formation for each injury model (microCT reconstruction for woven bone; day 7 defect; day 10 stress fracture; day 14 femoral fracture). Days post-injury (PID) are subsequently labeled.   Figure S1) (Einhorn and Gerstenfeld, 2015;Buettmann et al., 2019).

Ulnar stress fracture
Right ulnas had a stress fracture generated as previously described (Martinez et al., 2010;Buettmann et al., 2019). Briefly, a non-displaced (partial) stress fracture was made in the ulnar mid-diaphysis via fatigue loading by cyclic compression on a material testing machine (Instron, DynaMight 8841). Right forelimbs were loaded at a calibrated peak force of 3.1 N to a 50% increase in cyclic displacement from the 10th cycle of loading. Previous work has shown that loading to this average cyclic displacement level in similarly aged wildtype C57BL/6J mice produces a reproducible non-displaced crack on the compressive surface (Buettmann et al., 2019). Mice were allowed to heal for 10 days post-injury (PID 10), when the woven bone response, predominantly formed via periosteal intramembranous ossification, is maximal (Supplementary Figure S1) (Uthgenannt et al., 2007;Martinez et al., 2010).

Tibial cortical defect
The right tibia was prepared as previously described (Kim et al., 2007;Liu et al., 2018). Briefly, a 0.78 mm monocortical circular defect was made using a #68 sterilized drill bit attached to a Dremel tool (Bosch Tool Group, Model 395). It was centered on the anterior medial cortex of the tibia and was located 4.3 mm from the tibial plateau. Following drilling, the cortical defect was irrigated with sterile saline with the wound closed using 5-0 nylon sutures (McKesson, #1034511). Mice were allowed to heal for 7 days post-injury (PID 7), when the woven bone response, formed via intramembranous ossification, encompasses the entire localized marrow space (Supplementary Figure S1) (Uthgenannt et al., 2007;Martinez et al., 2010).

Frozen histology
Injured and contralateral uninjured limbs were harvested at previously mentioned timepoints (transverse femoral fracture-PID14, ulnar stress fracture-PID10, tibial cortical defect-PID7) and immediately fixed in 4% paraformaldehyde (Electron Microscopy Sciences; #15710) for 24 h. A small subset of transverse femoral fracture femurs (injured + contralateral) were also harvested at days 5 (pulse TMX dosing strategy) and day 7 (continuous TMX dosing strategy) to investigate Osx + and DMP1 + lineage cells in the rapidly expanding periosteum and mesenchyme before robust woven bone formation. All specimens underwent standard decalcification for 14 days (14% EDTA, pH 7.0) and subsequent tissue processing (30% sucrose infiltration) followed by embedding and freezing in O.C.T.
Compound (Tissue-Tek ® ; #25608-930). Sections were cut longitudinally at a thickness of 5 µm using the Leica CryoJane Tape-Transfer System and stored at −80°C until use.

Contralateral uninjured femur analysis
40X images were randomly taken from each cortical diaphyseal quadrant (ROIs: anterior-proximal; anterior-distal; posterior-proximal; posterior-distal) from uninjured D7 and D14 continuous TMX and vehicle mice from each Cre_ ERT2 line. We did not see any differences in TdTomato expression (Cre activation) between uninjured D7 and D14 images. Images were blinded and manually counted for TdTomato positive (TdT + ) osteocytes, periosteal labeled surface and endosteal labeled surface using the FIJI (Schindelin et al., 2012) ROI manager and cell counter plug-in. TdTomato positive (TdT + ) cells were normalized to total number of osteocytes or endosteal/periosteal bone surface length for their respective indices. Indices for all four cortical ROIs were averaged on each specimen for final data statistical analysis. 20x images from the femoral mid-diaphysis and distal femoral growth plate were also captured to qualitatively determine relative targeting of skeletal muscle, marrow cells, and chondrocytes based on cellular morphology and anatomical location (Supplementary Figure S2).

PID14 femur fracture analysis
Woven Bone (Intramembranous Region): Any tissue between the skeletal muscle and cortical bone was considered callus tissue. 40X images were randomly taken from two woven bone regions in the callus (~2.5-3 mm peripheral to the fracture site), one on the anterior side of the bone and the other on the posterior side. Images were blinded and manually counted for TdTomato positive (TdT + ) osteocytes within woven bone (Wo.B). Osteocytes were counted as any cell within the woven bone (Wo.B) tissue, excluding the bone surface and adjacent marrow spaces (marked by clusters of overlapping cells). The multilayered outline of cells encompassing the perimeter of the woven bone (i.e. expanded periosteal perimeter) was also manually counted for TdT + cells using the FIJI (Schindelin et al., 2012) ROI manager and cell counter plug-in. Both Frontiers in Physiology frontiersin.org indices were normalized to total Wo. B osteocytes and callus perimeter length, respectively. TdT + cellular area was also computed automatically by FIJI as per previous methods and normalized to total callus area (Wang et al., 2019;Shihan et al., 2021). In brief, TdT + cell area was counted automatically by collecting data only on the red channel (split channel function), thresholding to make the image binary (threshold 190), and calculating the thresholded area (particle analysis-no restrictions on size/circularity). Cartilage (Endochondral Region): 40X images were randomly taken from two cartilage regions anterior and posterior to the fracture site away from the mineralizing woven bone front. TdT + cartilage cellular area was also computed automatically by FIJI as per exact methods listed for the woven bone region and normalized to total cartilage area. TdT + cartilage cells per total cartilage cells were calculated for the same images by splitting the red and blue channels, and using particle analysis to automatically count the ratio of TdT + to DAPI + cells. In brief, TdT + cells were counted by binary thresholding (threshold 190), discretizing overlapping cells by watershed analysis, and running particle analysis (size: 20-200 microns; circularity: 0.2-1.0). DAPI + cells were counted by binary thresholding (threshold 150), discretizing overlapping cells by watershed analysis, and running particle analysis (size: 20-200 microns; circularity: 0.2-1).

PID10 stress fracture analysis
To complement the woven bone analysis for femoral fracture mice, the periosteal stress fracture callus was also analyzed. For this, 10X images were taken that were centered at the stress fracture crack line of the compressive region of the callus (this ROI encompassed the majority of the callus). Images were blinded and manually counted for TdT + positive osteocytes within woven bone (TdT + Wo.B Osteocytes) and TdT + callus perimeter similar to methods used for the femoral woven bone analysis. TdT + Wo. B cells were calculated for the same images by splitting the red and blue channels and using particle analysis to automatically count the ratio of TdT + to DAPI + cells within the 10X callus region, regardless of location. This cell population included the total number of TdT + woven bone lining cells, woven bone marrow cells and osteocytes. In brief, TdT + cells were counted by binary thresholding (threshold 150), discretizing overlapping cells by watershed analysis, and running particle analysis (size: 20-200 microns; circularity: 0.2-1). DAPI + cells were counted by binary thresholding (threshold 20), discretizing overlapping cells by watershed analysis, and running particle analysis (size: 20-200 microns; circularity: 0.2-1).

PID7 tibial cortical defect analysis
To determine if contributions of pre-existing Osx + and DMP1 + lineage cells differ following marrow-derived intramembranous bone repair, a small number of mice were given cortical defect injuries following pulse TMX regimens. (Cortical defect experiments were not performed under the continuous TMX protocol.) 10X images were taken centered at the PID7 cortical defect site around the anterior medial surface of the tibia and used to investigate TdT + cells expression within the intramedullary woven bone.

Statistics
Quantitative outcomes of TdTomato cellular expression per tissue area (woven bone or cartilage), per perimeter (callus or bone), or per cell number were analyzed within each inducible Cre line (Osx or DMP1). Due to the smaller sample sizes used (n = 2-4), data normality was first assessed by Q-Q plots and assumed to be normal if not deviating significantly from a straight diagonal line. Depending on outcome, data was compared by unpaired t-test or ANOVA (normally distributed) or Mann-Whitney or Kruskal-Wallis to test for the significant effects of tamoxifen dosing (continuous, pulse, vehicle) in GraphPad Prism Pro Version 9 (La Jolla, CA). The type of statistical test for each figure is noted in the legend. Direct statistical comparisons between Cre lines were avoided due to potential confounding technical differences in Ai9 recombination efficiency between Osx Cre_ERT2 and DMP1 Cre_ERT2 constructs, which may not reflect accurate changes in biology. Mouse sex was not tested as an independent variable because our study wasn't adequately powered to compute male and female differences (so they were pooled for analysis). For added clarity, data points from male and female mice are represented on graphs as diamonds and circles, respectively as noted in each figure's caption. Post-hoc Tukey's (parametric -ANOVA) or Dunn's (nonparametric -Kruskal-Wallis) were used to determine significance differences between individual groups after accounting for multiple comparisons corrections. Statistical significance was defined as p < 0.05 and trending values were denoted as p < 0.10. Data are presented as mean ± SD with individual sample sizes for each outcome denoted as data points in each graph and in the figure caption.

Results
Inducible Osx Cre_ERT2 under continuous dosing targets a higher percentage of femoral cortical bone cells in uninjured bones compared to inducible DMP1 Cre_ERT2 Uninjured contralateral femurs were first assessed for TdT + cells in intracortical osteocytes and bone surfaces at the middiaphysis following continuous TMX administration for 4 weeks. Overall, Osx TMX ;Continuous femurs showed greater targeting of cells compared to DMP1 TMX ;Continuous femurs in Frontiers in Physiology frontiersin.org 06 each bone component analyzed ( Figure 3). For example, Osx TMX ; Continuous femurs had 98% of osteocytes labeled TdT + compared to 72% in DMP1 TMX ;Continuous femurs. In addition, Osx TMX ; Continuous femurs had 91% and 85% of the periosteal and endosteal surfaces labeled, whereas DMP1 TMX ; Continuous femurs had 66% and 77% of the periosteal and endosteal surfaces labeled, respectively ( Figure 3). The majority of TdT + labeling was attributed to tamoxifen induction as expected, as both Osx TMX ;Continuous and DMP1 TMX ;Continuous femurs had significantly increased TdT + labeling in all investigated cortical compartments versus respective vehicle-treated controls (p < 0.05). In the absence of TMX, the periosteal surface and endosteal surface had negligible non-inducible recombination ("leakiness") in either Cre_ERT2 line, however leakiness was readily apparent in intracortical osteocytes. For example, 9.4% of osteocytes were TdT + in Osx VEH femurs while in DMP1 VEH femurs 21% of osteocytes were TdT + (Figure 3).
Qualitative assessment of TdT + cell labeling outside the cortical diaphyseal bone in the marrow, skeletal muscle and primary spongiosa was also performed (Supplementary Figure   S2). Osx TMX ;Continuous and DMP1 TMX ;Continuous femurs both showed minimal TdT + expression in marrow cells (Supplementary Figure S2; Panel 1). Notably, both DMP1 TMX ; Continuous and DMP1 VEH femurs showed robust TdT + expression in skeletal muscle cells, indicative of non-inducible recombination at this site (Supplementary Figure S2; Panel 2). Looking at the distal femoral growth plate, a place undergoing endochondral ossification similar to the fracture callus, Osx TMX ; Continuous femurs showed greater targeting of growth plate chondrocytes (white arrows) and trabecular bone within the primary spongiosa compared to DMP1 TMX ;Continuous femurs (Supplementary Figure S2; Panel 3). In summary, these results demonstrated that DMP1 VEH femurs had greater non-inducible TdTomato expression and hence leakiness in multiple tissue compartments, notably intracortical osteocytes and skeletal muscle, compared to Osx VEH . However following a 4-week period of TMX dosing, Osx Cre_ERT2 caused Cre activation in a greater number of bone cells compared to DMP1 TMX ; Continuous mice, such as intracortical osteocytes, periosteal and endosteal lining cells, and growth plate chondrocytes.

FIGURE 3
Osx Cre_ERT2 has greater diaphyseal cell targeting than DMP1 Cre_ERT2 in uninjured femurs (osteocytes and bone lining cells). 2.5X (scale bar 1 mm) and 40X images (scale bar 50 µm) were randomly taken from each cortical diaphyseal quadrant (Black ROIs) and used for quantification of Cre specificity from uninjured D7 and D14 continuous TMX and vehicle femurs from each Cre_ERT2 line. TdTomato positive (TdT + ) osteocytes (white arrow) were normalized to total number of osteocytes (TdT + and TdT − cells -orange arrow). TdTomato positive endosteal (TdT + Ec. S) and periosteal bone surface (TdT + Ps. S) were normalized to total bone surface length (BS). Data presented as mean ± SD with n = 3 per group. Mouse sex of each data point is represented by shape (circle-female; diamond-male). Effects between continuous and vehicle dosing within each inducible Cre line were compared by Unpaired t-test *p < 0.05; **p < 0.005; ***p < 0.0005 Abbreviations: Ma = Marrow; Ct. B = Cortical bone; Mu = Skeletal muscle.
Frontiers in Physiology frontiersin.org 07 Pulse-chase labeling reveals that preexisting Osx + but not DMP1 + lineage cells and their progeny give rise to most intramembranous woven bone osteocytes following femoral fracture The callus from the fractured femurs was next analyzed for TdTomato expression in woven bone regions at the callus periphery, known to primarily undergo intramembranous ossification, and revealed a large contribution of pre-exisiting Osx + but not DMP1 + lineage cells. With pulse dosing, Osx TMX ;Pulse callus had noticeably increased TdT + stained woven bone area compared to Osx VEH control (16% vs. 0.02%; Figure 4A) but still nearly 3-fold less staining less than Osx TMX ;Continuous calluses (16% vs. 44%; Figure 4A). Notably, TdT + osteocytes were significantly more abundant in Osx TMX ;Pulse calluses (74%) than Osx VEH controls (0.47%) although less abundant than Osx TMX ; Continuous femurs (99%, p < 0.05; Figure 4A). Lastly, Osx TMX ; Pulse femurs also had more TdT + labeled cells lining the perimeter of the intramembranous woven bone callus compared to Osx VEH (24% vs 0%) but this only reached signficance versus vehicle in the Osx TMX ; Continuous group (66% vs 0%, p < 0.05; Figure 4A). Taken together,
Frontiers in Physiology frontiersin.org 08 these results indicate that pre-existing Osx + lineage cells and their progeny (identified by pulse-chase labeling) make up about a fifth of intramembranous callus tissue (16%), which is two to three-fold less than the amount labeled by continuous dosing (44%), which captures both pre-existing and newly differentiated Osx + lineage cells and their progeny. Notably, the majority of intramembranous woven bone osteocytes are derived from pre-existing Osx + lineage cells and their progeny (74%).
Comparing Osx and DMP1 Cre_ERT2 models, it appears that most osteocytes within woven bone come from pre-existing Osx + lineage cells and their progeny and will acquire DMP1 + expression as evidenced by the similar labeling of osteocytes between Osx TMX ; Pulse and DMP1 TMX ; Continuous calluses (74% versus 87%). This is further supported when looking at early timepoints of fracture healing such as PID5 and PID7 in pulsed and continuous fracture calluses, respectively (Supplementary Figures S3, S4). For example, by PID5 a greater extent (i.e. longitudinal length) of the expanded periosteum is labeled by pre-existing Osx + lineage cells than DMP1 + lineage cells (Supplementary Figure S3) resulting in a greater proportion of pre-existing and newly-derived Osx + lineage cells and their progeny compared to DMP1 + lineage cells and their progeny within woven bone tissue at PID7 (Supplementary Figure S4). Overall, our results indicate that pre-existing Osx + lineage cells and their progeny, but not DMP1 + lineage cells (and their progeny), contribute to early Newly-derived but not pre-existing Osx + (panel (A)) and DMP1 + (panel (B)) lineage cells and their progeny make up cartilage callus following femoral fracture (endochondral region). 2.5X (Overview; scale bar 1 mm) and 40X images (cartilage -scale bar 50 μm; black/white ROI boxes) were randomly taken and averaged from two regions of interest centered in cartilage regions adjacent to the PID14 fracture site (Fx). TdTomato positive callus area were calculated from automated thresholding for TdT + area within each field of view. Cartilage TdTomato positive cells (TdT + Cg cells -white arrow) were thresholded, counted and normalized to total number of DAPI + chondrocytes (TdT + and TdT − cells -orange arrow). Data presented as mean ± SD with n = 2-4 per group. Mouse sex of each data point is represented by shape (circle-female; diamond-male). & Significant Tamoxifen Effect by 1-WAY ANOVA. #p < 0.10, *p < 0.05; **p < 0.005; Significantly Different by Tukey Post-Hoc.
Frontiers in Physiology frontiersin.org 09 woven bone formation in the fracture callus both by lining new woven bone surfaces and becoming embedded osteocytes.
Newly-derived but not pre-existing Osx + and DMP1 + lineage cells and their progeny make up cartilage callus following femoral fracture TdTomato expression in multiple cartilage regions immediately adjacent to the femoral fracture site was averaged to evaluate the role of pre-existing versus newly-derived Osx + and DMP1 + lineage cells and their progeny in endochondral ossification at 14 days post-fracture. Overall, we saw little evidence of pre-existing Osx + or DMP1 + lineage cells contributing to cartilage formation. For example, with pulse dosing, Osx TMX ; Pulse calluses had non-signficant TdT + stained cartilage callus tissue (0.3% versus 0.0%) and cartilage cells (2.3% versus 0.1%) compared to Osx VEH control ( Figure 5A). However, with continuous tamoxifen dosing there were trending increases in Osx TXM ;Continuous TdT + stained callus tissue (6.2 versus 0.3%) and cells (48% versus 2.3%; p < 0.10)
Frontiers in Physiology frontiersin.org compared to Osx VEH control ( Figure 5A) indicating the majority of Osx + lineage cartilage cells are newly-derived following fracture. Similarly, with pulse dosing, DMP1 TMX ; Pulse calluses had minimal cartilage callus area (0.1% versus 0.0%) and cartilage cells (0.7% versus 0.0%) targeted compared to DMP1 VEH but this was significantly enhanced with continuous tamoxifen expression ( Figure 5B). While continuous TMX dosing resulted in significant cartilage labeling compared to vehicle controls in both Cre lines, Osx TMX Continuous femurs, on average, targeted approximately 10-fold more chondrocytes compared to DMP1 TMX Continuous femurs (48% versus 6.0%). These data indicate that pre-existing Osx + and DMP1 + lineage cells and their progeny give rise to minimal chondrocytes in the fracture callus. However, it appears that a large portion of total chondrocytes become Osx + lineage cells once formed in the fracture callus between PID7 and PID14, with an even smaller population of chondrocytes becoming DMP1 + lineage cells near sites of endochondral ossification.
Pre-existing Osx + but not DMP1 + lineage cells and their progeny contribute a significant but small portion of periosteal woven bone osteocytes following ulnar stress fracture The ulnar stress fracture model was utilized in each Cre_ ERT2 line (Osx and DMP1) with pulse and continuous TMX dosing to further assess the role of each osteoblast cell lineages' contribution to periosteal woven bone intramembranous repair. These results partially mirrored the findings in the intramembranous region of the femoral fracture callus and suggest that pre-existing Osx + but not DMP1 + lineage cells and their progeny contribute significantly more to woven bone formation following stress fracture. For example, Osx TMX ; Pulse stress fracture calluses had signficantly increased TdT + cells within the woven bone regions of the stress fracture callus compared to Osx VEH control (20% versus 0%; p < 0.05; Figure 6A). However, the overall TdT + cell population was significantly less in Osx TMX ; Pulse calluses compared to Osx TMX ; Continuous Calluses (20% vs 82%; p < 0.05; Figure 6A). Stratifying TdT + cells based on location, the majority of TdT + cells in Osx TMX ; Pulse stress fracture calluses were embedded woven bone osteocytes (24% of Wo. B osteocytes TdT + ; p < 0.05 compared to 0% in Osx VEH ) but not callus peripheral cells in the expanded periosteum (4.6% peripheral cells TdT + ; p > 0.05 compared to 0% Osx VEH ). With continuous TMX dosing, Osx TMX; Continuous calluses showed a significant elevation in TdT + targeting of woven bone osteocytes (95% of Wo. B osteocytes TdT + ; p < 0.05) and expanded callus periosteum (72% of peripheral cells TdT + ; p < 0.05) compared to Osx TMX ; Pulse and Osx VEH groups ( Figure 6A). These data indicate that pre-existing Osx + lineage cells and their progeny make up a small but significant portion of total Osx + lineage cells in intramembranous callus tissue following stress fracture, mainly in the form of woven bone osteocytes. In addition, the majority of intramembranous callus cells acquire Osx + lineage cell specification after injury.
Pre-exisiting Osx + but not DMP1 + lineage cells and their progeny significantly contribute to cells in the intramedullary woven bone following tibial cortical defect To investigate the contribution of Osx + and DMP1 + lineage cells and their progeny in another widely used model of bone repair, a tibial cortical defect was created in Osx TMX ; Pulse and DMP1 TMX; Pulse mice. With TMX pulsing, TdT + signal was strongly present in the majority of woven bone cells, including woven bone osteocytes, woven bone lining cells and injured marrow surrounding the defect in Osx TMX ; Pulse but not DMP1 TMX; Pulse mice ( Figure 7). Osx VEH and DMP1 VEH defects showed minimal non-inducible expression. In all, this suggests that the majority of woven bone cells following cortical defect arises from pre-existing Osx + and their progeny but not preexisting DMP1 + lineage cells and their progeny. Frontiers in Physiology frontiersin.org

Discussion
We investigated the contributions of pre-existing versus newly-derived Osx + and Dmp1 + lineage cells and their progeny to regenerated tissues in three preclinical models of bone injury using inducible Osx Cre_ERT2 Ai9 and DMP1 Cre_ERT2 Ai9 mice. Using two different tamoxifen dosing regimens: 1) pulse-labeling with washout (4 weeks) before injury or 2) biweekly dosing before (2 weeks) and during bone injury healing, we found across injury models (femoral fracture, ulnar stress fracture, tibial cortical defect) that pre-existing Osx + lineage cells and their progeny, but not pre-existing DMP1 + lineage cells and their progeny, contributed a significant amount of total TdT + labeled tissue area and cells versus respective vehicle controls (Figure 8). These results support our first hypothesis and demonstrate that pre-existing Osx + lineage cells and their progeny but not DMP1 + lineage cells and their progeny are a significant source of woven bone forming osteoblasts and osteocytes following bone injury. In addition, continuous tamoxifen administration significantly increased labeling within each inducible Cre line. For example, Osx Cre_ ERT2 showing significantly higher targeting of callus tissue with continuous TMX dosing across all scenarios compared to DMP1 Cre_ERT2, supporting our second hypothesis ( Figure 8). Importantly, these results suggest that preexisting Osx + lineage cells and their progeny are likely critical for postnatal injury-induced bone formation, although their contribution varies based on skeletal site and the type of bone injury.
In the femoral fracture callus at day 14 post-injury, the specific Cre model and tamoxifen dosing regimen led to differential targeting of cells based on callus region. Comparing pulse to continuous dosing allowed us to see the maximum contribution of Osx + or DMP1 + lineage cells and their progeny to bone healing (pre-existing and newlyderived) within each Cre line. Using this methodology, we saw when comparing Osx TMX ; Pulse to Osx TMX ; Continuous, that the majority of Osx + lineage osteocytes (74% versus 99%) and some callus border cells (24% versus 66%) and virtually no chondrocytes (2.3% versus 48%) were derived from preexisting Osx + lineage cells and their progeny ( Figure 4A and Figure 5A). Comparing DMP1 TMX ; Pulse to DMP1 TMX ; Continuous, we saw that out of all DMP1 + lineage cells involved in femoral fracture, only a small percent of preexisting DMP1 + lineage cells become osteocytes (26% versus 87%), and virtually none become callus border cells (0.5% versus 36%) or chondrocytes (0.7% versus 6.0%; Figure 4B and Figure 5B). Our DMP1 TMX ; Pulse results are similar to previous results by Root et al., whom utilized the DMP1 CreERT2; Ai9 mouse crossed to the 2.3 kb Col1 Cre thymidine kinase (tk) mouse (Visnjic et al., 2001) during transcortical fracture healing in 8 week old mice (Root et al., 2020). In this study, the authors used ganciclovir administration (GCV) for 16 days prior to fracture to eliminate the proliferating 2.3Col1 tk + osteoblast lineage cells, which overlap significantly with pulse-labeled DMP1 + lineage cells (Matic et al., 2016), leaving only transcortical DMP1 + lineage cells prior to injury (Root et al., 2020). Tracing of these pre-labeled transcortical DMP1 + lineage cells following fracture for 7 days revealed minimal contribution of DMP1 + lineage cells to periosteal woven bone, although all DMP1 + lineage cells lining woven bone were 2.3Col1 GFP + , suggesting that they were bone-forming osteoblasts (Root et al., 2020). Similarly, our data also suggest minimal contributions of pre-existing DMP1 + lineage cells (even those that may not be targeted by 2.3Col1 tk) and their progeny to woven bone following transverse fracture.
Looking at Osx and DMP1 CreERT2 lines collectively in our results indicate that pre-existing Osx + osteoprogenitor lineage cells and their progeny at the time of fracture readily become woven bone callus lining cells, woven bone forming osteoblasts and the majority of embedded osteocytes, whereas pre-existing DMP1 + lineage cells and their progeny are mainly absent from fracture callus tissues (a few become woven bone lining osteoblasts and osteocytes) (Figure 8). Furthermore, the higher percentage of overall Osx + lineage cells compared to DMP1 + lineage cells in intramembranous callus tissue area (44% versus 19%) and chondrocytes (48% versus 6.0%), but similar overlap in the percentage of osteocytes (99% versus 87%) targeted under continuous dosing regimens suggests that Osx Cre_ERT2 targets a wider population of bone cells that eventually go on to become DMP1 + lineage concurrent with woven bone formation and matrix embedding (i.e. osteocytogenesis in woven bone) (Figure 8). This wider targeting of osteoblast lineage cells using Osx Cre_ ERT2 over DMP1 Cre_ERT2 is supported by previously published works that DMP1 is expressed at the mature osteoblast and osteocyte stages of differentiation during matrix mineralization and osteocyte cell embedding (Maes et al., 2007;Lu et al., 2011;Powell et al., 2011;Kim et al., 2012;Kalajzic et al., 2013;Matic et al., 2016;Shiflett et al., 2019).
Based on the continuous dosing regimen labeling preexisting and newly-derived cells and their progeny, we also found that cells will acquire Osx + lineage specification and to a lesser degree DMP1 + lineage specification within sites of endochondral ossification at day 14 in the femoral fracture callus (Figure 8). A limitation of this work is that we did not use co-staining to better characterize the identity of these newlyderived Osx + or DMP1 + lineage cells observed near the cartilage to bone transition zone (e.g., Collagen type II or Collagen X staining). Another limitation is that our study was underpowered to detect differences in these TdT + cell populations between mouse sexes. As emerging data suggests that mouse sex may differentially regulate the response to tamoxifen (Ceasrine et al., 2019) and lead to biological changes in fracture healing Frontiers in Physiology frontiersin.org 13 (particularly cartilage formation) (Haffner-Luntzer et al., 2021), future research is needed to determine if mouse sex significantly alters Cre specificity during bone healing. The similar trends in Cre specificity seen between males and females in our data suggest that mouse sex effects are subtle compared to the tamoxifen dosing regimen and Cre construct used for inducible cell targeting. Non-etheless, the anatomic location of these newly-derived Osx + and DMP1 + lineage cells and their progeny in both mouse sexes, within the chondrocyte transition zone (near vasculature), are in line with other reports showing that Osx + lineage cells labeled continuously during fracture healing can demonstrate a hypertrophic chondrocyte phenotype (labeled by Collagen X) (Hu et al., 2017;Buettmann et al., 2019). In addition, DMP1 mRNA has previously been shown via in situ hybridization to be weakly expressed in a small number of hypertrophic chondrocytes in the growth plate (Lu et al., 2011) and during fracture repair (Toyosawa et al., 2004). These results, along with our own, are consistent with the trans-differentiation of chondrocytes to osteoblast lineage cells as proposed by others (Bahney et al., 2015;Hu et al., 2017). Our use of pulse-labeling strategies extends these prior results and indicates that chondrocytes likely do not arise from pre-existing Osx + or DMP1 + lineage cells following bone injury. Therefore, researchers studying conditional gene deletion postnatally during transverse fracture repair would minimize targeting of cartilage cells with Osx Cre_ERT2 or DMP1 Cre_ ERT2 mice by using a similar pulse dosing strategy. However, our results differ from Mizoguchi et al., which showed that Osx + lineage cells labeled at postnatal day 5 (P5), can become fracture callus chondrocytes following bone injury nearly 15 weeks later (Mizoguchi et al., 2014). Overall, this suggests that there is a critical time-window between birth and 8 weeks postnatally in which pre-existing Osx + osteoprogenitor cells are bipotent in vivo.
In order to complement our femoral fracture results, we tested the requirement of pre-existing and newly-derived Osx + and DMP1 + lineage cells and their progeny to contribute to stress fracture repair. This model heals predominantly by intramembranous ossification (Martinez et al., 2010) and has not been extensively explored in the literature using Cre reporter mice. Our results largely mirror the woven bone results seen at day 14 of healing in the intramembranous region of the femoral fracture callus, with pre-existing Osx + but not DMP1 + lineage cells and their progeny significantly contributing to callus woven bone cells based on changes from each Cre lines respective vehicle controls ( Figure 6). These results reinforce that the stress fracture model largely mirrors the intramembranous processes in the femoral fracture model as we previously reported (Wohl et al., 2009). However, what was striking is that the overall percentage of total Osx + lineage Wo.B osteocytes labeled in pulse versus continuous dosing was much lower in the stress fracture (~20% total Osx + lineage Wo.B osteocytes came from pre-existing Osx + lineage cells) compared to femoral fracture (~75% total Osx + lineage Wo.B osteocytes came from pre-existing Osx + lineage cells). In contrast, osteocytes expressing DMP1 + cell lineage between pulse vs. continuous labeling were relatively unchanged (~20-25% total DMP1 + lineage Wo.B osteocytes came from pre-existing DMP1 + lineage cells) between full fracture and stress fracture repair. These findings suggest, that pre-existing Osx + lineage osteoprogenitors and their progeny contribute less to total callus area and cellularity in the less traumatic ulnar stress fracture than the femoral fracture model. The overall result that pre-existing Osx + lineage cells contribute more to callus cells than DMP1 + lineage cells with higher degrees of bone damage are consistent with previous reports using anabolic tibial loading at graded force levels (Harris and Silva, 2022), and may potentially reflect the smaller overall cellularity and decreased proliferative processes in stress fracture versus full fracture injuries as previously shown (Coates et al., 2019). However, it may also reflect changes in anatomic location (ulna versus femur) or slight differences in analysis regions between the two fracture models used in the current study (i.e. majority of callus used to analyze stress fracture vs. callus periphery in transverse fracture).
Despite these differences in pre-existing Osx + lineage cell recruitment, the full fracture and stress fracture model also show some striking similarities in the types of cells targeted between both inducible Cre drivers. For example, in both models, Osx TMX ; Continuous but not DMP1 TMX ; Continuous labeling results in a strong TdT + signal within woven bone marrow spaces known as sites of progenitor cell and blood vessel invasion (Figure 4 and Figure 6) that support bone healing (Hausman et al., 2001;Lu et al., 2006;Tomlinson et al., 2013). Furthermore, Osx TMX ; Pulse labeling shows much weaker TdT + signal compared to Osx TMX ; Continuous calluses at these woven bone marrow sites. Maes et al. has demonstrated previously that osteoblast precursors labeled instantaneously by Osx Cre (but not Collagen 1 Cre), can take on a pericyte-like profile and co-invade woven bone spaces in the fracture callus, thereby supporting angiogenesis and subsequent bone formation (Maes et al., 2010). This concept was further supported by Buettmann et al., where using continuous dosing in Osx Cre_ERT2 VEGFA fl/fl mice led to decreased femoral fracture and stress fracture angiogenesis and subsequent woven bone formation (Buettmann et al., 2019). Our pulse labeling strategy expands upon these results and suggests that pre-existing Osx + lineage cells and their progeny, due to reduced TdT + targeting of Wo.B marrow cells in femoral fracture and ulnar stress fracture, likely do not co-invade with vasculature (neither do more mature DMP1 + lineage cells). Thus, if Osx Cre_ERT2 VEGFA fl/fl mice were pulse-dosed with TMX (rather than continuously dosed as previously performed in Buettmann et al., 2019), we hypothesize that femoral fracture and ulnar stress fracture healing would not be impaired.
Lastly, we showed that pre-existing Osx + but not DMP1 + lineage cells and their progeny make up a majority of intramedullary woven bone tissue following monocortical defect. In particular, Osx TMX ; Pulse showed TdT + cells encompassing the majority of woven bone surfaces, osteocytes and even adjacent marrow, whereas these sites were largely void of TdT + expression in DMP1 TMX ; Pulse defects ( Figure 7). These results indicate that, at 8 weeks age, pre-existing Osx + lineage cells but not pre-existing DMP1 + lineage cells and their progeny significantly contribute to intramedullary bone formation following cortical defect. Although the exact bone compartment contributing to this differential TdTomato expression is unknown, work by Colnot suggests that both endosteal and marrow derived cellular niches act locally to play a large role in monocortical defect healing (Colnot, 2009). Therefore, it is likely that the pre-existing Osx + but not pre-existing DMP1 + lineage cells contributing to defect labeling are derived from the endosteum or marrow niche. Although continuous labeling revealed similar endosteal ( Figure 3) and minimal marrow (Supplementary Figure S2; Panel 1) targeting in Osx Cre_ERT2 and DMP1 Cre_ERT2 mice, pre-existing lineage cells at these sites were not quantitated in pulse-labeled uninjured specimens, which is a limitation of the current work. Other reports indicate that later pulse labeling (14 days postnatal or after) in Osx Cre_ERT2 and DMP1 Cre_ERT2 labels vascular associated reticular marrow cells and endosteal bone-lining cells that decrease in number over time (Powell et al., 2011;Kim et al., 2012;Park et al., 2012). For example, Matic et al. demonstrated that DMP1 + lineage endosteal bone lining cells decrease by 50-75% 3 weeks following tamoxifen induction (Matic et al., 2016). Therefore, it is possible that the differential Osx + and DMP1 + lineage cell labeling in the intramedullary woven bone seen in our study is due to a preferential decline in DMP1 + over Osx + lineage endosteal cells during the 4 weeks between pulse labeling and the cortical defect creation. Another possibility is that Osx Cre_ ERT2 targets a marrow or endosteal lineage cell population with higher regenerative capacity overall compared to DMP1 Cre_ERT2. This differential Cre specificity would be in line with previous reports showing that peri-vascular stromal Osx + lineage cells in the marrow have high regenerative capacity following bone injury (Park et al., 2012;Mizoguchi et al., 2014). Future studies, using duallabeling strategies, to determine the instantaneous degree of overlap between Osx + and DMP1 + lineage cells in various bone compartments, would be particularly informative.
In all, we have shown in the current study that pre-existing postnatal Osx + lineage cells and not pre-existing DMP1 + lineage cells and their progeny contribute significantly to cells populating woven bone in multiple widely used preclinical models of bone injury. This study underscores the importance that pre-existing Osx + lineage cells play in bone regeneration, especially for early woven bone formation, and suggest that bone targeting therapies to improve healing might target this particular cellular subset. Furthermore, this work provides a tissue and cellular atlas for inducible Cre targeting using the Osx Cre_ERT2 and DMP1 Cre_ERT2 models during bone healing, thereby providing a framework for researchers using these widely available tools in future studies.

Data availability statement
The original contributions presented in the study are included in the article/Supplementary Material, further inquiries can be directed to the corresponding author.