Regenerative and Anti-Inflammatory Potential of Regularly Fed, Starved Cells and Extracellular Vesicles In Vivo

Background: Mesenchymal stem/stromal cells (MSC) have been employed successfully in immunotherapy and regenerative medicine, but their therapeutic potential is reduced considerably by the ischemic environment that exists after transplantation. The assumption that preconditioning MSC to promote quiescence may result in increased survival and regenerative potential upon transplantation is gaining popularity. Methods: The purpose of this work was to evaluate the anti-inflammatory and regenerative effects of human bone marrow MSC (hBM-MSC) and their extracellular vesicles (EVs) grown and isolated in a serum-free medium, as compared to starved hBM-MSC (preconditioned) in streptozotocin-induced diabetic fractured male C57BL/6J mice. Results: Blood samples taken four hours and five days after injection revealed that cells, whether starved or not, generated similar plasma levels of inflammatory-related cytokines but lower levels than animals treated with EVs. Nonetheless, starved cells prompted the highest production of IL-17, IL-6, IL-13, eotaxin and keratinocyte-derived chemokines and induced an earlier soft callus formation and mineralization of the fracture site compared to EVs and regularly fed cells five days after administration. Conclusions: Preconditioning may be crucial for refining and defining new criteria for future MSC therapies. Additionally, the elucidation of mechanisms underpinning an MSC’s survival/adaptive processes may result in increased cell survival and enhanced therapeutic efficacy following transplantation.


Introduction
The healing of bone fractures produced by severe trauma remains a problem for contemporary medicine and comes at a high worldwide cost. Furthermore, aging and age-related chronic diseases such as diabetes exacerbate the condition, resulting in decreased osteogenic differentiation and bone quantity, directly impacting bone healing ability in affected individuals [1]. Bone healing is a dynamic process that includes inflammation, granulation tissue growth, intramembranous and endochondral ossification and, finally, remodeling of the resulting woven bone to a lamellar structure until the original shape is restored. The healing process is orchestrated by a complicated sequence of inflammatory, angiogenic, osteo/anabolic and osteo/catabolic mediators, which are only partially understood [2]. Adult mesenchymal stem/stromal cells, acknowledged as the

MSC Isolation from Human Bone Marrow
Human bone marrow samples from the iliac crest of four healthy donors were used to isolate and expand in vitro hBM-MSC. The BM was taken from donors after informed consent was obtained in accordance with Galway University Hospital's requirements under protocol Reference 02/08, approved by the institutional Clinical Research Ethical Committee. Mononuclear cells were counted using a hemocytometer and isolated from the BM by directly plating them into T175 flasks (Corning, NY, USA) at a density of 2.25-3 × 10 5 /cm 2 in a chemically defined serum-free medium (SFM) (https://patents.google.com/patent/WO2015121471A1) composed of α-minimum essential medium (α-MEM, Gibco, Grand Island, NY, USA), lipoprotein 40 µg (Sigma-Aldrich, Saint Louis, MO, USA), dexamethasone 50 nM (Sigma-Aldrich), ascorbic acid 2-phosphate 100 µM (Sigma-Aldrich), human serum albumin 1% (Baxter, Deerfield, IL, USA), insulin-transferrin-selenium 1% (ITS, BD Biosciences, Oxford, CA, USA), transforming growth factor β (TGF-β) 20 ng/mL and fibroblast growth factor 2 (FGF-2) 10 ng/mL (Thermo Fisher Scientific, Waltham, MA, USA). Cells were cultured at 37 • C, 5% CO 2 until passage 3 (P3) under normoxia (21% O 2 ) with a regular twice weekly medium change and cryopreserved in liquid nitrogen. To obtain the desired number of cells that had to be injected, hBM-MSC were thawed around 80 days before transplantation and cultured until P6, compensating for a cell death ratio of 15-20 percent during the subsequent starvation. Following P6, "starving" hBM-MSC were left in the same medium with no exposure to fresh medium for 70 days under normoxia. After that period, starving hBM-MSC were washed in phosphate-buffered saline (PBS, Gibco), detached using 0.25% trypsin (Gibco) for 15 min, counted and resuspended in PBS at 5000 cells/µL. Regularly fed cells were cultured in SFM until P3 and cryopreserved at 5000 cells/µL in liquid nitrogen until in vivo injection.

Extracellular Vesicle Isolation
EVs were separated from P3 regularly fed cells as follows: sub-confluent hBM-MSC cultured in SFM were washed three times in PBS and cultured in basal medium (αMEM, Gibco) for 72 h. Conditioned medium (CM) was harvested after 72 h and processed immediately through two subsequent centrifugation steps [33]. This clarified CM was differentially ultra-centrifuged (Optima XPN-100, Beckman-Coulter, Brea, CA, USA) at 10,000× g for 40 min at 4 • C and, subsequently, twice at 100,000× g for 2 h at 4 • C to pellet EVs [33]. EVs were resuspended in PBS and stored at −80 • C for further experiments or until in vivo injection.

Nanoparticle Tracking Analysis
EV samples were characterized and quantified by Zetaview (Particle Metrix GmbH, Inning am Ammersee, Germany) nanoparticle tracking analysis (NTA), equipped with a sample cell and two lasers (488 nm and 640 nm) and Zetaview 8.05.14_SP7 software. After calibration with 100 nm polystyrene beads, samples were diluted in filtered PBS and injected into the sample cell using a 1 mL syringe. Size distribution analyses of 11 different positions were performed for each sample on at least three different EVs preparations.

Transmission Electron Microscopy
EV samples were fixed in 2% paraformaldehyde and adsorbed for 10 min to formvar-carbon-coated copper grids. Grids were negatively stained with 2% uranyl acetate for five min at room temperature (RT). Stained grids were embedded in 2.5% methylcellulose for improved preservation and air dried before examination. Electron micrographs were taken with a transmission electron microscope (TEM) (HT7800 series, Tokyo, Japan) equipped with Megaview 3 digital camera and Radius software (EMSIS, Münster, Germany).

Differentiation Assays
Starved hBM-MSC at P6 and regularly fed P3 cells from the same donor were differentiated in triplicate through the osteoblastic, adipocyte, and chondroblastic lineages as described previously [17]. Osteogenesis and adipogenesis of the hBM-MSC were assessed using Alizarin Red S (Sigma) at day 17 for osteogenesis and Oil Red O (Sigma) at day 19 for adipogenesis. Chondrogenesis was assessed by Safranin O and Alcian Blue (pH 2.5) staining; dimethyl methylene blue (DMMB) assays, followed by DNA quantification with PicoGreen (Molecular Probes, Eugene, OR, USA) at 21 days, were used to quantify glycosaminoglycans.

Protein Array
The C-Series human growth factor antibody array C1 was used to compare 1 mL each of SFM recovered after starvation and fresh SFM (RayBio, Peachtree Corners, GA, USA) in duplicate. X-ray films were then used to detect horseradish peroxidase signals and quantified with Fiji ImageJ software (version 1.51r; NIH, Bethesda, MD, USA).

Fracture Model
For the in vivo studies, regularly fed (P3) and starved hBM-MSC (P6) and EVs from regularly fed cells (P3) were tested in vivo for the ability to mediate repair of a femoral fracture in a diabetic mouse model. The local National University of Ireland Galway Animal Briefly, male C57BL/6J mice were obtained from Charles River laboratories, UK, at six weeks old. Nine week old mice were intraperitoneally injected with streptozotocin (STZ; 50 mg/kg, Sigma) in Hank's balanced salt solution (HBSS, Sigma) to induce the diabetic condition. Injections were administered once a day for five days, and control animals were injected with the same volume of HBSS. Blood samples were obtained before the STZ injection to test baseline glucose levels, as well as weekly for three weeks to determine the onset and persistence of the diabetic state. All animals with no signs of pain and a glucose level ≥ 13 mM for three consecutive weeks were classified as hyperglycemic and enrolled in the study. After three weeks (day 0) animals were anesthetized with isoflurane, the joint surface was exposed, and an intramedullary stabilizing pin, a 27 G syringe needle, was inserted into the marrow cavity and confirmed using an X-ray machine, as previously described [21]. Mice femurs were positioned on the fracture device [35], a three-point bending guillotine, under the center of the sliding weight before being dropped onto the leg, causing the fracture [21]. An X-ray was taken immediately after the fracture, to ensure that the location of the intramedullary pin had not altered and that the transverse fracture had occurred. Following anesthesia, animals were given 0.05 mg/kg buprenorphine and were separately housed [21].

Cells and EVs Administration
Under isoflurane anesthesia, cells or EVs were administered on day two post fracture and randomized to the treatment group; the precise position of the injection into the fracture site was confirmed using an X-ray apparatus. A single dosage of starved and regularly fed hBM-MSC (2.5 × 10 5 in 50 µL in PBS, Gibco) was tested, as were EVs produced from the same number of regularly fed cells and supplied through local injection at the fracture site. Mineralized material was monitored radiographically at specified time-points (day seven: n = 5 to 6 per group and day 23: n = 6 to 8 per group). Control animals were injected with the same volume of PBS (day seven: n = 5 per group and day 23: n = 6 to 8 per group).

Animal Sacrifice and Tissues Collection
Animals were sacrificed by CO 2 inhalation and cervical dislocation on the seventh and 23rd days post fracture. Muscle and soft tissues were carefully removed from harvested femurs before wrapping the bones in PBS-soaked gauze (Gibco). Blood samples were collected from the tail vein in 0.5 M EDTA (Gibco) and rapidly frozen at −20 • C. Samples were taken on day two, four hours after injection, and day seven post fracture from diabetic and non-diabetic animals, injected and non-injected with hBM-MSC and EVs.

Micro-Computed Tomography
Micro-computed tomography (µCT) imaging was performed using a Scanco µCT100 (SCANCO Medical AG, Brüttisellen, Switzerland) with mineral matrix samples scanned in PBS within a 15 mm sample holder. A sample length of 6.78 mm spanning the fracture area was selected and scanned within a 15.1 mm field of view, with an X-ray power of 70 KeV through a 0.1 mm aluminum beam hardening filter and an integration time of 560 ms generating scan data with voxel size of 7.4 µM.
Weekly calibration checks were carried out by scanning a phantom containing five rods of varying known mineral densities (0, 100, 200, 400 and 800 mg HA/ccm) and three coordinate calibration pins for x, y and z calibration. All quality control checks during this study indicated that the accuracy of the density and coordinate calibrations were within the recommended range provided by Scanco to denote appropriate instrument calibration. Slice data were reconstructed automatically using Scanco density, x, y and z coordinate algorithms and a cylindrical volume of interest within each scan volume was defined for 3D evaluation. New mineral matrix calluses and original bone in the µCT slices were separated from other components present in the sample (i.e., marrow and PBS) using threshold values of 123.8 mg HA/ccm for new mineral matrix and 430.6 mg HA/ccm for original bone to create 3D reconstructions of the fractured regions. Mineral matrix regeneration was quantified using the Fiji ImageJ and the BoneJ plug-in [36]; briefly, µCT images were cropped to the desired size, binarized and analysed.

Statistical Analysis
Experiments were conducted at least in triplicate and repeated twice at a minimum. Data were compared using SPSS software version 22.0 (IBM, New York, NY, USA). All results are presented as the mean ± SD. Animal sample size was n = three to sixteen per group and by time-point. Statistical significance was evaluated by Student's t-test, one-way analysis of variance (ANOVA) followed by Tukey's, Fisher's or Bonferroni's post hoc test, as appropriate. Paired t-test was used to compare samples taken from related groups. Significance was set to p ≤ 0.05 *, +, **.

Starved and Regularly Fed Cells Characterization
Regularly fed hBM-MSC used in this study were isolated from fresh human BM and cultured in a serum-free, xenogen-free chemically defined culture medium until P3, and then stored in liquid nitrogen until in vivo injection. Prior to transplantation, a parallel aliquot of P3 hBM-MSC was thawed, allowed to expand until P6 and exposed to total starvation for 70 days under normoxic conditions in order to produce the required number of cells to be injected ( Figure 1A). Starvation led to a considerable reduction in cell number (about 15 to 20%) and different morphological and functional changes compared to the control hBM-MSC. By day 70, a more homogeneous population of fibroblast-like cells, finely interconnected, was present in the culture [17] ( Figure 1B).

EVs Characterization
EVs from P3 regularly fed cells were characterized, according to MISEV2018 [37]. TEM analysis on the isolated vesicles was performed to check the EVs' morphology ( Figure 3A) showing morphological and dimensional heterogeneity. A wide-ranging size distribution was also confirmed by NTA, displaying a mean EV size of 133.6 ± 4.74 nm ( Figure 3B). The expressions of EV surface markers CD63 and CD81 and additional vesicular markers syntenin-1 and flotillin-1 were confirmed by immunoblot ( Figure 3C), while no CD9 expression was detected. EVs were stained with CFDA-SE to discriminate intact vesicles from debris and membrane fragments ( Figure 3D). Using a mixture of fluorescent dimensional beads, specific size gates were considered (≤100 nm, from 100 to 160 nm and from 160 to 900 nm) ( Figure 3E, left panel). The percentage of CFDA-SE events falling in the three gates showed that P3 regularly fed hBM-MSC-derived EVs were mostly in the gates ≤100 nm and from 100 nm to 160 nm, confirming the average size previously observed by NTA. The expressions of CD81 and CD63 were also validated by flow cytometry ( Figure 3E, right panels), as well as CD9's low expression.

Diabetic Fractured Mouse Model Shows Increased Inflammation and Reduced Mineral Matrix Content
To validate our main hypothesis in vivo, we adopted a well-established preclinical model of delayed fracture healing [21,32] using the experimental plan depicted in Figure 4A. Diabetic induction with STZ was successful, resulting in a mean of plasma glucose level of 23.9 ± 5.3 mM in comparison to control animals (injected with HBSS) having 11.2 ± 0.57 mM (p ≤ 0.05) ( Figure 4B). Diabetic mice also showed a reduced but steady weight throughout the study, 26.34 ± 1.17 g for STZ-treated animals and 32.3 ± 1.92 g (p ≤ 0.05) for control animals ( Figure 4C). X-ray scans taken immediately after the creation of the mice's femur transverse fractures ( Figure 4D) and those taken two days later ( Figure 4E) directed the procedure for injecting cells into the fracture site. Additionally, X-rays were used to track the healing of fractures at days seven ( Figure 4F) and 23 ( Figure 4G). The results indicated that diabetes had a negative effect on bone regeneration only at the earlier time-point (seven days). All animals treated with STZ had reduced mineral content at the fracture site compared to untreated animals at this time-point, as seen from the µCT-analyzed parameters bone volume/total volume (BV/TV(%)) (p ≤ 0.05), connectivity (number of trabeculae) (p ≤ 0.05), connectivity density (number of trabeculae per unit volume, Conn.D (mm −3 )) (p ≤ 0.05), bone surface (BS(mm 2 )) (p ≤ 0.05), trabecular thickness mean (Tb. Th. Mean (mm)) (p ≤ 0.05), trabecular thickness max (Tb. Th. Max (mm)) (p ≤ 0.05) and trabecular spacing max (Tb. Sp. Max (mm)) (p ≤ 0.05), but not for the trabecular spacing mean (Tb. Sp. Mean (mm)) ( Figure 4H). No differences were seen in the mineral content in any µCT measurement assessed between diabetic and non-diabetic mice at the longer time-point (23 days) ( Figure 4I).

Starved Cells Enhance Anti-Inflammatory Effects and Favor Earlier Soft Callus Formation and Mineral Matrix Deposition
Blood samples taken from the animals two days after fracture (four hours after cell and EVs injections) revealed similar levels for starved and regularly fed cells for IL-3, IL-4, IL-5, IL-9, IL-12(p70), GM-CSF, IFN-γ, KC, MCP-1, MIP-1α, RANTES and TNF-α levels; however, their levels in starved and regularly fed cells were lower than in mice treated with EVs (p ≤ 0.05 * with respect to EVs, + with respect to reg fed cells,) ( Figure 5A-C). At the day two time-point, only IL-17 was greater in the starved cells than in both the EVs and regularly fed cells (p ≤ 0.05), whereas IL-1α, IL-1β, IL-2, IL-10 and IL-13 were lower in the starved cell mice than in animals injected with the other two treatments (p ≤ 0.05 * with respect to EVs, + with respect to reg fed cells) ( Figure 5A-C).    A,B) Graphs comparing the inflammatory cytokines in plasma at the day two time-point (sample size three to eight animals/group, one-way ANOVA followed by Tukey's post hoc test). (C) Radar graph highlighting the differences between the different treatments at day two (sample size three to eight animals/group, paired t-test). (D,E) Graphs comparing the inflammatory cytokines at day seven (sample size three to eight animals/group, one-way ANOVA followed by Tukey's post hoc test). (F) Radar graph highlighting the differences between the different treatments at the day seven time-point (sample size three to eight animals/group, paired t-test). (G) Heat map showing direct comparison of the inflammatory cytokines within the different treatments and the two time-points. (H-J) Radar graph highlights the differences between animals treated with cells (H), EVs (I), starved cells (J) at the different time-points. The data are normalized to fractured diabetic animals who had only PBS injection and presented as mean fold change ± SD. p ≤ 0.05 * respect EVs, + cells. Abbreviations: regularly fed cells (cells); starved cells (starved); extracellular vesicles (EVs); fold change (F.C.); interleukin (IL-1α, 1β, 2, 3, 4, 5, 6, 9, 10, 12(p40), 12(p70), 13, 17); granulocyte colony-stimulating factor (G-CSF); granulocyte-macrophage colony-stimulating factor (GM-CSF); interferon-γ (IFN-γ); keratinocytes-derived chemokine (KC); monocyte chemotactic protein 1 (MCP-1); macrophage inflammatory protein-1α, 1β (MIP-1α and MIP-1β); regulated on activation, normal T cell expressed and secreted (RANTES); tumor necrosis factor-α (TNF-α).
At seven days, although several cytokines were significantly lower in the starved cell-injected animals than in the regularly fed or EV-injected animals ((IL-1α, IL-12(p70), IL-17, GM-CSF, IFN-γ, MIP-1α, MIP-1β and RANTES, p ≤ 0.05 * with respect to EVs, + with respect to reg fed cells)), the expression of IL-6, IL-13, eotaxin and KC were higher after the injection of the starved cells, compared to the animals treated with EVs and regularly fed cells ( Figure 5D-F). When the data from the two time-points were compared between animals injected with regularly fed cells, it was seen that IL-1α, IL-6, IL-9, IL-10, IL-12(p70), IL-17, G-CSF, KC, MCP-1, MIP-1α and RANTES reached their secretion peak at the seventh day time-point with IL-2, IL-3, IL-13 and TNF-α peaking at the second day time-point (* p ≤ 0.05) ( Figure 5G,H).
To evaluate quantitative variations in the microarchitecture of a newly formed mineral matrix, µCT analyses were performed on mouse femurs injected with hBM-MSC, EVs, and diabetic fractured and non-diabetic fractured controls at two different time-points. After seven days, there appears to be denser cortical lightly mineralized material adjacent to the fracture site highlighting the soft callus presence as seen in the 3D reconstructions (blue) ( Figure 6A) and in Figure 6B-E with decreasing levels (blue) starting from starved ( Figure 6C) to EVs ( Figure 6D) and regularly fed ( Figure 6E) as compared to diabetic fracture controls ( Figure 6B). Quantitative analysis of µCT sequential images confirmed that starved hBM-MSC were able to promote greater mineral matrix deposition than EVs and regularly fed cells on the seventh day post fracture, (p ≤ 0.05, * with respect to EVs, + with respect to reg fed cells) ( Figure 6F,G). To evaluate quantitative variations in the microarchitecture of a newly formed mineral matrix, μCT analyses were performed on mouse femurs injected with hBM-MSC, EVs, and diabetic fractured and non-diabetic fractured controls at two different time-points. After seven days, there appears to be denser cortical lightly mineralized material adjacent to the fracture site highlighting the soft callus presence as seen in the 3D reconstructions (blue) ( Figure 6A) and in Figures 6B-E with decreasing levels (blue) starting from starved ( Figure 6C) to EVs ( Figure 6D) and regularly fed ( Figure 6E) as compared to diabetic fracture controls ( Figure 6B). Quantitative analysis of μCT sequential images confirmed that starved hBM-MSC were able to promote greater mineral matrix deposition than EVs and regularly fed cells on the seventh day post fracture, (p ≤ 0.05, * with respect to EVs, + with respect to reg fed cells) ( Figure 6F-G). and Tb. Sp. Max (mm). Data were normalized to diabetic animals with femoral fracture but receiving PBS injection. Results are presented as mean fold change ± SD. p ≤ 0.05 * with respect to EVs, +cells (sample size four to five animals/group, one-way ANOVA followed by Fisher's post hoc test). Scale bars: 100 μm, 1 mm. Abbreviations: micro-computed tomography (μCT); regularly fed cells (cells); starved cells (starved); extracellular vesicles (EVs); fold change (F.C.); bone volume/total volume (BV/TV); connectivity density (Conn.D (mm −3 )); bone surface (BS(mm 2 )); trabecular thickness BV/TV ratio, connectivity, Conn.D (mm −3 ) compared to regularly fed cell-and EV-injected animals and BS (mm 2 ) compared to EV-treated cells. No significant difference was seen between the treatments for trabecular Tb. Th. Mean (mm), Tb. Th. Max (mm), Tb. Sp. Mean (mm) and Tb. Sp. Max (mm). Data were normalized to diabetic animals with femoral fracture but receiving PBS injection. Results are presented as mean fold change ± SD. p ≤ 0.05 * with respect to EVs, +cells (sample size four to five animals/group, one-way ANOVA followed by Fisher's post hoc test). Scale bars: 100 µm, 1 mm. Abbreviations: micro-computed tomography (µCT); regularly fed cells (cells); starved cells (starved); extracellular vesicles (EVs); fold change (F.C.); bone volume/total volume (BV/TV); connectivity density (Conn.D (mm −3 )); bone surface (BS(mm 2 )); trabecular thickness mean (Tb. Th. Mean (mm)); trabecular thickness max (Tb. Th. Max (mm)); trabecular spacing mean (Tb. Sp. Mean (mm)); trabecular spacing max (Tb. Sp. Max (mm)).
At the 23 day time-point, a substantial amount of cortical and trabecular newly formed mineralized matrix (grey) and the presence of a lightly mineralized soft callus (blue) were evidenced as seen in Figure 7A-D; however, no significant difference in any of the investigated parameters was seen across all the tested treatments ( Figure 7E,F and the diabetic fracture control ( Figure S2A-F).  Data are normalized to diabetic animals that had a femoral fracture but only received a PBS injection. The results are presented as mean fold change ± SD (sample size five to eight animals/group, one-way ANOVA followed by Fisher's post hoc test). Scale bars: 1 mm. Abbreviations: micro-com- Data are normalized to diabetic animals that had a femoral fracture but only received a PBS injection. The results are presented as mean fold change ± SD (sample size five to eight animals/group, oneway ANOVA followed by Fisher's post hoc test). Scale bars: 1 mm. Abbreviations: micro-computed tomography (µCT); regularly fed cells (cells); starved cells (starved); extracellular vesicles (EVs); fold change (F.C.); bone volume/total volume (BV/TV); connectivity density (Conn.D (mm −3 )); bone surface (BS(mm 2 )); trabecular thickness mean (Tb. Th. Mean (mm)); (p ≤ 0.05); trabecular thickness max (Tb. Th. Max (mm)); trabecular spacing mean (Tb. Sp. Mean (mm)); trabecular spacing max (Tb. Sp. Max (mm)).

Discussion
When a fracture occurs, it impacts not only the bone tissue but also the nearby soft tissues, the local vasculature and the bone marrow. As a result, a hematoma develops, which is characterized by a limited supply of oxygen and nutrients and acute inflammation in the early stages [38]. Neutrophils are among the first inflammatory cells to colonize the fracture site, generating pro-inflammatory and chemotactic mediators that drive the recruitment of a large group of inflammatory cells, mainly monocytes and macrophages [39]. Fracture hematoma and inflammatory lesions often resolve within a few days to a week following the fracture, and are replaced at the soft callus stage by developing neo-vasculature and granulation tissue rich in mesenchymal cells [40]. Subsequently, chondroblasts develop the soft callus via cartilage apposition two to four weeks after the injury, providing quick structural support to the fracture. Finally, the soft callus serves as a scaffold for endochondral bone formation [40] and osteoprogenitor cell differentiation, followed by the deposition of woven bone on the cartilage leading to the hard callus phase months after the initial injury [40].
Inflammatory cells produce a variety of cytokines, influencing cells in both positive and negative ways and the various phases of bone regeneration [2,25].
At later stages many different studies have demonstrated that inflammatory and chemotactic mediators, such as TNF-α, IL-4, IL-6, IL-13, IFN-γ, KC and MCP-1 released from macrophages, lymphocytes and eosinophils, also stimulate the recruitment of fibroblasts, mesenchymal stem cells (MSC) and osteoprogenitor cells from their niches [25,[41][42][43]. For example, it was seen that mice missing the TNF-α receptor gene have a significant delay in the chondrocyte differentiation [42] and endochondral ossification [44]. Interleukin 6 (IL6) has been similarly implicated in bone healing delay as well as the lack of reduced fracture healing [32].
Such a hyper-inflammatory milieu was recapitulated in the diabetic STZ-induced animal models used in this research as well as by McCauley J. et al. [32].
MSC have long been proposed as a biological treatment in immunotherapy and regenerative medicine due to their multilineage differentiation capacity and immunological regulation. The cellular environment surrounding MSC or the pathological conditions, however, can negatively regulate or restrict the potential of the MSC, as in the case of fractures in diabetic conditions. This environment leads to a sharp decline in the injected cell viability [10] or a reduced activity of the resident MSC as a result of the activation of apoptotic [7] signaling brought on by the depleted nutrient and oxygen status as well as the hyper-inflammatory state. It is also true that an MSC's behavior is influenced by and adapts [10,17] to external stimuli; therefore, the success of mesenchymal stem/stromal cells as immunotherapeutic agents or in regenerative medicine may depend on understanding how to activate or inhibit their adaptive processes before or during therapeutic scenarios.
In our study, we subjected our cells to an intense preconditioning procedure, which, as demonstrated in our previously published paper, results in a significant reduction in the number of cells during starvation and may increase the presence of waste/toxic molecules such as lactate and ammonia [45]. This likely contributed to increasing the osmolarity and ionic strength of the culture medium [17]. Although we did not measure the level of ammonia in our previous research, we found that lactate transiently increased in the culture medium until day 27 before decreasing towards the end of the starving period [17]. Recent research demonstrated that lactate supplementation can decrease both lactate and ammonia levels in culture [46]. These results, along with the presence of albumin, both as messenger and protein, in starvation-induced conditions [17], which is known to be a lipid and metal ion transporter with antioxidant and buffering properties, imply that the starvation process activates specific cellular adaptive mechanisms that lessen the negative effects.
CD146 reduction is not a surprising event that was reported to be dependent on the culture conditions and aging leading to reduced osteoblastic activity [47]. The fact that our culture medium is composed of α-MEM, which was not seen to induce CD-146 reduction, even at higher passages (P8) [47], suggests that the starvation process is the cause of its decrease and, as a consequence, a reasonable reduction in the hBM-MSC osteoblastic potential. On the contrary, a comparison of the results obtained from P6 starved cells and P3 regularly fed cells showed no reduction in their osteoblastic capability.
In addition, the analysis of the pro-inflammatory cytokines and chemokines revealed that, compared to EV-treated animals, starved cell-and regularly fed cell-injected animals had a negligible influence on pro-inflammatory cytokine synthesis at both time-points (such as IL-1α, IL-1β, TNFα and IFN-γ), thus corroborating a putative inhibitory effect for these cells on pro-inflammatory cytokine production, which has been widely and previously described for MSC [3,55].
Nonetheless, the highest levels of IL-17 and IL-6 at the earlier time-point, as well as IL-13, KC and eotaxin at the seventh day time-point, suggest that starving cells outperform both regularly fed cells and EVs ( Figure 8). Indeed, recent findings suggest that an increased and early presence of IL-17-producing γδ T cells promotes the osteoblastic differentiation of progenitor cells [56] via two temporally defined mechanisms: during the early inflammatory cascade via C/EBPβ, independent of the canonical Wnt pathway and β-catenin signaling; later, during the reparative/remodeling phase, IL-17 acts via C/EBPδ [57], which could be consistent with the observations of increased IL-17 plasma concentrations in mice treated with starved cells. recruitment [32] and vascular regeneration by enhancing monocyte adhesion and density around the developing arterial collaterals, leading to successful fracture healing [43,49]. The fact that KC increases following the injection of starved cells implies that its availability in animals on day seven post fracture supports its beneficial effect on cell recruitment and revascularization. Furthermore, the increased osteoclast recruitment and resorption of bone fragments at the fracture site is attributed to eotaxin, which acts as a natural antagonist for the C-C motif chemokine receptor 2 and an agonist for CCR5 [32,58]; therefore, the increased eotaxin presence, favoring osteoclast recruitment and resorption of the bone fragments during the initial stages of fracture healing, positively influences bone regeneration ( Figure 8). Figure 8. Graphic outlining the effects of hBM-MSC-starved cell injection on bone healing, as well as the released cytokines and growth factors involved in the early phases of fracture healing. Results demonstrate that the starvation process induces the production of different growth factors such as IGFBP2, 3, 6, HGF, TGFβ2, PDGFα, VEGFR2 and IGF2, which induce the secretion of cytokines such as IL-17 and repress the production of IL-1α, β, 2, 10 and 13 at the earliest time-point. At day 7 after fracture, starved cells favor the expression of IL-6, 13 and KC, increase the presence of mineralized Figure 8. Graphic outlining the effects of hBM-MSC-starved cell injection on bone healing, as well as the released cytokines and growth factors involved in the early phases of fracture healing. Results demonstrate that the starvation process induces the production of different growth factors such as IGFBP2, 3, 6, HGF, TGFβ2, PDGFα, VEGFR2 and IGF2, which induce the secretion of cytokines such as IL-17 and repress the production of IL-1α, β, 2, 10 and 13 at the earliest time-point. At day 7 after fracture, starved cells favor the expression of IL-6, 13 and KC, increase the presence of mineralized matrix and, presumably, the soft callus formation at the fracture site, with a decrease in the expressions of IL-1α, 12(p70), 17, GM-CSF, IFN-γ, MIP-1α, β and RANTES. However, at 23 days post fracture, no difference was seen within the tested conditions. Abbreviations: streptozotocin (STZ); bone volume/total volume (BV/TV); connectivity density (Conn.D (mm −3 )); bone surface (BS(mm 2 )); trabecular thickness mean (Tb. Th. Mean (mm)); trabecular thickness max (Tb. Th. Max (mm)); trabecular spacing mean (Tb. Sp. Mean (mm)); trabecular spacing max (Tb. Sp. Max (mm)); interleukin (IL-1α, 1β, 2,6, 10, 12(p70), 13, 17); granulocyte-macrophage colony-stimulating factor (GM-CSF); interferon-γ(IFN-γ); keratinocytes-derived chemokine (KC); macrophage inflammatory protein-1α, 1β (MIP-1α and MIP-1β); regulated on activation, normal T cell expressed and secreted (RANTES); insulin growth factor binding protein 6 (IGFBP6, 3, 2); hepatocyte growth factor (HGF); insulin growth factor 2 (IGF2); platelet-derived growth factor (PDGFα); transforming growth factor 2 (TGFβ2); vascular endothelial growth factor receptor 2 (VEGFR2).
It is widely recognized that macrophages can play a dual role in fracture healing, being either pro-inflammatory during the initial phases as a result of the classical activation pathway, via IFN-γ, or anti-inflammatory through the alternative activation pathway, via IL-4 and IL-13 [2,43]. In fact, the alternative activation pathway promoting macrophage activity, collagen deposition and tissue homeostasis has been demonstrated to favor fracture healing [48]. Therefore, here, it may be suggested that the induced secretion of IL-13, by starved cells, or IL-4, by EVs, plays a role in alternative macrophage polarization favoring tissue homeostasis and matrix secretion [32,48], accelerating the development of a soft callus, and thus contributing to early fracture stabilization. Local vascular regeneration and infiltration within the fracture callus are just as important as inflammation resolution or bone regeneration. KC is a chemokine that is released by dendritic cells and its presence at the fracture site early during the healing process promotes neutrophil recruitment [32] and vascular regeneration by enhancing monocyte adhesion and density around the developing arterial collaterals, leading to successful fracture healing [43,49]. The fact that KC increases following the injection of starved cells implies that its availability in animals on day seven post fracture supports its beneficial effect on cell recruitment and revascularization. Furthermore, the increased osteoclast recruitment and resorption of bone fragments at the fracture site is attributed to eotaxin, which acts as a natural antagonist for the C-C motif chemokine receptor 2 and an agonist for CCR5 [32,58]; therefore, the increased eotaxin presence, favoring osteoclast recruitment and resorption of the bone fragments during the initial stages of fracture healing, positively influences bone regeneration ( Figure 8).
Such positive effects could suggest that starved cells, in addition to inducing a quicker soft callus formation, which is per se already interesting because of the evidence of a delayed soft callus formation in diabetic conditions [30], also favor the faster deposition of mineralized material that translates into an improved and earlier fracture stabilization at seven days post fracture. In fact, Kayal et al. demonstrated that under diabetic conditions [34,38], the delayed fracture healing is related to an increased chondrocytes' apoptosis, via the TNF-α/FOXO1 axis, and a premature cartilage resorption due to an increased osteoclastic activity [27,30]. These results support previous findings, revealing that the starvation approach had the most favorable influence on growth factor release, mineral matrix deposition and anti-inflammatory activity over EVs and regularly fed cells derived from the same donor. Further studies based on histological analysis will be needed to fully understand the response to the proposed cell therapy. However, on the downside, such favorable benefits appear to be limited to the early phases of fracture healing, as the fraction of mineralized matrix in normo-glycemic, diabetic and diabetic-treated animals was comparable on day 23, which is consistent with previous findings during fracture healing [21,30] (Figure 8).

Conclusions
This study demonstrated that the long-term starvation of hBM-MSC enhances the secretion of trophic factors, improving the early stage of fracture healing in diabetic mice by decreasing pro-inflammatory cytokine expression and increasing anti-inflammatory cytokine production, angiogenesis and progenitor cell recruitment, leading to an earlier soft callus formation, mineralization and stabilization of the fracture. Our findings are consistent with those by Zhang J. et al. [5], who demonstrated that hypoxia preconditioning improves fracture healing and suggest, as confirmed by Moya A. et al. [10], that starvation increases cell survival, indicating that preconditioning studies are needed to improve the regenerative and anti-inflammatory effects of grafted cell survival after transplantation.
This work supports the idea that the in vitro crosstalk between the cells and stimuli is an important factor that has yet to be deeply studied to fulfill MSC's therapeutic potential, thus pointing out that in vitro preconditioning of stem/stromal cells may be crucial for refining and defining new criteria for future cell therapy applications. That said, a better understanding of stem/stromal cell adaption and survival processes may lead to increased cell viability and, thus, an improvement in therapeutic efficacy after transplantation.
Since the main concern about using stem cells in clinical settings is the need for cell expansion outside their quiescent niches, we previously suggested in our published research [17] that the MSC culture in a xenogen-free medium would be the preferred choice for future MSC in vitro amplification due to its reduced nutritional content. In addition, we suggest that deep studies of the starvation approach at different time-points, both in normoxia or hypoxia, and for different periods, will also reveal the survival, regenerative and anti-inflammatory properties of the cells. Importantly, the therapeutic potential of the MSC also has to be explored by testing the properties of the EVs retrieved from starved cells.
In summary, there is still more that can be done to fully comprehend and appreciate the preconditioned MSC's and EV's whole potential, and this may also be revealed by pushing them to their limit, even though this is not strictly necessary.
Supplementary Materials: The following supporting information can be downloaded at: https: //www.mdpi.com/article/10.3390/cells11172696/s1, Figure S1: Analysis of the newly formed mineralized tissue five days after cells (starved and regularly fed) and EVs injections at the fracture site; Figure S2: Analysis of the newly formed mineralized tissue at 21 days after cell and EVs injection at the fracture site.  Data Availability Statement: Data are available from the corresponding author upon request. In data are available from the corresponding author upon request.