Altering the Architecture of Tissue Engineered Hypertrophic Cartilaginous Grafts Facilitates Vascularisation and Accelerates Mineralisation

Cartilaginous tissues engineered using mesenchymal stem cells (MSCs) can be leveraged to generate bone in vivo by executing an endochondral program, leading to increased interest in the use of such hypertrophic grafts for the regeneration of osseous defects. During normal skeletogenesis, canals within the developing hypertrophic cartilage play a key role in facilitating endochondral ossification. Inspired by this developmental feature, the objective of this study was to promote endochondral ossification of an engineered cartilaginous construct through modification of scaffold architecture. Our hypothesis was that the introduction of channels into MSC-seeded hydrogels would firstly facilitate the in vitro development of scaled-up hypertrophic cartilaginous tissues, and secondly would accelerate vascularisation and mineralisation of the graft in vivo. MSCs were encapsulated into hydrogels containing either an array of micro-channels, or into non-channelled ‘solid’ controls, and maintained in culture conditions known to promote a hypertrophic cartilaginous phenotype. Solid constructs accumulated significantly more sGAG and collagen in vitro, while channelled constructs accumulated significantly more calcium. In vivo, the channels acted as conduits for vascularisation and accelerated mineralisation of the engineered graft. Cartilaginous tissue within the channels underwent endochondral ossification, producing lamellar bone surrounding a hematopoietic marrow component. This study highlights the potential of utilising engineering methodologies, inspired by developmental skeletal processes, in order to enhance endochondral bone regeneration strategies.


Introduction
The goal of tissue engineering is to replace or regenerate damaged tissues through the combination of cells, three-dimensional scaffolds, and signalling molecules [1,2]. Bone tissue engineering has, thus far, generally focussed on the direct osteoblastic differentiation of mesenchymal stem cells (MSCs), in a process resembling intramembranous ossification [3]. However, the success of this approach to bone regeneration has been hampered by insufficient blood vessel infiltration, preventing the necessary delivery of oxygen and nutrients to the engineered graft [4,5]. Recently an endochondral approach to bone tissue engineering, which involves remodelling of an intermediary hypertrophic cartilaginous template [6][7][8][9][10][11][12][13][14], has been proposed as an alternative to direct intramembranous ossification for bone regeneration using MSCs. Chondrogenically primed bone marrow derived MSCs have an inherent tendency to undergo hypertrophy [15], an undesirable attribute in MSC-based articular cartilage repair therapies, but one which may be harnessed for use in endochondral bone tissue engineering strategies. Hypertrophic chondrocytes are equipped to survive the hypoxic environment a tissue engineered graft will experience once implanted in vivo [14]. Furthermore, cells undergoing hypertrophy release pro-angiogenic factors, such as vascular endothelial growth factor, to facilitate the conversion of avascular tissue to vascularised tissue [16]. The endochondral approach to bone regeneration may therefore circumvent many of the issues associated with the traditional intramembranous method; however, in order to be used to repair large bone defects, this approach first requires strategies to engineer scaled-up hypertrophic cartilage. This may be challenging, as recent attempts to generate large hypertrophic constructs using MSC seeded collagen scaffolds have demonstrated the formation of a core region devoid of cells and matrix [12].
Such 'scaling up' of engineered grafts, and the associated issue of core degradation, is a well-documented challenge in the field of tissue engineering [17]. Strategies to promote nutrient supply and waste removal include the use of bioreactors [18][19][20][21], modification of scaffold architectures [22,23], or a combination of both [24][25][26]. An alternative approach might be to recapitulate the mechanisms adopted during skeletogenesis to provide nutrients to the cartilaginous templates. During normal bone development, canals within the developing hypertrophic cartilage have been identified [27,28]. These canals play an important role in supplying nutrients to the developing cartilage [29]. Furthermore, endochondral ossification is dependent on neo-vascularisation of the cartilaginous template, with the in-growth of blood vessels via these cartilage canals identified as a key event during bone development [29]. Finally, these canals act as conduits for the migration of osteogenic cells into the cartilaginous template which in turn lay down new bone matrix [30]. Executing a developmental engineering paradigm [31], directed at mimicking the structure and function of the cartilage canal network, may therefore be a powerful tool in endochondral bone tissue engineering strategies.
Inspired by the cartilage canal network observed during endochondral skeletogenesis, we hypothesised that the introduction of channels into hypertrophic cartilaginous constructs would firstly facilitate extracellular matrix synthesis and the in vitro development of the construct, and secondly would accelerate mineralisation and vascularisation of the graft once implanted in vivo. To assess the influence of scaffold architecture on in vitro graft development, bone marrow derived MSCs were encapsulated in agarose hydrogels either containing an array of microchannels (termed 'channelled'), or in non-channelled (termed 'solid') controls, and cultured long-term (up to 10 weeks) to undergo hypertrophic chondrogenic differentiation. Furthermore, to test the efficacy of channelled architectures to enhance mineralisation and vascularisation in vivo, channelled and solid constructs were also subjected to a shorter period of hypertrophic priming (6 weeks) prior to subcutaneous implantation in nude mice. Constructs were harvested at 4 and 8 weeks postimplantation to test our hypothesis that graft architecture would influence mineralisation and vascularisation of the engineered hypertrophic tissue.

Experimental Design
The first phase of this study investigated the in vitro development of engineered cartilaginous constructs undergoing long-term hypertrophic chondrogenesis. MSCs were encapsulated in solid and channelled hydrogels, cultured in chondrogenic conditions (further details below), for a period of 5 weeks. Thereafter constructs were switched to hypertrophic conditions for an additional 5 weeks, resulting in a total in vitro culture period of 10 weeks. The second phase of this study investigated the potential of engineered hypertrophic cartilaginous constructs to undergo mineralisation and vascularisation in vivo. MSCs were encapsulated in solid and channelled constructs and cultured in chondrogenic conditions for a period of 5 weeks, as per phase 1. Thereafter constructs received an additional week in hypertrophic conditions (6 weeks total in vitro priming) prior to subcutaneous implantation in nude mice. Constructs were harvested at 4 and 8 weeks postimplantation.

Cell Isolation and Expansion
Porcine bone marrow derived MSCs (4 months old) from the femoral shaft were isolated aseptically and expanded according to a modified method for human MSCs [32] in high glucose Dulbecco's modified eagle's medium (DMEM) GlutaMAX supplemented with 10% v/v foetal bovine serum and 100 U/mL penicillin -100 mg/mL streptomycin (all Gibco, Biosciences, Dublin Ireland) at 20% pO 2 . Following colony formation, MSCs were trypsinised, counted, seeded at density of 5610 3 cells/cm 2 , and expanded to passage 1 (P1). MSCs underwent 15 population doublings prior to encapsulation within hydrogels.

Cell Encapsulation in Solid and Channelled Agarose Hydrogels
At the end of P1 MSCs were suspended in 2% agarose at a density of 20610 6 cells/mL. The agarose cell suspension was cast in a stainless steel mould to produce regular solid (non-channelled) cylindrical constructs (Ø5 mm63 mm). Channelled cylindrical constructs were fabricated utilising a pillared polydimethylsiloxane array structure [24] inserted into a cylindrical Tufset mould to produce a unidirectional 463 channelled array in the longitudinal direction with diameters of 500 mm and a centre-centre spacing of 1 mm.

In vivo Subcutaneous Transplantation
In phase 2 of the study, following 6 weeks in vitro priming, MSCseeded solid and channelled constructs were implanted subcutaneously into the back of nude mice (Balb/c; Harlan, UK) as previously described [34]. Briefly, two subcutaneous pockets were made along the central line of the spine, one at the shoulders and the other at the hips, and into each pocket three constructs were inserted. Nine constructs were implanted per each experimental group. Mice were sacrificed at 4 and 8 weeks post-implantation by CO 2 inhalation. In addition, three acelluar solid and three acellular channelled hydrogels were implanted as controls and harvested at 8 weeks post-implantation. These controls were not primed in chondrogenic and hypertrophic culture conditions before transplantation. A total of 7 mice were used for the experiment. The animal protocol was reviewed and approved by the ethics committee of Trinity College Dublin.

Biochemical Analysis
The biochemical content of constructs was analysed at each time point. Prior to biochemical analysis, constructs were sliced in half, washed in phosphate buffered saline (PBS), weighed, and frozen for subsequent assessment. Half of each construct was digested with papain (125 mg/mL) in 0.1 M sodium acetate, 5 mM L-cysteine-HCL, 0.05 M ethylenediaminetetraacetic acid (EDTA), pH 6.0 (all from Sigma-Aldrich) at 60uC and 10 rpm for 18 h. DNA content was quantified using the Hoechst Bisbenzimide 33258 dye assay, with a calf thymus DNA standard. Proteoglycan content was estimated by quantifying the amount of sulphated glycosaminoglycan (sGAG) using the dimethylmethy-lene blue dye-binding assay (Blyscan, Biocolor Ltd., Northern Ireland), with a chondroitin sulphate standard. Total collagen content was determined by measuring the hydroxyproline content, using a hydroxyproline-to-collagen ratio of 1:7.69 [35,36]. The other half was digested in 1 M hydrochloric acid (HCL) (Sigma-Aldrich) at 60uC and 10 rpm for 18 h. The calcium content was determined using a Sentinel Calcium kit (Alpha Laboratories Ltd, Uk). 3-4 constructs were analysed biochemically per each in vitro experimental group. 4-6 constructs were analysed biochemically per each in vivo experimental group.

Histology and Immunohistochemistry
At each time point samples were fixed in 4% paraformaldehyde overnight, dehydrated in a graded series of ethanols, embedded in paraffin wax, sectioned at 8 mm and affixed to microscope slides. Samples harvested at 8 weeks post-in vivo implantation were decalcified in EDTA for 3-4 days prior to wax embedding. The sections were stained with haematoxylin and eosin (H&E), 1% alcian blue 8GX in 0.1 M HCL to assess sGAG content with a counter stain of nuclear fast red to assess cellular distribution, picro-sirius red to assess collagen distribution, and 1% alizarin red to assess calcium accumulation (all Sigma-Aldrich). Collagen types I, II and X were evaluated using a standard immunohistochemical technique; briefly, sections were treated with peroxidase, followed by treatment with chondroitinase ABC (Sigma-Aldrich) in a humidified environment at 37uC to enhance permeability of the extracellular matrix. Sections were incubated with goat serum to block non-specific sites and collagen type I (ab6308, 1:400; 1 mg/ mL), collagen type II (ab3092, 1:100; 1 mg/mL) or collagen type X (ab49945, 1:200; 1.4 mg/mL) primary antibodies (mouse monoclonal, Abcam, Cambridge, UK) were applied for 1 h at room temperature. Next, the secondary antibody (Anti-Mouse IgG biotin conjugate, 1:200; 2.1 mg/mL) (Sigma-Aldrich) was added for 1 h followed by incubation with ABC reagent (Vectastain PK-400, Vector Labs, Peterborough, UK) for 45 min. Finally sections were developed with DAB peroxidase (Vector Labs) for 5 min. Positive and negative controls were included in the immunohistochemistry staining protocol for each batch. Terminal deoxynucleotidyl Transferase (TdT) cells were identified on the fully automated IHC Leica BOND-MAX (Leica biosystems). Briefly, sections were incubated with a mouse monoclonal TdT Clone SEN28 ready to use primary antibody (PA0339) (Leica biosystems). Heat mediated antigen retrieval was performed with epiptope retrieval solution 2 (AR9640) for 10 min (Leica biosystems). Thereafter, visualisation of the antibody was performed with the Bond Polymer Refine Detection kit (DS9800) (Leica biosystems). Heat mediated antigen retrieval was performed on CD31 sections with sodium citrate solution (Sigma-Alrich) for 20 min, and incubated with goat serum, biotin, and avidin, to block non-specific sites, and the CD31 (ab28364, 1:50; 0.2 mg/ mL) primary antibody (Rabbit polyclonal, Abcam, Cambridge, UK) was applied overnight at 4uC. Thereafter sections were treated with peroxidase followed by the secondary antibody (ab97051, goat polyclonal secondary antibody to Rabbit IgG, 1:200, 1 mg/mL, Abcam, Cambridge, UK) for 1 h, ABC reagent for 45 min, and developed with DAB peroxidase for 5 min. (both Vector Labs). Histological images are taken from two representative constructs.

Micro-computed Tomography
Micro-computed tomography (mCT) scans were carried out on constructs using a Scanco Medical 40 mCT system (Scanco Medical, Bassersdorf, Switzerland) in order to quantify mineral content and assess mineral distribution. In phase 1 of the experiment, constructs were scanned at the end of the 10 week in vitro culture period. In phase 2, constructs were scanned at 4 and 8 weeks post-implantation. Constructs were scanned in PBS, at a voxel resolution of 12 mm, a voltage of 70 kVp, and a current of 114 mA. Circular contours were drawn in an anti-clockwise direction around the periphery of the constructs. Additionally, circular constructs were drawn in a clockwise direction around the perimeter of the channels, so as to exclude the area with the channels from the analysis. A Gaussian filter (sigma = 0.8, support = 1) was used to suppress noise and a global threshold corresponding to a density of 172.6 mg hydroxyapatite/cm 3 was applied. This threshold was selected by visual inspection of individual scan slices so as to include mineralised tissue and exclude non-mineralised tissue. 3D evaluation was carried out on the segmented images to determine mineral volume and to reconstruct a 3D image. The variance of mineralisation with depth through the constructs was analysed qualitatively by examining sections ,0.75 mm of depth from the top of the construct (quarter section), and ,1.5 mm of depth from the top of the construct (centre section). 10 slices were compiled per section, corresponding to a thickness of 120 mm.

Statistical Analysis
All statistical analyses were carried out using Minitab 15.1. Results are reported as mean 6 standard deviation. The number of constructs analysed per group are provided in the figure legends. Groups were analysed by a general linear model for analysis of variance with groups of factors. Tukey's test was used to compare conditions. Anderson-Darling normality tests were conducted on residuals to confirm a normal distribution. Nonnormal data was transformed using the Box-Cox procedure. Any non-normal data which the Box-Cox procedure could not find a suitable transformation for was transformed using the Johnson procedure. Significance was accepted at a level of p,0.05.

Chondrogenically Primed MSCs can be Stimulated in vitro to Produce a Calcified Cartilaginous Tissue within Solid and Channelled Hydrogels
MSCs were encapsulated in solid and channelled hydrogels and were subjected to chondrogenic culture conditions for a period of 5 weeks, followed by an additional 5 weeks in hypertrophic culture conditions, resulting in a total in vitro culture period of 10 weeks. Solid MSC-seeded hydrogels accumulated significantly more sGAG compared to channelled MSC-seeded hydrogels after 5 weeks (1.3160.06 vs. 0.9460.08%ww, p,0.0001) and 10 weeks (0.7060.01 vs. 0.5860.04%ww, p = 0.0304) of in vitro culture, see Fig. 1(A). Both solid and channelled constructs show a significant reduction in sGAG after 10 weeks compared to their levels at 5 weeks. Both collagen and calcium accumulation increased with time in culture for both solid and channelled constructs, with greater levels of collagen observed in solid constructs after 10 weeks of culture (1.5860.09 vs. 1.2760.15%ww, p = 0.03), see Fig. 1(B). Channelled constructs accumulated significantly more calcium than solid constructs after 10 weeks in culture (3.7260.61 vs. 2.5860.15%ww, p = 0.0035), see Fig. 1(C). mCT analysis confirmed the enhancement of mineralisation in channelled constructs, see Fig. 1(D). Both solid and channelled constructs mineralised preferentially around their periphery, with a reduction in mineralisation observed with depth through the constructs.

Channelled Architectures Accelerate in vivo Mineralisation of MSC-seeded Hydrogels
MSC-seeded solid and channelled hydrogels underwent 5 weeks of culture in chondrogenic conditions, followed by an additional week of culture in hypertrophic conditions, prior to subcutaneous implantation in nude mice. Constructs were harvested at 4 and 8 weeks post-implantation. The DNA content of solid constructs 4 and 8 weeks post-implantation were lower than pre-implantation levels, whereas the DNA content of channelled constructs was higher after 4 weeks in vivo compared to pre-implantation levels, and further increased after 8 weeks in vivo, see Fig. 2 Histological analysis of constructs, pre-and post-implantation, was carried out using alcian blue, picro-sirius red and alizarin red staining to assess the spatial distribution of sGAG, collagen, and calcium deposition respectively, see Fig. 3. Prior to implantation, solid and channelled constructs stained positively and homogenously for alcian blue and picro-sirius red, whereas both constructs only stained positively for alizarin red around their periphery. At 4 weeks post-implantation, solid and channelled constructs stained weakly for alcian blue with channels beginning to show positive staining for nuclear fast red. An increased number of channels stained positive for nuclear fast red at 8 weeks postimplantation. Tissue that had filled the channels also stained intensely for picro-sirius red at 8 weeks post-implantation. At 4 weeks post-implantation, channelled constructs stained homogenously for alizarin red, whereas the staining in solid constructs was heterogeneous, with the core region of the engineered tissue staining negatively, indicating that it was devoid of mineral. H&E staining of acellular (i.e. MSC-free) hydrogels 8 weeks postimplantation demonstrated no bone or cartilage tissue formation in either solid or channelled constructs, though small levels of cellular infiltration were apparent within the channels of channelled constructs (data not shown).
mCT analysis was performed on MSC-seeded hydrogels at 4 and 8 weeks post-implantation to assess the spatial distribution of mineral and it's variance with depth through the construct, and also to quantify the volume of mineral in the construct. Mineral volume, as quantified by mCT, was significantly higher for channelled constructs as compared to solid constructs at both 4 and 8 weeks post-implantation (44.9966.54 vs. 28.5461.53% MV/TV, channelled constructs vs. solid constructs 8 weeks postimplantation, p = 0.014), see Fig. 4(A). At 8 weeks post-implantation, at a depth of ,0.75 mm from the tissue surface, solid constructs had a core region devoid of mineral, whereas a homogenous deposition of mineral was observed in the channelled constructs, see Fig. 4(B). At a depth of ,1.5 mm from the surface, the core region devoid of mineral in solid constructs had increased in area. While there was also evidence of a non-mineralised core region developing in channelled constructs, this area appeared smaller than the corresponding region in the solid constructs.

Channels Act as Conduits for Vascularisation and Provide a Milieu for Endochondral Bone and Marrow Formation
Upon retrieval from the back of nude mice, macroscopically the channels appeared to the filled with a reddish, well vascularised tissue, see Fig. 5(A,B). This vascularised tissue spread through the depth of the channels, see Fig. 5(C). In contrast, macroscopically the solid constructs did not appear to be vascularised, see Fig. 5(D). Histological and immunohistochemical analysis was carried out on constructs to further investigate vascularisation and de novo tissue formation within the channels, and to determine the pathway through which this tissue formation occurs. At 4 weeks postimplantation channels were stained for H&E and for CD31 to examine the presence of blood vessels and endothelial cells. TdT staining was also carried out to assess the proportion of primitive marrow cells in the cellular marrow. Blood vessel structures were identified by H&E and CD31 staining, see Fig. 5(E,F). Vascularisation did not appear to progress significantly from the channels into the calcified cartilage within the hydrogel. TdT staining at 4 weeks post implantation indicated a scanty positive staining as expected for normal marrow, see Fig. 5(G).
H&E staining of solid constructs did not demonstrate bone formation 4 weeks post-implantation, see Fig. 6(A). H&E staining of channelled constructs at 4 weeks post-implantation demonstrated woven bone formation within the channels, surrounding a marrow component consisting of a mixture of hematopoietic foci and marrow adipose tissue, see Fig. 6(B). At 8 weeks postimplantation no bone formation was evident in solid constructs, see Fig. 6(C), whereas within the channels of channelled constructs lamellar-like bone was evident, with the appearance of osteocytelike cells embedded within the bone matrix, which again surrounded a marrow component, see Fig. 6(D).
Prior to subcutaneous implantation (i.e. during the in vitro priming phase), channels had partially filled with a cartilaginous matrix containing MSCs which stained intensely for alcian blue and collagen type II and negatively for collagens type I and X, see    weakly for alcian blue, and very intensely for collagen type I, see Fig. 7(E,F). At this time point a decrease in the intensity of staining for collagen type II, as well as an increase in the intensity of staining for collagen type X, was also observed within channels as compared to pre-implantation levels, see Fig. 7(G,H). This suggests that the cartilage that had filled the channels in vitro supported endochondral bone formation in vivo.

Discussion
New vessel formation is critical for bone tissue regeneration. During long bone growth, canals within the developing hypertrophic cartilage facilitate vascular invasion and the migration of osteogenic cells, thus playing a key role in endochondral ossification. The aim of this study was to mimic this function in tissue engineered hypertrophic cartilaginous constructs. We hypothesised that introducing an array of channels into MSCseeded constructs would firstly facilitate the in vitro engineering of scaled-up hypertrophic cartilaginous tissues, and secondly, would accelerate the vascularisation and mineralisation of the engineered graft following implantation in vivo. During long-term in vitro culture, the presence of this array of channels lead to a reduction in total sGAG and collagen accumulation, whereas calcium deposition was enhanced in channelled constructs. In support of our hypothesis, the channels acted as conduits for blood vessel infiltration and their presence accelerated mineralisation of the engineered tissue in vivo. Finally, channels themselves provided a milieu for bone and marrow formation, with de novo bone being generated within these channels via the endochondral pathway.
Previous studies have investigated introducing nutrient channels into cartilaginous constructs in order to enhance matrix synthesis and the functional properties of tissue engineered cartilage [24]. For example, the introduction of a single nutrient channel into chondrocyte seeded agarose hydrogels results in superior mechanical properties and the formation of a more uniform fibrillar network whilst maintaining similar levels of sGAG and collagen content [22]. In the present study, scaffold architecture differentially regulated matrix synthesis, with solid constructs accumulating significantly more sGAG and collagen, and channelled constructs accumulating significantly more calcium. It may be that the channels provide a pathway for the diffusion of sGAGs and collagen out of the engineered tissue and into the media, and that calcium deposition is less influenced by such phenomena. Alternatively, or perhaps in conjunction, it may be that the altered rates of sGAG, collagen and calcium deposition in solid and channelled constructs is due to differences in the spatial gradients in regulatory cues that develop in these tissues. In solid hydrogels, the cellular consumption of oxygen at the periphery of the construct will result in the development of a low oxygen microenvironment within the core of the engineered tissue [37,38], which has been shown to enhance sGAG and collagen synthesis [33,39]. The introduction of channels into scaffolds may increase core oxygen levels [40], which has been shown to enhance hypertrophy and mineralisation of chondrogenically primed MSCs [33,41].
Previous attempts to engineer scaled-up cartilaginous grafts for endochondral bone repair have reported the development of a core region devoid of cells and cartilaginous matrix [12]. This phenomenon was not observed in our study. However, if the endochondral approach is to be implemented to treat critically sized bone defects, further scaling-up of in vitro cartilaginous grafts will be required, and at these very large scales nutrient diffusion limitations are likely to develop within the construct. It is perhaps only at these dimensions, and where core degradation would otherwise occur, that the introduction of channels would benefit the in vitro development of an engineered cartilaginous tissue.
Following in vivo implantation, biochemical and histological analysis of MSC-seeded constructs revealed a reduction in sGAG content and an increase in calcium accumulation, indicating a loss of the chondrogenic phenotype and progression down the endochondral route. Interestingly, while solid constructs maintained their pre-implantation collagen levels during the course of the in vivo implantation period, channelled constructs continued to accumulate collagen, which may be due in part to the formation of new tissue within the channels themselves. Both histological data and the DNA assay suggested that channels became highly cellularised once implanted in vivo. While we cannot rule out a role played by donor cells, as there were MSCs present within channels pre-implantation, previous studies investigating bone tissue engineering via the endochondral pathway have highlighted an important role played by recruited host derived cells in driving endochondral ossification [10,14]. It would appear that the channels in our tissue engineered hypertrophic cartilaginous constructs may be acting as conduits for cellular infiltration from the host, which are then playing a role in laying down new tissue within the construct.
In bone tissue engineering strategies, the prevention of core necrosis during in vitro development and following in vivo implantation is a significant challenge [5]. In the present study we utilised mCT to investigate the spatial variance of in vivo mineralisation with depth through the tissue and found that channelled architectures accelerated mineralisation, and furthermore, promoted the development of a more homogenous mineralised construct and limited the development of a core region devoid of viable cells and mineral. In addition to the inherent increase in nutrients and oxygen due to presence of the channels, they also facilitate blood vessel infiltration as evident by staining for CD31 positive endothelial cells, which in turn further increases oxygen levels and provides soluble factors and signals to promote osteogenesis [42].
Subcutaneous implantation of chondrogenically primed MSCs has been shown to produce endochondral bone containing hematopoietic marrow at 8 weeks post-implantation [8,9]. In the present study, high magnification images of channels at 4 weeks post implantation demonstrated deposition of an immature woven bone, being laid down by osteoblasts, which at 8 weeks postimplantation had developed into lamellar-like bone with osteocytelike cells evident within the bone matrix. Temporal analysis of chondrogenic extracellular matrix progression within channels indicated degradation of the sGAG/collagen type II rich-matrix following in vivo implantation, and a corresponding increase in the accumulation of collagen type I and collagen type X, suggesting that de novo bone found in the channels was being generated via the endochondral pathway. Moreover, at both in vivo time points there was evidence of hematopoietic foci and marrow adipose tissue, enclosed within the developing bone matrix. Given that it has previously been demonstrated that endochondral ossification is required for hematopoietic stem cell niche formation [43], this provides further validation that the mechanism through which bone is formed within the channels is indeed endochondral.
In the current study cartilaginous tissues were engineered using 4 month old, skeletally immature, porcine MSCs. Skeletally immature MSC seeded hydrogels have been shown to generate cartilaginous tissues with superior matrix deposition and mechanical properties as compared to skeletally mature MSC seeded hydrogels [44]. Further studies are required to confirm if the beneficial effect of channels, as shown in this study, would occur with the use of engineered tissues generated from a skeletally mature donor where the inherent regenerative capacity is not as high.
Bone tissue engineering via endochondral ossification has recently been demonstrated with collagen mesh scaffolds [12], on three-dimensional electrospun fibers [45], and with the bone void filler NuOss [46]. Utilising hydrogels as scaffolds for endochondral bone regeneration may be a particularly powerful tool in 'scaling up' tissue engineered grafts in order to treat defects of a clinically relevant size [38]. However, the chondrogenically primed MSC-seeded agarose hydrogels utilised in this study appeared to only partially support progression towards endochondral ossification, with the engineered tissue laid down in the main body of the hydrogel apparently locked within a hypertrophic, calcified cartilage state at 8 weeks post-implantation. Since the cartilaginous matrix within channels, which had filled up at least partially in vitro with scaffold-free tissue, had the ability to undergo full endochondral ossification in vivo it would appear that the agarose hydrogel acted as a barrier for vascularisation and the transition from mineralised cartilage to endochondral bone. This may be due to the agarose hydrogel undergoing minimal degradation during the 8 week in vivo period. A previous study investigating alginate hydrogels as scaffolds for bone tissue engineering demonstrated enhanced bone formation by accelerating the degradation properties of the hydrogel [47]. Future studies in our lab will examine the capacity of different naturally derived biodegradable hydrogels to produce endochondral bone in vivo, with the ultimate goal of scaling up anatomically accurate cartilaginous grafts as a paradigm for whole bone tissue engineering via endochondral ossification. However, all scaffolds or hydrogels will likely impede vascularisation to some degree, and adopting the channel strategy developed in this study is expected to accelerate the process of endochondral ossification regardless of the hydrogel used to tissue engineer hypertrophic cartilage grafts for bone regeneration.

Conclusion
Inspired by the cartilage canal network observed during endochondral skeletogenesis, this study demonstrated that tissue engineering channelled hypertrophic cartilaginous constructs accelerates mineralisation and vascularisation of the graft in vivo and shows promise for use in future endochondral bone regeneration strategies. The study reinforces the importance of optimising the architecture of engineered constructs targeting bone tissue regeneration, even if this is achieved via an endochondral pathway as opposed to the traditional intramembranous route.