Biological and molecular profile of fracture non-union tissue: current insights

Delayed bone healing and non-union occur in approximately 10% of long bone fractures. Despite intense investigations and progress in understanding the processes governing bone healing, the specific pathophysiological characteristics of the local microenvironment leading to non-union remain obscure. The clinical findings and radiographic features remain the two important landmarks of diagnosing non-unions and even when the diagnosis is established there is debate on the ideal timing and mode of intervention. In an attempt to understand better the pathophysiological processes involved in the development of fracture non-union, a number of studies have endeavoured to investigate the biological profile of tissue obtained from the non-union site and analyse any differences or similarities of tissue obtained from different types of non-unions. In the herein study, we present the existing evidence of the biological and molecular profile of fracture non-union tissue.


Introduction Materials and Methods
Introduction Bone healing is a complex but well-orchestrated physiological process which recapitulates aspects of the embryonic skeletal development in combination with the normal response to acute tissue injury [1,2]. It encompasses multiple biological phenomena and is margined by the combination of osteoconduction (scaffold formation), osteoinduction (timed cellular recruitment controlled by multiple signalling molecules) and osteogenesis (new bone forma-tion) [2][3][4][5]. In contrast to the scar formation, which occurs in the majority of other tissue types in adults, bone has the innate capability to repair and regenerate, regaining its former biomechanical and biochemical properties [6][7][8].
During the bone healing process, a well-regulated series of overlapping processes take place in the cortical bone, the periosteum, the bone marrow and the undifferentiated fascial tissue surrounding the fracture [10,12,13]. According to the histological appearance, two basic types of bone healing have been identified [6,7,11]. The primary (direct) healing pattern occurs when anatomical reduction is achieved, along with almost absolute stability [3,15]. The disrupted continuity of the bone in this type of healing is re-established with regeneration of the Harvesian system and the lamellar bone, with therefore no need of any remodelling [12,15]. On the contrary, the secondary (indirect) healing pattern that occurs in the vast majority of clinical cases depends to the formation of fibrocartilaginous callus [3,6]. This process can be broadly divided into five stages: that of inflammation, granulation tissue formation, soft callus formation (hyaline cartilage), hard callus formation (woven bone) and remodelling [6,9,11,14].
In more detail, following an injury the bone architecture is disrupted, as is the surrounding soft tissue continuity. Consequently, the local blood vessels are torn, a haematoma is formed and the coagulation cascade is activated [16]. This fracture haematoma contains cells that originate from the peripheral and intramedullary blood, as well as from the bone marrow [15]. They include inflammatory immune cells, neutrophils, monocytes and macrophages that are activated by the coagulation process; fibroblasts; and mesenchymal stem cells (MSCs) [6,16]. Prostaglandins, cytokines and other proteins are abundant in this environment and contribute to the formation of a complex microenvironment which has different effect on each cell population [6]. These mediators are known to increase cellular migration, proliferation, enhance osteogenesis, collagen synthesis and angiogenesis [6].
Subsequently, the necrotic or damaged pieces of bone are removed and the fracture haematoma is gradually replaced by granulation tissue [17]. The osteoprogenitor cells then proliferate and differentiate, leading to deposition of collagen and formation of soft callus. An increased vascularity and intense cell proliferation in the cambium layer of the periosteum is evident in this stage [13,17]. Bone formation then occurs by endochondral or intramembranous ossification. Initially, immature woven bone characterized by coarse collagen fibres arranged in a haphazard fashion is formed, but is then transformed to mature lamellar bone (remodelling) in a slow process [13,17]. During remodelling that could last several months to years after fracture, both osteoblast and osteoclast activity is intense, with bone resorption followed by appositional production of new bone by osteoblasts [17].
In vitro investigations to evaluate osteogenic activity include measurements of a number of secreted substances (proteins) including: alkaline phosphatase (ALP), osteonectin, osteopontin, osteocalcin and bone sialoprotein. Alkaline phosphatase is a key protein secreted by osteoblasts in response to osteogenic activity and represents a marker of the earlier stage of osteoblast differentiation [18]. Osteonectin, osteopontin and osteocalcin are non-collagenous bone matrix proteins, abundant in bone tissue [19]. They are thought to be of great importance in bone development, growth, turnover and fracture repair; along with osterix, as essential factor for osteoblast differentiation and bone formation, they represent markers of the later stage of differentiation [18][19][20]. Bone Sialoprotein, an extracellular matrix protein secreted by osteoblastic cells, has also been reported to modulate osteoblast differentiation and mineralization [21].
As already mentioned, the physiological sequence of fracture healing depends on numerous endogenous and exogenous factors [22,23]. If this sensitive balance is altered in any way, complications may arise, such as delayed union or non-union. The criteria for defining a non-union are not yet standardized [24]. FDA (Food and Drug Administration) defines a non-union as the incomplete fracture healing within 9 months following injury, along with absence of progressive signs of healing on serial radiographs over the course of three consecutive months [25]. In the United States alone, it is estimated that 5-10% of all fractures are complicated by non-union or delayed union [26], posing an enormous economic burden to the healthcare system [27]. The tibia and the femur are the most common long bones associated with the development of non-union [28,29].
According to the radiological and histological appearance, nonunions are characterized as: hypertrophic, usually resulting from insufficient fracture stabilization (extensive callus formation) [30]; and atrophic, where the fracture stabilization is adequate but there is localized dysfunction in biological activity (little callus formation and presence of a fibrous tissue-filled fracture gap) [30,31]. Synovial pseudarthrosis is considered as a different pathological entity, caused by inadequate immobilization with or without the presence of infection [32]. Moreover, non-unions can be characterized according to the presence of bacteria at the fracture site, as septic or aseptic nonunions [33].
It is generally accepted that the progression to a non-union in most cases represents a multifactorial process. Various risk factors have been implicated with compromized fracture healing, including: patient dependent factors such as age, gender, medical comorbidities (i.e. anaemia, diabetes and hormone disorders), smoking and administration of pharmacological agents (i.e. steroids, non-steroidal antiinflammatories, etc.); and patient independent factors such as the 'personality' of the fracture, presence of infection and adequacy of surgical technique [22,25,34].
The exact biological process leading to a non-union remains obscure and it is well accepted that any planned interventions to reverse this process should be well-timed and well-aimed to restore both biological and mechanical deficiencies [3,14,31,35]. It can be postulated that by gaining a better understanding of the underlying mechanisms leading to a non-union, both clinicians and scientists would be allowed to target specific pathways independently, tailoring treatment to each patient's individual requirements [11]. Therefore, the purpose of this review is to investigate the biological profile of tissue obtained from the non-union site and to analyse any differences or similarities of tissue obtained from different types of non-unions. Moreover, it aims to evaluate whether any interventions on the tissue obtained would influence in a positive aspect its biological characteristics and bone repair responses.

Eligibility criteria
Studies selected were original articles fulfilling the following inclusion criteria: (i) the tissue was obtained from a non-union site and examined or processed for defining its characteristics and properties; (ii) only tissue acquired from human subjects was included; (iii) articles were published in English language and (iv) the full text of each article was available. All studies that did not fulfil all eligibility criteria were excluded from further analysis, whereas no publication date restrictions were imposed.

Information sources
Studies were identified by searching the following resources/databases: PubMed Medline; Ovid Medline; Embase; Scopus; Google Scholar; and the Cochrane Library, to retrieve all available relevant articles. The terms used for the search included: non-union(s), nonunion(s), human, tissue, bone morphogenic protein(s) (BMP's) and MSCs. The identified articles and their bibliographies including any relevant reviews were manually searched for additional potential eligible studies.

Study selection
Two of the authors (M.P., I.P.) performed the eligibility assessment, in an independent, unblinded and standardized manner. Most citations were excluded on the basis of information provided by their respective title or abstract. In any other case, the complete manuscript was obtained, scrutinized by the two reviewers and included if fulfilling the eligibility criteria. Any disagreement between reviewers was resolved by consensus.

Extraction of data
Relevant information on author's name, publication year, patient demographics, site and duration of non-union, type of the non-union, characteristics and evaluation of tissue samples, culture properties, gene expression, protein expression and effect of additional interventions was carefully extracted.

Data analysis
All outcomes of interest were inserted in an electronic database and outcome of different studies were documented. The characteristics of tissue samples were then compared across different studies and the effect of any intervention was evaluated.

Macroscopic structure of non-union tissue
Urist et al. was the first to hypothesize the mechanism of nonunion based on its macroscopic and microscopic characteristics [53]. He reported that white soft tissue was interposed between the bone segments, a finding later supported by other authors [51], and explained this as fibrinoid degeneration of the connective tissue in the interior of the callus [53]. With regards to synovial pseudarthrosis, a yellow frond-like material was found interposed between the bone fragments, with clear serous fluid filling this space in aseptic cases, whereas in septic cases murky fluid was present [32].

Neuroimmunohistochemistry
Only one study performed neuroimmunohistochemical analysis revealing paucity or total lack of peripheral innervation in the nonunion tissue [48].  Analysis of vessel density Blood vessels were present in cases of hypertrophic non-unions, with a varying density (Table 6) [44,48,50]. When comparing however atrophic and hypertrophic non-union tissue, an interesting finding was that the number of fields containing no blood vessels, some blood vessels and hot-spots, was very similar [44]. This was also confirmed with immunohistochemistry studies, where no significant difference was evident in the median vessel count between atrophic/ hypertrophic non-unions and normal unions [44]. Finally, histological findings confirmed the presence of vascular tissue in both types of non-unions (Table 3) [19,40,44,46].

Electron microscopy
Two studies performed ultrastructural examination of the non-union tissue by the means of electron microscopy (Table 6) [32,50]. In a study by Quacci et al., it was found that the non-union tissue contained normal fibroblasts and chondrocytes [50]. In addition, Heppenstall et al. who examined synovial pseudarthrosis reported large amounts of surface fibrin and densely packed collagen [32].

Cell senescence
Bajada et al. was the only author to report on the cell senescence of non-union stromal cells [40]. According to his findings, from passage I onwards, many of the cells developed an appearance that was less bipolar and more spread along with the development of prominent stress fibres. Further passages lead to prolonged culture doubling times (phenotypic changes are consistent with the onset of cell senescence). When examining the proportion of SA-b gal positive cells, that was significantly greater in the non-union stromal cells  when compared to the bone marrow stromal cells, but that did not correlate with the patient's age, number of previous operative procedures or time between original fracture and operative management.

Osterix
Koga et al. has studied the effect of low-intensity pulsed ultrasound on non-union cells cultured with the presence of BMP-7 and reported no significant difference in the expression of osterix [18].

Osteonectin
Osteonectin expression was investigated by Lawton et al. [19]. Osteonectin was found to be strongly positive in non-cuboidal and induced osteoblasts of early woven bone, as well as cuboidal osteoblasts of later woven bone. Included osteoblasts and flattened lining cells on lamellar bone were only weakly positive, whereas endothelial cells were consistently negative.

Osteopontin
Lawton et al. investigated osteopontin expression during the different stages of repair [19]. Osteopontin was found to be weakly positive in non-cuboidal osteoblasts on early woven bone, and moderately positive in cuboidal osteoblasts on the surface of woven bone later in repair. Multinucleate resorptive cells were associated with a strong signal, in comparison with most flattened cells on the surface of lamellar bone and endothelial cells that were negative.

Bone Sialoprotein
Iwakura et al. studied the expression of Bone Sialoprotein under osteogenic conditions and found it to be higher in the non-union cells than under undifferentiated conditions in the human dermal fibroblasts (controls) [30].

Dickkopf-1 expression
The expression of Dickkopf-1 (Dkk-1) was studied by Bajada et al. [40]. According to his findings, both non-union and bone marrow Atrophic Samples largely consisted of fibrocartilaginous tissue that contained occasional bony islands. In some areas, the excised non-union tissue was well populated by fibroblastic cells, but other areas were largely acellular and consisted mostly of a collagenous extracellular matrix. Areas of vascularization were seen consistently and the presence of osteoclasts within absorption pits was also occasionally notable. After enzymatic treatment to extract cells and their plating out into monolayer culture, the majority of the adherent cells present were stromal in appearance, i.e. bipolar and fibroblastic. Occasional multinucleated osteoclasts were also seen in the early cultures, as were cells with a stellate (possessed multiple cytoplasmic processes) or dendritic appearance Bajada [43] Hypertrophic Fibrocartilaginous non-union with little evidence of new bone formation and no signs of infection Reed [44] Hypertrophic Specimens contained fibrous tissue, fibrocartilage, hyaline cartilage and bony islands. Areas of new bone formation by both endochondral and intramembranous ossification. Morphologically samples appeared well vascularized Reed [44] Atrophic Specimens contained fibrous tissue, fibrocartilage, hyaline cartilage and bony islands. Relatively few areas of new bone formation, predominantly via the endochondral route. Necrotic bone was more prevalent in the atrophic non-union group.
Morphologically samples appeared well vascularized Kloen [45] Not mentioned Delayed unions and non-unions: 11/21 specimens had foci of woven bone (having cuboid-shaped osteoblasts lining the osteoid, suggesting active bone formation) surrounded by large areas of fibrous tissue that was interspersed with areas of numerous blood vessels. Ten of 21 specimens had similar areas of fibrous tissue but lacked woven bone. Within the samples that contained woven bone, two patterns of bone formation were observed: (i) bone appeared to be forming directly from fibrous tissues; (ii) bone seemed to be forming from cartilage. Other observations included scattered lamellar bone fragments surrounded by osteoclasts and a paucity of lining osteoblasts. Some specimens also showed villous projections resembling synovial pseudarthroses with lining cells resembling synoviocytes Guerkov [46] Atrophic: 4; hypertrophic: 3 Mainly fibrous tissue with organized collagen bundles. No ossicles were seen in any of the sections examined. All sections from atrophic non-unions were oligocellular and contained few vessels, whereas those from hypertrophic non-unions were more cellular, with little evidence of cartilaginous tissue Lawton [19] Not mentioned (had callus) Human fracture callus: heterogeneous appearance with several of the elements of normal fracture healing (haematoma, fibrous tissue, woven and compact lamellar bone, and cartilage) being present in close proximity in any one section. Non-union gap: tissues consisted largely of vascularized fibrous tissue or avascular cartilage stromal cells secreted Dkk-1 into conditioned medium at comparable levels under control (i.e. non stimulated) conditions. However, Dkk-1 levels detected in stimulated non-union stromal cells conditioned medium were markedly and significantly greater than those found in stimulated bone marrow stromal cells cultures.

Gene expression
Several authors have examined the expression of different genes in the non-union tissue. A summary of their results is outlined in Table 11 [14,22,30,42,52] and Table 12 [47,49].
Western blot assay Western blot assay was used to detect the presence of specific proteins in the tissue under examination. Fajardo et al. investigated the presence of MMP's and reported that MMP-7 and MMP-12 were present in both non-union and mineralized callus tissue; however, the signal intensity of both enzymes was stronger in the non-union tissue [14]. In another study, he and his team examined the presence of BMP's [39]. His finding included: BMP-2 was present in both non-union and mineralized callus tissue; BMP-4 was detected in non-union samples but decreased in healing bone samples; BMP-7 was detected in the healing bone but was absent in the non-union samples.

Effect of interventions to the non-union tissue
Genetic predisposition to fracture non-union Several authors have investigated the theory of genetic predisposition to fracture non-union by analysing samples from peripheral venous blood [33,54] or bone callus [55], and comparing them with uneventful healing fractures. Numerous polymorphisms such as those of two specific SNPs (rs1372857, genotype GG and rs2053423, genotype TT) were identified to be associated with an increased risk of developing non-union [33,55,56].

Discussion
Non-unions represent a significant public health problem and have been associated with devastating consequences for the patients, their family and the society as a whole [57]. The mechanism behind the progression of a fracture to a non-union state is multifactorial and as a consequence the treatment can be very challenging. The treatment of non-unions has evolved over the years from prolonged immobilization [53] to the use of biological stimulation and polytherapy. Such a strategy attempts to address all the elements of a compromized fracture healing response [3,31]. With regard to the macroscopic appearance of non-unions, a common finding is the interposition of soft tissue between the bone fragments [51,53]. In aseptic non-unions, this tissue is whiter in colour, occasionally surrounded by clear fluid, compared to infected non-unions where this tissue becomes more yellowish and frequently surrounded by murky fluid [32]. The experience of the authors confirms the above findings and in fact the macroscopic appearance of the non-union tissue is used as an additional marker for confirming/ suspecting an underlying septic process.
Regarding the culture characteristics of the non-union tissue, there was an inconsistency in the reported findings. This may be because of the different types of non-union tissue examined (i.e. atrophic and hypertrophic), as well as because of the different topography of the non-unions from where samples were obtained. Finally, the expression of several genes was reported to be different in non-union tissue and controls [14,22,30,39,42,52], a finding suggesting that such differences may contribute to the pathogenesis of non-unions.
Several similarities were reported in the histological analysis of atrophic and hypertrophic non-unions. The main types of tissues involved include fibrous, cartilaginous and connective tissue in varying degree [30,40,43,44,46,50]. In atrophic non-unions, bony islands were not always present [30,40,43,44,46,50], whereas necrotic bone was more prevalent [44]. Generally, the cellular density of atrophic non-unions was lower compared to hypertrophic nonunions, while some areas were completely acellular [40,46]. This suggests a different cellular background, which may correspond to the higher failure rate following revision surgery of atrophic nonunions [31].
More importantly, Iwakura et al. showed that tissue derived from hypertrophic non-unions contains MSC's [30], a finding later confirmed by Koga et al. [18]. Similarly, Bajada et al. reported the presence of biologically active cells in atrophic non-union tissue, largely CD34/CD45-negative, CD105-positive, with the potential to differentiate to osteoblastic, adipogenic and chondrocytic lineages [40].
In contrast to the common preconception that atrophic nonunions are relatively avascular and inert [44,58], several authors have confirmed the vascularity of the atrophic non-union tissue [19,32,40,44,46,48,50]. In addition, Reed et al. reported no significant difference in the vessel density between atrophic non-unions, hypertrophic non-unions and healing fractures [44]. This biological finding may be of importance, as it suggests that treatments targeting to the enrichment and restoration of local angiogenesis could be applied as an effective treatment modality in the clinical setting.

Bone production
Predominantly via the endochondral route [44] Bone formation by both endochondral and intramembranous ossification [44] Cells Generally oligocellular [46]; some areas acellular [40] More cellular [46] Fibroblastic: majority of cells [40] Osteoclasts: occasionally [40] Bipolar cells: majority of cells [40] Cells with a stellate (possessed multiple cytoplasmic processes) or dendritic appearance [40] Fibroblast-like [30] Vascularization Well vascularized [40,44]; few vessels [46] Well vascularized [44] 698 intra-operative samples can be negative [37,38]. A possible explanation for this phenomenon could be the presence of biofilms (bacteria adhere on implants and tissues around the fracture site, forming matrix-enclosed communities), which are resistant to "normal" concentrations of systemic antibiotics [37]. Palmer et al. and Gille et al. have reported the benefit of utilizing molecular based techniques to identify these infections [37,38]. This can be very important, as distinguishing between septic and aseptic non-union is essential for determining the course of treatment. However, limitations of their use in clinical practice include: the fact that single-primer PCR can only Fajardo [14] Hypertrophic MMP-7 and MMP-12 were found to be stained within the substance of the non-union tissue and not localized within a particular cell type or cellular component. Both enzymes were likewise not visualized in the bone callus specimens Kwong [35] Aseptic non-unions, only fractures with areas of cartilage were chosen There was a significant reduction in BMP-2 and BMP-14 expression in cartilaginous areas of non-healing fractures compared to healing fractures, but no statistical differences in the endogenous expression of noggin and chordin (BMP inhibitors) Fajardo [39] Hypertrophic BMP-7: absent in the non-union specimens but present in the fracture callus specimens. BMP-2: positive immunostaining was restricted consistently to the fibrous tissue of the non-union tissue Kilian [52] Atrophic Immunostaining appeared in close vicinity to immature osteoid trabeculae. EDB+ fibronectin immunostaining was negative for scFvL19 antibody Reed [44] Hypertrophic No statistically significant difference in median vessel counts between atrophic, hypertrophic and normal unions Reed [44] Atrophic No statistically significant difference in median vessel counts between atrophic, hypertrophic and normal unions Kloen [45] Not mentioned The most consistent expression was that of BMP-2, BMP-4, and BMP-7 in the osteoblasts lining the newly formed osteoid. The staining was cytoplasmic and, in certain specimens, was specifically located in the Golgi apparatus, illustrating local production of BMP. No correlation between the location of the delayed union or non-union and staining. In the areas of dense fibrous tissue the presence of staining for all BMP isoforms tested was the same as or less than that in the areas close to bone at all time-points after the fracture. Expression of Type IA, Type IB, and Type II BMP Receptors: positive staining was observed in the osteoblasts lining the ossified tissue, in the areas near the ossification sites, and in the fibrous tissue. As observed for the BMP antibodies, there was a trend towards decreased staining in areas remote from bone formation. There was no clear trend between a decreased percentage of positive staining and an increased duration of the non-union. Expression of pSmad1: in the osteoblasts lining the areas of reactive bone formation as well as in osteoclasts, fibroblast-like cells and chondroblast-type cells Lawton [19] Not mentioned (had callus) In normally healing fractures, mature osteoblasts on woven bone were negative for MGP mRNA, but positive for osteonectin, osteopontin and osteocalcin mRNA molecules. In non-unions, osteoblasts displayed a novel phenotype: they were positive for MGP mRNA, in addition to osteonectin, osteopontin and osteocalcin mRNA molecules Lawton [47] Not mentioned (had callus) In areas of new bone covered by plump osteoblasts, the matrix was either stained uniformly or in a superficial zone, indicating the presence of collagen type III. Fibrous tissue in the fracture gap was also immunostained positively detect one target organism [37]; concerns for oversensitivity with regard to clinical relevance [37,59] and associated cost implications. Cell senescence is known to play an important role in healing and tissue regeneration [60]. In essence, the senescence of adult stem cells or more differentiated cells present in the non-union tissue may represent one of the main mechanisms of the loss of the regenerative potential, leading to healing impairment [60]. As already mentioned, Bajada et al. reported that an increased proportion of non-union stromal cells were senescent when compared to bone marrow stromal cells, which did not correlate with the patient's age [40]. However, the pathways leading to this genomic damage and the contribution of several factors (such as repeated cellular replication and the consequent cell stress [40]) are yet to be determined.
Bone morphogenic proteins are some of the major signalling molecules, promoting the differentiation of MSC's into chondrocytes or osteoblasts [12,13]. Kloen et al. reported evidence of ongoing BMP signalling in the non-union tissue, where endogenous BMP's, their receptors and molecules involved in their signal transduction were present in the tissue [45]. Moreover, others have suggested that imbalance in the expression of BMP's and their inhibitors Drm (gremlin), follistatin, noggin and chordin, might account for the impaired bone forming ability [35,39]. When the non-union tissue was cultured in the presence of exogenous BMP, the MSC's differentiated into functional osteoblasts, with an increased bone nodule formation [41,49]. Treatments regulating concentrations of BMP's have already been used in clinical practice with encouraging results (such as BMP-2 and BMP-7 [31]). Future research is needed to investigate the effects of similar agonist molecules or their inhibitors.
Matrix metalloproteinases are proteases that play an important role in bone remodelling and bone repair. When the MMP's or their Reed [44] The number of fields containing no blood vessels, some blood vessels and hotspots was very similar in the atrophic and hypertrophic non-union groups Not applicable Santavirta [48] Samples mostly consisted of vascularized connective tissue of varying density Not applicable Quacci [50] A lot of blood vessels were present in the tissue, often appearing free of blood and occluded by thrombi at different organization stages Fibroblasts and chondrocytes found in the non-union tissue seemed normal, with a good secretion apparatus. The cell membranes were able to produce matrix vesicles. Hydroxyapatite crystals could be observed in the cell matrix or inside matrix vesicles Heppenstall [32] Not applicable (5 patients) Large amounts of surface fibrin. Some cells had profuse rough endoplasmic reticulum and resembled fibrocytes or Type B synovial lining cells. Some of these cells contained prominent lipid droplets and intermediate filaments. There were also phagocytic cells with vacuoles containing granular and cellular debris, resembling to Type A lining cells or monocyte-macrophages. Surrounding the cells were some necrotic cells, clusters of apatite crystals and occasional clumps of collagen fibres infiltrated with more fibrin-like material. Deeper was more densely packed collagen Table 7 Cell surface protein expression Author Cell surface protein expression (flow Cytometry) Koga [18] Strongly positive for the MSC's related markers CD29, CD44, CD105 and CD166 but negative for the hematopoietic markers CD14, CD34, CD45 and CD133 Iwakura [30] Positive for MSC's related markers CD13, CD29, CD44, CD90, CD105 and CD166, but negative for hematopoietic markers CD14, CD34, CD45 and CD133 Bajada [40] Less than 1% of NUSC and BMSC were immunopositive for CD34 and CD45, while 78% AE 14% (mean AE SD) of NUSC and 92% AE 7% (mean AE SD) of BMSC were immunopositive for CD105 MSC: mesenchymal stem cells; NUSC: non-union stromal cells; BMSC: bone marrow stromal cells.   inhibitors are disrupted, disorders of fracture healing may occur [14]. In a study by Fajardo et al., MMP-7 and MMP-12 genes were reported to be significantly up-regulated within the tissue of hypertrophic non-unions [14]. When the hypertrophic non-union tissue was examined in vitro, it was found that the same proteins directly bounded to and degraded BMP-2, a highly osteoinductive agent [14]. This action of the MMP's may be responsible for the impaired fracture healing in the case of hypertrophic non-unions, even though the same finding may not correlate to atrophic fracture non-unions.
Several reports suggest that low-intensity pulsed ultrasound treatment stimulates bone healing, although the mechanism behind this remains obscure [61,62]. When applying low-intensity pulsed ultrasound in non-union cells cultures, it was found that there was a significant effect on the osteogenic differentiation rather than proliferation of non-union tissue cells [18]. In addition, growth factor synthesis and release was stimulated [46]. The use of low-intensity pulsed ultrasound can therefore improve union rates and accelerate the healing process.
Dickkopf-1 is a secreted protein acting as an antagonist of the Wnt signalling pathway, suppressing fracture repair by inhibiting osteogenic differentiation [40,63]. Bajada et al. has compared the levels of Dkk-1 in atrophic non-union stromal cells and bone marrow stromal cells, reporting an increased secretion by the non-union cells, associated with reduced osteoblastic differentiation [40]. When they treated the bone marrow stromal cells with recombinant human Dkk-1 or conditioned medium from the non-union cells, the effect on osteogenic differentiation remained inhibitory [40]. This finding suggests that Dkk-1 may play an important role in the development of nonunions, however further research is needed to shed more light on the underlying mechanism of an increased Dkk-1 production by nonunion cells.
Another important element of progression to non-union that needs to be discussed is genetic predisposition. Several authors have investigated this theory by analysing samples from peripheral venous blood [33,54], and bone callus [55] and comparing them with uneventful healing fractures. Numerous polymorphisms such as those of two specific SNPs (rs1372857, genotype GG and rs2053423, genotype TT) were identified to be associated with an increased risk of developing non-union [33,55,56].
The herein study has some limitations. First, it excludes studies involving experimental animal models. However, the outcome of such studies should be treated with caution, as they cannot be translated directly to the clinical scenarios. Second, there is an inherent inconsistency in defining non-union, and as such the timing of tissue harvesting would be slightly different, which might be responsible for some of the differences reported among similar studies. Moreover, as the term MSC's is fairly recent, studies performed in earlier years used a different terminology for the same cells, such as osteoprogenitors, skeletal stem cells, etc. As a result, their findings could not be compared to those of more recent studies.
Strengths of the study include the systematic approach of analysing the results and the detailed careful analysis of the data obtained. Collectively, this manuscript presents our current understanding of the molecular and cellular pathways that can be involved in the development of non-union. Direct recommendations to be applied in the clinical setting cannot be safely made with the available evidence. We deem essential that a widely accepted definition of the timeframe for non-unions should be set allowing an earlier intervention in such cases. The conceptual frame of the "diamond concept" for a successful fracture healing response should be considered in cases where bone repair is desirable [5]. Cellular therapies and inductive molecules with scaffolds have a role to play in future treatment strategies, as would do tissue engineering approaches [64]. Although still under intense investigation genetic therapy could be another treatment option in the foreseeable future.

Conclusion
In conclusion, failure of fracture healing and progression to nonunion represents a not uncommon clinical complication carrying devastating consequences. The histopathological appearance of nonunion tissue between atrophic and hypertrophic non-union indicates that both types of non-unions are not avascular and contain a potentially active population of MSC's. Pathways believed to be involved in their pathogenesis include an imbalance in the expression of BMP's and their inhibitors, and an up-regulated expression of several substances such as that of the MMP's and Dkk-1which can block  [22] Genes expressed more than two times than in normal tissue: CDO1; PDE4DIP; COMP; FMOD; CLU; FN1; ACTA2; TSC22D1 Not applicable Fajardo [14] MMP-7 and MMP-12 mRNAs were significantly elevated in the non-union tissue when compared with local mineralized callus from the same site MMP-7 and MMP-12 were the only enzymes (of 53 examined) significantly elevated in non-union tissue when compared with local mineralized callus from the same site Iwakura [30] Not applicable It showed the expression of mRNA of Col II, Col X, SOX9 and aggrecan chondrogenic conditions after a 21-day induction. Under adipogenic conditions after a 21-day culture period, it showed the expression of LPL and PPAR-g2 (higher than under undifferentiated conditions in the control group) Fajardo [39] BMP gene expression in healing bone displayed several up-regulated genes between the two tissues BMP antagonist genes (DRM, follistatin, noggin): increased in non-union tissue when compared to fracture callus tissue. BMP receptors (R1A, R1B, R2): expressed but did not demonstrate any significant differences. BMP-4: up-regulated in non-union tissue when compared to the fracture callus tissue. RNA levels of the BMP antagonists Drm/Gremlin, follistatin and Noggin: up-regulated in the non-union tissues. BMP-7: increased in the fracture callus tissue Hofmann [42] Gene terms significantly overrepresented in human non-union osteoblast cultures: skeletal development; response to wounding; organ morphogenesis; vasculature development; proteinaceous extracellular matrix; extracellular space; cytokine activity; glycosaminoglycan binding; growth factor activity; insulin-like growth factor binding. Genes significantly down-regulated in human non-union osteoblast cultures: IGF-2, FGF-1, FGF-receptor 2 (FGF-R2), BMP-4, TGF-b2, PDGF, Wnt-induced proteins (WISP2 and 3), b-catenin and prostaglandin E2 receptor EP4 Confirmed the results of the microarray, especially regarding the down-regulation of some genes involved in osteoblast differentiation and bone metabolism Kilian [52] Not applicable In qualitative and quantitative RT-PCR, EDA+ fibronectin mRNA was detectable at low levels. in none of the seven non-union samples, EDB+ fibronectin mRNA transcription was detected by qualitative and quantitative PCR  Lawton [47] Not applicable Signal for procollagen type I mRNA over fibroblasts and over osteoblasts on woven bone was uniformly strong in most non-unions and normal fractures Not applicable Non-unions: in the zone of new bone formation and the interface zone, a population of surface and included osteoblasts was strongly positive for the procollagen type III mRNA signal; osteoblasts in the old zone were usually negative, while the gap zone contained osteoblasts only rarely; fibroblasts were frequently positive in the gap zone and interface. Normal fractures: procollagen type III mRNA was seen in the very early granulation tissue, where most of the positive cells were mesenchymal spindle cells (a cell population that includes osteoblast precursors; osteoblasts were in the vast majority negative; small areas of fibrous tissue in which fibroblasts were either negative or weakly positive Boyan [49] BMP (bovine or dog) There was no stimulation of Type I collagen message in the non-union fibrocartilage cells. Non-union periosteal cells were found to be more strongly activated by BMP The increase in mRNA levels of Type II collagen was not significant compared to controls Not applicable Table 13 Comparison between atrophic/hypertrophic non-union tissue

Type of analysis Atrophic Hypertrophic
Histology Table 4 Immunohistochemistry/vessel density No difference in the median vessel count between atrophic/hypertrophic non-unions [44] Cell surface antigen profile CD 105 [40] CD13, CD29, CD44, CD90, CD105, and CD166 [30] Cells formed a uniform monolayer of elongated cells that had few cellular extensions [46] Also consisted of elongated cells, but the cells were more cuboidal, having cellular extensions in a multilayer [46] Cell proliferation No significant effect of pulsed electromagnetic field stimulation [46] ALP activity No differences between cultures from atrophic or hypertrophic non-unions [46] Osteocalcin Low levels [46] Low levels [46]; higher than in human dermal fibroblasts [30] Mineralization assay Reduced compared to bone marrow stromal cells [40] Higher than haematoma cells [30]; lower than human osteoblasts (normal healing) [42] ª 709 Table 14 Effect of interventions the BMP and Wnt pathways respectively. Immerging evidence also support a genetic predisposition in this patient group.