Biological and molecular profile of fracture non‐union tissue: A systematic review and an update on current insights

Abstract Fracture non‐union represents a common complication, seen in 5%–10% of all acute fractures. Despite the enhancement in scientific understanding and treatment methods, rates of fracture non‐union remain largely unchanged over the years. This systematic review investigates the biological, molecular and genetic profiles of both (i) non‐union tissue and (ii) non–union‐related tissues, and the genetic predisposition to fracture non‐union. This is crucially important as it could facilitate earlier identification and targeted treatment of high‐risk patients, along with improving our understanding on pathophysiology of fracture non‐union. Since this is an update on our previous systematic review, we searched the literature indexed in PubMed Medline; Ovid Medline; Embase; Scopus; Google Scholar; and the Cochrane Library using Medical Subject Heading (MeSH) or Title/Abstract words (non‐union(s), non‐union(s), human, tissue, bone morphogenic protein(s) (BMPs) and MSCs) from August 2014 (date of our previous publication) to 2 October 2021 for non‐union tissue studies, whereas no date restrictions imposed on non–union‐related tissue studies. Inclusion criteria of this systematic review are human studies investigating the characteristics and properties of non‐union tissue and non–union‐related tissues, available in full‐text English language. Limitations of this systematic review are exclusion of animal studies, the heterogeneity in the definition of non‐union and timing of tissue harvest seen in the included studies, and the search term MSC which may result in the exclusion of studies using historical terms such as ‘osteoprogenitors’ and ‘skeletal stem cells’. A total of 24 studies (non‐union tissue: n = 10; non–union‐related tissues: n = 14) met the inclusion criteria. Soft tissue interposition, bony sclerosis of fracture ends and complete obliteration of medullary canal are commonest macroscopic appearances of non‐unions. Non‐union tissue colour and surrounding fluid are two important characteristics that could be used clinically to distinguish between septic and aseptic non‐unions. Atrophic non‐unions had a predominance of endochondral bone formation and lower cellular density, when compared against hypertrophic non‐unions. Vascular tissues were present in both atrophic and hypertrophic non‐unions, with no difference in vessel density between the two. Studies have found non‐union tissue to contain biologically active MSCs with potential for osteoblastic, chondrogenic and adipogenic differentiation. Proliferative capacity of non‐union tissue MSCs was comparable to that of bone marrow MSCs. Rates of cell senescence of non‐union tissue remain inconclusive and require further investigation. There was a lower BMP expression in non‐union site and absent in the extracellular matrix, with no difference observed between atrophic and hypertrophic non‐unions. The reduced BMP‐7 gene expression and elevated levels of its inhibitors (Chordin, Noggin and Gremlin) could potentially explain impaired bone healing observed in non‐union MSCs. Expression of Dkk‐1 in osteogenic medium was higher in non‐union MSCs. Numerous genetic polymorphisms associated with fracture non‐union have been identified, with some involving the BMP and MMP pathways. Further research is required on determining the sensitivity and specificity of molecular and genetic profiling of relevant tissues as a potential screening biomarker for fracture non‐unions.


| INTRODUC TI ON
Bone healing is a complex biological process aiming at restoring the affected area to its pre-injury levels. This is achieved through repair and regeneration of the cellular and extracellular components, regaining its former biochemical and biomechanical properties. 1,2 Successful bone healing requires the orchestrated interaction between the biological (cellular, signalling molecules and extracellular matrix) and mechanical environments. 3 Moreover, according to the 'Diamond Concept', other parameters that are considered essential for a successful healing include the local vascularity and the patient's biological fitness and comorbidities. 4 The definition of non-union has been inconsistent in the literature. The FDA (Food and Drug Administration), however, defines non-union as incomplete fracture healing within 9 months following injury, coupled by the lack of progression in radiological signs of healing over the course of three consecutive months. 5 Despite the advancement in both the understanding of fracture healing and some of the pathways that regulate it, the rates of fracture nonunion remain largely unchanged over the years. To date, fracture non-union remains common, occurring in 5%-10% of the 850,000 fractures seen yearly in the UK. 6 This poses a significant direct and indirect socioeconomic burden through prolonged medical treatments and productivity losses. 6 Further understanding of the biological processes and underlying mechanisms, along with their interactions, leading to fracture non-union need to be elucidated in order to reduce this risk.
We have previously published a systematic review outlining the biological and molecular profile of 'non-union tissue'. 1 Nevertheless, one critically relevant and important aspect not previously considered because of the scarce evidence at the time was the relevance of tissues harvested from sites away from the non-union site, such as peripheral blood and bone marrow products. Moreover, the accelerated improvement in laboratory techniques over the last decade also meant the biological and molecular understanding of the multiple pathways involved in bone healing is everchanging. Consequently, the herein study provides an up-to-date review on the knowledge that has been acquired in this important clinical condition. We aim to summarize the current evidence on (i) macroscopic and microscopic characteristics; (ii) cellular characteristics and function (cell surface protein expression, morphology, viability, proliferation, senescence, mineralization and alkaline phosphatase [ALP] activity); (iii) molecular characteristics (protein, mRNA, miRNA and gene expression) of non-union tissue and relevant tissues; (iv) differences between atrophic and hypertrophic non-unions; (v) effect of intervention(s) on non-union tissue and relevant tissues; and (vi) genetic predispositions to fracture non-union.

| MATERIAL S AND ME THODS
This systematic review was conducted according to the PRISMA guidelines. 7 Our protocol was similar to that of our previous publication, with the only difference being the addition of other types of tissues not harvested from the non-union site ('relevant tissue') in our inclusion criteria. 1 We define 'relevant tissue', as bone marrow or peripheral blood derived products, investigated to identify associations with progression to non-union. The reason for including studies assessing relevant tissue was due to the growing body of evidence demonstrating the correlation of these tissues with the occurrence of non-union, which we feel could be helpful to guide clinicians in their practice. tissue to contain biologically active MSCs with potential for osteoblastic, chondrogenic and adipogenic differentiation. Proliferative capacity of non-union tissue MSCs was comparable to that of bone marrow MSCs. Rates of cell senescence of non-union tissue remain inconclusive and require further investigation. There was a lower BMP expression in non-union site and absent in the extracellular matrix, with no difference observed between atrophic and hypertrophic non-unions. The reduced BMP-7 gene expression and elevated levels of its inhibitors (Chordin, Noggin and Gremlin) could potentially explain impaired bone healing observed in non-union MSCs. Expression of Dkk-1 in osteogenic medium was higher in non-union MSCs. Numerous genetic polymorphisms associated with fracture non-union have been identified, with some involving the BMP and MMP pathways. Further research is required on determining the sensitivity and specificity of molecular and genetic profiling of relevant tissues as a potential screening biomarker for fracture non-unions.

K E Y W O R D S
non-union(s), nonunion(s), fracture, human tissue, mesenchymal stem cell(s), mesenchymal stromal cell(s)

| Eligibility criteria
The inclusion criteria were as follows: (i) tissue obtained from the non-union site and processed for defining its characteristics and properties, OR studies assessing tissue relevant to non-union as defined above ('relevant tissue'); (ii) only tissue acquired from human subjects was included; (iii) articles were published in English language; (iv) the full text of each article was available; and (vi) for nonunion tissue, articles published between August 2014 (date of our previous publication) and 2 October 2021; for relevant tissue, no publication date restrictions were imposed. Studies that did not fulfil the eligibility criteria were excluded from further analysis.

| Search strategy and information sources
Adhering to our previously published protocol, the following da-

| Study selection
Two of the authors (MP and JV) performed the eligibility assessment independently, in an unblinded, standardized manner. Title and abstract sift were conducted first, followed by review of full text by MP and JV. Only studies fulfilling the eligibility criteria were included. Data of each eligible study were independently extracted by MP and JV, with results checked by the third author (IP). Any disagreement between reviewers was resolved by consensus, and if necessary, the senior researcher (PVG) was consulted.

| Extraction of data
Information on author, year of publication, patient demographics, non-union site, the duration and type of non-union, characteristics of non-union tissue (macroscopic/microscopic), cellular characteristics and functions (cell surface protein expression, morphology, viability, proliferation and cellular senescence), molecular characteristics (gene expression, protein expression) and effect of additional interventions were all carefully extracted.

| Data analysis
Outcomes of interest as mentioned in 'Extraction of data' section were inserted in an electronic database. Wherever possible, each characteristic of tissue samples was compared across different studies. We also evaluated the effect of any interventions documented in these studies. Qualitative results were summarized and presented in tables, whereas quantitative results are presented with p values if stated by the study. Statistical comparison was not made between studies, due to the heterogeneity in terms of study methodologies observed in each of these in vitro studies.

| Studies characteristics
The study characteristics of the non-union tissue and relevant tissue are outlined in Table 4.  Non-union was defined based upon radiographic and clinical examination, with minor variations between studies. Samples of non-union tissue and relevant tissue were mostly obtained during the surgical treatment of non-unions.

| Macroscopic characteristics of nonunion tissue
The macroscopic structure of non-union tissue was only assessed by Han et al.'s study, whereby tough scars surrounding the site of fracture non-union were identified. 14 The same team also described bony sclerosis of the fracture ends and complete obliteration of the medullary canal, with fibrous connections found between the fracture fragments. 14

| Immunohistochemistry
The immunohistochemical findings of non-union tissue and relevant tissue are summarized in Table 7. 8,[13][14][15][16]18,19 BMPs were present in non-union tissue. 8,14 Interestingly, Han et al. found BMP to be locally generated by non-union tissue. 14 Additionally, BMP antagonists were also found to be present in both normal and non-union tissue alike. 16 ALP and SMAD2/3 were both found to be increased in scaphoid non-union tissue. 13  In terms of relevant tissue, peripheral PIGF levels were found to be higher in non-union patients, with an initial surge followed by a rapid decline. Both TGF-ß2 20 and IL-17 19 on the contrary were reported to be lower in non-union patients.

| Analysis of vessel calibre, area and density
Blood vessels were present in cases of hypertrophic non-unions, with a varying density (Table 8). 8,13,16 Only one study assessed vessel density in atrophic non-unions, reporting a 2.4-fold increase when compared against that of induced periosteal membrane (control group). 8 However, both vessel calibre and median area were smaller in non-union tissue in this study. 8 All these reaffirms histological findings whereby vascular tissue was found to be present in both atrophic and hypertrophic non-unions. 11,12,14,16

| Morphology, viability, proliferation and cellular senescence
The (i) cell morphology, viability and proliferation of non-union tissue; and (ii) the effect of non-union serum on proliferation of BM-MSCs are outlined in Table 10. 8,[10][11][12]17,19 Overall, non-union MSCs were found to have comparable proliferative capacities and viability to that of BM-MSCs. 8,10,11,12,17 On the contrary, non-union serum was found to have a negative effect on MSC proliferation. 19 Comparing the cell senescence rates of non-union MSCs and those of bone marrow MSCs, Vallim et al. found no difference between the two groups. 11

| Mineralization and Alkaline phosphatase (ALP) activity assay
The outcomes of mineralization assay for non-union tissue are outlined in Table 11. [10][11][12][13]24,26 The findings of the four studies which evaluated ALP activity and its mRNA expression are outlined in

TA B L E 4 Study characteristics of non-union tissue and relevant tissue
Author  Han 14 Not mentioned Delayed union and non-union areas comprised a mix of different types of tissues: fracture fragments and surrounding tissues were mainly subject to fibrosis, in which the formation of new blood vessels could be seen, and a small amount of woven bone could be seen nearby. In these woven bones, Gergen Bauer's cells grew along the osteoid as cubes, suggesting active bone formations. A large number of cartilage cells existed in the intramedullary tissues, and there was no new bone and neovascularization. Bone marrow occlusion was observed, and in the fibrous tissue of adjacent bone and the gap of bone fractures, there were internal cartilage ossifications and fibrous ossifications. Scattered lamellar bone fragments were observed in some samples; these fractures were surrounded by osteoclasts, and there was a lack of osteoblasts.

Isolation of tissue
Wang 10 Not mentioned There were no significant differences in the morphology of atrophic / hypertrophic non-union tissues. They included MSCs, fibrocartilage cells and hyaline chondrocytes. Some sections showed very few bone islands. BMP-2-positive cells were present in both hypertrophic and atrophic non-union tissue.
Schwabe 16 Not mentioned The tissue was a very heterogeneous mixture of fragments of lamellar bone, immature and hypertrophic cartilage, unorganized fibrous tissue and newly formed woven bone. Independent of the group, bone apposition and resorption were seen in the tissue samples. Differences between the groups were not obvious.

Atrophic Hypertrophic
Type of tissue  18 Not mentioned PIGF was higher in non-union patients, reaching significance at Days 1 and 3 (p < 0.05); but less marked at Day 5 (p = 0.09). PIGF displayed initial massive surge followed by rapid decline in non-union patients. TGF-beta 2 appeared higher in union group (not statistically significant). Levels of MCP-1 and IL8 showed no clear difference between non-union and union groups.
El-Jawhari 19 Atrophic IFNγ, TNFα and IL-1 levels similar between non-union, union and control arms. However, lower levels of IL-17 detected at later stages of fracture healing (vs. union and control arms) Schira 13 Atrophic ALP reached higher levels in scaphoid non-unions as opposed to cancellous bone. Likewise, immunofluorescence for phosphorylated SMAD2/3 revealed increased activity in scaphoid non-unions.
Han 14 Not mentioned The depth of BMP-2 staining in the cytoplasm increased with increasing proximity to the new bone formation region, and there was some staining of the Golgi apparatus, showing that BMP-2 was locally generated. A wide variety of cells, including epithelial cells, smooth muscle cells around the small blood vessels, fusiform fibroblast-like cells and chondrocyte cells, showed positive staining in the fibrous tissues, indicating osteogenesis. There was no difference in the immunostaining of fibrous tissue between the samples with and without new bone. There was no positive BMP staining in the extracellular matrix or the fibrous tissue space. Sub-parts of view, fracture fragments were mainly fibrotic tissues and BMP-2 staining was negative. In the surrounding tissues, especially in the sticking scars and posted plate scars, neovascular and woven bone filled in a lot of the fibrous tissues, and in the vicinity, there were stained cells, indicating BMP-2 expression. There was a small amount of cartilage with positive staining in the cytoplasm, without expression in fibrous tissues of the closed medullary cavity. DCN expression was extensive in the interstitial fracture fragments. There was no positive staining of cartilage cells in the medullary cavity. DCN expression in the sticking scars was close to perivascular. The rate of expression of BMP-2 was highest in the posted bone scar group, and was low in the bone ends and canal content group (p < 0.05). There was no significant difference between the other two groups. The fracture fragment group had the highest DCN expression, with significant differences from the other two groups; the least significant difference analysis showed that between the fracture fragment group and the other two groups, p < 0.05; between the other two groups, p > 0.05 Wang 15 Atrophic/ hypertrophic The mean optical density of BMP-2 was 0.154 ± 0.041 in hypertrophic non-union tissue, 0.137 ± 0.037 in atrophic non-union tissue, there was no significant difference between the 2 groups (p > 0.05). The mean optical density of BMP-2 was 0.148 ± 0.040 in the 20-to 35-year-old group, 0.142 ± 0.040 in the 35-to 50-year-old group, 0.146 ± 0.056 in the more than 50-year-old group, there was no significant difference among the three groups (p > 0.05). The mean optical density of BMP-2 was 0.145 ± 0.037 in the 9-12 months group, 0.147 ± 0.0400 in the 13-24 months group, 0.145 ± 0.054 in the more than 24 months group, there was no significant difference among the 3 groups (p > 0.05).
Schwabe 16 Atrophic Bone morphogenic antagonists were demonstrated in non-union and control tissue.

Analysis of vessel density
Cuthbert 8 2.4-fold increase in non-union tissue when compared against induced membrane tissue. Both calibre and median internal vessel area of bloods vessels in NU tissue were smaller compared to induced membrane. Schira 13 Angiogenesis in scaphoid non-unions is similar to cancellous bone. Blood vessels and endothelial cells were detected by immunohistochemical staining of PECAM-1 in non-unions and controls revealing similar levels of angiogenesis in both tissues.

Schwabe 16
Histology: Vessels were present in all investigated samples without a difference between the tissue from non-union and control patients. Immunohistochemistry: well vascularized but also unvascularized areas with no difference between the non-union and the control tissue. Studies on relevant tissue have also investigated genetic predisposition to fracture non-union and identified numerous polymorphisms and genotypes associated with the increased risk of developing non-union (Table 13). [21][22][23]25,[27][28][29]

| DISCUSS ION
Fracture non-union represents a significant public health problem with detrimental socioeconomic costs. In addition to productivity losses, the direct treatment cost of established non-union in the UK has been estimated to be in the regions of £7,000 and £79,000 per person, dependent on its complexity. 40 With multiple pathophysiological factors influencing its progression, fracture nonunion remains a challenging condition to treat. 41 The improved understanding of its pathophysiology has seen the evolution with the treatment of non-unions, from prolonged immobilization in the 1950s 42 to the modern techniques of biological stimulation and polytherapy. 43 The commonest macroscopic appearance of non-unions is soft tissue interposition between fracture fragments. 14 In terms of histological analysis, several similarities exist between atrophic and hypertrophic non-unions. Firstly, fibrous, cartilaginous and connective tissues were historically reported to be the tissue types common to both atrophic and hypertrophic non-unions. 32,33,34,36,45,46 Studies included in this systematic review 11,13,16 confirm these findings. Secondly, bony islands were not always present in both atrophic 15,32,33,34 and hypertrophic non-unions. 15,32,34,36,45,46 Thirdly, whilst fibroblast-like cells account for the majority of the population in both atrophic and hypertrophic non-unions, 11,13,33,36 MSCs were still present in both tissues. 15 However, several differences also exist. Atrophic non-unions contain a mixture of lamellar and woven bone, 16 with TA B L E 1 2 ALP activity and ALP related mRNA expression

Type of analysis Atrophic Hypertrophic
Histology Table 6 Immunohistochemistry SMAD2/3 revealed increased activity in non-unions 13 Close vicinity to immature osteoid trabeculae 35  Higher than haematoma cells 36 Very low mineralization potential and significantly lower than 'normal' human osteoblasts 37 Under osteogenic conditions, mineralization was significantly higher than that of fracture haematoma cells, in contrast to undifferentiated conditions 36

SDF-1, VEGF, BMP-2 present in non-unions
Note: As only reporting on studies published after our original review 1 would provide an incomplete picture of the differences between atrophic and hypertrophic non-unions, we include all relevant data regardless of publication date. differentiation and has therefore been extensively studied given its important role in the field of bone regeneration. 52,53 Interestingly, studies have reported evidence of BMP signalling and generation in non-union MSCs, 8,14,49 with no difference in BMP expression between atrophic and hypertrophic non-unions. 15  Chordin and Gremlin knockdown. 10 Furthermore, they also demonstrated Chordin knockdown to rescue the osteogenic ability of nonunion cells. 10 Taken altogether, these findings support the idea of imbalance expression between BMP and its inhibitors driving the pathophysiology of impaired bone healing observed in non-union MSCs. 16,39,56 Matrix metalloproteinases (MMP) are important key player, which modulate bone remodelling and repair. Disruption to either MMP or their inhibitors could result in disorders of fracture healing. 38 In vitro studies on hypertrophic non-union tissues have found MMP to bind directly and degrade BMP-2, known to be an osteoinductive molecule. 38 Furthermore, non-union tissues were found to have an upregulation of MMP-7, MMP-9 and MMP-17 genes. 13,38 All these findings highlight the potential role of MMP as one of the key players in the pathogenesis of fracture non-union.
Although Dkk-1 is well known as an antagonist of the Wnt signalling pathway inhibiting osteogenic differentiation, 33,57 Dkk-1 expression by non-union tissue has only been investigated by two studies, reporting similar expression when compared against BM-MSC 33 and healthy cancellous bone. 13 However, release of Dkk-1 by atrophic non-union MSCS cultured in osteogenic conditions was higher than that of BM-MSCs. 33 Whilst this study suggests the potential role of Dkk-1 in the pathophysiology of non-union, further research is still warranted to better understand the mechanism of action which Dkk-1 plays in causing non-union.
There has been emerging evidence over the recent years on the genetic predisposition of fracture nonunion. 19 Secondly, heterogeneity with the definition of non-union, timing of tissue harvest and laboratory assays may all account for the different results reported in studies. Lastly, the abbreviation/term MSC is only more recently used in this field, which could be referred to as mesenchymal stem cells or mesenchymal stromal cells. 58 As such, historical studies using alternative terms such as 'osteoprogenitors' and 'skeletal stem cells' were excluded as authors felt it does not guarantee the accuracy of comparison made. There are several strengths of this systematic review. This includes the systematic approach on both screening and analysis of the findings from current literature. Furthermore, this systematic review provides an up-to-date understanding on the biological profile of non-union tissue and relevant tissue at a cellular and molecular level. Due to the huge heterogeneity in available evidence,

TA B L E 1 5 Effect of interventions
we are unable to recommend any direct clinical application. The complex pathophysiology of non-union requires the treating clinician to consider the interaction between biological, physiological and molecular components of the 'diamond concept' of bone healing. 59 Cellular therapies with osteogenic cells and osteoinductive molecules, osteoconductive scaffolds and tissue engineering are treatment strategies which holds great promise. 60,61 Although still in its early stages, further work on the molecular and genetic profiling of relevant tissue such as patient's serum could serve as an advantageous screening and predictive tool of fracture non-union.

CO N FLI C T S O F I NTE R E S T
All authors declare no conflict of interest.