Background

Articular cartilage is a highly specialised tissue acting as a shock absorber, enabling synovial joints to articulate with low frictional forces. Due to its avascular, aneural and alymphatic state, it has a limited repair potential [1]. Surgical options to manage damaged articular cartilage include arthroscopic debridement [2–5], bone marrow stimulation techniques [6–8], chondrocyte implantation [9–13], osteochondral autografts (mosaicplasty) [2, 14, 15], osteochondral allograft [16–18] and, in the presence of osteoarthritis, joint replacement [19].

Bone marrow stimulation techniques, such as microfracture, are the most frequently used method in clinical practice for treating small symptomatic lesions of the articular cartilage [6–8]. However, the resulting tissue has shown to be a mixed fibrocartilage tissue [20–22] with varying amounts of type II collagen [8, 21, 23, 24] and inferior to native hyaline cartilage. Fibrocartilage is vulnerable to shear stresses and prone to breaking down over time [20]. Subchondral osseous overgrowth has also been reported after microfracture [25, 26]. Osteochondral grafts can lead to donor site morbidity and healing seams at the recipient site [27, 28]. Autologous chondrocyte implantation (ACI) [9, 10] and its later evolution, matrix-induced autologous chondrocyte implantation (MACI), offered great promise with 80% of patients showing good or excellent results at 10 years [29] but at best results in hyaline-like repair and has experienced complications such as graft failure, periosteal hypertrophy and delamination [30, 31]. In addition, it has also been reported that cells may lose their phenotype during expansion [32, 33].

There is therefore a growing interest in regenerative medicine, which can broadly be thought of as two main types: cell therapy, where cells are injected directly into the blood or into tissues, and tissue engineering, where cell-scaffold combinations are used to repair or regenerate tissues.

Stem cells are cells that have the ability to divide and develop into many different cell types in the body and can be categorised as pluripotent and multipotent. Pluripotent stem cells are often harvested from embryonic sources and can develop into any type of cell in the body whereas multipotent stem cells are generally taken from adults and can divide and develop into a more limited range of cell types. When stem cells divide, the new cells can either remain stem cells or develop into a new type of cell with a more specific function (Table 1).

Table 1 Table describing the three main properties of stem cells
Full size table

Mesenchymal stem cells (MSCs) are a form of multipotent cells that may offer an alternative to cartilage repair techniques not hampered by availability and donor site morbidity.

The introduction of stem cell therapies into clinical practice however is a form of translational research, which as per any “bench-to-bedside” pathway now has enormous governance issues [34, 35] and is highly regulatory across four phases (Table 2) and by the Tissues and Cells Directive (2004/23/EC) https://www.hta.gov.uk/policies/eu-tissue-and-cells-directives.

Table 2 Description of the different phases of clinical trials
Full size table

Many published reviews on cartilage repair only list human clinical trials [13, 36–46], underestimating the wealth of basic sciences and animal studies that are precursors to future research and may be relevant in clinical practice further down the line. In addition, true translation would imply that all of the clinical studies would have supporting pre-clinical data.

We therefore set out to perform a systematic review of the literature to assess the translation of stem cell therapy to explore what research had been carried out at each of the stages of translation from bench-top (in vitro), animal (pre-clinical), and human studies (clinical) and assemble an evidence-based cascade for the responsible introduction of stem cell therapy for cartilage defects. In particular, we wanted to focus on the key burning questions pertaining to cartilage repair such as cell source, dosage (how many cells should be used), requirement for scaffolds and the role for extrinsic growth factors.

Main text

Search methodology

This review was conducted in accordance to PRISMA guidelines [47] using CINHAL, MEDLINE, EMBASE, Scopus and Web of Knowledge databases from 1st January 1900 to 30th June 2015.

The keywords used in the selection were “(“mesenchymal stem cells”[All Fields] OR “mesenchymal stem cells”[MeSH Terms] OR “mesenchymal”[All Fields] OR “stem cells”[All Fields] OR “Stem Cells”[MeSH Terms] OR “MSC”[All Fields]) AND (“Articular Cartilage”[MeSH Terms] OR “articular”[All Fields] OR “cartilage”[All Fields] OR “cartilage”[MeSH Terms]) AND (“healing”[All Terms] OR “repair”[All Terms] OR “Regeneration”[MeSH Terms] OR “regeneration”[All Fields] OR “tissue engineering”[MeSH Terms] OR “tissue engineering”[All Fields]) AND (“defect”[All Terms]) AND (“chond*”[All Terms])”.

All review and non-English studies were excluded. For analysis, only original research studies were included. Any duplicates were excluded. Initially, KM and JS independently screened studies’ title and abstract. Those included had the full text reviewed. Any disparities were discussed with the senior author (AJG). The references of eligible studies were also searched and included where relevant.

Unpublished trial databases (e.g. ClinicalTrials.gov) were reviewed as the grey literature using popular search engines, including Google. The keywords used for registered clinical trials in clinical trial databases were “stem cells”, “cartilage” and “orthopaedics”.

Eligible studies were drafted into tables tabulating the key data.

Results

The initial search identified 2880 study articles, of which 239 were included for analysis. The PRISMA flow diagram is shown in Fig. 1.

Fig. 1
figure 1

Flow chart of literature search used for the review

Full size image

In vitro studies

MSC source

A list of cell sources used in the in vitro studies is shown in Table 3. The commonest being human MSCs (66%) followed by rabbit MSCs (15%). The majority of the studies used bone marrow-derived MSCs (63%) followed by adipose tissue (33%). Two studies used commercial cell lines [48, 49].

Table 3 Cell species and cell sources
Full size table

Scaffold

Within the in vitro studies, 26 different types of natural scaffold and 9 types of synthetic scaffolds were identified with a further 18 different types of hybrids, the most popular being a fibrin-polyurethane scaffold (Table 4).

Table 4 Types of scaffolds
Full size table

Growth factors

The commonest used growth factors were TGF-β and the bone morphogenetic protein (BMP) family. A list of growth factors used can be seen in Table 5.

Table 5 Number of in vitro studies using different growth factors
Full size table

Cell seeding and passage

There was wide heterogeneity in cell seeding density and there appeared to be no standard form of measurement. Li et al. [50] examined three different seeding densities: 2, 5 and 10 × 106 cells/scaffold, and found that scaffolds seeded with 5 × 106 cells per scaffold induced the highest chondrogenesis; however, other groups [51–53] found that a higher seeding density results in better chondrogenesis.

Apart from 26 studies which did not state cell passage number, most studies used MSC of an early passage, anything between uncultured fresh (passage zero (P0) and five times passaged cells (P5). One study used cells of P6 [54], and another study used cells between P4 and P7 [48]. No relationship was apparent between chondrogenesis and number of passages.

Length of study

The length of each in vitro study can be seen in Table 6. The majority of studies were short-term models; 27 studies (25%) ended between 1 and 2 weeks, 35 studies (33%) ended between 2 and 3 weeks and 15 studies (14%) ended between 3 and 4 weeks.

Table 6 Length of studies
Full size table

Method of assessment

A range of techniques was used to assess chondrogenesis within the in vitro studies. These techniques consisted of histology, immunohistochemistry, qPCR, biochemical analysis, imagery and mechanical testing. The techniques used are summarised in Table 7.

Table 7 Types of techniques used to assess chondrogenesis of MSCs
Full size table

Animal studies (pre-clinical)

One hundred eleven animal studies were included of which 109 were controlled laboratory studies, one was a pilot study [49] and one was a longitudinal case study on a race horse [55]. The commonest animal studied with 59 studies was rabbit (53%). The different species of animals studied is shown in Table 8.

Table 8 Different species of animals used to assess reparative effect of MSCs on cartilage defect
Full size table

Defect

The size of the defect varied from 2 to 25 mm2 in the smaller animals and from 1 to 64 mm2 in the larger animals. All but two studies [56, 57] used the knee for defect creation.

Stem cell type

Bone marrow-derived stem cells were used in 84 studies (75%). Thirteen studies (11%) used adipose stem cells [54, 58–69], six (5%) used synovia [70–75] and three (2%) used periostium-derived MSCs [76–78]. Three studies (3%) used embryonic stem cell-derived MSCs [79–81] whereas 2 studies (2%) used muscle-derived MSCs [82, 83]. One group showed promising results of allogenic MSCs in a rabbit model when compared to autologous cells, although numbers were small [84, 85]. Another used compared autologous chondroprogenitor cells and allogenic chondroprogenitor cells against controls in an equine model and reported that repair tissue quality in the allogenic cell group was not superior to that in the control (fibrin only) group and also showed poorer radiographic changes in the allogenic group [23].

Cell culture, dose and delivery

There was much variation in the number of cells implanted and the number of cell passages from 3–10 or more [79, 86].

The number of cells varied from 4 × 103 – 1 × 1010. The majority of studies used between 106 and 108 cells. Some did not specify the number of cells implanted. Two studies suggested that improved chondrogenesis occurs with a higher implanted cell number [75, 87], although others suggested that the high cell numbers increase the risk of synovitis [75] and synovial proliferation [88].

The cells were transplanted into the defect both as cell therapy (injection directly into the joint) (17 studies, 15%) or by tissue engineering (cell-scaffold combinations) (94 studies, 85%). Fifteen studies [49, 65, 72, 75, 81, 86, 89–97] used a mixture of solutions prepared from hyaluronic acid [65, 92, 94–97], phosphate buffer solution [91], plasma [75], basal medium with chondrogenesis [89], collagen acid [93], sodium alginate [86] or a growth factor medium [90]. Two studies used MSCs only [49, 72].

Scaffold

Ninety-two studies (82%) used a scaffold. The material used was a synthetic polymer either collagen based, fibrinogen glue or a synthetic protein (e.g. rHuBMP-2) in 62 (56%) studies (Table 9).

Table 9 Table showing the types of scaffold used in animal studies
Full size table

Growth factors

Thirty-two studies (29%) assessed the effect of growth factors on MSC chondrogenesis. Seventeen out of 38 (44%) used TGF-β1/3 (Table 10), the majority of which show a positive effect on chondrogenesis.

Table 10 Table showing growth factors used in animal studies
Full size table

Associated procedures

Ten of the studies compared MSC treatment against other surgical modalities such as debridement [55], microfracture [49, 91, 96, 98, 99] and mosaicplasty [77, 100–102].

Outcome measures

There were a variety of outcome measures used to analyse the results of the studies. The majority of studies (79%) used evidence of hyaline-like cartilage as being a positive outcome (Tables 11 and 12).

Table 11 Outcome measures used in animal studies (some studies used more than one outcome measure)
Full size table
Table 12 Analysis technique used on repaired tissue
Full size table

Human studies (clinical)

Thirty-one published studies by 15 different groups looked at clinical applications of MSCs. One used allogenic stem cells [103] and the rest autologous stem cells. The types of studies can be seen in Tables 13 and 14.

Table 13 Number of publications for each study type and phase
Full size table
Table 14 Summary of the published clinical studies
Full size table

There were 52 unpublished clinical trials, majority of which are early phase studies (I–II; 63%) and only 5 trials were phase II/III. Table 15 shows a summary of these clinical trials.

Table 15 Clinical trials (unpublished/on-going) registered in ClinicalTrials.gov
Full size table

Defects

The majority of studies (42%) used MSCs to treat knee osteoarthritis [103–115]. The rest of the studies looked at knee cartilage defects except for two which studied the ankle talar dome [116, 117]. One study used MSCs to treat knee osteoarthritis (OA), knee OA and ankle OA [112].

Of the knee cartilage defects, the patients were heterogeneous with varying defect sizes and locations, including the patellae [118–121], patella-femoral joints [122, 123], femoral condyle [113, 119–121, 123–132], trochlear [119–121] and tibial plateau [121]; and several had multiple defect sites [105, 120, 123, 128].

Previous treatment and associated procedures

The majority of patients who received MSC treatment had undergone previous arthroscopy [103, 104, 118, 119, 122, 124, 130], failed debridement [113, 118, 119, 121–123, 125, 127, 131] or bone marrow stimulation [114, 116, 117, 126].

Cell harvest source

Twenty-one studies (68%) used bone marrow-derived MSCs from the anterior or posterior superior iliac spine [103–105, 109, 111–113, 115–118, 120, 122–128, 130–132]. Five studies (18%) used adipose-derived MSCs [106–108, 110, 114], two studies (7%) used synovium-derived MSCs [129, 133] and two studies (7%) used peripheral blood progenitor cells collected by apheresis [119, 121].

Cell stage

Twenty studies (61%) culture-expanded their cells [103–105, 107–113, 115, 118, 120, 122–126, 129, 133], whereas 11 studies (39%) used fresh concentrated stem cells from bone marrow [116, 117, 127, 128, 130–132], fat tissues [106, 114] or peripheral blood [119, 121] in a one stage-procedure. In studies using bone marrow concentrate, approximately 60 ml of bone marrow aspirate was harvested and concentrated down to a volume of 2–4 ml before use [116, 117, 127, 130–132]. In studies using culture-expanded cells, the majority used cells from early passages, P1–P3 [103, 105, 109, 110, 112, 113, 115, 118, 120, 122–125, 129]. One study reported the use of cells at a late passage (P5) [104] ,and five studies did not specify a passage number [107, 108, 111, 126, 133].

Thirteen studies (42%) confirmed the phenotype of cells before clinical application [105, 108–110, 112, 115, 119, 120, 122–125, 129]. Commonly used surface markers to select MSCs were CD29, CD44, CD73, CD90 and CD105. Also CD14, CD34 and HLA-DR were used to eliminate non-MSCs.

Cell dose and delivery

The number of cells applied (dose) varied from 2–57 million for bone marrow-derived MSCs [103–105, 109, 111–113, 118, 120, 122–125, 129] and from 1.2–100 million for adipose-derived MSCs [107, 108, 110, 114]. For synovial MSCs, 8–77 million cells were used [129, 133], and for peripheral blood progenitor cells, 20 million cells were used [119]. Also, the methods for implantation varied from arthroscopic implantation (35%) [107, 108, 116, 117, 127, 128, 130–133], intra-articular injection [103–106, 109–112, 114, 115, 119, 121, 123] or open surgery (29%) [113, 118, 120, 122–126, 129].

In the cell therapy studies, the cells were suspended with a variety of different co-stimulators, including hydroxyapatite (HA) [106, 119, 121, 123], platelet rich plasma (PRP) [106, 114] and platelet lysate [104]. Some studies also administered multiple injections of stem cells [119, 121] and/or further injection of HA [115, 119, 121, 123], PRP [106, 114] or nucleated cells [104] following a stem cell injection.

The most frequently used scaffolds were type I collagen of porcine or bovine origin [113, 118, 122, 124, 126, 129], followed by ascorbic acid sheet [120, 123] and platelet-rich fibrin glue mixture [108, 125].

Rehabilitation

Early continuous passive motion was employed in 14 studies [113, 117–122, 124–127, 129–131]. Six studies did not report details on post-operation rehabilitation [104–106, 109, 116, 132]. Three studies aimed for full weight bearing very early by week 4 [107, 108, 122] whereas 11 studies (40%) aimed for full weight bearing by the 6th–8th week [113, 117–121, 124, 125, 127, 131, 133]. No study addressed the effect of rehabilitation on the quality of the repair.

Outcomes

Most commonly used outcome measures for treatment efficacy were radiological (77%) [103–106, 109–112, 115–117, 119, 121, 123–125, 127–134] and arthroscopic assessment (61%) [107, 108, 113, 116–122, 124–126, 130–133]. Most commonly used patient-reported outcomes are International Knee Documentation Committee (IKDC) score (36%), followed by a visual analogue scale (VAS) pain (39%) and Tegner activity scale (29%).

Adverse effects

None of the studies reported any severe adverse effects related to the MSC treatment. Two group reported minor adverse events including mild pain and effusion after the injections, which persisted for no more than 7 days [103, 114].

Conclusions

There is a growing fascination with the role of mesenchymal stem cells in cartilage repair.

As early as the 1950s, Pridie showed fibrocartilaginous repair through subchondral drilling [135–137]. Initially, Pridie drilling was reported as a treatment for osteoarthritis [135, 138] and was often associated with many additional procedures such as synovectomy and trimming of osteophytes.

Since Pridie’s initial experiments, the process of marrow stimulation techniques or exposure of mesenchymal stem cells from cancellous bone has changed its guise on several occasions.

Ficat in 1979 described “Spongialization” in which the cancellous bed was exposed in 85 patients with chondral lesions of the patella with encouraging results [139]. Johnson et al. [140] described abrasion arthroplasty and encouraged its use especially in younger patients [141, 142]. Other authors had less positive outcomes [143–146]. Dandy wrote an entertaining article on abrasion arthroplasty where he highlighted that at least in the treatment of osteoarthritis, its effects could relate to the arthroscopic washout, rest or even the placebo effects of the charismatic surgeon [147]. The final evolution of marrow stimulation was the term “Microfracture” enabled by commercially manufactured bone picks used to breach the subchondral bone [8]. Marrow-stimulating technique procedures, in particular microfracture, are now considered the first-line treatment for full-thickness cartilage lesions and have demonstrated good to excellent results in 60–80% of patients [148, 149].

Cartilage repair has evolved from marrow stimulation techniques through to chondrocyte transplant and now stem cells at rapid pace. An ideal translational pipeline would demonstrate how in vitro data was used to inform a pre-clinical model, which would later form a phase I/IIa first-in-man study and subsequently a phase III clinical trial. This would of course be the safe and responsible method by which novel therapies are brought to the market.

This systematic review is the first of its kind to explore the full spectrum of evidence from in vitro studies, through animal studies to human clinical trials, and yet, we found little evidence of connectivity between in vitro, animal and then human work. In fact, we did not find a single group that had carried out and reported studies in all three categories.

Indeed, even from groups, which showed a seemingly hierarchical approach to translation, discrepancies became apparent. For example, Saw et al. from Korea used a pre-clinical goat model to repair cartilage defects using HA plus bone marrow-derived cells [150] and then moved into a first-in-man study, but in doing so, elected to change from bone marrow aspirate to peripheral blood and justified this change because it was easier to harvest peripheral blood than marrow [151].

There are several sources of cells that have been used in cartilage repair including bone marrow, peripheral blood, synovium, adipose tissue and umbilicus (Table 14) without any clear evidence of superiority of one over the other.

One stage vs. two stages

As two stage procedures involving cell culture are expensive and cumbersome, there is an increasing push towards a single stage stem cell treatment. In this situation there is some supportive pre-clinical data [91, 95, 98, 152–154], but there does not appear to be a pre-clinical study that directly compares bone marrow concentrates against cultured MSCs.

Several groups have reported the use of bone marrow concentrates in clinical practice [116, 117, 127, 128, 130–132], in which the buffy coat is used containing the nucleated cells, of which a few will be stem cells.

Briefly, the patient has approximately 60 mL of bone marrow harvested from the iliac crest which is then spun down in a cell centrifuge (SmartPrep, Harvest Technologies Corp., USA, or IOR-G1, Novagenit, Mezzolombardo, TN, Italy) to provide 6 mL of concentrate containing nucleated cells. A small amount of the nucleated cells are then placed onto a hyaluronic acid membrane (Hyalofast, Fidia Advanced Biopolymers, Italy) or collagen membrane (IOR-G1, Novagenit, Mezzolombardo, TN, Italy) as a scaffold, which is then arthroscopically placed into the cartilage defect which had been pre-prepared using a burr or drill. The construct is then held with a platelet gel obtained from a harvest of 120 mL of patient’s venous blood taken the day before surgery (Vivostat system, (Vivolution, Denmark)) [118]. The results of the first 30 patients have been reported as showing improvements in MRI and arthroscopic appearance as well as clinical scores at 3 years follow-up [118].

This new technique is of course an evolution of the autologous matrix-enhanced chondrogenesis (AMIC) which used the stem cells from the adjacent marrow (and not pre-harvested bone marrow concentrates) within either collagen patches [155–157] or polyglycolic acid–hyaluronan-based scaffolds [158, 159].

There has also been a further step taken to avoid bone marrow harvest in which peripheral blood has been used in knee chondral lesions. In an RCT, arthroscopic subchondral drilling was followed by postoperative intra-articular injections of hyaluronic acid (HA) with and without peripheral blood stem cells (PBSC). Fifty patients were studied and randomised 1 week after surgery to receive either 8 injections of HA or 8 injections of HA plus PBSC. Those that underwent PBSC received stimulation with filgrastim, which contains recombinant human granulocyte colony-stimulating factor prior to harvest [106, 151]. At 18 month follow-up, they reported no adverse effects and improved MRI findings in the PBSC group compared to HA alone, took biopsies of 16 of the 25 patients in each group and claimed better tissue morphology in the PBSC group, as graded by the International Cartilage Repair Society Visual Assessment Scale II. Interestingly, however, the same group’s pre-clinical used bone marrow aspirates and not peripheral blood [150].

Autologous vs. allogenic

There is an increasing interest in allogenic cells to avoid donor site morbidity and to reduce cost. The pre-clinical data with regards to allogenic cells is conflicting. One group showed promising results of allogenic MSCs in a rabbit model when compared to autologous cells, although numbers were small [160, 161]. Another group compared autologous chondroprogenitor cells and allogenic chondroprogenitor cells against controls in an equine model and reported inferior repair in the allogenic cell group [23]. Despite conflicting pre-clinical data, human studies using allogenic cells began in Korea in 2009. A phase I/IIa study to assess safety and efficacy of a combination of human umbilical cord blood-derived mesenchymal stem cells and sodium hyaluronate (CARTISTEM® (MEDIPOST Co., Ltd., Korea)) was performed in knee chondral defects (NCT01041001). A parallel phase 3, open-label, multi-centre RCT comparing CARTISTEM® and microfracture in knee chondral defects was carried out in Korea and the USA (NCT01733186). Results are still pending.

Another area of huge controversy is the actual dose of cells that should be used. In vitro between 50,000 cells/mL and 100 billion cells/ml have been studied. In pre-clinical animal studies, this ranged from 1000 to 1 billion cells/mL, and in human studies, the reported range has been 1.2 million cells/mL–24 million cells/mL.

It remains unclear what the most appropriate cell dose should be, with some groups reporting that a higher cell number leads to a better repair [52, 71, 87, 95, 162–164], but Zhao et al. [99] highlighted the limitation to cell saturation and survival, and thus, there may be a top limit to cell number that can be used to aid repair.

A multitude of methods for cell delivery have also been adopted, from direct joint injection or embedded in a plethora of scaffolds, such as type I collagen gels of porcine or bovine origin, ascorbic acid sheets or fibrin glues (Table 14).

In vitro and in pre-clinical studies, a plethora of growth factors have been studied including TGF-β1 and TGF-β2 and BMP-7 but none of these have been included in human clinical trials (Table 5).

It is clear that the relationship between cell passage, cell dose, the use of scaffolds and growth factors and the efficacy of MSC treatment is still to be established.

Future

There is no question that the field of cartilage repair accelerates at rapid pace, and it is clear that the single stage procedures are likely to win over two stage procedures to save costs and reduce the burden on both provider and the patient. The reduction of donor site morbidity is a further driver helping direct progress.

The concept of cell banks of allogenic cells clearly meets all of the above criteria, but the lack of good supporting pre-clinical and long-term safety and efficacy data does little to pacify potential pitfalls of this direction. The fact that the phase 3 RCT of allogenic umbilical stem cells was allowed to be registered (NCT01041001) before the same group registered their phase I/IIa safety study (NCT01733186) intimates that sometimes clinical pace exceeds that of the regulators to lay down new ground.

Tools are likely to be introduced to the operating theatre that might improve the efficacy of treatment, such as fluorescence-activated cell sorting (FACS) machines which can isolate MSCs from the buffy coat of bone marrow aspirate by their cell surface markers. At present, this technology is expensive and complicated and ways to reduce cost and make the process simple are required before they could enter the operating theatre.

Induced pluripotent stem cells (iPSCs) are adult somatic cells that have been genetically reprogrammed to an embryonic stem cell-like state by being forced to express genes and factors important for maintaining the defining properties of embryonic stem cells [165].

These cells show unlimited self-renewal, and some in vitro studies have shown chondrogenic differentiation by iPSCs from human chondrocytes biopsied from osteoarthritic knees [166] and cartilage formation from human neural stem cells [167]. However, this work is at a very early stage, and aside from the ethical considerations, much research into control of cell phenotype and cell fate to alleviate concerns for cancer risk are required before this technology is ready to move into the pre-clinical and clinical realms.

In conclusion, this review is a comprehensive assessment of the evidence base to date behind the translation of basic science to the clinical practice of cartilage repair. We have revealed a lack of connectivity between the in vitro, pre-clinical and human data and a patchwork quilt of synergistic evidence. It appears that the drivers for progress in this space are largely driven by patient demand, surgeon inquisition, and a regulatory framework that is learning at the same pace as new developments take place. We strongly recommend funding body commission studies that have a clear translational purpose in order to drive the science towards patient benefit.