Which substances loaded onto collagen scaffolds influence oral tissue regeneration?—an overview of the last 15 years

Collagen scaffolds are widely used for guided bone or tissue regeneration. Aiming to enhance their regenerative properties, studies have loaded various substances onto these scaffolds. This review aims to provide an overview of existing literature which conducted in vitro, in vivo, and clinical testing of drug-loaded collagen scaffolds and analyze their outcome of promoting oral regeneration. PubMed, Scopus, and Ovid Medline® were systematically searched for publications from 2005 to 2019. Journal articles assessing the effect of substances on oral hard or soft tissue regeneration, while using collagen carriers, were screened and qualitatively analyzed. Studies were grouped according to their used substance type—biological medical products, pharmaceuticals, and tissue-, cell-, and matrix-derived products. A total of 77 publications, applying 36 different substances, were included. Collagen scaffolds were demonstrating favorable adsorption behavior and release kinetics which could even be modified. BMP-2 was investigated most frequently, showing positive effects on oral tissue regeneration. BMP-9 showed comparable results at lower concentrations. Also, FGF2 enhanced bone and periodontal healing. Antibiotics improved the scaffold’s anti-microbial activity and reduced the penetrability for bacteria. Growth factors showed promising results for oral tissue regeneration, while other substances were investigated less frequently. Found effects of investigated substances as well as adsorption and release properties of collagen scaffolds should be considered for further investigation. Clinical relevance: Collagen scaffolds are reliable carriers for any of the applied substances. BMP-2, BMP-9, and FGF2 showed enhanced bone and periodontal healing. Antibiotics improved anti-microbial properties of the scaffolds.


Introduction
Collagen is an important biomaterial which is frequently applied in oral and maxillofacial tissue engineering and regenerative medicine approaches that aim to restore, maintain, or enhance damaged or substantially reduced tissues [1,2]. Different fields in dentistry face specific objectives referring to hard and soft tissue regeneration, and collagen has been successfully implemented into various treatment strategies [2,3]. As the main component of extracellular matrix (ECM), collagen is essential for the structural maintenance of oral tissues and plays an important role in numerous cellular processes [4,5]. In tissue engineering application, collagen exhibits excellent biocompatibility and low antigenicity [6]. As collagen is biodegradable and shows poor mechanical properties, non-enzymatic or enzymatic cross-linking procedures using glutaraldehyde or formaldehyde for example can be performed to modify collagen membranes. Consequently, these procedures prevent rapid degradation and enhance mechanical properties [7,8]. Collagen membranes are used in guided tissue regeneration (GTR) and guided bone regeneration (GBR), which are methods used in oral surgery as augmentative techniques. Both GBR and GTR are based on the principle to cover periodontal epithelium and bone as well as to maintain the tissue defect space in order to exclude epithelium and gingival connective tissue due to their faster growing properties compared with hard tissue. While GTR focuses on regenerating soft tissue of the periodontal attachment, GBR, in addition, is aiming to support new bone tissue growth [9]. Favorable properties of a membrane include biocompatibility, impermeability for epithelial cells or bacteria, and adequate mechanical properties to allow proper handling during surgery and stiffness for space maintenance and less pressure to the bone graft during healing. Also, the time of resorption should allow the bone graft to regenerate, before degradation of the membrane causes the loss of the stated properties. In the beginning of GTR development, a bacterial filter consisting of cellulose acetate was placed over the treated area to serve as an occlusive membrane [10]. Aiming to improve clinical application, further studies investigated the use of nonresorbable expanded polytetrafluorethylen (e-PTFE) in membranes, which could also have been reinforced by titanium. Although mechanical stability can be enhanced using titanium, a second surgical entry is required to remove the membrane [11]. The next generation of barrier membranes were focused on resorbable materials, including collagen, to avoid a surgical removal [12]. Further advances aimed to improve the regeneration process and wound healing, and to serve as local drug carriers. To release specific substances at the wound site, collagen membranes can be loaded with various agents, such as antibiotics, growth factors, or adhesion molecules [13]. Progress in the field of tissue engineering has resulted in the development of innovative scaffold designs that act as a biological platform to enable reparation and restoration of injured tissues during the healing process. Collagen can easily be used to build 3D scaffolds and sponges, and can also serve as a carrier for the delivery of substances, such as growth factors or bone morphogenetic proteins, aiming to enhance healing and support regeneration [14].
In the past years, the number of publications investigating the role of drug-loaded collagen scaffolds in tissue engineering has increased, indicating a growing interest in this topic. The primary aim of this review is to give an overview of different cell-free regenerative approaches which utilize collagen as a biomaterial carrier by summarizing existing studies from the past 15 years. At the time of creating this publication, there was no existing literature reviewing this topic. The secondary aim was to provide information about the outcome of those studies according to their influence on oral tissue regeneration. This information could allow suggestions of future applications in dentistry and oral surgery or the need for further research.

Study design
A study protocol for a qualitative literature review was developed by all authors based on recommended methods [15]. The focused PICO (problem, intervention, comparison, outcome) question was: Does loading of substances onto collagen scaffolds influence oral tissue regeneration parameters in vitro and/or in vivo, compared with unloaded control? The whole review process was designed according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) Statement [16].

Risk of bias and quality assessment
The risk of bias and the quality of the included studies was assessed independently by two authors according to the revised recommendation of the CONSORT statement. After scores were calculated, an overall estimation risk of bias (low, all criteria met; moderate, one or more criteria partly met; or high, one or more criteria were not met) was performed for each selected study. Any disagreement between the reviewers during study selection, data extraction, and quality assessment was resolved by discussion until consensus among all authors was reached. For in vitro and in vivo studies, no further risk of bias was assessed as no validated bias risk assessment tool was available.

Search results and study characteristics
The initial search procedure using PubMed (n = 248), Scopus (n = 56), and Ovid Medline® (n = 20) led to 324 search results overall. After discarding of duplicates, the total findings were reduced to 259 unique journal articles. Following the inclusion and exclusion criteria, 182 studies had to be excluded leading to a total of 77, which were further investigated and later synthesized in this review (Fig. 1).
Out of these 182 excluded studies, 4 were review articles, 22 were using cell-based approaches, and 28 used collagen scaffolds but did not load them with their substance as they mostly used a collagen membrane only as a cover. Another 53 articles were evaluating different aspects like mechanical testing, implant survival after alveolar socket preservation and implant placement, systemic therapy, and others. From 72 studies which utilized other biomaterials than collagen as drug carriers, 18 were actually using collagen but in a mixed combination with deproteinized bovine bone mineral (DBBM), a widely used type of scaffold in dentistry and oral surgery (Fig.  1). The timeline of the past 15 years reveals a steadily increasing number of scientific publications in this field. Starting from zero to three publications from 2005 to 2010, to around 10 publications per year in the last 5 years (Fig. 2).
In all included publications, in vitro and in vivo testing was performed in 34% (n = 32) and 55% (n = 52), respectively. Clinical testing of loaded collagen was carried out in only 11% (n = 10) of the publications, where BMP-2, doxycycline, alendronate, simvastatin, and platelet-rich fibrin have been tested in humans.

Drug types and types of collagen scaffolds
All of the 77 included studies were assigned to subgroups according to their defined category: biological medical products, pharmaceuticals, or tissue-, cell-, and matrix-derived products ( Table 1).
The numerical values show how often the specific substance was used in one of the 77 studies In total, 36 different drugs were utilized as therapeutic agents in 77 included publications. Biologicals were the most frequently used substance-category (n = 60), followed by pharmaceuticals (n = 27) and tissue-, cell-, and matrixderived products (n = 10). BMP-2 (n = 33), followed by FGF2 (n = 5) and BMP-9 (n = 4) together with SDF-1 (n = 4) were to most often used substances. Looking at the types of collagen scaffolds which were utilized as carrier materials, sponges, or matrices were used in 50, membranes in 23, and gels in only 4 studies (Tables 2, 3, and 4).

Risk of bias and quality of the included studies
Sixty-seven out of 77 included studies were conducting in vitro and in vivo testing which are regarded as the lowest level of evidence. Ten out of 67 pre-clinical studies did not use unloaded collagen scaffolds as control. These studies utilized different substance dosages as positive control [40,63,68,82], other substances [64,85], blood clot formation after surgical interventions [33,39], or no control group at all [23,66]. Four out of ten clinical studies were case series with five to ten patients [46][47][48]89], whereas one was a pilot study with 12 patients [90] and five were clinical trials with 20 to 64 patients [45,49,69,71,76]. Eight out of ten clinical studies did not use an unloaded collagen scaffold as control [45-49, 69, 89, 90]. A CONSORT-based quality analysis of the ten included clinical studies showed a high risk of bias due to missing sample size calculation, unreported or missing methods of randomization and allocation concealment, incompleteness of followup, and lack of masking.

Biological medical products
Biopharmaceuticals or biological medical products are manufactured pharmaceutical drugs which are semisynthesized or extracted from biological sources. This drug category was most commonly applied in the included studies ( Table 1). The following table lists all studies which used different biological medical products loaded onto different types of collagen scaffolds (Table 2).

Multifunctional growth and differentiation factors
& Bone morphogenetic protein-2 (BMP-2) Collagen membranes loaded with BMP-2 at a concentration of 10 ng/mL increased human osteoblast proliferation up to 5 days post-seeding. Osteoblast differentiation markers, such as osterix, collagen I, and osteocalcin, were increased and mineralization of primary osteoblasts was pronounced (Miron R.J. et al. 2013 [17]). Another study used BMP-2 loaded collagen gel as control, comparing a designed chitosan-alginate-hydroxyapatite (CAH) scaffold as carrier for BMP-2. Release kinetics revealed a faster release from collagen compared with CAH (84 to 64% within 7 days). Alkaline phosphatase (ALP) activity and Alizarin red staining were significantly increased in mesenchymal stem cells (MSC) cultured with loaded collagen compared with unloaded CAH. Other results are not included in this review when collagen was not used as the substance carrier and BMP-2 was loaded in combination with MSC in the vivo part of this study (He X. et al. 2014 [18]). When loading BMP-2 at a low (10 ng/mL) or high (100 ng/mL) concentration onto collagen membranes, the substance was fully absorbed and slowly released within 10 days. Seeding murine ST2 stromal bone marrow cells onto these scaffolds revealed that ALP mRNA levels were significantly induced at 3 days and bone Records identified through database searching (Pubmed: n=248, Scopus: n=56, Ovid: n=20) n=324 Additional records identified through other sources n = 0

Records after duplicates removed n=259
Records screened n=259

Records excluded n=0
Full-text articles assessed for eligibility n=259 Studies included in qualitative synthesis n=77    [20]). Elastin-like polypeptide-collagen hydrogels were loaded with BMP-2 and doxycycline together. Release of BMP-2 assayed by ELISA showed a linearity over 7 days. Adhesion, proliferation, and migration of human adiposederived stem cells on hydrogel surfaces was higher for drugloaded samples compared with unloaded gels (Pal P. et al. 2019 [21]). In vivo, collagen sponges were soaked with 100, 500, or 1000 μg/mL BMP-2 and used in a calvarial defect model in rabbits. After 1 month, height and amount of newly generated tissue were significantly higher at any concentrations of BMP-2. The amount of mineralized bone was only significantly increased in the 1000 μg/mL group compared with control (Hasegawa Y. et al. 2008 [22]). Another study in 2008 used gene transfer in bone marrow stromal cells (BMSC) to promote BMP-2 production. Human BMP-2 complementary DNA (cDNA) was subcloned into the plasmid pcDNA3.1 and BMSCs were transfected with it. Results showed that the level of expressed and secreted BMP-2 was significantly higher in transfected cells than in untransfected cells. Collected cultured medium from transfected BMSCs induced the expression of ALP and osteocalcin in nontransfected cells. When incorporating BMP-2-cDNA3.1 complexes into collagen sponges and implanting them into rabbit calvarial defects at both 4 and 8 weeks, experimental groups showed significantly more newly formed bone. No negative control was utilized (Seol Y.J. et al. 2008 [94]). In a middiaphyseal defect model of diabetic rats, collagen sponges loaded with BMP-2 were inserted. Qualitative microradiography and assessment showed that BMP-2 increased new bone formation after 6 weeks. Histomorphometric analysis revealed a significantly higher area of new bone formation.
Histologically, platelet endothelial cell adhesion molecule-1 staining at 3 weeks was higher in BMP-2 treated rats. All defects in the control group without BMP-2 had nonunion, so mechanical testing could only be executed in the BMP-2 group (Azad V. et al. 2009 [24]). Another study used criticalsized calvarial defects as orthotopic model and created subcutaneous pouches in rats as ectopic model. The aim was to assess the effect of collagen sponges loaded with different doses of BMP-2 (25, 50, 100, and 200 μg/mL). Bone formation was enhanced at all doses of BMP-2 compared with control. After 2 weeks, bone formation could be observed in the ectopic model which was less at 8 weeks post operatively but showed advanced bone remodeling and maturity (Lee J.H. et al. 2010 [25]). In 2011, a study combined BMP-2 together with glycosaminoglycans (GAG) of biglycan (BGN) on         [30]). When a GBR procedure was conducted at the buccal side of edentulous maxillary alveolar ridges in dogs, onlay grafts containing bovine hydroxyapatite incorporated with collagen matrix were used and covered by a collagen membrane. In one group, the graft was loaded with BMP-2; in the other group, BMP-2 was loaded onto the collagen membranes. Newly formed bone could be observed in both groups without significant differences. Only the bone-to-residual bone substitute contact ratio was significantly lower in the group where only collagen membranes carried BMP-2 (Chang Y.Y. et al. 2015 [31]). The sequential treatment using SDF-1 before BMP-2 loaded onto collagen scaffold showed an enhanced osteoblastic differentiation, cell migration, and the highest degree of new bone formation in calvarial defects in mice (Hwang H.D. et al. 2015 [32]). When different carrier systems were compared in vivo, collagen sponges, biphasic calcium phosphate blocks and collagenated biphasic calcium phosphate blocks were loaded with BMP-2.
Critical-sized calvarial defects in ten rabbits were filled with loaded materials. Histological and histomorphometric analyses were performed at 2 or 8 weeks after surgery. At early stage, new bone formation was significantly less when collagen sponges were used and dimensional stability seemed to be greater in the other groups at both points in time (Jung I.H. et al. 2015 [33]). Lee et al. [34] investigated the effect of the combination of SDF-1 and BMP-2 loading onto collagen scaffolds. BMP-2 induced ectopic and orthotopic bone regeneration, whereas SDF-1 did not have a beneficial effect ( [37]). In a study comparing onlay-type grafted human freeze-dried cortico-cancellous bone blocks (FDBB) and deproteinized bovine bone with collagen (DBBC) loaded with BMP-2 (2.5 μg), buffer and BMP-2 loaded collagen sponges acted as control. Scaffolds were placed in rat calvarial defects and 2 or 8 weeks post-surgery, histologic and histomorphometric analyses were performed. At both time points, all BMP-2 groups showed significantly higher new bone area and bone density compared with control. BMP-2 loaded FDBB and DBBC showed significantly higher total augmented area than loaded collagen, but highest bone density could be observed in BMP-2 loaded collagen sponges (Lee J.S. et al. 2016 [38]). Another study in 2016 assessed the effect of limited-dose BMP-2 (1.5 μg) loaded biphasic calcium phosphate (BCP) scaffolds on healing of rabbit calvarial bone defects and compared them with BMP-2 loaded collagen sponges. The control group received no filling of the defect at all. At 2 or 8 weeks post-operative, both groups showed extensive new bone formation, which was significantly greater after 8 weeks in all groups. In the loaded collagen group, bony collapse was found of the upper defect. New bone formation and the size of the augmented area were significantly greater when using loaded BCP compared with collagen (You H. et al. 2016 [39]). In a study which was aiming to evaluate a novel BCP scaffold, collagen membranes were used as control. Scaffolds were loaded with 20, 50, or 100 μg of BMP-2 and applied in cylindrical bone defects which were created bilaterally in the femurs of rats. No unloaded collagen membrane as a negative control was used. BMP-2 was gradually released within 14 days from collagen compared with a rapid release from BCP within 1 day. An increasing concentration of BMP-2 led to higher bone fill irrespective of the carrier system. Loaded collagen membranes at a BMP-2 concentration of 100 μg were promoting up to 30% bone volume over total volume (BV/TV), whereas loaded BCP reached up to 100%. TRAP staining and osteoclast numbers were less in defects receiving BMP-2 loaded collagen scaffolds compared with loaded BCP (Zhang Y. et al. 2016 [40]). BMP-2 combined with alendronate (ALN) led to enhanced osteoblast differentiation and mineralization of mesenchymal stromal cells. Resorption markers showed a higher expression while TRAP positive cells were reduced. In vivo, bone mass and density was increased demonstrating a possible positive effect of ALN together with BMP-2 (Cho T.H. et al. 2017 [41]). In 2018, Durham et al. [42] investigated the effects of collagen sponges loaded with BMP-2 in critical-sized defects in mice. BMP-2 led to greater healing compared with control emanating from the surgical margin. While BMP-2 treatment led to an increase in cell presence within the healing defect there was, however, no regional orientation of craniofacial stem cells or vascularization (Durham E.L. et al. 2018 [42]). To determine the effect on BMP-2 soaked collagen membranes on peri-implant dehiscence effects, implants were placed in five dogs per group where the buccal bone was removed afterwards, resulting in a defect. This defect was filled with bone substitute material and covered with BMP-2 or saline collagen membranes. BMP-2 loaded scaffolds did not improve bone regeneration of this peri-implant bone defects compared with non-soaked membranes (Sun Y.K. et al. 2018 [43]). As the release of growth factors mainly depends on scaffold degradation, another study aimed to improve this condition by using scaffolds in combination with gene transfection using non-viral vectors. Nanohydroxyapatite-collagen scaffolds were loaded with BMP-2-encoding plasmid DNAfunctionalized calcium phosphate nanoparticles, naked plasmid DNA, or BMP-2 solution, and implanted into subcutaneous tissue of the back of rats. Collagen membranes loaded with either of these substances in different concentrations were used as control. The amount and its releasing period of BMP-2, as well as alkaline phosphatase activity were significantly greater in scaffolds loaded with BMP-2-encoding calcium phosphate nanoparticles. Released BMP-2 could be observed on the macrophages or multinuclear giant cells surrounding the scaffold fragments. Tartrate-resistant acid phosphatase (TRAP)-or osteocalcin-positive sites were occurring in all groups and a mineralization area was observed in nanoparticle loaded collagen membranes (Tenkumo T. et al. 2018 [44]). In the first clinical phase II study, which evaluated the safety and efficacy of BMP-2, maxillary sinus floor augmentations were performed in patients who received rhBMP-2 via an absorbable collagen sponge, at concentrations of 0.75 or 1.50 mg/mL. Unloaded bone grafts but not collagen sponges were used as control. CT analysis 4 months after treatment and 6 months post-functional loading revealed similar bone dimension increases but significantly lower bone densities in BMP-2 groups. Bone biopsies at implant placement confirmed normal bone formation (Boyne P.J. et al. 2005 [45]). In another clinical study, five patients with mandibular bone defects have been reconstructed with BMP-2 soaked collagen sponges alone or with bone marrow cells and allogenous cancellous bone chips. No unloaded sponges were used as control. Three cases were showing bone regeneration of the defect site. In two patients the grafts failed but were successfully reconstructed using autogenous bone from the iliac crest (Carter T.G. et al. 2008 [46]). In a case series, seven molarsite implants were placed in six consecutive patients. All sites were augmented with BMP-2 at 1.50 mg/mL loaded onto a collagen sponge after extraction to enhance bone regeneration prior to implantation. No negative controls were utilized. All sites achieved primary stability at implantation and were functionally loaded. No implants were lost or developed complications during the healing period of 3 to 8 months (Levin B.P. et al. 2012 [47]). Another case series of ten patients implanted collagen sponges loaded with 1.4 mL BMP-2 into the surgical defect after root resection. Panoramic radiographic and clinical assessments after 3 months revealed sufficient bone regeneration. No unloaded scaffolds were used as comparative control (Kumar M.S. et al. 2016 [48]). In a 3-month clinical trial, two BMP-2 delivery systems for alveolar ridge preservation were tested. A total of 32 patients received a collagen sponge soaked with 1.05 mg rhBMP-2 and another 32 patients received β-tricalcium phosphate and hydroxyapatite scaffold immersed in BMP-2. No unloaded sponges were utilized as control. The loaded scaffolds were placed into the extraction sockets. Dimensional bone changes at 25, 50, and 75% of extraction socket level and histomorphometric results of bone biopsies before implant placement showed no significant differences (Jo D.W. et al. 2019 [49]).
& Bone morphogenetic protein-6 (BMP-6) In critical-sized supra-alveolar periodontal defects in 11 Beagle dogs, collagen sponges were soak-loaded with BMP-6 at 0.25, 1, and 2 mg/mL and implanted into the defects. After 8 weeks, the lowest concentration of BMP-6 showed the best results compared with control. New bone regeneration and cementum height were significantly enhanced (Chiu H.C. et al. 2013 [50]).

& Bone morphogenetic protein-7 (BMP-7)
An in vivo study was combining SDF-1 with different growth factors (BMP-7, VEGF, PDGF) to evaluate the regenerative effect in critical-sized calvarial bone defects in mice. BMP-7 (1 μg) and SDF-1 (1 μg) alone or combined were loaded onto collagen membranes. Maximal bone formation was found in the BMP and the BMP + SDF-1 group at 2 and 4 weeks. After 4 weeks, the defects were almost completely healed. SDF-1 did not show a significant additional effect when combining with BMP-7 (Jin Q. and Giannobile W. 2014 [51]).
& Bone morphogenetic protein-9 (BMP-9) Collagen membranes loaded with BMP-9 fully absorbed the substance and slowly released it for up to 10 days. When seeding murine ST2 stromal bone marrow cells onto collagen membranes which were carrying a low (10 ng/mL) or high (100 ng/mL) dose of BMP-9, ALP mRNA levels were significantly induced at 3 days and BSP at 14 days. ALP levels were significantly higher (5-fold increase at 3 days and 2-fold increase at 14 days) in groups where the scaffold was loaded with BMP-9 compared with BMP-2. In addition, mRNA levels of collagen1, BSP, and osteocalcin were significantly higher in BMP-9 compared with BMP-2 groups. Effects on cell attachment, proliferation, and on osteoblast differentiation were not significant. Alizarin red staining further showed that BMP-9 induced up to 3-fold more staining in comparison with BMP-2 (Fujioka-Kobayashi M. et al. 2016 [19], Fujioka-Kobayashi M. et al. 2017 [20]). In a calvarial defect model in rabbits, deproteinized bovine bone mineral (DBBM) was used in combination with a collagen barrier membrane to investigate bone formation using BMP-9. BMP-9 was loaded onto DBBM or the collagen membrane and results were compared with unloaded materials and empty bone defects. Loaded samples promoted significantly higher mineralized tissue volume, new bone height, and new bone area which were evaluated by histomorphometry 8 weeks post-surgery. Multinucleated giant cells could be observed in contact with DBBM surfaces, whereas collagen membranes were not. BMP-9 loaded collagen membranes showed better horizontal wound closure with a faster regeneration of new host bone, suggesting that GBR procedures might benefit more from loaded collagen barrier membranes compared with DBBM (Fujioka-Kobayashi M. et al. 2017 [52], Saulacic N. et al. 2017 [53]).

& Erythropoietin (EPO)
Due to the known proangiogenic effects of EPO, Schwarz et al. [54] soak-loaded this substance onto collagen sponges. In vitro evaluation of the activity of human gingival fibroblast showed no significant change. When placed in subcutaneous pouches on the back of rats, organization of collagen type III fibers were pronounced and biodegradation of the collagen sponge speeded up after 14 days (Schwarz F. et al. 2014 [54]).
& Fibroblast growth factor 2/β (FGF2/FGFβ) D'Mello S. et al. [55] used collagen scaffolds as carrier to deliver polyethylenimine-pDNA nano-sized complexes encoding the FGF2 protein in BMSCs. Recruitment and attachment of BMSCs to scaffolds could be observed using scanning electron microscopy. Complex-loaded scaffolds significantly increased cell proliferation and expression of FGF2 compared with controls (D'Mello S. et al. 2015 [55]). FGF2 loaded onto collagen membranes covering biphasic calcium phosphate bone grafts in calvarial defects in rabbits led to an increase of new bone formation and the amount of residual graft material decreased regardless of different FGF2 concentrations (Lee S.H. et al. 2017 [56]). When FGF2 was loaded onto collagen sponges, in vitro testing of the activity of human gingival fibroblast showed an increase which was not significant. Implanted into subcutaneous pouches on the back of rats, organization of collagen type III fibers was pronounced and biodegradation of the scaffold speeded up after 7 days (Schwarz F. et al. 2014 [54]). In another study, collagen hydrogel fabricated from bovine type I collagen was mixed with FGF2 and injected into sponge-form collagen. Those scaffolds were implanted into class II furcation defects in dogs. At 4 weeks, healing of alveolar bone including cementum-like tissue and periodontal attachment was significantly increased (Momose T. et al. 2016 [57]). Collagen sponges impregnated with 5 μg of plant-derived rhFGF2 (p-FGF2) protein or E. coli-derived rhFGF2 (e-FGF2) were used to fill a calvarial defect in mice. Both groups, containing differently derived FGF2, enhanced bone healing to higher levels than without FGF2. Proliferation, mineralization, and osteogenic marker expression in MC3T3-E1 cells were also similar, suggesting the usefulness of a plant-based expression system for the production of biologically active rhFGF2 (Poudel S.B. et al. 2019 [58]).

& Human growth hormone (HGH)
Collagen sponges soak-loaded with HGH were surgically placed in subcutaneous pouches on the back of rats. Biological activity of human gingival fibroblasts in vitro and organization of collagen type III fibers was insignificantly pronounced. Degradation of the scaffold was accelerated (Schwarz F. et al. 2014 [54]).

& Interferon gamma (IFN-γ)
To evaluate the tissue response to IFN-γ by investigating a possible decrease in the differentiation of myofibroblasts, collagen scaffolds with or without IFN-γ were implanted submucoperiosteally in the palate of rats. Histological and immunohistochemical analyses were performed at 1, 2, 4, 8, and 16 weeks after implantation. IFN-γ caused a major numerical reduction in myofibroblasts and a faster influx of host cells (Jansen R.G. et al. 2011 [59]).

& Platelet-derived growth factor (PDGF)
Non-viral vectors in gene therapy can be a safe and efficient way to deliver growth factors. In a study by Elangovan et al. [60], the aim was to develop such a delivery system for bone regeneration using collagen scaffolds and polyethylenimine (PEI)-plasmid DNA complexes encoding PDGF-B. Low cytotoxicity could be observed and transfection efficacy testing in human BMSCs revealed significant proliferation in vitro. Complex loaded scaffolds showed significantly higher new bone volume to total volume ratio in calvarial defects in rats (Elangovan S. et al. 2014 [60]). An in vivo study was combining SDF-1 with PDGF and other growth factors like BMP-7, VEGF evaluating the effect in critical-sized calvarial bone defects in mice. PDGF (0.5 μg) and SDF-1 (1 μg) alone or combined were loaded onto collagen membranes. At 2 and 4 weeks, there were no differences in bone mineral content among the groups containing PDGF and SDF-1 alone or in combination, and VEGF. Groups containing BMP-7 significantly showed the maximal bone formation (Jin Q. and Giannobile W. 2014 [51]). Soak-loaded collagen sponges with PDGF-BB showed a significant increase of human gingival fibroblast activity. When implanted in subcutaneous pouches on the back of rats, an ingrowth of collagen type III fibers was observed and biodegradation of the scaffold was significantly faster (Schwarz F. et al. 2014 [54]).

& Receptor activator of NF-κB ligand (RANKL)
To establish and test a local osteoporotic model without performing an ovariectomy, the mandibles of five beagle dogs were prepared by drilling a hole and implanting a collagen sponge with 20, 40, or 60 μg RANKL. The maximum bone density loss of about 40% could be observed at 40 μg RANKL. Dental implants combined with injectable βtricalcium phosphate microsphere bone grafts loaded with BMP-2 were placed. The group with the lowest dosage of 5 μg BMP-2 showed the highest increase in bone areas and bone to implant contact. These changes in bone regeneration were statistically not significant (Chang A.R. et al. 2017 [61]).
& Stromal cell-derived factor 1 (SDF-1)/C-X-C motif chemokine 12 SDF-1 sequentially followed by BMP-2 soaked onto collagen sponges showed significantly enhanced osteoblastic differentiation and cell migration in vitro. In calvarial defects in mice this sequential treatment showed the highest degree of new bone regeneration. SDF-1 alone did not cause any changes (Hwang H.D. et al. 2015 [32]). A study by Lee et al. investigated the effects of simultaneous SDF-1 and BMP-2 treatment on bone formation. Collagen sponges with or without BMP-2 were loaded with different doses of SDF-1 (0.1, 0.5, or 1 μg). SDF-1 had no additive effect on osteoblastic differentiation and cell migration of mesenchymal cells, or capillary tube formation of endothelial cells compared with BMP-2 or SDF-1 alone. Loaded scaffolds were implanted into subcutaneous pockets and critical-sized calvarial defects in mice. SDF-1-only-treated implants did not lead to significant in vivo bone formation and SDF-1 treatment did not improve BMP-2-induced ectopic and orthotopic bone regeneration (Lee C.H. et al. 2015 [34]). To investigate the effects of SDF-1 on bone regeneration in periodontal diseases, collagen membranes loaded with SDF-1 were implanted into established bone defects on the buccal side of bilateral mandibles in rats. The results showed that SDF-1 recruited hostderived mesenchymal stem cells and hematopoietic stem cells to the wound area. CD11b+ inflammatory cell response was significantly reduced. Furthermore, SDF-1 led to an increase in vascular formation, promoted early osteoclastogenesis, accelerated scaffold degradation, and improved the quality and quantity of regenerated bone (Liu H. et al. 2015 [62]). In an in vivo study in mice, SDF-1 was loaded in different dosages onto collagen scaffolds containing BMP-7, VEGF, or PDGF, and implanted into critical-sized calvarial bone defects for 2 and 4 weeks. Part of this study, where SDF-1 was loaded onto demineralized bone matrix instead of collagen for further in vivo testing, is not being considered in our review. Biopsies were examined using micro-CT (μCT) and histology. SDF-1 promoted calvarial critical-sized bone defect healing like VEGF and PDGF, without having significant additional effects when combining it with BMP-7, VEGF, and PDGF (Jin Q. and Giannobile W. 2014 [51]).
& Transforming growth factor beta (TGF-β) Collagen gels loaded with 100 or 500 ng of TGF-β3 demonstrated an initial large release, followed by a sustained release over 14 days which was determined by ELISA measurements. Proliferation of mink lung epithelial cell was not influenced by TGF-β3 conditioned media, but alkaline phosphatase activity was decreased. No unloaded control was used (Premaraj et al. 2006 [63]). Collagen membranes soaked in TGF-β1 at a concentration of 10 ng/mL allowed human osteoblasts to adhere and increased osteoblast proliferation at 3 and 5 days post-seeding (Miron R.J. et al. 2013 [17]). In another in vitro study, the authors were investigating if collagen membranes adsorb growth factors from bone conditioned medium. Recombinant human TGF-β1 loaded onto collagen membranes served as control, which significantly provoked gene expression of TGF-β target genes. No negative controls were utilized (Caballe-Serrano J. et al. in 2017 [64]).

& Vascular endothelial growth factor (VEGF)
An in vivo study was combining SDF-1 with VEGF and growth factors BMP-7 and PDGF to evaluate the effect in critical-sized calvarial bone defects in mice. VEGF and SDF-1 either alone or combined were loaded onto collagen membranes. At 2 and 4 weeks after surgery, there were no differences in bone mineral content among the groups containing VEGF and SDF-1 alone or in combination, and PDGF. Groups containing BMP-7 showed the maximal bone formation (Jin Q. and Giannobile W. 2014 [51]). Kim J.H. et al. developed a composite scaffold consisting of VEGF loaded onto mesoporous silica nanoparticles, which were then incorporated within a type I collagen sponge. This composite material demonstrated good biocompatibility in a subcutaneous tissue model in rats. VEGF was sustainably released over the test period of 28 days and cell proliferation was increased. Testing with the chick chorioallantoic membrane (CAM) model showed significantly increased numbers of blood vessel complexes compared with scaffolds without VEGF (Kim J.H. et al. 2016 [65]).

Pharmaceuticals
Pharmaceuticals are fully synthesized drugs which are aiming to influence physiological pathways. In this review, included studies used a wide variety of pharmaceuticals. Those substances were loaded onto collagen scaffolds aiming to influence oral hard and soft tissue regeneration. Some studies also evaluated if certain substances improved properties of the scaffold which indirectly could influence tissue regeneration (Table 3).

Antibiotics
The use of antibiotics, such as amoxicillin or tetracycline, loaded onto collagen membranes, demonstrated a lower penetration as well as a decreased adhesion of Streptococcus mutans and Aggregatibacter actinomycetemcomitans (Cheng C.F. et al. 2009 [66], Cheng C.F. et al. 2015 [67]). Doxycycline-containing collagen hydrogels showed a gradual time-dependent drug release over 5 days, and indicated efficacy against four bacterial strains (Escherichia coli, Pseudomonas aeruginosa, Streptococcus sanguinis, and methicillin-resistant Staphylococcus aureus) (Anderson T.R. et al. 2015 [68]). Elastin-like polypeptide-collagen hydrogels loaded with doxycycline together with BMP-2 showed an initial burst release within the first 10 h, followed by a gradual release of doxycycline over 7 days, measured by spectrophotometry. Adhesion, proliferation, and migration of human adipose-derived stem cells on hydrogel surfaces was greater and all drug-loaded hydrogels showed bioactivity against P. aeruginosa, S. sanguinis, and E. coli (Pal P. et al. 2019 [21]). In a clinical setting, 30 patients with at least one pair of contralateral interproximal intrabony defects were treated with open flap debridement and placement of doxycycline containing gel in one group. In the other group, exposed root surfaces were etched using 24% EDTA (EDTA) before gel application. No negative control was carried out. Samples of gingival crevicular fluid of up to 21 days after the intervention were analyzed using high-performance liquid chromatography. In both groups, doxycycline was faster released in the first 3 days, followed by a sustained release over 21 days. After root surface etching, local doxycycline availability of doxycycline was higher (Gamal A.Y. et al. 2012 [69]).

Bisphosphonates
Alendronate (ALN) combined with BMP-2, loaded onto collagen sponges, enhanced osteoblast differentiation and mineralization of mesenchymal stromal cells in vitro. In a calvarial defect model in 80 rats, bone-forming and bone-resorbing markers showed a higher expression, while TRAP positive cells were reduced. Bone mass and density was also higher, suggesting a positive additive effect of ALN together with BMP-2 (Cho T.H. et al. 2017 [41]). In vivo, alendronate loaded on collagen membranes reduced the surface resorption of onlay bone grafts in a ridge augmentation model in eight pigs. However, in five cases, necrosis of the overlying periosteal tissues was observed macroscopically (Möller B. et al. 2014 [70]). In a clinical study with 20 patients, collagen sponges were soaked with 20 mg/mL alendronate and placed into extraction sockets of incisors and premolars. Collagen sponges alone led to a 22.8% and loaded scaffolds to a 44.4% decrease in alveolar ridge resorption (De Sarkar A. et al. 2015 [71]).

Corticosteroids
Collagen membranes loaded with dexamethasone demonstrated a burst-like initial release, followed by a sustained release over 28 days. In a calvarial defect model in 48 male Sprague-Dawley rats, bone volume and quality of new bone formation was increased (Piao Z.G. et al. 2015 [72]).

Hypoxia mimetic agents (HMA)
HMAs such as deferoxamine (DFO), dimethyloxalylglycine (DMOG), or L-mimosine (L-MIM) loaded onto collagen barrier membranes showed a burst-like release in the first hours. In osteoblast-and osteoclast precursor cells (RAW 264.7 and MC3,T3-E1) DFO and DMOG inhibited cell proliferation and metabolic activity. Osteoclastogenesis was significantly reduced by DFO, suggesting a possible positive effect on oral tissue regeneration (Edelmayer M. et al. 2017 [73]). In human gingival fibroblasts, they did not significantly influence cell viability or proliferation (Al-Habbal D. et al. 2018 [74]). In both studies, DFO and DMOG led to an increase in the expression of vascular endothelial growth factor.

Statins
Simvastatin that was locally administered via collagen sponges at two different concentrations (0.5 mg/50 mL or 2.2 mg/50 mL) in a calvarial bone defect model in 64 Wistar rats showed no differences in radiographic densitometry and histometric analyses after 30 days. After 60 days, the group with lower concentrations even showed reduced bone formation, concluding that there was a detrimental effect on bone regeneration (Calixto J.C. et al. 2011 [75]). In a clinical splitmouth setting, collagen membranes loaded with 1.5 mg simvastatin were placed into intrabony defects of 60 patients, while the contralateral site was filled with an unloaded scaffold. Loaded scaffolds significantly increased OPG and decreased RANKL levels in the gingival crevicular fluid, demonstrating a potential role of simvastatin in periodontal regeneration (Suthanthiran T. et al. 2019 [76]).

& Calendula officinalis flower extract
Collagen scaffolds were loaded with calendula officinalis flower extract and implanted in full-thickness skin wounds in 39 rabbits. Results revealed an inferior wound healing and clinical outcome when compared with the non-loaded control. Collagen cross-linking was increased, leading to a decreased degradation behavior of the scaffolds (Millan D. et al. 2016 [77]).

& Cyanoacrylate
Evaluating new methods for membrane fixation, a study applied cyanoacrylate on collagen membranes in a rabbit calvarial GBR model. After 8 weeks, little amounts of cyanoacrylate adhesive remained on the collagen membrane, causing an inflammatory reaction. There was no difference in bone regeneration (Song W.K. et al. 2018 [79]).

& Glycosaminoglycan/hyaluronic acid
In 2011, a study combined glycosaminoglycans (GAG) of biglycan (BGN) together with BMP-2 and collagen scaffolds as a carrier. In C2C12 myogenic cells, biglycan and BMP-2 combined loaded collagen scaffolds induced BMP signaling and function at most. In vivo, loaded scaffolds were implanted in critical-sized mandibles defects of 16 Sprague-Dawley rats. Bone formation was significantly increased when BMP-2 was combined only with de-glycanated BGN, indicating that the GAG component of BGN functions as a suppressor for the BGN-assisted BMP function (Miguez P.A. et al. 2011 [26]). Collagen sponges that were loaded with 1% hyaluronic acid gel demonstrated significantly better bone fill of critical-sized calvarial defects in 32 adult Wistar rats (de Brito Bezerra B. et al. 2012 [80]). Collagen membranes soaked with hyaluronic acid showed nearly 80% adsorption and a gradual release over the observation period of 10 days. When implanted in the scalp pouch of diabetic rats, hyaluronic acid increased the residual collagen content in the diabetic group and delayed membrane degradation significantly (Eliezer M. et al. 2019 [81]).
& S-NO-HSA/human albumin (nitric oxide donor) S-nitroso human serum albumin (S-NO-HAS) is a donor of nitric oxide which is a mediator involved in bone regeneration. Collagen sponges soaked with 100 mL S-NO-HAS (5%, 20%) demonstrated higher bone formation in a calvarial augmentation model in 40 rabbits compared with the positive control group using human albumin (5%, 20%). No negative controls were used (Fügl A. et al. 2014 [82]).

& Mussel adhesive protein
Aiming to improve membrane fixation in GBR procedures, collagen membranes were loaded with mussel adhesive protein and compared with cyanoacrylate. Results suggested excellent biocompatibility of mussel adhesive protein loaded collagen membranes in guided bone regeneration of calvarial defects in rabbits. No difference in bone regeneration was found (Song W.K. et al. 2018 [79]).

Bone conditioned medium (BCM)
In vitro, Caballe-Serrano et al. [64] generated bone conditioned medium (BCM) by placing demineralized harvested bone chips in cell culture medium. Aqueous soaked collagen membranes were placed in BCM to see if these scaffolds adsorb grow factors activating TGF-β signaling. Collagen membranes with BCM decreased the expression of adrenomedullin and pentraxin 3 and increased the expression of interleukin 11 and proteoglycan 4 in gingival fibroblasts. BCM did not influence the viability or morphology of the fibroblasts (Caballe-Serrano J. et al. in 2017 [64]). In a calvarial defect model in rats, collagen membranes were soaked in bone conditioned medium or culture medium as control and implanted in the defect site. No unloaded control was used. The control group showed significantly more bone than the BCM group. Scanning electron microscopy and histology showed that BCM might shift bone formation toward the formation of bony islands rather than new bone connected to the host bone (Kuchler U. et al. 2018 [85]).

Cell conditioned medium (CCM)
Cultured conditioned media from human bone marrowderived mesenchymal stem cells (MSC-CM) was collected to investigate its influence on endogenous stem cell mobilization and bone regeneration. Insulin-like growth factor-1, vascular endothelial growth factor, transforming growth factor-β1, and hepatocyte growth factor were present in the MSC-CM. When MSC-CM was cultured with rat bone marrow-derived stem cells, migration, proliferation, and expression of alkaline phosphatase, osteocalcin, and Runx2 was increased. Collagen sponges soaked with MSC-CM were grafted into rat calvarial bone defects and showed early bone regeneration (Katagiri W. et al. 2013 [86]).

Platelet-rich plasma (PRP)
PRP preparations from donor rats were soak-loaded on absorbable collagen sponges and surgically implanted into contralateral critical-sized calvaria osteotomies in 18 rats. Histologic and histometric analysis after 4 or 8 weeks showed that bone formation was highly variable in PRP sites as well as in controls. Radiographic analysis did not show significant differences to the control group. Conclusively, PRP preparation did not enhance bone regeneration in vivo   [87,88]).

Platelet-rich fibrin (PRF)
In a clinical case series of seven patients with alveolar ridge atrophy of different etiologies, a novel augmentation protocol using a combination of xenogeneic bone substitutes covered with PRF loaded onto collagen carriers was carried out. No autologous bone transplantation or periosteum splitting was performed. The gap between the flap margins was bridged using the collagen matrix loaded with liquid PRF. No unloaded collagen matrices were used as control. The whole augmented site was finally covered by either a PTFE-based membrane or sterile latex. After 4 to 8 months, open healing resulted in newly formed soft tissue without any signs of scar formation or fibrosis. Biopsies during implant insertion histologically showed well-vascularized newly formed bone that incorporated the BSM granules (Ghanaati S et al. 2019 [89]). A retrospective clinical pilot study with 12 patients utilized injectable platelet-rich fibrin (iPRF) preparations loaded onto collagen sponges as a graft for sinus augmentation. No negative controls were carried out. Minimum and maximum residual subantral bone heights were 2.0 and 9.8 mm. After 6 months, panoramic radiographic measurements showed significant new bone formation at all sites. Implants of 10.0 or 11.5 mm length were placed successfully (Gülsen U. et al. 2019 [90]).

Platelet secretome (PSEC)
In a study by Mozgan E.M. et al. [91], secretome of washed platelets (washed PSEC) and secretome of unwashed platelet preparations (unwashed PSEC) were lyophilized onto CBM. Supernatants were collected by adding and replacing medium after 1 to 48 h. Mitogenic response of human periodontal fibroblasts was induced by the supernatants and showed the highest levels of total protein, TGFβ1, and PDGF-BB in the first hour. These effects decreased rapidly in subsequent time points resulting in continuously declining release kinetics (Mozgan E.M. et al. 2017 [91]).

Enamel matrix derivative (EMD)
EMD demonstrated efficient adsorption properties on absorbable collagen sponges. Nearly 100% of the amelogenin proteins could be found after loading. Release kinetics showed a gradual release for up to 10 days. EMD significantly induced a 1.5-fold increase in cell attachment and resulted in a 2-6-fold increase in mRNA levels of runt-related transcription factor 2 (Runx2), collagen1a2, alkaline phosphatase, and bone sialoprotein. Alizarin red staining was pronounced when combined with ACS (Miron R.J. et al. 2017 [92]). In vivo, two different EMD carriers (Emdogain®/Osteogain®) soaked on separate absorbable collagen sponges (ACS) were applied upon periodontal wound regeneration in chronic class III furcation defects in monkeys. A significantly higher amount of total adsorbed amelogenin was found for ACS-loaded-Osteogain® when compared with Emdogain®. Combined with open flap debridement (OFD), both substances resulted in higher amounts of connective tissue attachment and bone formation, compared with treatment with OFD + ACS and OFD alone. Osteogain® showed higher new attachment formation, cementum, and new bone area (Shirakata Y. et al. 2017 [93]).

Discussion
Enhancing regenerative properties of scaffolds for oral tissue engineering remains an important study objective. The timeline of the past 15 years of scientific literature in this field ( Fig.  1) together with the high number of 77 included journal articles and the large variety of 36 different substances which have been used, demonstrate the importance of this topic (Tables 2, 3, and 4). One of the reasons might be the common use of collagen biomaterials in oral and maxillofacial surgery as well as in periodontal treatment. GBR or GTR procedures often require collagen scaffolds as tissue substitute or to cover bone grafts. Biocompatibility, impenetrability for epithelial cells and bacteria, degradation behavior and physical or architectural properties which allow mechanical stability are important elements for scaffolds. Also, adequate adsorption of loaded drugs and their diffusion as well as controlled release, are factors which are indispensable [95]. Improving regenerative capabilities of these scaffolds by additionally adding regenerative substances might seem to be the next logical step. Today, the clinical use of drug loaded scaffolds in oral and maxillofacial surgery is still limited as only few substances are clinically approved and therefore available.

Literature search and excluded studies
Out of 259 unique search results of all three database findings, 182 publications were excluded. The major reason was that in 72 of these studies, other biomaterial carriers than collagen were used in their investigations. These studies were matching the inclusion criteria and were therefore part of the initial findings. This might be explained due to the often-used method of assessing the effect of other biomaterials on the gene regulation of collagen, the collagen synthesis or the secretion level. In 28 studies, loading of the collagen scaffold did not take place. In these cases, bone substitute materials were loaded and unloaded collagen membranes were only used to cover the grafted site. As previously stated, only studies utilizing collagen as carrier material were taken into consideration for being included in this review. Since cell-based therapies are not approved for the clinical use for oral applications yet, the authors chose not to include this alternative approach, leading to 22 studies which had to be excluded.

Types, adsorption, and release kinetics of collagen scaffolds
Looking at the types of collagen scaffolds which were utilized as carrier materials, sponges or matrices were used in 50, membranes in 23, and gels in only 4 studies. The selection of the scaffold type might often have depended on the planned clinical application. In a clinical setting, sponges or matrices are usually used to fill soft or hard tissue defects, while membranes are used as a cover and gels for defects which require a more flowable consistency of the carrier. In our review, we found that nine out of ten clinical studies used sponges or matrices, while the remaining one was using membranes. Scaffold selection in in vitro testing was evenly distributed using sponges or matrices in 15 and membranes in 13 publications. In vivo experiments more often used sponges or matrices (n = 37) than membranes (n = 15). Gels were only used for in vitro testing (Tables 2-4). Adsorption behavior was not assessed in gels. In membranes, mussel adhesive protein was completely adsorbed and cyanoacrylate adhesive remained, showing an inflammatory reaction [79]. Sponges fully absorbed BMP-2 and BMP-9 [20], whereas hyaluronic acid was nearly 80% adsorbed [81] and amelogenin proteins were adsorbed by almost 100% [92]. Out of all 77 included articles, 18 studies investigated release properties of different substances from collagen carriers [18, 20, 21, 35, 36, 40, 41, 44, 63, 65, 68, 69, 72-74, 81, 92]. Release kinetics of sponges or matrices showed an initial burst-like release phase of BMP-2 and BMP-9 within the first 8 to 24 h and subsequently a slower continuous phase from deeper layers over 10 days [20,41]. Hyaluronic acid and amelogenin were gradually released over 10 days and VEGF over 28 days [65,81,92]. Collagen membranes released BMP-2 fast in the first 24 h, followed by a gradual release over 28 days [40,44]. Doxycycline was faster released in the first 3 days, followed by 21 days of sustained release. Dexamethasone, DMOG, and DFO were bust-like released from membranes in the first hours and while dexamethasone was continuously released for 28 days, DMOG was not detectable after 24 h and DFO rapidly declined after a period of 48 h [72][73][74]. PSEC burst-like released in the first hour, then declining over the 48-h observation period [91]. Collagen gels were tested for their release kinetics of BMP-2, TGFβ3, and doxycycline. One study demonstrated a burst-like release of BMP-2, followed by a gradual release over 28 days [18]. Another study showed a gradual release from the beginning for the period of 7 days [21]. Also, doxycycline release kinetics were different in three studies. While it was gradually released in one study for 5 days [68], another study demonstrated a burst-like release in the first hours, followed by a gradual release over 7 days [21]. In a clinical setting, doxycycline was faster released from collagen membranes within the first 3 days followed by a sustained release over 21 days [69]. The release of TGFβ3 from gels revealed to be characterized by an initial burst-like phase with a delayed release for up to 14 days [63]. Release periods were always the maximum time of observation, while the release was never complete at that time point unless otherwise stated. In vivo, collagen sponges were completely resorbed after 8 weeks [27]. Studies trying to establish a slower and more controlled release were successful by using nanofibers or gene transfection applying non-viral vectors, whereas heparin-conjugated fibrin gel did not improve release kinetics [35,36,44].

Biological medical products
Growth and differentiation factors, especially bone morphogenetic proteins, were by far the most frequently applied substances. BMPs belong to the transforming-growth-factor-β superfamily and comprise 13 BMP family members according to the current nomenclature. BMP-2, BMP-4, BMP-6, BMP-7, and BMP-9 play an important role in ossification [96], while BMP-2, BMP-6, and BMP-9 are most potent in inducing differentiation of stem cells into osteoblasts [97]. Recombinant BMP-2 and BMP-7 are officially approved and available for clinical use. Although there are specific bone-related indications for treatment, up to 85% of usage is off-label and also the costs are very high [98]. Current clinical applications in oral and maxillofacial surgery are tooth extraction socket healing [99,100], mandibular bone reconstruction [46,101,102], alveolar reconstruction for cleft repair [103,104], or maxillary sinus augmentation procedures [105,106]. In this review, studies were investigating the effects of BMP-2, BMP-6, BMP-7 and BMP-9. All types of bone morphogenetic proteins increased osteoblast differentiation markers and improved hard and soft tissue healing in general and when compared with other substances in vivo [17,18,22,24,25,27,28,33,35,36,39,40,[42][43][44]. Previous studies have shown that adverse effects can occur, especially if the local concentrations of BMPs are too high. Side effects are ranging from inflammatory complications to osteoclast activation, induction of adipogenesis and bone cyst formation and wound complications [107][108][109][110]. In a clinical study, BMP-2 soaked collagen sponges in some cases supported the regeneration of mandibular bone defects [46]. Additional systemic PTH therapy or SDF-1 in combination with BMP-2 did not have any significant effects on tissue regeneration [30,51]. The application of pulsed electromagnetic fields to stimulate bone regeneration resulted in increased bone volume and density and improved the microarchitecture of the regenerated bone in groups with lower concentrations of BMP-2 (2.5 and 5 μg), suggesting its ability to reduce BMP-2 doses to avoid adverse effects of high concentrations. Schwartz et al. previously reported that PEMF enhanced the osteoinductive effects of BMP-2 on mesenchymal stem cells, which might be an interesting aspect to investigate in further studies [111]. In supraalveolar periodontal defects in dogs, BMP-6 at the lowest applied concentration loaded onto collagen sponges showed the best results in bone regeneration and cementum height [50]. Previous studies have shown that the effects of BMP-9 on ossification are the strongest, followed by BMP-2 and BMP-7 [96]. In a study evaluating the additional effect of SDF-1 on local BMP-7 treatment utilizing collagen membranes, BMP-7 increased maximal bone formation without beneficial effects of SDF-1 [51]. Comparing BMP-9 to BMP-2 loaded collagen scaffolds, BMP-9 showed similar release kinetics but induced significantly greater ALP, colla-gen1 mRNA, BSP, and osteocalcin levels in murine ST2 stromal bone marrow cells. Cell attachment, proliferation, and osteoblast differentiation did not differ, but alizarin red staining was three times higher compared with BMP-2 [19,20]. In vivo, BMP-9 loaded onto collagen membranes enhanced bone regeneration parameters and showed faster horizontal wound closure [52,53]. When erythropoietin, which is well known for its proangiogenic effect, or human growth hormone or fibroblast growth factor 2 were loaded onto collagen sponges and implanted in subcutaneous pouches on the back of rats, scaffold degradation was faster and organization of collagen fibers was pronounced [54]. Collagen membranes loaded with FGF2, covering biphasic calcium phosphate bone grafts in vivo, led to an increase in new bone formation and decreased graft residuals [56]. Aiming to develop a new way of delivery, plasmid DNA complexes encoding for FGF2 carried by collagen scaffolds increased proliferation and FGF2 expression in BMSCs [55]. In a periodontal model, FGF2 mixed with collagen hydrogel and injected into furcation defects of dogs, bone healing and periodontal attachment were significantly increased [57]. Submucoperiosteally in rat palates IFN-γ loaded collagen scaffolds reduced the number of myofibroblasts and supported the influx of host cells [59]. PDGF soak-loaded on collagen sponges increase human gingival fibroblast activity and scaffold degradation [54]. In a rat calvarial defect model, PDGF increased bone mineral content similar to SDF-1 or VEGF alone [51]. Evaluating an efficient way to deliver growth factors, plasmid DNA complexes loaded onto collagen scaffolds, encoding PDGF, were shown to be effective and significantly enhanced new bone volume to total volume ratio in vivo [60]. Aiming to establish a local osteoporotic model without ovariectomy, collagen sponges loaded with RANKL were successfully used to cause bone loss [61]. SDF-1 is essential for BMP-2 induced osteogenic differentiation. In vivo, SDF-1 alone had no effect in a study but increased bone mineral content in vivo comparable with VEGF or PDGF alone in another study [51]. The sequential treatment of SDF-1 after BMP-2 increased bone regeneration and enhanced differentiation and migration of osteoblasts in vitro [32]. In combination with BMP-7, SDF-1 showed no additional effect [51]. In a periodontal defect model in rats, SDF-1 increased migration of MSCs, reduced inflammation, accelerated scaffold degradation, and improved bone regeneration [62]. TGF-β1 soaked collagen membranes increased human osteoblast adhesion and proliferation [17]. VEGF is a potent inducer of angiogenesis, which is a key factor to repair damaged tissue. Loaded onto collagen membranes, VEGF significantly increased bone mineral content after 2 and 4 weeks in vivo, comparable with SDF-1 and PDGF [51].

Pharmaceuticals
In this particular subgroup, 16 different substances have been used in 24 studies. Evaluating the effect of antibiotics loaded onto collagen membranes and hydrogels, findings of the included studies demonstrated a lower penetration and decreased adhesion of different types of bacteria [66][67][68][69]. Bisphosphonate (alendronate) loaded collagen scaffolds enhanced osteoblast differentiation and mineralization of MSCs in vitro [41] and significantly reduced bone loss of the alveolar ridge after tooth extraction clinically [71]. While alendronate also reduced surface resorptions of onlay bone grafts in vivo, in five of eight pigs, necrosis of the overlying periosteal tissues could be observed, suggesting that high local co nc en tra tion s m i gh t be d isa dva nt age ou s [ 70 ]. Corticosteroids, glycosaminoglycan, or hyaluronic acid, human serum albumin, parathyroid hormone, and progranulin enhanced bone regeneration in vivo in rat or rabbit calvarial defects [26,72,80,82]. Hypoxia mimetic agents showed diverse effects on osteoblast and osteoclast precursor cells as well as VEGF expression in vitro [73]. Statins and extracts of the calendula officinalis flower loaded on collagen sponges or gels, respectively, had adverse effects on wound healing in bone defect models in vivo [75,77]. However, a clinical study using simvastatin loaded membranes for periodontal regeneration showed a decrease in RANKL and an increase in OPG in gingival crevicular fluid, which might demonstrate a potential role of simvastatin in periodontal regeneration [76]. As collagen membranes often need to be fixated as part of GTR or GBR procedures, suturing and screw or tag placement might be necessary, which might be complicating the procedure. Also, additional interventions are needed as sutures and especially screws or tags might need to be removed. Application of resorbable surgical adhesives appear to be more convenient as cyanoacrylate has been used for decades to replace sutures, and also closures of sinus membrane perforations during sinus lift procedures [112] or gluing autologous bone grafts onto recipient beds after demineralization seem to be successful [113]. Aiming to use surgical adhesives instead, cyanoacrylate or mussel adhesive proteins loaded onto collagen membranes demonstrated that small remains of cyanoacrylate were still existent after 8 weeks, causing inflammation, while mussel adhesive protein showed excellent biocompatibility [79]. Due to previous studies, showing that cyanoacrylate is not fully adsorbed and causing adverse reactions [114,115], mussel adhesive protein might have its advantages and therefore calls for further investigations.

Tissue-, cell-, and matrix-derived products
In 10 out of 77 included studies, investigators applied products derived from tissue, cells, or matrices. Bone conditioned medium loaded on collagen membranes was influencing TGF-b target genes but did not improve the formation of new bone in vivo [64,85]. Conditioned medium of human bone marrow-derived mesenchymal stem cells increased activity, migration, and proliferation of rat bone-marrow-derived stem cells [86]. Collagen scaffolds were also successfully used as carriers of platelet-rich plasma, proved to preserve essential growth factors [116], and the use of platelet secretome induced mitogenic response and the expression of growth factors in human periodontal fibroblasts [91]. Enamel matrix derivate demonstrated a positive effect in cell attachment, osteoblast differentiation, and proliferation [92]. In vivo, bone formation and connective tissue formation were increased when evaluating the effect of EMD on periodontal regeneration [93]. Platelet concentrates or "platelet-rich plasma" (PRP) were already introduced in the 1990s, aiming to promote local healing at surgical sites [117]. PRP requires the addition of anticoagulants during blood collection while PRF is obtained by centrifugation without anticoagulants and is therefore completely autologous. This fibrin matrix contains platelets and leukocytes as well as a variety of growth factors [118]. In a clinical case series, a novel augmentation protocol for alveolar ridge atrophy, using xenogeneic bone substitutes covered with PRF loaded onto collagen carriers, was performed. Results showed newly formed soft tissue without any signs of scar formation or fibrosis and well-vascularized newly formed bone [89]. A retrospective clinical pilot study aimed to use PRF without any hard tissue substitute for sinus floor augmentation. Injectable platelet-rich fibrin (iPRF) was prepared and loaded onto collagen sponges as a graft which were then implanted. Panoramic radiographic measurements showed significant new bone formation [90]. Follow-up data or three-dimensional measurements were missing and would be recommended for future studies.

Limitations
In our review, included literature most often conducted in vivo (55%) and in vitro (34%) experiments, whereas only 11% conducted clinical testing with patients. The evidence level of in vitro and animal experiments is regarded to be low, even though these studies are forming the basis for studies in humans. New substances and methods need to be evaluated in an in vitro and in vivo setting, before employing design and results from these studies. Out of ten clinical studies, four were case series with five to ten patients [46][47][48]89], one was a pilot study with 12 patients [90], and five were clinical trials with 20 to 64 patients [45,49,69,71,76]. Contrary to our research question, we did not exclude studies without unloaded controls. The reason was that we aimed to provide a wider scope and to discuss articles which primarily focused on release properties without using negative controls. Articles which did not use negative controls were distinctly pointed out in our review. Only two out of ten clinical trials-which applied statins or bisphosphonates-used unloaded collagen scaffolds as control [71,76], whereas the other eight studies in humans did not [45-49, 69, 89, 90]. The absence of negative controls leads to results which might show the feasibility and dose-dependent changes of certain methods but do not allow any further conclusions. Overall, the evidence level of included publications was low and the risk of bias was high. One reason for the low number of clinical studies might be the fact that specific clinical applications require other bone substitute materials which have not been included in this review. Another reason might be the low number of clinically approved substances and therefore a lot of studies still focus on pre-clinical testing before clinical tests may follow. This review shows that clinically available BMP-2 was more often investigated. BMP-2 has been investigated in 33 studies, while other growth factors or substances which demonstrated positive effects on tissue regeneration were evaluated in only one to five studies. Today, the only stem cell-based treatment that is routinely reviewed and approved is hematopoietic stem cell transplantation. Cell-based therapy approaches might be a paradigm shift that may provide alternative therapeutic solutions for several diseases. Currently, many preclinical and clinical studies are conducted to approve cell-based therapies for diverse applications. As this approach is still under debate and the use in oral medicine is still not approved, we did not include studies which were focusing on these substances. While there are other bone substitute materials which are used for oral tissue regenerative applications, we focused on collagen scaffolds as they are commonly used and have been established in many applications as the gold standard. Another commonly used biomaterial for oral tissue engineering and regeneration is deproteinized bovine bone mineral alone or in combination with collagen. Evaluation of all available biomaterials in our review would have gone beyond the scope. A further review might be needed to address this aspect on its own.
Our research question primarily focused on the effects of substance loaded collagen scaffolds on oral tissue regeneration compared with unloaded scaffolds. Nevertheless, we did not exclude studies which were not using unloaded collagen scaffolds as control or even used loaded collagen scaffolds as a control for other loaded biomaterials. This way, we were able to include five additional journal articles which investigated release properties of carriers which did not use an unloaded scaffold [63,68,73,74,91]. The use of substance loaded bone substitute materials for applications in guided bone or tissue regeneration is still in an early phase of clinical use und needs further trials. A wide range of different bone substitute materials are widely used in clinical applications, whereas the availability of drug-loaded scaffolds is still low. Our review of the past 15 years of scientific experiments applying loaded collagen scaffolds includes 77 publications. These studies investigated 36 different substances. These facts might indicate that widespread investigations are aiming to find candidates to improve the regenerative potential of collagen scaffolds. Testing of all substances were performed in a pre-clinical setting (in vitro, in vivo), resulting in a low evidence level. Only BMP-2, doxycycline, alendronate, simvastatin, and platelet-rich fibrin were tested in humans, while the study design was qualitatively poor and had a high risk of bias. Overall, BMP-2 was most frequently investigated and low doses of BMP-2 showed positive effects on oral regeneration in all studies. BMP-9 might have more pronounced effects compared with BMP-2, even at lower concentrations. FGF2 was tested in five studies, demonstrating an increase in bone and periodontal healing. Collagen scaffolds loaded with antibiotics improved the scaffold's anti-microbial activity and even reduced the penetrability for bacteria. Other biological medical products, pharmaceuticals, and tissue-, cell-, and matrix-derived products showed positive or diverse effects on hard tissue and soft tissue regeneration. Because of the low number of studies for each of these substances, results might not be reliable and need to be further investigated. Collagen scaffolds of any type have proven, in any of the included experiments, that they are functional and clinically acceptable carrier materials for any kind of substance without having any negative effect by itself.

Conclusion
Collagen scaffolds of any type have proven to be capable carriers for all 36 substances which were utilized in the 77 included publications. Adsorption and release kinetics of all assessed substances showed clinically favorable results. These properties were also influenced to optimize the burst-like initial release and prolong the release. Biological medical products were by far investigated most often. Especially BMP-2 was investigated most frequently and showed positive effects on oral regeneration in all 33 studies where it was applied on different types of collagen scaffolds. BMP-9 showed more pronounced effects compared with BMP-2, even at lower concentrations. FGF2 demonstrated an increase in bone and periodontal healing. Pharmaceuticals were used in 24 studies to be loaded onto collagen carriers. Collagen scaffolds loaded with antibiotics improved the scaffold's anti-microbial activity and even reduced the penetrability for bacteria. Hyaluronic acid might prolong collagen membrane stability. Tissue-, cell-, and matrix-derived products were investigated in only 10 out of 77 included articles. Enamel-matrix derivate and PRF showed promising results pre-clinically and clinically, respectively. In general, other biological medical products than BMPs, pharmaceuticals, and tissue-, cell-, and matrix-derived products were investigated much less frequently. Outcomes of those studies demonstrated positive or diverse effects in terms of influencing hard tissue and soft tissue regeneration. Due to the low number of studies, those results do not allow conclusive assumptions. Ten out of 77 studies tested doxycycline, alendronate, simvastatin, and platelet-rich fibrin clinically in humans. Five of them were prospective clinical trials while only two clinical studies, which investigated the effects of loaded alendronate and simvastatin, used unloaded collagen scaffolds as control. Because of the low number of investigations for each substance, besides BMP-2, and the low level of evidence of studies using these substances, results might not be reliably reflecting the substances' influence on oral regeneration and need further investigation. In addition, more prospective clinical trials using appropriate study designs and higher sample sizes are needed. Still, findings of investigated substances, as well as physical properties (adsorption and release) of collagen scaffolds should be considered for further investigation and optimization. This review provides an overview of the last 15 years of scientific studies using drugloaded collagen scaffolds for oral tissue regeneration and should be considered for further research in this field.
Funding Information Open access funding provided by Medical University of Vienna.

Compliance with Ethical Standards
Conflict of Interest The authors declare that they have no competing interests.
Ethical approval This article does not contain any studies with human participants or animals performed by any of the authors.
Informed consent For this type of study, formal consent is not required.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.