Naturally, the inoculation of prostheses with autologous cells at the time of surgery represented the first step toward “living” implants.

Given the insurmountable thromboembolic complications of artificial hearts and the conclusion of deBakey et al. (109) and Wesolowski et al. (108) in 1964 that prosthetic blood surfaces remain non-endothelialised in humans it was foreseeable that first efforts toward tissue incorporation would be made at leading centers for artificial heart research. Accordingly, it was Ghidoni of deBakey's group who inoculated fragments of autologous skeletal muscle into the velour lining of left ventricular bypass pumps leading to foci of proliferating spindle-shaped cells but no endothelium after 2 weeks (139)—the longest possible implantation period before these early pumps failed. The attached tissue stripes sowed central necrosis and the concept was abandoned in favor of using cultured cells.

Fifteen years later, Mid-Western vascular surgeons from Indianapolis (140) and Ann Arbor (141) took up the idea of autologous single-stage cell inoculation and applied it to Dacron prostheses calling it “Endothelial Cell Seeding.” In dogs, seeding of venous endothelial cells successfully generated a mid-graft endothelium (141). Yet, when carried over into clinical trials, venous endothelial cell-seeding failed to achieve any improvement to the patency of the up to 60 cm long, medium-diameter ePTFE prostheses used for femoro-popliteal bypass grafts (142). It became obvious that seeding densities were too low, aggravated by a reduced reproductive capacity of endothelial cells in smokers (143–145) and patients with hyperlipidaemia (146, 147)- two conditions present in the vast majority of patients. Attempts to increase harvest efficiencies by using microvascular endothelial cells from adipose tissue (148, 149) or bone marrow (150, 151) again led to the successful creation of a functional endothelium in animal experiments (151–153). Yet, microvascular seeding raised concerns regarding unrestricted subendothelial proliferation (154) and when it was eventually clinically tested it also showed no significant improvement over controls (155). Typical for tissue engineering in general, single-staged seeding was rediscovered 25 years after it was first described (156, 157) and by a few researchers is still carried over into the present time (158–161). The main novelty was the replacement of the bio-stable PET and PTFE scaffolds by bio-degradable materials (156). Using again bone marrow as a cell source the concept was experimentally tested in vascular grafts in dogs (151), sheep (162), pigs (163), and mice (160) followed by a clinical pilot study in children (157). When seeded tube grafts were implanted as extracardiac total cavo-pulmonary connections (TCPC) significant graft stenoses developed in more than a quarter of all patients (159, 164). Experimental work suggested that higher seeding densities may mitigate graft stenosis (160, 161). Alternatively, pharmacological interventions were tested emulating the previously tried ACE inhibitor therapy (165) with modern angiotensin II receptor blockade (166). Eventually, single stage seeding of degradable scaffolds was carried over to heart valves and tested in acute (167) and semi-acute (168) sheep implants. Like 37 years before (140), cells were embedded in a fibrin matrix deep within the polymeric meshwork at the time of implantation. When bone marrow cells were seeded on supramolecular polymers, long term results were discouraging with valves failing due to tissue overgrowth and leaflet fusion (158). When the same scaffolds were used as unseeded valves, results were significantly better (158).

Eventually, the ultimate exploit of single-staged cell inoculation was in myocardial regeneration where it endured the longest. This was partly because the clinical drivers were not surgeons and as such least likely to have been exposed to the disappointing previous efforts. At the same time, the enthusiasm for single-staged myocardial cell inoculation was certainly also carried by its concurrence with the early days of stem cell research and the accompanying fascination with the potential of these cells. Early experimental reports describing startling myocardial regeneration in mice after MI (169) triggered a global stampede toward myocardial regeneration work resulting in clinical trials being reported within a year (170, 171) just to be thoroughly disproven 2 years later (172). Fifteen years after the first gung-ho period, meta-analyses of clinical trials with adult stem cells still showed conflicting results ranging from minimal to no therapeutic benefits (173, 174). The greatest shortcoming of the initial approach was the naïve misconception of myocardium as a gel-like structure rather than the dense sponge which particularly the venous side of the vasculature constitutes. As such, rather than engrafted, the injectates were largely washed out and the few remaining cells did not find a matrix environment to survive (175). Recent approaches toward defeating this Achilles heel have been the alternative use of compact 100 μm thick myocardial muscle bundles grown from human induced pluripotent stem cells (hiPSC) (176) or human embryonic stem cells (hESC) (177) or the injection of large cell agglomerates in form of “micro tissue” spheres (178, 179) (Figure 8) and the use of injectable biomaterial scaffolds to entrap and nurture the cells after delivery which has started to achieve incremental improvements (181, 182). Apart from embedding transplanted cells in a matrix that contains engraftment and pro-angiogenic cues, the second benefit of injecting such gels into myocardial infarcts is as a space-holder arresting an otherwise deleterious downward spiral of remodeling (183, 184) (Figure 9). Moreover, the initial hope to facilitate the trans-differentiation of progenitor cells to cardiomyocytes has given way to a more modest expectation allowing for anti-fibrotic and anti-inflammatory outcomes on the basis of a paracrine effect regardless of the cell types inoculated (188). In a translation context, exploiting the components of a paracrine mechanism, particularly exosomes is attractive due to reduced regulatory requirements and their off-the-shelf nature (189). Notwithstanding, the efficacy of this approach will again rest on efficient delivery perhaps most likely through controlled release from injectable scaffolds or sophisticated targeting approaches (190).

In hindsight, the trajectory of myocardial regeneration, however, hasn't been different from other areas of cardiovascular tissue engineering. Although much basic knowledge had been gained in the three preceding decades, this latest embodiment of the earliest approach to tissue engineering repeated many of the mistakes and equally fell for the temptations of its predecessors: hype at the beginning rather than thorough basic research; premature translations into clinical procedures; the unbridled belief in stem cells alone ignoring the embedding environment, the need for neo-angiogenesis and the importance of biomechanics and the resulting push-back by public opinion. The latter was highlighted by the inability of the first sufficiently powered phase III clinical trial to achieve clarity with respect to bone marrow derived cells (BAMI) to recruit more than 12.5% of required patients (189). Like with other fields of cardiovascular tissue engineering, the short-cut temptation single staged cell-inoculation held may also have slowed down the clinical realization of “true” engineering approaches in myocardial regeneration. Promising developments from myocardial muscle bundles to micro-tissue spheres would have otherwise seen a much more proactive clinical translation. Given the clinical need for a successful solution, the current “death-valley” will need to be crossed with determination and in the absence of glory. While acceptance of the complexity of the pathobiology involved has been a first step toward a more scientific approach, the issue of misleading animal models also hampers success in this area of cardiovascular tissue engineering. As convenient as a murine fresh infarction model may be, the majority of patients in need of myocardial regeneration are not suffering from acute but chronic ischaemia. The “replacement fibrosis” of acute infarction (191) is distinctly different from the extracellular matrix environment building up (192) in chronic ischaemia (193) where a highly cellular interstitial environment, enriched in matricellular proteins and capable of transducing growth factor responses may be required for structural and functional recovery (193). Similar to other areas of cardiovascular tissue engineering, focused and carefully considered models like Wolfgang Schaper's chronic ischemic heart model of the 1970 (194) gave way to the more expedient acute ligation models with their exclusive focus on acute ischaemia (195, 196).

The idea of sufficiently multiplying autologous cells in vitro before creating confluent functional cell layers on the blood surface of an implant again goes back to the artificial heart research of the 1960s. Adapting a method of culturing human endothelial cells that was first described in 1963 (197) it was once more Baylor College that successfully pioneered the in-vitro culture of autologous cells on the silastic membranes of circulatory assist devices (23) followed by the in-vivo proof of the antithrombogenicity of such a lining (32). At the height of pneumatically driven artificial hearts shear stress resistant confluent endothelial monolayers were created on the displacement membranes of a variety of TAHs (198–201). However, this very first tissue engineering approach utilizing cell culture never materialized as a game changer in clinical practice and quietly faded away as modern assist-devices made it obsolete.

A decade later, the group in Vienna–intimately involved in the initial clinical trials with single stage venous endothelial cell seeding of vascular prostheses (202, 203)- introduced mass-culture of autologous endothelial cells (79) to create shear-stress resistant confluent monolayers (204–206) prior to implantation. Preclinical trials confirmed the persistence and non-thrombogenicity of such tissue engineered grafts in a non-human primate model (207). An attempt to replicate this concept using cryopreserved multidonor allogenic endothelial cells in order to create “off the shelf” grafts was unsuccessful in the same senescent non-human primate model (208) (Figure 10). A clinical feasibility study was successfully done in 1989 (210) and a randomized clinical trial showed autologous in-vitro endothelialisation to dramatically improve the 3-year patency of femoro-popliteal and femoro-distal bypass grafts (211). Based on this evidence the group adopted this cell culture-based method as a routine procedure for all patients who had no saphenous vein available with excellent clinical outcome (212). Although a significant proportion of patients had distal reconstructions–a patient group with particularly poor prognosis–the 8 (213), 9 (214), and 12 year (215) patency rate of these grafts was continuously superior even to vein grafts. These excellent results were confirmed in more than 300 patients with a follow-up period of up to 17 years (80) (Figure 11). By then, some early limitations of this two-staged method had also been resolved. The initially experienced failure of some patients' own endothelial cells to proliferate into mass cultures had been overcome once risk factors like serum levels of lipoprotein a and triglycerides had been identified (146) and addressed (147). While delays between cell harvest and availability of the tissue-engineered grafts remained the biggest obstacle for acute clinical indications, improved mass culture techniques based on in-situ procedures and very-low density plating that cut out cell passaging had distinctly shortened this lag phase (79). Similarly, clinically used attachment matrices had been optimized (205, 206) culminating in an RGD-enriched, engineered lattice providing significantly enhanced shear stress resistance for the endothelium (216). The final proof of success came with the demonstration of a confluent endothelium on mid-segments of clinically explanted in-vitro endothelialised grafts from as early as 30 days (217) to almost 4 years after implantation (218) (Figure 11) and the histological proof of genuine arteriosclerotic changes more than 10 years after implantation (80) (Figure 12).

At the height of in-vitro endothelialisation of vascular grafts, the concept was also applied to bioprosthetic heart valves. Glutaraldehyde detoxification made it possible to grow endothelial cells to confluence on bioprosthetic leaflets (219–223). Such in-vitro lined tissue valves maintained endothelial integrity in the heterotopic primate model (224, 225) and even showed some degree of mitigation of tissue calcification (225, 226). However, when the extent of residual immunogenicity of conventional bioprostheses and its role in tissue calcification became apparent (227–229) better immune masking either through decellularization (230–233) and/or higher crosslink efficiency (234, 235) became the primary goal rather than a living endothelium. Also still in the 1980s Libby and Birinyi (236) and Pober et al. (237) had shown that under such circumstances an endothelium would augment rather than suppress an inflammatory process (236, 237) as the exposure of ECs to inflammatory cytokines, such as IL-1 or TNF-α, would cause the ECs to bind manifold more leukocytes (238) due to the induction of endothelial leukocyte adhesion molecules (ELAMs) (239). As such, an endothelium on conventional bioprosthetic leaflets even after detoxification would continually be proinflammatory (236). Attempts to reduce immunogenicity by decellularizing allografts (240–246) and xenografts (242, 247–251) before lining them with cultured endothelial cells were followed by clinical trials with in-vitro endothelialised, decellularized allografts (250, 252–254). When the in-vivo performance of equally decellularized matrices compared in-vitro endothelialisation with decellularization only, there was no difference (255) leading over to the later approach of non-vital decellularized native valve implants (256).

While the first two decades of tissue engineering were firmly tied to permanent scaffolds, this began to change in the 1980s. Long before anyone else, cardiovascular surgeons in Groningen had the idea of making the scaffolds gradually disappear while allowing functional tissue to develop (257–261). Pioneering this concept as early as in 1983 (261) they pre-empted key elements of the subsequent era like biodegradable polymer grafts inoculated with cultured smooth muscle cells (262). Yet, the catchy term “tissue engineering” —not coined but popularized–several years later by Joseph Vacanti, a surgeon-scientist at Boston's Children's Hospital and Robert Langer, a chemical engineer at the MIT (263) contributed to the lasting association of Boston with this approach. When they began their efforts, their underlying concept had a holistic rather than a cardiovascular claim. Indicative of the broader scope of their approach, the initial clinical problem they strove to resolve was curing diabetes through pancreatic islet transplantation (264) followed by attempts to re-grow cartilage (265). Toward, the mid 1990s, the “Boston group” had expanded this concept to heart valves (104) and shortly later also included vascular grafts (266) co-culturing embedded autologous arterial and venous cells within degradable scaffolds made of polyglactin/PGA fabric. Replacing a single valve leaflet in a sheep (104) confirmed the concept but also the previously shown superiority of autologous tissue engineered implants over their allogenic counterparts (208). However, at explantation leaflets were thicker and stiffer than native leaflets and did not have the delicate, differentiated microanatomy required for stress-reduction during opening and closing (104, 267, 268). The phenomenon of fibrosis and tissue shrinkage associated with a macrophage driven resorption process was seen in all animal models including senescent primates (269). This healing mode was well-known from the tissue reaction to resorbable surgical sutures made of the same polyglycolic- (PGA) and polylactic-acids (PLA) or polycaprolactone (PCL) materials which had been clinically used for decades (270). Attempts to minimize this cicatrition in favor of more mature valvular tissue included addressing biomechanics (271–273) as well as inadequate cell growth in the depth of the scaffolds through pulsatile bioreactors (274). Polymers were also iterated using natural polyesters like Polyhydroxyalkanoate (PHA) (271, 272) and Poly-4-Hydroxybutyrate (P4HB) (275). At the same time, the issue of the cell source also remained a challenge for two-stage culture-based approaches. Attempts to optimize the ease of harvest and the yield followed again in the footsteps of the previous era by either trying allogenic cells (276) or different autologous cell sources like bone marrow cells (275), mesenchymal stem cells (277), umbilical cord blood derived EC progenitor cells (278), prenatally harvested progenitor cells (279), human induced pluripotent stem cells (hiPSCs) (280) endothelial progenitor cells (EPC) (92) including reprogrammed and reconditioned cells (281). Different from previous eras, however, significantly more resources and internationally attracted manpower had characterized the “Boston era.” Eventually, fellows of John Mayer's group carried the program to other institutions in the early 2000s: Simon Hoerstrup to Zurich (282) [later on also collaborating with Frank Baaijens' group in Eindhoven (283)], Toshiharu Shin'oka to Tokyo (156, 157) and Christopher Breuer to Yale (later Ohio State) (160, 284). While other researchers like Laura Niklason at Duke (later Yale) (285, 286) also started their programs initially on the basis of bioreactor-based vital implants using degradable scaffolds (163), the group in Montreal went a step further. In an attempt to avoid all synthetic scaffolds and the associated problems of shrinkage and fibrosis, L'Heureux and Auger created polymer-free neo-vessels consisting of adventitial fibroblatsts, medial smooth muscle cells and an endothelium (287–291) Their extreme modern pendent are 3D printed pure stem-cell grafts (292). In clinical pilot trials with L'Heureux's AV-shunts graft dilatation was observed (288). In line with the general desire to cut out an autologous culture step and offer a product off the shelf their next generation of tissue-cultured vascular grafts was still autologous but devitalized to be storable (293) followed by a further step toward a commercial product by using devitalised allogenic cells (294). This step away from vital implants stood at the beginning of a general return to the concept of in vivo tissue regeneration of non-vital scaffolds originally pursued from 1968 onwards for artificial hearts (10, 11, 24, 34–36).

In retrospect, of all the vital tissue engineering concepts pursued, in-vitro endothelialisation was the first and for decades the last embodiment that was successfully translated into clinical practice (Figures 10–12). The successful pilot program ceased when new regulations prohibitively tightened the circumstances under which a patient's tissue may be processed as part of an implantable device outside the operating room. Although this hurdle would be surmountable by establishing integrated facilities and ISO-compatible cleanroom productions the combination of the inconvenience of not having an “off the shelf” product readily available with a protracted process that excludes acute interventions makes a revival of “vital” two-stage tissue engineering unlikely—regardless of whether the scaffold is permanent or temporary.

Ironically, two diametrically opposed tissue engineering concepts of the past decades—decellularized allo and xenograft valves and pericardia on the one hand (242, 296, 297) and vital implants of cultured cells on degradable scaffolds on the other (104)—ended in congruence when the latter added a decellularization step. Now, both concepts bet on the in-situ re-population of allogenic or even xenogenic antigen-reduced, cell-derived matrices. A main challenge of this approach is the fine line between remnant immunogenicity (298–300) and avoidance of crosslinking to allow repopulation with host cells. The fact that a cell-free matrix can still be immunogenic has been shown (301) not least in the catastrophic failure of Synergrafts in children (302). Furthermore, interstitial cells seem to be needed to maintain the microarchitecture of the extracellular matrix (303). While the microarchitecture of the extracellular matrix of allografts is well-preserved in transplant recipients in whom the donor cells survived, the typical acellularity found in allografts from non-immune suppressed recipients is associated with a loss of microarchitecture and hyalinisation of the matrix (303).

Although the verdict is still outstanding whether such constructs will eventually get sufficiently repopulated in humans, one may in the meantime draw parallels from the experience with auto- and allograft heart valves. There, three basic facts emerged: (a) cells can remain vital in the depth of a transplanted native valve provided they are autologous and were already there at the time of implantation (304–306); (b) the survival of allogenic cells in the depth of transplanted native valves is only possible in the immune-suppressed context of heart transplantation but otherwise allograft heart valves become rapidly acellular (307); and (c) in spite of presumably representing the optimal extracellular matrix structure of a native valve, allografts remain acellular after years of implantation (303, 307). These observations indicate that in humans, even the destination matrix of a native decellularized aortic valve may remain acellular unless it already contains autologous cells at the time of implantation as is the case in autografts. Theoretically, this absence of cell ingrowth into conventional allografts could either be secondary due to post implantation matrix disturbances resulting from the death of co-transplanted cells or primarily reflect the mitigated trans-anastomotic tissue outgrowth in humans. However, even in larger animal models, the distal leaflet sections remained mostly acellular (308). Therefore, the question as to the source of cellular re-population cannot be evaded. Since all decellularized matrices present at least a temporary barrier for transmural endothelialisation, and significant fall-out healing from the circulation has remained an unfulfilled pipedream for decades (107, 108, 113, 116, 309), the partially successful “in-situ tissue regeneration” observed in animals may have its most likely explanation in the vigorous trans-anastomotic outgrowth potential of adjacent intimal tissue in sheep that does not exist in man. Hence, no convincing evidence going beyond facile case reports (308) has been presented yet that unambiguously demonstrated a successful autologous in situ re-population of decellularized native heart valves in patients. Early reports rather highlighted the unresolved issue of remnant immunogenicity of the decellularized matrix in the absence of crosslinking. CryoLife had reported partial recellularization of the distal conduit but an absence of cellularity in the leaflet after up to 11 years in vivo (310). This was followed by the clinical debacle with their decellularized xenografts in children (302). “Leaflet repopulation” was also reported during clinical use of the Matrix PTM line of valves, another decellularized xenograft, but the repopulating cells generally appeared to be inflammatory rather than phenotypically appropriate valve cells (302). Whether non-cardiac-derived decellularized matrices like the “CorMatrix” made from porcine small intestinal mucosa would allow to revisit the non-crosslinked xenograft concept has been decisively answered in a recent study where more than half of the valves failed (311). Together with reoperation rates in humans when used as patches (312) clinicians have been calling for a moratorium on such non-crosslinked decellularized xenografts (313, 314). Notwithstanding the conclusion that non-crosslinked decellularized xenografts are showing dangerous inflammation and failure, the excellent clinical performance of decellularized allografts over their conventional counterparts (315–317) makes it likely that the concept will prevail even if it loses the attribute of “guided tissue regeneration” upon more vigorous patho-histological analyses under strictly defined criteria.

While clinical studies with decellularized natural heart valves (318–320), pericardium (296) and vascular grafts (321) have been reported for more than two decades decellularized in-vitro grown matrices have been tested for equally long–both as heart valves and as vascular grafts–but so far clinical trials have only been performed with vascular grafts. The concept was pioneered by Niklason's group in 2003 (322) decellularizing allogenic PGA/smooth muscle cell (SMC) constructs that were cultured in a bioreactor for 2 months. With the initial focus on dilatation resistance and patency (323) and animal implants mostly reflecting trans-anastomotic outgrowth (324) the key question as to the repopulation of the decellularized matrix in man had to wait for clinical studies. Starting in 2012 (325) sixty “Humacyte®” access grafts (325, 326) were implanted as part of a phase 2 trial and 16 tissue samples ranging from 16 to 200 weeks were histologically analyzed showing sometimes a dense repopulation of the media with α-actin positive cells (326). While showing that the decellularized matrix can be populated with mesenchymal cells, the countless wall punctures of access grafts make it difficult to ascertain whether the healing mode was transmural per se or required the puncture holes. A subsequent clinical femoro-popliteal bypass study eventually provided sufficiently long (28 cm) unpunctured grafts (327) to potentially clarify this issue. Of three obtained specimens, one was at the anastomosis and as such disqualified; one was at a perforation site leaving one specimen where the low-density cell infiltrates could have been evidence for a transmural repopulation. Likewise, the luminal irregularities after 24 months, reflected in angiographic diameter variations between 2.9 and 5.8 mm could alternatively be interpreted as signs of tissue formation or thrombotic appositions. Similarly, the described lack of an endothelium may have been due to the preceding attempts of balloon thrombectomies (327) or to the a'priori absence of an endothelium due to a lack of significant trans-anastomotic outgrowth in man and the absence of transmural endothelialisation. As such, what still remains to be proven is whether a healing mode proposed by Grabenwöger's group in Vienna as early as in 1998 (328, 329) whereby radial micro-holes greatly augmented the repopulation of decellularized matrices will eventually be required to also see surface endothelialisation on such in-vitro grown, decellularized matrices. It would, however, not simply be a punching-out of holes but a fine-titration between too broad ingrowth spaces that may lead to premature connective tissue maturation preventing further capillary sprouting (Figure 13) (329) and insufficient ingrowth spaces. As an alternative concept to decellularization, L'Heureux's group clinically pursued de-vitalisation through airdrying. Initially, the devitalised grafts were still partially re-vitalised by in vitro lining with the patient's own endothelial cells (293). This hybrid approach was followed by completely devitalised non-endothelialised allogenic grafts (“Lifeline Grafts®”) (294) eventually leading to woven textile structures using yarns of devitalised in vitro grown tissues (333). No repopulation, however, was ever reported in any of these grafts. Leaning on L'Heureux's earlier work (290), Syedain's group in Minneapolis using fibrin gels to create scaffold-free cell-produced decellularized vascular grafts (334) provided the most convincing repopulation evidence yet. They implanted 15 cm long AV shunts in middle-aged non-human primates (Papio Anubis) that remained unpunctured throughout. Within large animal models, Papio Anubis seems to come closest to the senescent Papio Ursinus model that was previously shown to reflect the trans-anastomotic outgrowth-stoppage (102) seen in humans (107). The study convincingly demonstrated the repopulation of the mid-graft media with mesenchymal cells that likely originated from the adventitial side (334). Yet, as it did not show traces of transmural angiogenesis and a largely absent endothelium in the mid-graft region it also highlighted once more the dependence of endothelialisation on trans-anastomotic outgrowth on these matrices—something one cannot expect in humans to occur.

Corresponding with a proven wall-repopulation but an unresolved endothelialisation mode were the suboptimal clinical results with these decellularized, bioreactor-grown vascular grafts. “Lifeline” grafts were disappointing though they were few. “Humacyte,” in contrast, had 60 access grafts implanted at 6 centers (325, 326) and 20 femoro-popliteal grafts at 3 centers (327). Primary patencies were low both in the access grafts and in the femoro-popliteal grafts. In access grafts where the 1-year primary patency is known to be between 43 and 57% for ePTFE grafts (58, 59), it was 28% for the “Humacyte” grafts (325). In above-knee femoropopliteal bypass grafts with good run-off the 2 year primary patency was 58% (327) compared to 73% in ePTFE grafts for a similar patient group in Frank Veith's multicentre study (37) and 80% for in vitro endothelialised 6 mm ID grafts in comparable patients (80). In vitro endothelialised 7 mm grafts even had a 2 year primary patency of 88% in spite of also including patients with distal reconstructions with poorer run-off (80).

During the past decade the concept of decellularized cell-produced matrices was also carried over into heart valves. The group in Zurich decellularized their well-established tissue engineered constructs of cultured human fibroblasts on a degradable scaffold (335). Using the senescent baboon model (335) and the sheep model—first for transcatheter pulmonary valve replacements (336–338) and later for orthotopic TAVIs (339)—in depth repopulation of such a matrix with mesenchymal cells was demonstrated. The patchy endothelial cover of the leaflets (158, 339), however, again hinted at the likely dependence of surface repopulation on the trans-anastomotic outgrowth seen in sheep. Furthermore, leaflet shrinkage as a consequence of repopulation emerged as an inevitable consequence of early tissue maturation (335, 336, 338) but could be addressed by shape compensation during production (340). Syedain's group in Minneapolis also tested their scaffold-free, fibrin gel-based cell-derived decellularized matrix in surgically implanted aortic valve replacements in sheep (341). Similar to others (158), repopulation with α-SMA and Vimentin positive cells was primarily seen at the leaflet base while distal leaflets remained poorly cellularized and only sporadically endothelialised.

As such, decellularized in-vitro grown matrices have successfully shown their ability to get repopulated by mesenchymal host cells primarily in areas of tissue contact with the host. The sub-optimal clinical performance of vascular grafts as well as the decreasing endothelial coverage with increasing distance from the nearest anastomosis, however, suggest again a dependence of the neointima formation on the trans-anastomotic outgrowth occurring in animals.

The idea that synthetic implants contain morphogenic cues to stimulate the in situ regeneration of key cardiovascular structures again goes back to the early days of tissue engineering. There, the goal was to induce the in-vivo formation of a non-thrombogenic “neointima” on artificial hearts (10, 11, 24, 32, 34–36) and vascular grafts (342) merely on the basis of iterations of porosity and surface structures. Clinical success never materialized for the same reasons it hasn't materialized half a century later when biomechanics and molecular biology provided the tools to recapitulate refined biophysical and biochemical properties of the target tissues. (343–345). The reason for the continual absence of a clinical translation of this concept lies in decades of experimental cross-purpose design between a biologically possible tissue regeneration response and one that is actually feasible at the intended site and in the intended host environment. Although we have already highlighted that this has been the overarching dilemma of all tissue engineering approaches of the past decades, it is particularly pertinent for non-vital synthetic implants as their clinical performance depends much more on the in-situ formation of functional tissue (346) than any other concept. As such, an eventual break through of “in vivo” generation of functional tissue in synthetic scaffolds will not depend on whether the cues provided replicated nature best but whether they were based on appropriate assumptions regarding the host response.

After it had been shown that the cellular source of “fall out” healing was predominantly from endothelial precursor cells in the bone marrow (347, 348) and that these endothelial progenitor cells (EPCs) express amongst other ligands the vascular endothelial growth factor receptor VEGFR-2, as well as the CD133 and CD34 antigens (349–351) several attempts were made to augment the homing of these cells onto “non-vital” synthetic grafts by immobilizing anti-ligand antibodies on the surface (352, 353) including antibodies against chemotaxis stimulating receptors on stem cells (354). Yet, even with sporadic studies showing an enhanced recruitment of circulating cells (349, 355), without a broad consensus to focus on endothelialisation and cell population from the circulation, this approach will continue to remain an unrealised promise. As such, transmural ingrowth and endothelialisation remain the most realistic modes of tissue regeneration for synthetic scaffolds.

For permanent scaffolds to achieve successful transmural healing, ingrowth permissive spaces were shown to be a sine qua non. The minimal dimensions of such spaces were well defined at 80–400 μm² (102, 356) optimally even being as large as 5,000–6,000 μm² (357). Therefore, if transmural endothelialisation and healing is the goal, this requirement would exclude the use of ePTFE of <60μm internodal distance, woven Dacron and most of the electrospun polyurethane grafts unless their scaffold structures were manipulated to increase the interfibrillar spaces (357–360). Yet, making transmural endothelialisation of synthetic scaffolds a reliable occurrence also requires the suppression of the surface compaction with interstitial thrombus (107) (Figure 14). Therefore, functionalising the scaffolds with pro-angiogenic signals (361–368) must not only be seen as quantitative augmentation of angiogenesis but as the attempt to outpace the build-up of an inhibitory barrier. Encouragingly, however, thin walled ePTFE of 60–90 μm IND (102, 123, 135) whose porosity was known to provide the necessary vascular ingrowth spaces (102) was shown in non-human primates to successfully achieve transmural endothelialisation without the need for any ingrowth augmentation (102, 123, 125, 369). Regrettably, the clinical confirmation of this observation was thwarted by regulatory concerns leading to a last minute addition of a very low-porosity wrap on the outside (370) of PTFE grafts turning high-porosity into low-porosity prostheses. One of the many puzzles in tissue engineering is why the “real” clinical study never happened since, as it would have certainly provided the entire field with much needed “translational” energy.

If the complexity of a clinical in-vivo environment makes it already difficult to accomplish transmural endothelialisation (371) in non-degradable scaffolds one needs to reckon with additional challenges in degradable scaffolds. There, it is less an issue of whether sufficient ingrowth spaces are pre-existent or progressively “open up” during degradation but whether the inflammatory process accompanying the break-down of the scaffold introduces additional obstacles. While “harnessing the natural inflammatory process” has been the underlying principle behind the degradation and replacement of a temporary scaffold with newly formed tissue (371–375) its effect on transmural endothelialisation has not been tested yet. Some data on electrospun heart valves confirm the presence of intramural vessels and mesenchymal cells corresponding with tissue contact (Figure 15) but a clear distinction from a trans-anastomotic origin cannot be made (158). If the latter was the case, even the intramural cell population of the scaffold would likely be different in humans. The proximity of a trans-anastomotic endothelium to a pro-inflammatory environment would trigger the active recruitment of inflammatory cells from the circulation (236, 237). The known trans-differentiation potential of such actively recruited mononuclear blood cells (85–91) and their ability to participate in the intramural cellularisation process and neo-tissue formation of a synthetic scaffold (87) makes it likely that the entire healing pattern would thus be distinctly different in humans. Ironically, a successful and rapid replacement of a biodegradable scaffold by mature mesenchymal tissue may even prematurely terminate endothelial migration (376) thereby thwarting transmural surface endothelialisation all together. While the iteration of scaffold chemistry (371, 377) and microstructure (371, 378) would allow the titration of the degradation process (371, 379) the actual clinical requirements against which to titrate remain again poorly defined in the absence of better suited animal models. Whether inserted as infra-renal interposition grafts in rats (260, 377, 380) or comparably short interposition grafts in rabbits (381, 382), dogs (383, 384), mice (385), sheep (386, 387) or pigs (388–390): trans-anastomotic outgrowth has again most likely overpowered any other form of healing in these studies on degradable scaffolds. Even if tubular structures contained valves, trans-anastomotic neointimal outgrowth provided the most probable cell source for the leaflet population observed (158, 379). In the expected absence of a neo-intima reaching the distal leaflets in humans, this may lead to situations where leaflet degradation precedes tissue formation. Clinical pilots added to this concern. Based on the equilibrium between degradation of modern supramolecular polymers originally developed in Eindhoven (379, 391) and tissue formation observed in the sheep (386) an early clinical feasibility/first in man study was commenced with pulmonary valves implanted into children. The developing post-implantation coaptation deficiency of leaflets in 11 out of 12 valves (392) hinted at a healing behavior not seen in the sheep. Without histological proof to the contrary, a lack of tissue formation in the distal leaflets will always seem the most likely explanation given the known healing deficiencies in humans. The subsequent corrective modifications of the polymer degradability (392) may have retrospectively led to well-functioning polymeric heart valves but may still defy the claim of eliciting “endogenous tissue restoration.” In a parallel clinical trial using the same material for extracardiac TCPCs (393) the close contact of the tubes with their surrounding tissue may eventually lead to successful transmural healing. However, the likelihood of obtaining sufficient representative samples within a meaningful time frame is very low in both trials: RVOT revisions increasingly happen endovascularly and TCPCs only yield samples in the rare occasion of a transplantation or death many years later. Long term animal implants published since have confirmed in similar supramolecular polymer valves that even in an animal model with excessive trans-anastomotic outgrowth like the sheep cellularisation was mainly restricted to the leaflet base while the distal leaflets remained poorly cellularised with only patchy endothelium (158).

As such, any tissue engineering approach that justifies the word “engineering” will require animal models which allow a rational and mechanistic exploration of key components of the “endogenous tissue restoration” expected at the clinical destination site. A myriad of fascinating modern scaffold materials for instance has been investigated under in-vivo conditions which did not emulate key cornerstones of the human situation ranging from thermoplastic polymers such as degradable polyurethanes (394); poly-e-caprolactone (PCL)–alone or in combination with poliglecaprone (395), with polydioxanone (396) or with Gelatine (397)- “supramolecular”: ureic-pyrimidone-modified polycaprolactone (158, 398) also in combination with ureic pyrimidone PEG (398) to thermoset polymers such as polyglycerol-sebacate (PGS) (399) or nanofiber scaffolds of naturally occurring polymers such as polysaccharides (cellulose) (400–402); hyaluronic acid (343); silk (385, 403), collagen (404), Elastin-like blends (405, 406), fibrin (407), or collagen-elastin (408). For none of these materials do we have an answer to two of the critical questions asked at the beginning: (1) as scaffolds, do they allow transmural endothelialisation or alternatively facilitate true fall-out endothelialisation and (2) given the absence of trans-anastomotic neointimal outgrowth in man: is scaffold degradation (eg., by blood borne inflammatory cells) balanced against neo-tissue formation to prevent a premature structure-loss in patients? The same questions need to be asked with regards to the effect of incorporated/grafted bioactive molecules such as VEGF (361, 366, 367, 409), NO (364), TGF-b (410), SFD-1 (411), and many others (368, 412) not to mention the various ingrowth gels whether they are from natural proteins (388, 407, 413, 414) or fully synthetic (415); functionalised, (416–422) and/or potentially cell selective (423). For gels, however, an important second purpose may emerge as “space-holders” further facilitating transmural endothelialisation. In this role, their presence in the interstices of the scaffolds may also prevent the build-up of impenetrable thrombus near the blood surface until the gels get replaced by ingrowing tissue.