Over the last decade, transcatheter aortic valve implantation (TAVI) has become an accepted alternative technique to surgical valve replacement for over 150,000 patients worldwide [5, 6]. Despite minor issues related to the implantation of the device, this non-invasive technique is cost-effective and provides increased comfort to patients, relative to traditional surgical valve implantation. In a fast-growing market, where TAVI-related survival rates depend highly on the patient’s initial health, one can expect that less-critical patients could be treated successfully with TAVI in the coming years.
Currently, the valve material used in the TAVI procedure is biological tissue, such as bovine or porcine pericardium. This material has been extensively used in open heart valve surgery over the last decades and remains the gold standard. However, it is well known that biological material undergoes structural degradation. In particular, studies about TAVI have shown that, once assembled inside the metallic stent and crimped at low diameter for catheter insertion, the biological material undergoes additional stress and is already degraded prior to implantation [1, 13, 28]. Actually, no work has been published yet, which proves that the durability of the biological valve material is jeopardized with the crimping process. But a first long-term durability study about TAVI first generation devices has been recently presented and shows high rates of valve degeneration after a few years . The structural degradation of biological tissue emphasizes the interest in investigating the potential of synthetic leaflet material alternatives. Textile material could be a potential alternative to replace biological valve leaflets. This material is strong and flexible and can be obtained in various configurations: non-woven, woven or knitted. Several research projects have focused over the last years on investigating the potential of fibrous scaffolds for valve tissue engineering. In particular, the potential of non-woven constructions has been investigated in order to produce leaflet material [2, 9, 16], but with no convincing results regarding long-term durability.
With respect to woven material, the potential of textile polyester [polyethylene terephthalate (PET)] has also been investigated in vitro in recent years in various studies. Woven textile constructions have outstanding folding and resistance properties and as a result, these materials are easy to crimp and insert, even in low-profile devices . Moreover, woven materials are discontinuous, mitigating the risk of catastrophic rupture. Rupture propagation is isolated to the single filament and does not propagate to neighboring filaments instantaneously. Decades ago, Teixeira et al.  showed that the tear resistance of woven fabrics actually increases with the yarns’ movement freedom within the fabric construction. Heim et al. showed that woven textile materials could resist up to 200 million cycles in vitro under accelerated cyclic loading . Recently, textile and biological tissue were compared in terms of mechanical and hydrodynamic performance. Results show that the behavior of textile is close to the behavior of biological tissue. However, in order to consider textile materials as potential TAVI leaflet materials, one must show that their performances in vivo is at least comparable to those of biological tissue, which remains the gold standard.
One of the limits of textile materials is their roughness due to yarn crossing and surface discontinuity. Basically, the standard foreign body reaction (FBR), which is characterized by encapsulation of the synthetic leaflets with fibrotic tissue, is largely influenced by the topography of the leaflet surface. Being porous, textile material may be specifically prone to calcification and fibrotic tissue ingrowth. This behavior may depend on the size and number of the pores as well as on the topography of the yarns that are involved in the textile construction.
With a thick fibrotic tissue layer or high calcium rates on the leaflet surface, the movement of the leaflet would become limited and stenosis would be generated.
The purpose of this work was to carry out the first ever preliminary investigation study (based on a limited number of animals involved) of the performances of textiles as valve leaflet material in vivo. Valve prototypes made from various textile constructions were implanted in sheep models and the material behavior was observed over a 6-month period.
Materials and methods
Textile as leaflet material has not been studied in vivo yet, but several groups have already studied the interaction mechanisms that occur when a fibrous construction is in contact with biological tissues. The results reveal that the tissue ingrowth in the fibrous construction depends largely on the fiber size [19, 21, 25] and the pore size [8, 15] as well as the roughness of the textile material [4, 14]. Woven materials can be produced with various roughnesses and surface topography. Both characteristics depend on the woven yarn used and the fabric construction, which is considered. Yarns can be obtained from one or several filaments. Actually, monofilament yarns are characterized by a smoother surface but are mechanically stiffer, while multifilament yarns are rougher but more flexible. Regarding the woven construction, yarns can be crossed in various ways such as plain weave, twill or satin. The construction pattern depends on how many times a weft yarn goes over or under a warp yarn in the weaving process.
In this preliminary work, three plain weave woven PET fabrics were considered: multifilament calendered, multifilament non-calendered (Cardial, St Etienne, France) and monofilament (Sefar, Heiden, Zwitzerland). The calendering process consists of passing the fabric under rollers at a high temperature (above the glass transition temperature) and at high pressure in order to thin the material and close the pores. Figure 1 shows the differences among the fabrics used in the study.
These constructions were chosen in order to vary the roughness and porosity at the yarn level (between the monofilament and the multifilament samples) as well as at the fabric construction level (between the calendered and the non-calendered samples).
Before implantation, the textiles used were characterized by their physical characteristics such as thickness, roughness, permeability and porosity in order to discuss the influence of the construction parameters on the behavior observed in vivo.
Characteristics of the textile used
Standard equipment was used to assess the physical properties of the fabric samples used in this study.
Leaflet thickness was measured with a dedicated instrument (C110T, Kroeplin GmbH, Schlüchtern, Germany). The applied force during the measurement was 0.5 gf/cm2, which corresponds to the sensitivity of the device used.
Surface roughness was assessed in order to study the correlation between the topography and the interaction with the biological tissues. Roughness, given as a standard mean deviation (SMD) value in micrometers, was measured with a Kawabata Evaluation System instrument (KES-FB4, Kato Tech Co., Ltd., Kyoto, Japan).
In order to assess the permeability of each fabric, square textile specimens (2 cm2) were placed under a static water column applying a 120-mm Hg of pressure over the specimen. Flow across the fabric was measured (l/min) over a 3-min period of time.
Porosity was defined according to following definition: the ratio of the open space in a fabric to the total volume of the fabric, expressed as percentage. It is given by the following equation
M=mass per unit area of porous material (g/cm2)
L=thickness of porous material (cm)
ρ=density of solid material (1.38 g/cm3 for PET).
Textile valve manufacturing
All the fabric valves were obtained from a tubular textile membrane according to a shaping process already described in a previous work . The design of the valve was obtained from the geometry of the mold used to shape the cusps (Figure 2). The mold is characterized by three cylindrical shapes (same radius as the valve radius), which are prolonged with triangular surfaces for leaflet coaptation purposes.
All the valves were assembled to commercially available 23-mm biological valve rings (Sorin, Milano, Italy) with an Ethicon monofilament suture yarn (Prolene 5-0) for classic surgery implantation purposes. In this study, two valve specimens from each considered fabric were prepared and sterilized with ethylene oxide (EtO).
Valve implantation procedure
After cardiopulmonary bypass, the prototypes were implanted in the mitral position in juvenile sheep models (Laboratoire de Recherches Biochirurgicales de la Fondation Alain Carpentier, Paris, France). Six sheep weighing 30 kg on average were used as animal models. The mitral position was preferred to the aortic position to limit the risks of the surgical procedure, while respecting, however, the physiological pressure conditions. The experimental procedure was performed in accordance with the French regulation (agreement number MESR No. 01425.04) after the agreement of the local animal ethics committee.
After 6 months, all the animals were euthanized and the explanted valves were analyzed at two levels: (1) at macroscopic level and (2) at histological level, in order to observe the fibrotic capsule on the leaflets and to evaluate the calcium deposits.
Explanted valve analysis
After macroscopic examination, implanted materials were dissected and at least 0.5 cm of surrounding tissue was excised and gently rinsed in saline before utilization. For histological analysis, samples were fixed in a 4% paraformaldehyde solution, dehydrated and embedded in paraffin. Height-micron-thick sections were obtained with a microtome HM3555 (Thermo Fisher Scientific, Villebon sur Yvette, France). Sections were stained with hematoxylin phloxine saffron (HPS) in order tovisualize cells’ nuclei (blue), cytoplasm (pink) and connective tissue (yellow). Digital sections were acquired with a NanoZoomer 2.0-RC C 10730 (Hamamatsu, Japan). Sections were analyzed with the NanoZoomer software in order to determine the fibrous capsule thickness in three different zones (Figure 3): (1) leaflet insertion zone to the aortic wall, (2) leaflet central zone and (3) leaflet end zone (free leaflet).
An average fibrotic tissue thickness value was calculated from five values regularly measured along each zone length from each section. Finally, considering three leaflet sections for each valve, between 15 values and 30 values could thus be considered to assess the average fibrotic thickness in each zone. Actually, in some cases, images could not be utilized as the cutting was not clean enough.
Regarding the calcium rates in the valve leaflets, it was assessed by air-acetylene flame spectrophotometry (VARIAN AA240) after lyophilization and total hydrolysis of the samples. The results are given in micrograms per milligrams of leaflet material. A minimum of three values could be defined for each valve and an average value could be calculated.
Characteristics of the leaflet
One animal from the multifilament valve group died within a few hours from surgical procedure complications, and was excluded from the study. All other animals survived the procedure and were euthanized within a 6-month period.
Fibrotic tissue ingrowth
Figure 4 presents typical paraffin-cut sections (in radial direction) of explanted valve leaflets for all three fabric constructions considered in this study: sample A (MULTI), sample B (MULTICALEND) and sample C (MONO).
In the figures, fibrotic tissue appears in pink while the fabric filaments can be identified in between, in white. Several observations can already be made at a qualitative level. First, the results reveal that, whatever the construction, the level of fibrotic tissue is not spread regularly on the leaflet surface. It appears to be thicker in the leaflet insertion zone to the aortic wall. Second, it is seen that the tissue ingrowth level depends largely on the fabric topography and is globally limited for the multifilament construction. These observations are confirmed at a quantitative level with the fibrosis thickness values that are presented in Figure 5.
When comparing the MULTI and MULTICALEND samples, one can observe that the calendering process tends to transform the textile material into a more-fibrosis-promoting substrate, especially in zones 1 and 2. Despite similar fabric constructions (50 yarns in weft and 75 yarns in warp), the average thickness of the fibrotic layer is 605 μm on the non-calendered substrate vs. 2086 μm in zone 1 and 296 μm vs. 110 μm in zone 2. The difference is less significant in zone 3.
Regarding the MONO and MULTI samples, the results reveal that the monofilament construction generates twice as much fibrotic tissue as the multifilament one (1368 μm vs. 605 μm on average in zone 1).
A closer look at the interface zone between the fabric and the biological tissues shows that the fibrotic tissue penetrates the fibrous construction (Figure 6) in all samples. The phenomenon is, however, limited in the calendered material. Actually, in this case the tissue seems to remain at the surface of the textile, which behaves more like a non-porous membrane.
Levels of calcification
The levels of calcification measured on the valves are plotted in Figure 7. The results show large differences among the samples. One can observe that the calcium level is minimum for the monofilament substrate, with 7.5 μg/mg on average, and maximum for the multifilament, with 58.6 μg/mg. Moreover, from Figure 6, one can observe that calcium penetrates between fibers in the MULTI sample, while the phenomenon is limited in the MULTICALEND sample as fibers are closer together in the construction. Regarding the MONO sample, calcium deposits were not apparent.
In this preliminary work, we observed the in vivo interaction between biological tissues and three different fabric constructions used as valve leaflet material within a 6-month implantation time. Due to the limited number of valves used in this work, it is difficult to establish some statistics about the observed results. However, the results provide some interesting trends about the influence of the fabric topography on the tissue ingrowth and calcium dosage observed in textile used as valve leaflet material.
Regarding the fibrotic tissue formation, when comparing the MONO and MULTI samples, the multifilament material appears to be a less-fibrosis-promoting substrate compared to the monofilament material. Actually, the average fibrotic tissue layer is 380 μm thick for the MULTI sample and 1081 μm for the MONO sample on average in the three zones, as can be seen in Figure 5. This difference is mainly due to the size of the fibers involved in the fabric construction. The monofilament yarn diameter is 37 μm while the fiber diameter in the multifilament yarn is 10 μm. Basically, the curvature radius is reduced in filaments characterized by smaller diameters. It has been shown in the literature that fibroblastic cells, which induce fibrosis, have more difficulty spreading on fibers characterized by a small curvature radius. Sanders et al., studying the fibrous encapsulation of single fibers in rat models, demonstrated that below a threshold fiber diameter around 6 μm, the proliferation of fibroblasts is reduced on the fiber while it tends to increase when larger diameters are used [20, 21]. The fibrous capsule thickness increases more significantly, especially above 11 μm. It seems that a curvature threshold effect triggers cell signaling, which limits fibroblast proliferation below 11 μm. This phenomenon has been observed in other studies as well. Considering PMMA electrospun scaffolds, Liu et al. brought out that cells tend to align along fibers larger than 1 μm, limiting their proliferation . Actually, the only way for fibroblasts to fix on smaller fibers is to spread along them in a one-dimensional (1D) configuration. The cell is not able to wrap around the fiber because the process requires exaggerated deformation energy. This appears to be true for fibers above 1 μm. Under that size, the phenomenon is reversed. Hodgkinson compared nanofibrous and microfibrous silk scaffolds and showed that cell proliferation is increased on nanofibers because cells can wrap around the fibers and connect with neighboring cells bridging the pores of the fibrous network . These results are confirmed by Tian et al.  who investigated the behavior of aligned poly(glycolic)acid (PGA) fiber mats made from fibers between 500 nm and 10 μm. The results show that NIH3T3 fibroblasts adhere better to fiber with nanoscale diameters than to 10-μm fibers. This is consistent with other literature findings showing that cells growing on the smaller fibers (nanoscale) exhibit an architecture that mimics cells growing on a glass surface [17, 23, 27]. Conversely, while proliferation is limited on fibers between 1 μm and 10 μm, the curvature of the 37-μm monofilament (MONO sample) is large enough for the fibroblast to spread in a nearly two-dimensional (2D) shape, as was demonstrated by Edwards et al. . The authors tested PEEK-woven constructions made from monofilament (100 μm) and multifilament material (10 μm) in contact with fibroblasts and showed that fibroblasts tend to easily focalize on the monofilament surface due to the large curvature radius.
As can be seen in Table 1, the monofilament material is also characterized by limited roughness compared to the multifilament (5.2 μm vs. 17.7 μm). Various works in the literature underline how roughness can limit cell proliferation. Kim et al.  studied the encapsulation of titanium-coated epoxy surfaces implanted subcutaneously in rat models and showed that textured surfaces exhibited thinner fibrous encapsulation levels compared to polished ones. This result is in accordance with other works performed on striped surfaces. Brydone et al.  showed that fibroblasts have difficulties to bridge stripes separated with a 12 μm distance, in particular, when the depth of the grooves is increased. It is assumed in this work that fibroblasts need to be guided to proliferate. On striped surfaces, they tend to stay at the top of the stripes and extend along them, which limits the proliferation. This result is confirmed by Kolind et al.  who studied various textured surfaces. Regarding the monofilament, the limited roughness of the substrate may explain part of the observed proliferation in this work. Moreover, with respect to pore size, the monofilament material is characterized by small pores. It appears that fibroblasts can thus bridge the filaments in that material, connect to surrounding cells and spread on the surface in a 2D fashion as was described by other groups [8, 15].
Other features like fiber alignment seem to influence the fibrous capsule development as well. Cao et al.  tested in vitro the fibrotic encapsulation of aligned fiber and random polycaprolactone (PCL) nanofibrous scaffolds. They reported a decrease of around 40% of the capsule thickness in the random mat, suggesting that bridging between fibers is easier in an aligned fibrous construction. Zhong et al.  obtained similar results with aligned scaffolds. A yarn like the one used in the MULTI sample can be considered as an aligned fibrous construction. However, the 10 μm filament being characterized by a low curvature radius, the curvature effect described previously, becomes predominant despite the alignment and proliferation being limited on the multifilament material.
It is also interesting to observe that the porosity is not as influential on tissue ingrowth as pore size. Despite the largest porosity in the MULTI sample (71% vs. 56% for the MONO sample), the material undergoes the lowest tissue ingrowth. Actually, as can be seen in Figure 6, with the size of the pores being larger in the monofilament substrate, the permeability of the MONO sample is larger (3920 ml/cm2/s vs. 790 ml/cm2/s). This facilitates the proliferation of fibroblasts . The fibrotic tissue penetrates trough the thickness of leaflet material, creating a bridging between both sides of the leaflets. On the contrary, the multifilament is less permeable with no apparent pores, which limits the bridging effect even if fibrotic tissue tends to penetrate between fibers, as can be observed in Figure 6.
The phenomenon seems to be reversed in the calendered substrate. Actually, the calendering process closes the pores in the material, which leads to limited permeability (261 ml/cm2/s). Less fibrosis should go through the thickness and proliferate. However, the calendering process tends to melt the fibers together and flatten the yarn, as can be seen in Figure 1. This reduces the roughness of the substrate (5.2 μm vs. 17.7 μm for the non-calendered substrate), which tends to be considered as a flat membrane with a very large curvature radius. Fibroblasts can spread more easily on such a surface (1204 μm on average in Figure 5).
One remarkable result is that whatever the fabric, what is considered the fibrotic layer is thicker on the insertion site of the leaflet (zone 1). Actually, this is the zone where the valve is fixed to the ring and the ring to the aortic root. Fibrotic tissue formation is initiated in that wounded zone first. The phenomenon is particularly visible in the MULTICALEND sample, characterized by a 2086-μm fibrosis thickness in zone 1. Proliferation is easier in that zone compared to the leaflet length which is more mobile. This is consistent with the fact that fibroblasts develop more static substrates.
With regard to calcium deposits, it is observed from the comparison that there is a good correlation between the roughness values that were measured and the calcification level that was identified on the explanted valves. One can observe in Figure 7 that the average calcium level is minimum (7.5 μm/mg) for the monofilament characterized by the smallest roughness value (SMD=5.2 μm). On the contrary, the maximum value (58.6 μm/mg) is obtained for the roughest multifilament substrate (SMD=17.7 μm). Calendered material is located in between (32.3 μm/mg) with SMD=8.4 μm. One advantage of the calendered material is that the space between the fibers is reduced with the calendering process. It is, therefore, more difficult for calcium to penetrate between the fibers, as is observed in Figure 7. Globally, compared to biological tissue, textile presents the advantage of being characterized by only extrinsic calcification as a synthetic material. However, the roughness of the material offers specific locations that behave as potential calcium nucleation sites, as was described by various authors [18, 22]. This process may stiffen the leaflet and jeopardize the lifetime of the device.
In this work, the behavior of textile material used as valve leaflet material was studied in vivo in sheep models over a 6-month period. Three plain weave woven fabrics were compared: multifilament, calendered multifilament and monofilament. All animals were euthanized within the 6-months period. The results reveal that, whatever the considered substrate, fibrotic tissue develops mainly in the leaflet insertion zone level. Moreover, the non-calendered multifilament material is the less-fibrosis-promoting substrate. The monofilament material is characterized by a large filament curvature radius as well as large pore sizes, which tend to promote fibroblast proliferation. Regarding the calendered material, it appears that despite the reduced pore size, the calendering process flattens the yarns, creating some large flat welded zones providing the substrate with limited roughness. Basically, the material generates a foreign body reaction similar to that of a continuous membrane. With respect to calcification rates, the monofilament seems to present an advantage due to the limited amount of potential nucleation zones. Based on the obtained results, future work must focus on developing a hybrid textile material that combines good anti-fibrotic features while being calcification-resistant.
Amoroso NJ, D’Amore A, Hong Y, Rivera CP, Sacks MS, Wagner WR. Microstructural manipulation of electrospun scaffolds for specific bending stiffness for heart valve tissue engineering. Acta Biomater 2012; 8: 4268–4277. CrossrefPubMedWeb of ScienceGoogle Scholar
Brydone AS, Dalby MJ, Berry CC, Dominic Meek RM, McNamara LE. Grooved surface topography alters matrix-metalloproteinase production by human fibroblasts. Biomed Mater 2011; 6: 035005. PubMedWeb of ScienceCrossrefGoogle Scholar
Cao H, McHugh K, Chew SY, Anderson JM. The topographical effect of electrospun nanofibrous scaffolds on the in vivo and in vitro foreign body reaction. J Biomed Mater Res A 2010; 93: 1151–1159. PubMedWeb of ScienceGoogle Scholar
Cribier A, Eltchaninoff H, Bash A, et al. Percutaneous transcatheter implantation of an aortic valve prosthesis for calcific aortic stenosis, first human case description. Circulation 2002; 106: 3006–3008. CrossrefPubMedGoogle Scholar
Dvir D. Half of transcatheter heart valves show degeneration within 10 years of TAVI. EuroPCR 2016. Session: Tuesday 17 May 11.20-12.20, Late-breaking trials, registries and innovations. Google Scholar
Hodgkinson T, Yuan XF, Bayat A. Electrospun silk fibroin fiber diameter influences in vitro dermal fibroblast behavior and promotes healing of ex vivo wound models. J Tissue Eng 2014; 5: 2041731414551661. PubMedGoogle Scholar
Kiefer P, Gruenwald F, Kempfert J, et al. Crimping may affect the durability of transcatheter valves: an experimental analysis. Ann Thorac Surg 2011; 92: 155–160. CrossrefWeb of SciencePubMedGoogle Scholar
Kim H, Murakami H, Chehroudi B, Textor M, Brunette D. Effects of surface topography on the connective tissue attachment to subcutaneous implants. Int J Oral Maxillofac Implants 2006; 21: 354–365. PubMedGoogle Scholar
Kolind K, Dolatshahi-Pirouz A, Lovmand J, Pedersen FS, Foss M, Besenbacher F. A combinatorial screening of human fibroblast responses on micro-structured surfaces. Biomaterials 2010; 31: 9182–9191. CrossrefWeb of SciencePubMedGoogle Scholar
Kucinska-Lipka J, Gubanska I, Janik H, Sienkiewicz M. Fabrication of polyurethane and polyurethane based composite fibres by the electrospinning technique for soft tissue engineering of cardiovascular system. Mat Sc Eng C 2015; 46: 166–176. CrossrefWeb of ScienceGoogle Scholar
Li M, Guo Y, Wei Y, MacDiarmid AG, Lelkes PI. Electrospinning polyaniline-contained gelatin nanofibers for tissue engineering applications. Biomaterials 2006; 27: 2705–2715. CrossrefPubMedGoogle Scholar
Liao KK, Li X, John R, et al. Mechanical stress: an independent determinant of early bioprosthetic calcification in humans. Ann Thorac Surg 2008; 86: 491–495. PubMedWeb of ScienceCrossrefGoogle Scholar
Liu Y, Ji Y, Ghosh K, Clark RA, Huang L, Rafailovich MH. Effects of fiber orientation and diameter on the behavior of human dermal fibroblasts on electrospun PMMA scaffolds. J Biomed Mater Res A 2009; 90: 1092–1106. Web of SciencePubMedGoogle Scholar
Sanders JE, Cassisi DV, Neumann T, et al. Relative influence of polymer fiber diameter and surface charge on fibrous capsule thickness and vessel density for single-fiber implants. J Biomed Mater Res A 2003; 65: 462–467. PubMedGoogle Scholar
Sanders JE, Stiles CE, Hayes CL. Tissue response to single-polymer fibers of varying diameters: evaluation of fibrous encapsulation and macrophage density. J Biomed Mater Res 2000; 52: 231–237. PubMedCrossrefGoogle Scholar
Schoen F, Levy R. Pathological calcification of biomaterials. Biomaterials Science: an introduction to materials in medicine. 2012. Chapter II/4.5. Google Scholar
Takahashi Y, Tabata YJ. Effect of the fiber diameter and porosity of non-woven PET fabrics on the osteogenic differentiation of mesenchymal stem cells. J Biomater Sci Polym Ed 2004; 15: 41–57. CrossrefPubMedGoogle Scholar
Teixeira NA, Platt MM, Hamburger WJ. Mechanics of elastic performance of textile materials: Part XII: relation of certain geometric factors to the tear strength of woven fabrics. Text Res J 1955; 10: 838–861. Google Scholar
Vaesken A, Heim F, Chakfe N. Fiber heart valve prosthesis: influence of the fabric construction parameters on the valve fatigue performances. J Mech Behav Biomed Mater 2014; 40: 69–74. Web of ScienceCrossrefPubMedGoogle Scholar
Yang F, Murugan R, Wang S, Ramakrishna S. Electrospinning of nano/micro scale poly(L-lactic acid) aligned fibers and their potential in neural tissue engineering. Biomaterials 2006; 26: 2603–2610. Google Scholar
Zegdi R, Ciobotaru V, Noghin M, et al. Is it reasonable to treat all calcified stenotic aortic valves with a valved stent? Results from a human anatomic study in adults. J Am Coll Cardiol 2008; 51: 579–584. CrossrefPubMedWeb of ScienceGoogle Scholar
Zhong S, Teo WE, Zhu X, Beuerman RW, Ramakrishna S, Yung LY. An aligned nanofibrous collagen scaffold by electrospinning and its effects on in vitro fibroblast culture. J Biomed Mater Res A 2006; 79: 456–463. PubMedGoogle Scholar
About the article
Published Online: 2017-05-03
Published in Print: 2018-06-27
Research funding: This work was partly supported by the French ANR: ANR-12-EMMA-0001.
Conflicts of interest: The authors declare no conflicts of interest.
Informed consent: Informed consent is not applicable.
Ethical approval: The research related to the animal use complies with all the relevant national regulations and institutional policies.