Critical seeding density improves properties and translatability of self-assembling anatomically shaped knee menisci

Abstract

A recent development in the field of tissue engineering is the rise of all-biologic, scaffold-free engineered tissues. Since these biomaterials rely primarily upon cells, investigation of initial seeding densities constitutes a particularly relevant aim for tissue engineers. In this study, a scaffold-free method was used to create fibrocartilage in the shape of the rabbit knee meniscus. The objectives of this study were: (i) to determine the minimum seeding density, normalized by an area of 44 mm², necessary for the self-assembling process of fibrocartilage to occur, (ii) examine relevant biomechanical properties of engineered fibrocartilage, such as tensile and compressive stiffness and strength, and their relationship to seeding density, and (iii) identify a reduced, or optimal, number of cells needed to produce this biomaterial. It was found that a decreased initial seeding density, normalized by the area of the construct, produced superior mechanical and biochemical properties. Collagen per wet weight, glycosaminoglycans per wet weight, tensile properties, and compressive properties were all significantly greater in the 5 million cells per construct group as compared to the historical 20 million cells per construct group. Scanning electron microscopy demonstrated that a lower seeding density results in a denser tissue. Additionally, the translational potential of the self-assembling process for tissue engineering was improved though this investigation, as fewer cells may be used in the future. The results of this study underscore the potential for critical seeding densities to be investigated when researching scaffold-free engineered tissues.

1. Introduction

Fibrocartilage, a connective tissue found in various parts of the human body, frequently experiences degeneration, leading to conditions as diverse as osteoarthritis, difficulty eating and speaking, and back pain [1–3]. Fibrocartilaginous tissues, such as the knee meniscus, temporomandibular joint disc, and intervertebral disc, are responsible for withstanding significant mechanical loads in vivo, by virtue of the biochemical content of their native extracellular matrix (ECM). In particular, the collagens of the native ECM are responsible for conferring tensile strength and stiffness to these tissues, and the glycosaminoglycans (GAGs) are thought to provide compressive strength and stiffness [4, 5].

The knee meniscus represents a prime candidate for research in tissue replacement or regeneration. Meniscal lesions alone are responsible for nearly 1 million surgeries annually in the U.S. and Europe, and are the root cause of the most frequent procedures practiced by orthopedic surgeons [5, 6]. The knee meniscus displays a specific, functionally important wedge shape, as well as mechanical anisotropy between its circumferential and radial directions [7–9]. This very specific organization allows the knee meniscus to effectively transfer loading from the distal femur to the tibial plateau [10]. Thus, the biomechanical functions of the knee meniscus are inexorably tied to the structure of tissue. Importantly, the knee meniscus and all fibrocartilages display a fundamental inability to mount a healing response after structural damage is sustained due to trauma, aging, or disease, and thus represent ideal candidates for replacement by tissue engineering [10, 11].

Recently, our lab has pioneered a scaffoldless approach, termed the self-assembling process in tissue engineering, which relies upon cadherin mediated biophysical interactions that occur early during seeding of cells in a high density culture [12]. Similar studies, focused on engineering scaffold-free liver and optic tissues, have also reported all-biologic constructs displaying the presence of cadherins early during tissue culture [13, 14]. In the self-assembling process specifically, the use of a non-adherent seeding well and the upregulation of N-cadherin early during culture result in the minimization of free energy during tissue formation [15, 16]. Despite the fact that cells are not seeded within a biomaterial, the seeding well serves as a mold during this process, providing support to the developing neotissue and helping determine its final shape. This approach, first demonstrated for articular cartilage, has also been extended to the generation of fibrocartilage [17]. In particular, this has resulted in success via the creation of all-biologic constructs unhindered by intervening synthetic scaffolds [18–21]. Additionally, self-assembling constructs may lack the immunogenicity and potential toxic degradation byproducts of some biomaterial scaffolds. Finally, self-assembling constructs could avoid stress-shielding in vivo and have the potential for more seamless integration with native tissue [16]. However, despite these successes, additional investigations are necessary to understand and enhance structure-function relationships involving ECM content/organization, mechanical properties, and construct size and geometry.

The number of cells needed to form a construct is a non-trivial issue in tissue engineering and regenerative medicine. While one may infer that greater cell numbers may imply a larger tissue construct, this is not always the case [22, 23], and there is no guarantee that the properties of engineered tissue constructs scale linearly as construct size increases. Furthermore, several musculoskeletal tissues, including fibrocartilage, possess relatively low cellularity. From a translational perspective, generating a tissue construct requires a large number of cells, whose acquisition from a patient or donor may or may not be feasible, especially if an investigator is employing an autologous or allogeneic approach. Interestingly, altering seeding density has repeatedly been reported to change or improve the quality of tissue constructs [24–26]. Recent work in our laboratory has demonstrated the use of this for self-assembling articular cartilage [23], although the generation of shape-specific knee meniscus fibrocartilage and the relationship of its biochemical and biomechanical properties to seeding density remain unknown.

The objectives of this study were to (i) determine the minimum seeding density, normalized by an area of 44 mm², necessary for the self-assembling process of fibrocartilage to occur, (ii) examine relevant biomechanical properties of engineered fibrocartilage, such as tensile and compressive stiffness and strength, and their relationship to seeding density, and (iii) identify a reduced, or optimal, number of cells needed to produce this biomaterial. Self-assembling fibrocartilage constructs were seeded with varying numbers of cells in the shape of the native knee meniscus using non-adherent agarose molds of constant size and cultured for four weeks. At the end of culture, construct properties were assessed. It was hypothesized that (i) a threshold seeding density existed, below which the self-assembling process would not occur, (ii) construct biomechanical and biochemical properties would increase with greater seeding densities and eventually plateau, and (iii) constructs could be seeded with fewer cells than the historical density of 20 million per construct while possessing equivalent or greater biochemical and biomechanical properties.

2. Materials and methods

2.1 Cell isolation

Bovine articular chondrocytes and meniscal cells were harvested from the legs of four 8-week-old calves (Research 87) [27, 28]. Chondrocytes were obtained from the entire surface of the distal femur, and meniscal cells were obtained from the meniscus after trimming out the outer meniscal rim. The tissues were minced into approximately 1 mm³ pieces. Cartilage was digested in 0.2% (w/v) Worthington’s ‘collagenase type II′ enzyme mixture (Worthington) in base medium (Dulbecco’s modified Eagle’s medium (DMEM) (Invitrogen) with 3% (v/v) fetal bovine serum (FBS) (Benchmark), 1% (v/v) non-essential amino acids (NEAA) (Invitrogen), and 1% penicillin/streptomycin/fungizone (PSF) (Lonza) for 18 hours. Meniscal tissue was digested in a 0.25% (w/v) pronase (Sigma) in base medium for 1 hour, followed by a 0.2% (w/v) collagenase type II (Worthington) in base medium for 18 hours. After digestion, the tissues were washed and centrifuged four times using PBS and filtered through 70 μm cell strainers. Chondrocytes were frozen in DMEM with 20% (v/v) dimethylsulfoxide (DMSO) and 10% (v/v) FBS, and kept in liquid nitrogen until resuspension with freshly isolated meniscal cells for seeding, as described previously [12, 17].

2.2 Construct seeding and sample preparation

To generate self-assembling fibrocartilage, cells were seeded in non-adherent meniscus-shaped agarose wells formed from a rapid prototyped (ZPrinter) wedge-shaped ring mold, as described previously [29]. Meniscus-shaped ring wells were fabricated with major axis of 9.5 mm and minor axis of 7 mm, and an inner post with major axis of 4 mm and minor axis of 2.5 mm. Depth of each well was 14 mm, with an internal volume of approximately 200 μL. Approximate area of bottom surface of each meniscus shaped construct was 44 mm² [30]. Initial seeding densities were 1.25, 2.5, 5, 10, 15, and 20 million cells per construct. Each construct was seeded within 180 μL of liquid while keeping the dimensions of the well constant and varying total number of cells. To form each construct, chondrocytes and meniscal cells were seeded at a 1:1 ratio in 180 μL of a chondrogenic media formulation described previously [17]. The chondrogenic medium formulation was as follows: DMEM with GlutaMAX (Invitrogen), 100 nM dexamethasone (Sigma), 1% (v/v) PSF (Lonza), 1% (v/v) ITS+ premix (BD biosciences), 50μg/mL ascorbate-2-phosphate (Sigma), 40 μg/mL L-proline, and 100 μg/mL sodium pyruvate (Fischer Scientific). An additional 500 μL of medium was added four hours after seeding. The total volume of medium in each well was changed daily, following an internal pilot study. Media volume was 6 mL during the first week of culture and 2 mL thereafter. All cultures were kept at 37 °C with 10% (v/v) CO₂. After one week, tissue constructs were unconfined from their agarose wells and kept in free floating culture for an additional three weeks. During this time, constructs were cultured in a 24-well culture plate with 400 μL of agarose coated to the bottom of the well to prevent any potential adhesion of the constructs, or cells within the constructs, to the plate. After a total culture time of four weeks, construct gross morphology and wet weight were assessed before division of each construct into pieces for mechanical, biochemical, and histological analysis. Each construct was split in the same manner, with sections being taken from similar areas for each specific test. No heterogeneities were observed. All photographs of constructs for gross morphology, compressive testing, and tensile testing were taken using a Canon 100 mm f/2.8 macro lens.

2.3 Histology

Whole constructs were frozen at −20 °C in HistoPrep (Fisher Scientific) after culture, and additional tissue constructs were taken at days 1, 4, 7, 14, and 22 for comparison. Sections were fixed in formalin prior to staining with Safranin-O/Fast Green for GAG content and Picrosirius Red for collagen content. Other sections were fixed in acetone for immunohistochemistry (IHC) staining of collagen I and collagen II. IHC was performed with the Vectastain ABC and DAB Substrate Kits (Vector Labs) with rabbit anti-human collagen I (US Biologics) and rabbit anti-bovine collagen II antibodies (Cedar Lane Labs). The histological examination encompassed both the vertical (i.e. the full thickness) as well as the horizontal (i.e., encompassing both the major and minor axes) planes of the construct.

2.4 Biochemistry

Biochemical samples were weighed, lyophilized, weighed again for dry weights, and then digested in 125 μg/mL papain (Sigma) for 18 hours at 65 °C. Collagen content was quantified by a modified colorimetric hydroxyproline assay. GAG content was quantified by a Blyscan (Biocolor) assay kit. Cell numbers were quantified by a PicoGreen dsDNA reagent (Invitrogen), using a conversion factor of 7.7 pg DNA/cell.

2.5 Compression testing

Uniaxial, unconfined stress-relaxation was performed on 3 mm diameter punches in order to determine compressive properties. After compressive samples were removed from constructs, each punch was photographed to determine its area using ImageJ software (NIH). Punches were taken perpendicular to the surface of constructs, after which a wedge-shaped cutting jig was used to create flat samples for compressive testing. Samples were then placed in a phosphate-buffered saline (PBS) bath on an Instron uniaxial testing machine (Model 5565, Canton, MA). Sample thickness was determined using a custom program that compressed the sample until a 0.05 N load was reached, at which point the height was recorded. The ramp rate (0.025 mm/s) and height detection load (0.05 N) were deemed appropriate for the samples tested since the load displacement curve during the height detection phase did not show overshooting of the load. Errors in measuring height during this step were estimated to occur due to slight compression of the sample. However, a back-calculation using the stiffness values of the constructs found applied strain during the height detection test to be <4%, ranging from 1.7% to 3.9%. Prior to compressive testing, samples were preconditioned at a 5% strain for 15 cycles at 1 Hz. While submerged in PBS, samples were subjected to stepwise loading at a rate of 1% s⁻¹ to 10% and 20% strain using a non-porous platen. Ramp duration was chosen such that the material tested would not appear more elastic than is actually the case. Specifically, ramp duration during loading was found to be similar to one-eighth of relaxation time for a viscoelastic material as suggested [31]. Following the application of 10% and 20% strains, platen position was maintained for 200 s and 450 s respectively, after which the force vs. time curve displayed negligible change, signifying the sample had relaxed. A custom program in MATLAB was used to determine the relaxation modulus and instantaneous modulus of each tested sample using a Kelvin solid model described previously [21, 32, 33] and Eq. (1) and Eq. (2) for viscoelastic materials presented below:

$$\mathrm{\sigma }(\mathrm{t})=3/2\ast {\mathrm{E}}_{\mathrm{r}}({\mathrm{u}}_{\mathrm{i}})/\mathrm{z}\{1+({\mathrm{\tau }}_{\mathrm{\sigma }}/{\mathrm{\tau }}_{\mathrm{e}}-1)exp^[-(-\mathrm{t}-{\mathrm{t}}_{\mathrm{i}})/{\mathrm{\tau }}_{\mathrm{e}}]\}$$ (1)

$$\text{Ei}={\mathrm{E}}_{\mathrm{r}}-({\mathrm{\tau }}_{\mathrm{\sigma }}-{\mathrm{\tau }}_{\mathrm{e}}){\mathrm{E}}_{\mathrm{r}}/{\mathrm{\tau }}_{\mathrm{e}}$$ (2)

Where σ(t) represents strain at time t, E_r represents relaxation modulus, E_i represents instantaneous modulus, u_i represents strain at the start of step-wise compression, z represents tissue height, τ_σ and τ_e represent time constants, and t_i represents the time at which peak loading occurs during stress-relaxation.

2.6 Tensile testing

Samples were taken in the circumferential and radial directions of the meniscus-shaped constructs, and cut into dogbone shapes to ensure failure at the desired location. A premade cutting jig was used to uniformly attain a goal gauge length of 2.3 mm and goal sample width of 0.6 mm. Tests were conducted on dogbone-shaped specimens according to ASTM standards [34]; failure position was visually monitored during testing. Samples that did not fail at the mid-point were discarded. Grip-to-grip strain was not used, as gripping the constructs directly crushed them, yielding failure at the grips. Instead, constructs were glued to paper tabs, which were then gripped in the testing machine. All samples included in the analysis failed at the narrowest portion of the dogbone. Photographs of these samples were then used to measure the dimensions of the strained area using ImageJ. Tissue samples were adhered to strips of paper using cyanoacrylate glue, and secured to the grips of a tensile testing machine (TestResources). Samples were strained at a rate of 1% of gauge length s⁻¹, while displacement and load were recorded until failure of the sample at the dogbone. A relatively small strain rate was used such that the viscous contribution to tissue biomechanical properties would be relatively small, allowing for the viscoelastic samples to exhibit a linear region, as shown previously [35]. This allowed for the calculation of an equivalent Young’s modulus. Data were analyzed with a custom program in MATLAB to determine the Young’s modulus from the linear region of the strain curve and the ultimate tensile strength (UTS) at peak stress, as described previously [36]. Young’s modulus and UTS were calculated using Eq. (3) and Eq. (4) presented below:

$${\mathrm{E}}_{\mathrm{Y}}=(\mathrm{F}/\mathrm{A})/(\mathrm{\Delta }\mathrm{L}/\mathrm{L})$$ (3)

$$\text{UTS}={\mathrm{F}}_{max}/\mathrm{A}$$ (4)

Where F represents uniaxial force on the sample in the direction of testing, A represents initial cross-sectional area of the sample, ΔL represents the initial gauge length of the sample, L represents change in this gauge length, and F_max represents the peak force attained during loading.

2.7 Preparation of samples for scanning electron microscopy

Construct sections for scanning electron microscopy were stored at 4° C in cacodylate buffer with 5% (w/v) sucrose after construct culture. Sections were submerged in solutions with increasing ethanol concentrations up to 100% ethanol prior to preparation for imaging. Prior to imaging, sections were critical point dried and gold sputter-coated. For each sample in a group, 8 images were taken from various locations around the tissue, with quantification measurements averaging values for all images within in a group. For density measurements, the threshold function in ImageJ was used to determine percent area of ECM versus background in an image. Percent area was then averaged over all images to calculate ECM density. A Phillips XL30 TMP scanning electron microscope was used to image samples.

2.8 Statistical analysis

For each biochemical and biomechanical test, n = 7 samples were used. Results were analyzed with a one-factor ANOVA followed by a Tukey’s HSD post-hoc test when merited (p < 0.05). A two-factor ANOVA was used in tensile test results, with one factor being mechanical orientation, and the second factor being seeding density. In figures detailing results from biochemical and biomechanical assays, groups denoted without the same letter are statistically different. All data are presented as means ± standard deviations.

3. Results

3.1 Gross morphology

Constructs with seeding densities lower than 5 million cells did not form contiguous tissue that retained the form of its wedge-shaped mold when manipulated. Gross morphological images of constructs taken at the end of four weeks of culture are found in Figure 1. Thickness of constructs is shown in Table 1. Construct thickness was positively correlated with cell numbers per construct (as estimated by DNA content) as shown in Figure 1.

Figure 1. Initial seeding densities and resultant morphologies of self-assembling constructs.

In general, constructs (A) maintained a wedge-shaped profile, with a circumferential direction (dashed line) and radial direction (solid line), similar to the native rabbit knee meniscus (B). Construct morphologies during culture (C) demonstrates that constructs seeded with 1.25 or 2.5 million cells per construct did not produce continuous tissue. Each construct was initially grown in a ring shape, after which it was cut in half to create the two compartments of the knee meniscus. Additionally, the 5 million cells per construct group contracted slightly during culture. 20 million cells per construct represented historical seeding density. Final DNA content of constructs (D) indicates a relationship with initial seeding density. Plot of construct final DNA content versus final construct thickness (E). Groups lacking a letter in common are significantly different (p < 0.05).

Table 1.

Dimensions, weights, and hydrations of engineered menisci

Group (cells seeded)	Major diameter (mm)	Minor diameter (mm)	Thickness (mm)	Weight (mg)	Hydration (%)
1.25 million	†	†	0.13 ± 0.04D	1.79 ± 0.40D	84.9 ± 3.4B
2.5 million	7.58 ± 0.59B	4.92 ± 0.15B	0.16 ± 0.05D	4.04 ± 0.54D	87.4 ± 1.0B
5 million	13.98 ± 1.19A	6.18 ± 0.71B	0.65 ± 0.15C	26.49 ± 3.78 C	81.3 ± 2.1A
10 million	13.44 ± 0.81A	9.61 ± 1.39A	1.17 ± 0.27B	85.15 ± 5.61B	84.3 ± 2.0AB
15 million	13.35 ± 0.64A	9.83 ± 0.36A	1.22 ± 0.31B	83.87 ± 17.79B	86.2 ± 1.2B
20 million	13.62 ± 0.36A	9.98 ± 0.46A	1.64 ± 0.21A	102.50 ± 9.35A	87.5 ± 1.4B

Construct morphology was quantified after culture. Diameters of constructs seeded with 2.5 million cells were less than all other groups. Thickness and construct weight increased with greater cell numbers per construct. Tissue in the 5 million cells per construct group had less water content than all other groups except the 10 million cells per construct group. Groups lacking a letter in common are significantly different (p < 0.05). Diameters of constructs seeded with 1.25 million cells were not assessed († = constructs diameters distorted).

3.2 Histology

Histological sections stained for ECM biochemical content are found in Figure 2. Staining for GAG revealed a peak in Safranin-O staining at 5 million cells per construct. Similarly, Picrosirius Red staining for collagen showed a similar trend, with staining intensity greatest at 5 million cells per construct and decreasing as seeding density increased to 20 million cells per construct (historical density). All groups stained positive for collagen type I and collagen type II, as found in native meniscus fibrocartilage. Tissue from the 5 million cells per construct group maintained this trend of staining more intensely than other groups for Safranin-O and Picrosirius Red throughout various time-points (days 1, 4, 7, 14, and 22) during culture, as shown in Figure 3 and Figure 4.

Figure 2. Self-assembling constructs appear similar to native fibrocartilage under histological and immunohistochemical staining.

Tissue in all groups stained for collagen, GAG, and collagen types I and II, as found in native fibrocartilage. The 5 million cells per construct group displayed the most intense staining, followed by the 10 million cells per construct group. Staining became less intense in constructs seeded with more cells. Constructs seeded with 1.25 and 2.5 million cells produced thin tissues with fewer GAGS.

Figure 3. Self-assembling constructs seeded with 5 or 10 million cells produce abundant collagen in culture.

Picrosirius red staining for total collagen showed increasing production over time, with the greatest amount of collagen being produced within two weeks of construct seeding. Constructs seeded with 5 or 10 million cells displayed the most collagen throughout culture. Constructs seeded with 1.25 or 2.5 million cells produced thin tissue. Constructs seeded with 15 or 20 million cells (historical seeding density) produced large tissues, which displayed relatively sparse collagen staining.

Figure 4. Self-assembling constructs seeded with 5 or 10 million cells produce GAG more quickly in culture.

Safranin-O staining for GAG showed constructs seeded with 5 million or 10 million cells produced GAG more quickly than other groups, starting within day 7 of culture. GAG production increased over time in all groups, plateauing between day 14 and day 21 of culture. Constructs seeded with 1.25 or 2.5 million cells produced thin tissue, and GAG production began relatively later during culture. Constructs seeded with 15 or 20 million cells (historical seeding density) produced large tissues with relatively sparse GAG staining.

3.3 Biochemistry

Biochemical data for collagen and GAG content were calculated per wet weight as well as per construct, as shown in Figure 5. Values from the 5 million cells per construct group were significantly greater than values from the historical 20 million cells per construct group. It is notable that seeding 5 million cells per construct resulted in significantly lower construct hydration. Increasing seeding density produced greater amounts of GAG and collagen per construct until a plateau was reached with the 10 million cells per construct group. When looking at ECM normalized to wet weight, significant differences among groups of different seeding densities were noted for both GAG and collagen production, with constructs at a seeding density of 5 million cells per construct having the highest ECM production.

Figure 5. Biochemical content is greater in self-assembling constructs seeded with 5 or 10 million cells.

In terms of total ECM content, total collagen per construct (A) and total GAG per construct (B) increased with greater initial seeding densities until a plateau was reached at 10 million cells per construct. When normalized to wet weight, ECM content was greatest in constructs seeded with 5 or 10 million cells. Collagen per wet weight (C) was greatest in the 5 million cells per construct group, and GAG per wet weight (D) was greatest in the 5 million cells per construct group followed by the 10 million cells per construct group. Groups lacking a letter in common are significantly different (p < 0.05).

3.4 Biomechanical properties

Results of biomechanical tests are displayed in Figure 6. Mechanical properties of the 5 million cells per construct group were significantly greater than those of the historical 20 million cells per construct group. Constructs seeded with 5 million cells per construct had the highest Young’s modulus (2.4- to 11.8-fold of other seeding density groups) and UTS (2.7- to 6.7-fold of other seeding density groups) in both the circumferential and radial directions. In terms of anisotropy, constructs had higher Young’s moduli and UTS in the circumferential direction than in the radial direction. Constructs seeded with 5 million cells per construct exhibited the highest instantaneous (1.5- to 2.4-fold of other seeding density groups) and relaxation (1.3- to 2.4-fold seeding density groups) moduli at both 10% and 20% strain.

Figure 6. Neomenisci exhibit anisotropy, and constructs seeded with 5 million cells possess greater mechanical properties.

Tensile stiffness as assessed by Young’s modulus (A) was greater in both circumferential and radial directions for the 5 million cells per construct group than the historical 20 million cells per construct group. Similarly, UTS (B) was greatest in both circumferential and radial directions for the 5 million cells per construct group. Ring-shaped self-assembling constructs displayed mechanical anisotropy that recapitulated the anisotropy seen in native tissue. Across all groups, tensile properties measured in the circumferential direction of constructs were significantly greater than those measured in the radial direction (denoted by α andβ, respectively). Furthermore, compressive stiffness as measured by instantaneous (C) was greater in the 5 million cells per construct group than other groups. Additionally, compressive stiffness as measured by relaxation modulus (D) was greater in the 5 million cells per construct group than other groups. Groups lacking a letter in common are significantly different (p < 0.05). Constructs seeded with fewer than 5 million cells were not assessed († = not mechanically testable).

3.5 Scanning electron microscopy

When viewed under scanning electron microscopy, construct sections displayed a network of collagen fibers. The fiber network found in constructs from the 5 million cells per construct group appeared slightly denser than the fiber network imaged at higher seeding densities. When quantified, the 5 million cells per construct group displayed a fiber density of 58% by area, as compared to 54%, 54%, and 49% for the 10, 15, and 20 million cells per construct groups, respectively.

4. Discussion

This study investigated the effects of varying initial seeding density in a tissue-engineered, anatomically shaped, fibrocartilaginous biomaterial. Experimental results supported the hypotheses driving this study: (i) constructs seeded with fewer than 5 million cells were not fully assayable or capable of being manipulated, and did not form contiguous engineered tissue, (ii) collagen and GAG content per construct was greater at higher initial seeding densities, and (iii) histological, biomechanical, and biochemical assays demonstrated that a lower seeding density, as normalized to initial area, was superior to the historical seeding density of 20 million cells per construct. This is the first study to report the beneficial effects of lower seeding densities in scaffold-free self-assembling fibrocartilage, and the first to show that lower seeding densities can generate biomechanically superior, anisotropic, and scaffold-free engineered knee meniscus tissue.

In assessing construct mechanical anisotropy, it was found that circumferential tensile properties were greater than radial tensile properties across all groups tested. For example, the historical 20 million cells per construct group displayed a circumferential Young’s modulus 450% greater than the radial Young’s modulus. Similarly, in the 5 million cells per construct group, UTS in the circumferential direction was 150% greater than UTS in the radial direction. These findings are consistent with previously published data that growth of scaffold-free tissue around a central post can result in mechanical and organizational anisotropy [29, 37]. To understand why this occurs, it is useful to imagine developing neotissue as a pre-stressed collagen network. Collagen-rich tissues engineered with fibroblasts, chondrocytes, and/or fibrochondrocytes have all been reported to contract during culture [38–40]. Since each knee meniscus construct is grown along a central post, inwards contraction is instead redirected towards contraction along this central post, resulting in anisotropic circumferential fiber alignment. Anisotropy appears to have been generated in this fashion in the engineered tissues assayed. However, a significantly dense and large enough engineered tissue must be present for this effect to occur. Thus, towards enhancing anisotropy in a clinically relevant-sized tissue, future work could examine the relationship between tissue contraction, the variation of initial seeding density, and the use of ECM modulating agents such as chondroitinase ABC or lysyl oxidase.

The data from this study suggest that a minimum threshold seeding density exists, below which the self-assembling process will not take place. This was reflected in the gross morphology of constructs seeded at lower densities, with tissue that was thin, easily compromised, and unable to maintain its shape under its own weight. These differences in construct manipulability and morphology corroborate with earlier results reporting the poor quality of self-assembling articular cartilage constructs seeded at low densities [23]. In addition to these grossly observable differences, ECM content dropped dramatically in groups seeded with the fewest cells (1.25 and 2.5 million per construct), as compared to the best group (5 million cells per construct). The low seeding density constructs possessed less than 4% of the GAG per construct and less than 9% of the collagen per construct found in the superior 5 million cells per construct group. Additionally, final DNA content in these groups was just 3–15% of the final DNA content in the group seeded with 5 million cells. One reason for this absence of satisfactory tissue formation below a threshold seeding density may be a lack of cadherin-initiated downstream signaling. As demonstrated by earlier work, self-assembling cartilage displays high expression of N-cadherin within the first few days of culture after seeding, which allows for cell-to-cell attachment through energy minimization and tissue maturation reminiscent of native cartilage [15]. Close cell-to-cell contact mimics early stages of morphogenesis [41], and, thus, cadherin mediated downstream signaling may be crucial for satisfactory tissue formation to occur. Therefore, inadequate cell numbers below a threshold seeding density may lead to inferior construct properties partially due to a lack of cadherin-based signaling.

The results from this study increase the translational potential of self-assembling meniscus-shaped constructs. First, the results indicate that contiguous self-assembling fibrocartilage can be generated using lower seeding densities than previously demonstrated [29, 30, 42]. This overcomes a significant feasibility hurdle in the generation of a tissue-engineered meniscus, as isolation of large numbers of cells is impractical, and previous literature has reported that meniscus cells may alter their phenotype dramatically during passaging [10, 43]. Additionally, compressive properties of the tissue grown in this study, with values ranging from 100–600 kPa, are comparable to the values of compressive properties found in native meniscus [44]. Young’s modulus for constructs grown in this study, lacking any growth factor treatment, reaches 3 MPa, which approaches the lower end of tensile stiffness measured for native meniscus tissue [44]. Taken together, the increased mechanical and biochemical properties, along with the lower cell number requirements, highlight the significance of growing self-assembling and other engineered tissues at appropriate initial seeding densities.

The search for a critical seeding density per initial construct area among the groups tested, resulting in tissue with the best mechanical and biochemical properties, can be viewed as a constrained minimization. Although total biochemical content per construct increased with greater initial seeding densities, tensile properties, compressive properties, and biochemical content per wet weight were greatest in engineered tissue seeded with 5 and 10 million cells per construct. A functionality index for self-assembling engineered cartilage has been described previously, and can be calculated using these parameters [20]. Whereas a result of 1 indicates that a construct possesses values equivalent to native tissue, functionality index values of 0.69, 0.52, 0.49, and 0.39 were calculated for the 5, 10, 15, and 20 million cells per construct groups, respectively. The 5 million cells per construct group displayed the greatest functionality index. Moreover, minimizing further constraints relevant to translation (e.g., construct immunogenicity, variability of construct geometry) while maintaining or increasing this functionality index represents an area for future work.

In this study, segregated fibrochondrocytes and articular chondrocytes were used in co-culture. Although the engineered tissue displayed collagen type I, collagen type II, and GAG, as found in native tissue, the constructs did not recapitulate the zonal architecture of the knee meniscus. This leads to certain limitations, as the tissue derives specific functions from the biomechanical properties and thus ECM present in each zone. In particular, the inner zone of the meniscus has relatively greater GAG content and compressive properties, while the outer zone of the tissue has relatively greater collagen type I and tensile properties [10]. These differences result from the variations in cell types in the meniscus [35]. Prior work has used multiple cell types to recapitulate these zonal variations in self-assembling meniscus tissue [36], although this study did not examine the use of different seeding densities of separate seeding densities. In the future, these techniques and others promoting cell sorting or segregation should be explored in conjunction with various seeding densities to achieve zonally variant meniscus constructs with anisotropic mechanical and biochemical properties.

The correlation of final construct dimensions with initial seeding densities also holds implications for tissue engineers to create scaffold-free tissue of controlled thickness, and/or multi-layered constructs. Importantly, this may be due to limitations in nutrient supply, especially for the groups seeded with 10, 15, or 20 million cells per construct. Future studies should continue to examine the role of varying nutrient supply and/or frequency of replenishment in scaffold-free and scaffold-based engineered tissues. Additionally, due to the relationship between final construct dimensions and initial seeding density shown in this study, tissue engineers may be precluded from generating fibrocartilage with thicknesses of several millimeters or more. However, layering strategies may be adopted in the future to bypass this limitation and/or create multilayered constructs. Tissue engineers have created multi-layered constructs, including in cartilage [45, 46], but relatively little work exists in the literature detailing all-biologic, scaffold-free constructs with multi-layered architectures. This goal is particularly relevant since the knee meniscus possesses varying ECM organization at different depths within the tissue [47, 48]. Future research could employ zonal layering with various cell types to further recapitulate native tissue-level organization. In addition, several recent studies have explored agents that enhance engineered tissue by contracting and/or cross-linking the ECM [28, 49, 50], and these should be investigated further to increase ECM organization and integrate multi-layered tissue constructs seeded at various densities.

Conclusions

This study showed that self-assembling anatomically shaped knee meniscus fibrocartilage might be grown with a lower seeding density than used previously. In addition to this, constructs demonstrated anisotropy in the circumferential versus radial directions of the meniscus. Furthermore, constructs at a critical seeding density displayed increased mechanical and biochemical properties, and superior ECM organization as visualized by SEM. This is the first study to correlate a lower seeding density per area to superior properties in self-assembling meniscus constructs, the first study to examine the ECM structure of scaffold-free, self-assembling meniscus constructs by the use of SEM, and the first study to demonstrate a threshold seeding density exists for the self-assembling process of fibrocartilage. Future work should quantify the beneficial effects of various anabolic (e.g., TGF-β1) and catabolic (e.g., chondroitinase ABC) stimuli upon knee meniscus tissue engineered using this critical initial seeding density.

Supplementary Material

supplement. Supplementary Figure 1. Dogbone tissue samples for tensile testing.

Tensile samples (left) taken in the circumferential direction and (right) in the radial direction for strain-to-failure testing.

Figure 7. Tissue density is greatest in constructs seeded with 5 million cells and decreases with higher seeding density.

When imaged with SEM, constructs seeded with 5 or 10 million cells exhibited a denser tissue with clearly defined fibers. In comparison, constructs seeded with 15 or 20 million cells displayed a less dense extracellular matrix with a less prominent collagen network. When quantified, tissue density was greatest in constructs seeded with 5 million cells, and decreased afterwards, becoming significantly less in constructs seeded at the historical density of 20 million cells per construct. Groups lacking a letter in common are significantly different (p < 0.05). Constructs seeded with fewer than 5 million cells were not assessed († = not imaged with SEM).

Appendix A. Figures with essential color discrimination

Certain figures in this article, particularly Figures 2, 3, and 4, are difficult to interpret in black and white. Full color images for these figures can be found in the online version of this article.