Natural large-scale regeneration of rib cartilage in a mouse model

Case Study:.

UCSF’s Committee on Human Research–approved consent (CHR #H42089-22594) was obtained allowing for postoperative quantitative CTs to examine the regeneration at the graft harvest site. CT imaging was performed with a 16-row multi-slice CT scanner (Lightspeed, GE, Milwaukee, WI). The kVp and mA were based on patient weight. 3.75 mm image slice thickness was obtained and reconstructed to a 1.25 mm thickness. The CT image field of view was restricted to the resection site marked by a radio-opaque skin marker to limit radiation exposure. A bone density phantom (MINDWAYS, Austin, TX) was included. The DICOM dataset was analyzed and reconstructed with Amira (VSG/FEI, Burlington, MA) and InVivo (Anatomage, San Jose, CA) software. Our study complies with the World Medical Association Declaration of Helsinki, Ethical Principles for Medical Research Involving Human Subjects.

Mouse Procedures:

CD-1, FVB (Charles River), or MRL/MpJ (000486, Jackson Laboratories) mice (average of 53 days of age) were used for a survival surgery procedure in which a costal cartilage segment was excised with or without the perichondrium. The length of the resected cartilage varied from ~2–5 mm with an average of 4.10 mm (representing about 33% of the typical length of costal cartilages from ribs 8–10). The procedure took ~1 hour, could be easily carried out after practice, used commercially available micro-surgical tools, and was well-tolerated. For autograft surgeries, a flap incision was made in the intercostal muscles between rib 7 and 8, and a 4–5mm perichondrium strip from ribs 8, 9, or 10 was inserted either as an autograft or using tdTomato labeled (C57BL/6J, 007676, Jackson Laboratories) tissue into a C57BL/6J mouse. Further details for surgical procedures can be found in Supplementary Materials. To assess new calcium deposition, mice were given an intraperitoneal injection of calcein (Sigma C0875) 12.5mg/kg, 2 days before sacrifice and incorporation was assessed using a Zeiss Axio Observer fluorescent microscope. For label retention studies, ten-day-old female CD-1 mice were injected intraperitoneally with 50 mg/kg of 5-ethynyl-2’-deoxyuridine (EdU), twice daily for 3 days. This pulse was followed by a chase of 8 hours, 2 weeks, or 8 months. EdU was then visualized using Click-iT (Life Technologies) and labeled cells were distinguished from autofluorescing red blood cells because they were positive for a nuclear stain (Hoechst 33258). All procedures were carried out in accordance with approved protocols by the Institutional Animal Care and Use Committee (IACUC) at the University of Southern California (USC).

Tissue analysis and histology:

Rib cages were fixed in 95% EtOH and stained with alizarin red and alcian blue using a standard protocol (17) or fixed in 4% paraformaldehyde and decalcified (3–10 days in 10% EDTA) for histological analysis. Samples were prepped for paraffin or cryo-sectioning using standard protocols or for plastic sectioning using the Immuno-Bed Kit (Polysciences, Inc., #17324). Sections were stained with nuclear fast red, hematoxylin and eosin, alcian blue, using standard protocols or trichrome, (Gomori One Step, aniline blue, Newcomersupply 9176B). After antigen retrieval, sections were incubated in antibody for proliferating cell nuclear antigen (PCNA) (Vector Labs, VP-P980, 1:100) followed by detection with Alexa Fluor® 568 goat anti-mouse IgG (H+L) secondary (A11031, 1:300). Images were captured with a Nikon AZ100 macroscope or a Zeiss Axio Imager.

Micro-computed tomography (μCT):

Scans of the mouse rib cage were performed at the USC’s Molecular Imaging Center on a subset of the animals (80-293 days post resection). μCT scans were acquired with the InveonCT scanner (Siemens Medical Solutions USA, Inc., Knoxville, TN) with the following settings: 80kVp, 120uA, no filter, 451 projections covering 220 degrees with 4s/projection, bin=2, and a voxel size of 18.729 microns. Data were reconstructed using Cobra Reconstruction Software (Exxim Computing Corp., Pleasanton, CA). One sample was scanned with higher resolution on the μCT 50 scanner (Scanco Medical AG, Bruttisellen, Switzerland) using the following settings: 70kVp, 114uA, no filter, 2000 projections covering 360 degrees with 1.5s/projection, and a voxel size of 6.8 microns. Higher resolution data were reconstructed using the Scanco software. Datasets were loaded into Amira 5.3.1 (Visage Imaging, Inc., Berlin, Germany) or Osirix (http://www.osirix-viewer.com) for visualization and analysis. Images were segmented to measure mineralized cartilage volumes (mm³) in areas of resection. Measurements of each sample were performed by at least two people blinded to the strain and healing time.

Quantification and analysis:

To assess healing over time and across mouse strain, we constructed a mathematical model that considered physical and geometric parameters. To determine the resection volume and surface area should repair be complete and continuous, we used the resection length and the major and minor radii of the cut ends (clearly identifiable), as measured from reconstructed μCT scans, and assumed a tapered elliptical cylinder profile. We then used the measured repair volume and assumed a simple differential equation, dC/dt = (V_f-C) / (s_b(V_r/S_r)), where C is repair volume, t is time, V_f is final repair volume, V_r is the resection volume, S_r is the resection surface area, and s_b is a parameter defined in units of days/mm to express the inverse rate or ‘slowness’ of repair for each genetic background, b. The model was fit using the Bayesian modeling language JAGS via statistical software, R. Further details are explained in Supplementary Materials.

A case study in a human male

Although regeneration of the costal cartilage in humans has been described, the patients in these studies were either unusual (undergoing reconstructive surgery for pectus excavatum) or the radiographic imaging used was at low resolution (6, 7, 11). We therefore decided to determine if regeneration could be detected first-hand in our clinic. A 42 year old male required craniofacial reconstruction at UCSF’s Department of Oral & Maxillofacial Surgery, due to trauma. We harvested a portion of bone (8 cm) and cartilage (1 cm) from the right 6^th rib (Fig. 1A) and a CT scan immediately after surgery was used to estimate the resection site (Fig. 1B–B’). After ~6 months, a reconstructed CT scan from the same patient showed incomplete regeneration of the bone and regeneration of the costal cartilage with a distinct low-density region between the two segments likely representing the reformation of the costochondral joint (Fig. 1C–C’). While this imaging provides evidence for both bone and cartilage repair in a human and confirms previous and anecdotal reports, understanding the cellular and molecular mechanism necessitates histological and cellular analysis. This is best achieved by establishing an animal model.

Development of a mouse model

We devised a survival surgery procedure in mice that takes advantage of the observation that mouse and human thoracic anatomy are very similar (Fig. 2A–E) and that allowed for the removal of cartilage portions with and without the surrounding perichondrium (Fig. 3A–F’). Because we did not know if wild-type mouse strains would undergo cartilage repair at all, we decided to also include MRL/MpJ mice in our study. This line has an enhanced regenerative capacity in some tissues. With regard to cartilage, a punch wound in the ear elastic cartilage of MRL/MpJ mice led to full regeneration after 4–5 weeks, while a punch wound in the control strain remained intact (19). In addition, MRL/MpJ mice appear to have a better capacity to repair articular injuries, (20) and a lower immune response after a traumatic articular injury, suggesting that they may be protected from post-traumatic arthritis (21). To date, the capacity for costal cartilage repair in MRL/MpJ mice has not been investigated.

To test the role of the perichondrium, surgeries were divided into three groups: 1) surgeries in which only the costal cartilage was removed leaving the perichondrium remaining in the animal (n=33, Fig. 3A), 2) surgeries in which both rib and perichondrium were removed (n = 12, Fig. 3E), and 3) controls sacrificed shortly after surgery to assess technique (n = 7). Aside from the incision made to remove the cartilage, the remaining perichondrium was intact, and relatively undisturbed (still attached to surrounding tissues) (Fig. 3C). The vast majority of the cartilage had been removed although a few small fragments intimately attached to the perichondrium were occasionally left behind. The integrity of removed cartilage portion was also analyzed. It typically had a smooth exterior under surface illumination and histological analysis revealed no attached perichondrium (Fig. 3D–D’). Upon visual inspection of with-perichondrium portions, the cartilage surface had a distinct ragged appearance and an intact perichondrium was clearly evident in histological section (arrows and bracket, Fig. 3F, F’). Our human case study along with a few published reports indicated that healing might take on the order of several months (5, 6). Thus, within the cartilage-only surgeries, we examined approximately half at 1–3 months (average 95 days, n = 24). The other half were long-term recovery surgeries analyzed at 7–9 months (average 274 days, n = 13). The complete-removal surgeries had an average post-surgical recovery time of 117.5 days, n=19. Roughly half of the surgeries utilized the MRL/MpJ mouse strain (n = 26/56).

At birth, the costal cartilages of mice can be detected with alcian blue (Fig. 3G). However by sexual maturity, when the surgeries were performed (average age = 53 days), chondrocytes become entrapped in a mineralized matrix (Fig. 3G’). This mineralization was retained in older mice (9 months, Fig. 3G”) and at high magnification was evident as a spongiform matrix both by alizarin red staining and high resolution μCT (Fig. 3H, 3I) with the empty spaces likely occupied by resident hypertrophic chondrocytes. This mineralization greatly facilitated our ability to locate the resected region and additionally to quantify repair by μCT scanning (Fig. 3G”, arrowheads). This phenomenon is in contrast to humans whose costal cartilages only mineralize later in life. Indeed, the degree of mineralization can be used to determine the age of forensic specimens (18).

Surgeries in which both cartilage and the surrounding perichondrium were excised, failed to repair even after 9 months as assessed by whole mount skeletal staining and μCT (Fig. 4A–A’). Occasionally scattered mineralized fragments were evident but these were small and/or disorganized (data not shown). To our surprise, however, all cartilage-only surgery samples regardless of genetic strain, repaired—with the majority of new cartilage filling-in by 1–2 months followed by mineralization (Fig. 4B–B’). After 1 week post resection (PR) some alcian blue positive cartilage could be detected at the lateral edges of the cut ends (Fig. 4C). By 3 weeks PR, diffuse alcian blue staining was evident along the resected region (Fig. 4D) while with longer healing times alizarin red staining was also present but typically found in distinct patches of varying sizes (Fig. 4E; Fig. 5A,B). Samples with longer healing time-points correlated with alizarin red staining that spanned a larger portion of the resection (see Fig. 5). To definitively confirm that the repair tissue consisted of cartilage, we serially sectioned through a healed region identified by skeletal preparation to examine the cellular profile (Fig 4E). We found that the repair tissue did indeed consist of cartilage, extended throughout the resected region, and contained chondrocytes arranged in clusters typical of isogenous groups (Fig. 4G1–3). When comparing the new cartilage to unresected cartilage (Fig. 4F) we noticed that new cartilage contained abundant hypertrophic profiles that at times appeared larger than normal, and that the surrounding perichondrium was irregular and on average 2.10x thicker (Fig. 4G1–3). To ensure that the observed morphology was not an artifact, we also analyzed specimens prepared using standard histological techniques (example in Fig 4H–I’). We confirmed our initial findings and additionally observed that although thickened, the perichondrium histomorphology was normal—cells in the perichondrium were surrounded by matrix, had somewhat elongated nuclei, and were not hypertrophic (Fig. 4H’ vs. 4I’).

A thickened perichondrium could be due to an increase in matrix deposition (perhaps a kind of scar tissue). Alternatively, the thickening could represent an increase in the number of cells. These cells could be predominantly vascular, providing a supportive niche, alternatively the cells could be chondrogenic progenitors that expand in response to injury for the purpose of mediating repair. In support of this latter hypothesis, a thickened perichondrium (1.98x thicker) was also evident at earlier PR time points (compare Fig. 4J’–K’) and within the expanded zone, abundant cells were evident with pre-hypertrophic profiles in isogenous groups characteristic of chondrogenesis. Furthermore, cells in this location expressed PCNA and therefore were actively proliferating (compare Fig. 4L’ to Fig. 4M’). These observations suggest that an important response to injury involves the expansion of progenitors within the perichondrium and that these progenitors may mediate repair by differentiating into cartilage and filling-in the resected region. After the initial thickening, the perichondrium maintains its thicker dimensions, potentially housing now quiescent progenitors.

The new cartilage underwent mineralization (like flanking costal cartilages had during post-natal development) as evidenced by alizarin red staining (Fig. 4B, E, 5A, B). Specimens with long healing times had resections that were eventually completely mineralized (Fig. 4B–B’). In addition, we could use mineralization as a proxy for the extent of repair, and measured mineralization volume with μCT imaging in both wild-type and MRL/MpJ specimens (a random subset of the cohort was analyzed, 80–293 days PR healing times, n=28). To quantify the extent of repair, we used measurements from reconstructed μCT scans (examples shown in Fig. 4A’, B’) to developed a mathematical model that took into consideration healing via extension from the ends (appositionally) and/or filling-in (as though from progenitors located in the perichondrial sleeve) and incorporated a Bayesian statistical fitting procedure to account for sources of variation and uncertainty including measurement error, model misfit, and random variability between animals. Using this approach we found that although there may be differences early in the healing process, both wild-type and MRL/MpJ strains mineralized asymptotically to a similar degree, and they did so on a time-scale related to the geometric properties of the rib resection, especially the surface area to volume ratio suggesting that filling-in from the periphery was the main repair mechanism (Fig. 5C). Interestingly, although in some cases the repair could have a somewhat larger volume than the resected region, the new cartilage consistently maintained a cylindrical profile and did not exhibit extensive protuberances or a grossly enlarged callus suggesting that either the perichondrium constrained the repair or that another unknown mechanism regulates cartilage shape and size.

The perichondrium clearly plays a critical role in facilitating repair since removal of both the cartilage and the surrounding perichondrium failed to support cartilage re-growth even after ~9 months PR (Fig. 4A–A’). In addition, in our ‘cartilage only’ removal experiments, we found a strong correlation (Pearson correlation coefficient = 0.491, p = 0.008) between the amount of repair (% of area filled in, based on time) and the “thinness” (surface area/volume ratio) of the resected region, with thin, large surface area to small volume resections exhibiting the most complete degree of repair (Fig. 5D). These results suggest that the fastest repair occurs when the resected region has a small volume and is surrounded by abundant perichondrium.

The ability of cells to retain a DNA label has been used to identify slow cycling cell populations that in some contexts—hair follicle and bone marrow and auricular cartilage, correlate with the location of resident stem cells (14, 22, 23). To determine the location of label-retaining cells in costal cartilages, we pulse-chase labeled animals with EdU (24) starting at postnatal day 10. After a short chase (8 hours), very few cells were labeled in both the perichondrium and mature cartilage (data not shown). However after a 14 day and 8 month chase, labeled cells could be found both within the cartilage matrix and perichondrium (Fig. 6A–C’). After 14 days labeled cells within the cartilage matrix were clustered in small groups of 2–3 cells and located predominantly at the periphery (Fig. 6A,B, green arrows), although occasionally small clusters could also be found in the cartilage interior (not shown). The lack of labeled cells within the cartilage after a short chase and the small clusters found after a long chase suggests that new chondrocytes form preferentially from perichondrium progenitors at least at these post-natal stages. Even after an 8 month chase, rare label-retaining cells could be found in the perichondrium and occasionally presumably new chondrocytes (pairs indicated with green arrows) could be seen in the periphery (Fig. 6C, C’). The perichondrium has been shown to be comprised of distinct layers (inner and outer) which are hypothesized to have distinct functions (inner, appositional growth at the periphery; outer layer structural and vascular support) and have classically been visualized with light or electron microscopy but recently shown to express specific molecular markers (15, 25). Labeled cells in the perichondrium were counted and classified as inner or outer cells (Fig. 6B–B’). For the most part, classification was unambiguous as labeled cells were typically found either at the inner or outer extremes suggesting that the periochondrium has distinct populations of cells that are not highly proliferative. As has been seen in other contexts (14), we did find an enrichment of label-retaining cells within the perichondrium, however after the 14 day chase we found more labeled cells in the outer than inner layer (2.2 times more, p=0.018, paired T-test), while after 8 months labeled cells could only be found within the inner layer. While it is possible that label-retention represents cells that only divided once or twice and then differentiated, it is also possible that the label-retaining cells represent stem cells. Further lineage analysis will be needed to determine this more precisely.

We next tested the capacity of the perichondrium to support new cartilage generation by placing the perichondrium in an ectopic location (Fig. 6D). These strips were largely devoid of alizarin red staining indicating that our dissection technique did not carry along large pieces of adherent mature cartilage (Fig. 6E). We then chose a nearby intercostal muscle, created a glancing flap and sutured the perichondrium inside. By visual inspection, we observed no evidence of graft resorption or extrusion at 1, 2, and 3 weeks after surgery. We then assayed for the formation of differentiated cartilage with alizarin red staining. We found that regardless of the healing time (ranging from 2 weeks to 5 months) ectopic mature cartilage was evident (Fig. 6F). This ectopic cartilage could take up calcein (Fig. 6G) indicating new mineralization. Samples analyzed up to 2 months had large cartilage pieces but by 5 months only small fragments were present (data not shown) suggesting that the generation and retention of cartilage in this location was robust (n = 10/15) but limited over periods 2 months and longer. A tdTomato labeled perichondrium implantation also resulted in ectopic cartilage. Both mature and prehypertrophic cartilage cells were tdTomato positive indicating that the ectopic cartilage arose from donor rather than host tissue (Fig. 6H,H’). Furthermore, the ectopic tissue was proliferative as evidenced by PCNA expression (Fig. 6I, I’).

Time sequence of repair

In humans, regeneration of the costal cartilage has been observed but not rigorously documented or analyzed at the cellular level. Here, we establish a simple baseline model in a mammal that allows further studies of cartilage repair to proceed within the context of the enormous flexibility and low cost of the mouse genetic system. From this initial study, we propose a timeline for perichondrium-mediated cartilage repair over the course of 1–2 months (Fig. 7). Within the first week, chondrogenic progenitors within the periochondrium expand. Next we propose that these progenitors fill the resection zone by emerging both from the perichondrium lateral to the cut site and from the remaining perichondrium along the length of the resection. New chondrogenic progenitors proliferate to create isogenous groups, secrete cartilage matrix, and differentiate into mature chondrocytes with increasing hypertrophic profiles. Although the cartilage in the repair zone contains chondrocytes with variable hypertrophy and tends to be more nodular than normal (Fig. 4I) (how chondrocyte size and therefore cartilage size may be regulated is still to be fully determined (26)), the regenerated cartilage fully differentiates, producing a mineralized matrix as is typical of mouse costal cartilage. Thus, as in fracture repair where the surrounding connective tissue (periosteum) has been shown to play an important role (12, 27), cartilage repair in this context may also be mediated by cellular contributions from the surrounding connective tissue (perichondrium). Furthermore, the sequence of events observed here likely recapitulates some of the steps that occurred during development (expansion of progenitors, increased hypertrophy, mineralization)(28).

Use of mice

More generally, the establishment of this mouse model allows for an enormous variety of genetic, drug, and surgical interventions to be compared within transgenic and genetically altered backgrounds. For example, in this study we compare the repair capabilities of both wild-type and the mouse strain MRL/MpJ. Although we do not see an enhanced rate of repair in this background within the time frame examined (months), there may be differences in the repair rate within the first 2–4 weeks. Future studies will involve characterizing the early steps of cartilage repair to understand the cellular dynamics (infilling vs. appositional growth) and examining repair within the context of other loss-of-function or gain-of-function alleles, particularly for cartilage-inducing growth factors. Through these studies, the molecular nature of the repair competency of the rib could be elucidated.

Perichondrial progenitors

During development, the perichondrium is a major source of chondrogenic and osteogenic progenitors (27, 29, 30). During cartilage repair, the perichondrium may be a critical player and represents an important area of interest for developing new therapies. For example, in the intervertebral disc, the location of label-retaining, slow cycling cells as well as the expression of stem cell markers suggests the presence of a niche in the perichondrium adjacent to the edge of the annulus fibrosus (8). In addition, similar studies in the knee joint have suggested that a niche might exist in a similar perichondrium-like region, called the ‘groove of Ranvier’ (31) or in the articular surface layer which may have remnant characteristics of the perichondrium (9, 32). Label-retaining cells have been found in the perichondrium surrounding the elastic cartilage of the mouse auricle and perichondrial cells have been isolated from mouse and human auricle that have chondrogenic capacity (14, 33). In this study, we also find label-retaining cells within both the inner and outer layers of the periochondrium after a 2 week chase and within the inner layer after an 8 month chase (Fig. 6A–C). Furthermore the ectopic cartilage found after intramuscular perichondrium implantation contains both mature and new prehypertrophic cartilage cells and is derived from the donor tissue (Fig. 6E–I’) providing further support for the idea the perichondrium is the source of cells for repair rather than simply providing a scaffold for chondroprogenitors derived from some other location. Future studies using techniques to both isolate perichondrial cells and observe their potential in culture and to track their lineage in response to injury are needed. Fortunately genetically modified mouse lines have been successfully employed to address similar questions in the epidermis and this technology could be re-applied here (34).

General use of rib-derived or ‘rib-like’ cells in cartilage repair

Due to morbidity problems with harvesting autologous cartilage for grafting purposes, there is great interest in cell-based therapies to repair damaged cartilage. Currently the most common approach is to culture and transplant autologous articular chondrocytes for acute joint injuries repair (35). Unfortunately these cells can have limited viability, fail to proliferate, may undergo dedifferentiation, or even fail to integrate (36). Another potential source for cells is costal cartilage and this source has some advantages. Costal cartilage is a hyaline-type cartilage similar to the cartilage found in the articular region. Chondrocytes can be readily procured from the patient and may expand more efficiently in vitro than articular cartilage. Furthermore, costal cartilage that has dedifferentiated in culture can be converted into high quality hyaline cartilage (37). In addition, the use of costal cartilage cells has been shown to be a feasible option for articular cartilage repair in animal models (38, 39). However, costal cartilage as well as cartilage generated from other sources can have a tendency to undergo mineralization over time. For the purpose of articular cartilage defect repair, this may be good for reconstructing the osteochondral junction so that the articular surface does not delaminate. However this may present challenges for articular surface reconstruction which needs to remain unmineralized for best function. Mineralization poses less of a problem for the use of costal cartilage in craniofacial and trachea repair. Thus the use of costal cartilage and/or the use of banked stem cells derived from costal cartilage perichondrium may be an attractive option for various cartilage or even bone repair applications. A handful of studies have demonstrated a robust ability of rib perichondrium to facilitate the repair of articular defects in rabbits and sheep (40, 41). Given our findings, it may be beneficial to re-visit these injury studies using genetic techniques to track and lineage trace implanted cells. Furthermore developing efficient ways to isolate chondrocyte progenitors from rib periochondrium may have highly useful clinical applications.

The embryonic origin of the rib is distinct from that of the appendicular skeleton (somite vs. lateral plate mesoderm (42)). Thus one possible reason for the high regenerative capacity of ribs may be related to the developmental history experienced by rib vs. limb progenitor cells. One idea would be to mimic this history by transitioning generic pluripotent cells through a somite mesoderm-like induction process. For example, Nakayama and colleagues have shown that human pluripotent cells transitioned in this way have enhanced chondrogenic capacity when compared with human bone marrow-derived mesenchymal stem cells (43) and it will be interesting to determine if these cells share a common molecular profile with rib-derived progenitors. Although banking perichondrium-derived cells may be a feasible option, rib-like chondrogenic progenitors generated from induced pluripotent cells, may be easier to develop for clinical use. In either case, given our findings, it is attractive to imagine using these rib or rib-like progenitors in a variety of applications including acute cartilage injury repair, auricle and trachea reconstruction, rhinoplasty, and even myringoplasy (44, 45) and we hope future studies will investigate these possibilities.

In conclusion, here we present a simple method for the study of large segmental cartilage defects in a mammalian model. Similar to sporadic reports in humans (references (6, 7, 11) and this study, Fig. 2), we find that the mouse costal cartilage repairs well, within 1–2 months, and we have identified the perichondrium as a key player and the likely source of repair cells.