Cell‐associated type I collagen in nondegenerate and degenerate human articular cartilage

Chondrocytes with abnormal morphology are present in nondegenerate human cartilage suggesting dedifferentiation to a fibroblastic phenotype and production of a mechanically‐weakened matrix of unknown composition. We determined the relationship between in situ chondrocyte morphology, chondrocyte clusters, and levels of cell‐associated collagen type I. Chondrocyte morphology in fresh femoral head cartilage from 19 patients with femoral neck fracture and collagen type I labelling was identified with Cell TrackerTM fluorescence and immunofluorescence, respectively, in axial/coronal orientations using confocal microscopy with images analysed by ImarisTM. In axial images of grade 0 cartilage, 87 ± 8% were normal chondrocytes with a small (10 ± 6%) abnormal population possessing ≥1 cytoplasmic process. More normal chondrocytes (78 ± 11%) were collagen type I negative than those labelling positively (p < 0.001). For abnormal chondrocytes, 81 ± 14% labelled negatively for collagen type I compared to those labelling positively (19 ± 3%; p = 0.007; N(n)=11(3)). Overall, approximately 9% of the cells in normal cartilage labelled for collagen type I. With degeneration, the percentage of normal chondrocytes decreased (p < 0.001) but increased for abnormal cells (p = 0.036) and clusters (p = 0.003). A larger percentage of normal, abnormal and clustered chondrocytes now demonstrated collagen type I labelling (p = 0.004; p = 0.009; p = 0.001 respectively). Coronal images exhibited increased (p = 0.001) collagen type I labelling in the superficial zone of mildly degenerate cartilage with none in the mid or deep zones. These results show that collagen type I was identified around normal and abnormal chondrocytes in nondegenerate cartilage, which increased with degeneration. This suggested the presence of mechanically weak fibro‐cartilaginous repair tissue in otherwise macroscopically nondegenerate human cartilage which progressed with degeneration as occurs in osteoarthritis.


| INTRODUCTION
Osteoarthritis (OA) is a complex and painful whole joint disorder in which the increasing mechanical weakness of the degenerating cartilage plays a central role. The failure of the extracellular matrix (ECM) to bear load is pivotal in OA progression. However, the role of chondrocytes, the metabolism of matrix components (collagens, proteoglycans, and other minor components), and the sequence of the changes is not fully understood. This is of paramount importance to identify targets to slow cartilage loss and protect the painsensitive underlying bone. Although it is known that there are substantial changes to the collagen type and content with primary OA (e.g., Hollander et al., 1995), there are gaps in our knowledge particularly in relation to collagen type I. While collagen type I has an essential role in many types of connective tissues, its presence in hyaline cartilage in preference to type II collagen, leads to a mechanically-weakened fibro-cartilageneous "repair" tissue (Huey et al., 2012) and thus potentially accelerates the degenerative process.
There has been some controversy and contradictory evidence about the presence of collagen type I in human cartilage and the concept of a "phenotypic switch" for the synthesis of type II to type I collagen during cartilage degeneration. Early studies by Adam and Deyl (1983) using immunofluorescence suggested that only collagen type II was present in normal human femoral head cartilage obtained from necropsy samples. However, in freshly-resected OA cartilage from arthroplasty procedures, collagen types I and III were also reported to be present. Studies by Aigner et al. (1993) and Aigner et al. (1997) on the other hand, using in situ hybridisation could not detect collagen type I messenger RNA (mRNA) signal in normal or OA cartilage. Nevertheless, they reported "hardly any" collagen type I protein, but in agreement with others (Wotton & Duance 1994) have described the presence of type III collagen (a characteristic of dedifferentiated chondrocytes; Charlier et al., 2019) increasing with cartilage degeneration (Hosseininia et al., 2016).
More recently, Gebhard et al. (2003) and Miosge et al. (2004) reported increased levels of collagen type I expression with degeneration but because the ratio of type I to type II did not increase in OA, this suggested that there was no general shift to a dedifferentiated phenotype. Using laser capture microdissection, Fukui et al. (2008) reported that the expression of collagen type I, alpha-2 (COL1A2), was most enhanced in the superficial zone of OA cartilage. Brew et al. (2010) studying late-stage OA from knee arthroplasty operations described a marked increase in gene expression (but not protein levels) for collagen type I compared to age-matched controls.
They commented that this probably did not reflect a generalised change in chondrocytes to a proliferative or hypertrophic phenotype as seen, for example, in the growth plate. These and other studies (Eyre et al., 2006;Roberts et al., 2009) suggest that there is negligible collagen type I present in normal nondegenerate human cartilage, however, there is a marked increase in mRNA and protein levels with OA progression.
The increased production of collagen type I during cartilage degeneration strongly suggests a change in the differentiation status of some chondrocytes towards a fibroblastic phenotype (Charlier et al., 2019;Hall, 2019). The process may also be coupled with other events, for example, the development of a fibroblastic morphology, with associated increased polymerisation of G to F-actin, with reduced levels of aggrecan and Sox9 (SRY sex determining region Y)box 9 (Sox9) production (Martinez-Sanchez & Murphy, 2013). This phenotypic modification of chondrocytes leading to the production of a mechanically-incompetent ECM is recognised as an important mechanism that contributes to the loss of cartilage homeostasis in OA (Fukui et al., 2001).
During studies on the in situ morphology of fluorescentlylabelled chondrocytes within macroscopically normal (grade 0) human femoral head cartilage, a significant proportion of cells demonstrated a fibroblastic-like morphology with fine cytoplasmic processes (Karim et al., 2018). These morphological changes suggested that even in apparently nondegenerate cartilage, chondrocytes were undergoing dedifferentiation potentially producing collagen type I with the implication that microscopic areas of cartilage were becoming mechanically weakened. To clarify the situation, in the present work, the relationship between in situ human chondrocyte morphology within nondegenerate and degenerate human femoral head cartilage and levels of collagen type I has been investigated using fluorescent labelling of chondrocytes, immunofluorescence and confocal laser scanning microscopy (CLSM). The results demonstrated that cell-associated collagen type I was present in nondegenerate cartilage and increased with degeneration.
Importantly, chondrocytes of both normal and abnormal morphology produced this protein, suggesting an "uncoupling" between chondrocyte morphology and collagen type I production. 2.2 | Cartilage grading, sampling, fluorescent labelling of in situ chondrocytes, and CLSM Grading of the articular cartilage was determined by visual assessment of the macroscopic appearance, using the Osteoarthritis Research Society International criteria (Pritzker et al., 2006). Cartilage was assessed as grade 0 (nondegenerate) or grade 1 (mildly degenerate). In some cases, it was difficult to tell precisely if the cartilage was grade 1 or 2, and thus it was described as grade 1-2 (see Figure 1). Full-depth cartilage explants were harvested using 3-mm diameter biopsy punches (Kai Medical) (Styczynska-Soczka et al., 2020). Cartilage samples were then incubated (1.5 h; 21°C) with CMFDA (5-chloromethylfluorescein diacetate) Cell Tracker green (12.5 µM; Invitrogen) to label living cells. Explants were washed in phosphate-buffered saline (Invitrogen), fixed (formaldehyde 4% vol/vol; 30 min; Fisher), frozen in optimal cutting temperature freezing medium (CellPath) and then cut on the cryostat into 20-µm thick sections for either axial or coronal visualisation. Sections were then incubated in bovine testicular hyaluronidase (H3506; 1000 U/ml; Merck) for 30 min at room temperature to facilitate the primary antibody penetration. Following this, sections were incubated with the primary antibody for collagen type 1 (ab34710, Abcam) followed by fluorescent secondary antibody (Alexa Fluor 568, A-11031, Thermo Fisher Scientific) and imaged by CLSM using established methods (Karim et al., 2018).

| Chondrocyte morphology and collagen type I analysis
Axial and coronal view images were analysed using Imaris TM (Version 8.3.1; Bitplane). Chondrocyte morphology (green) and fluorescentlabelling of collagen type I (red) were analysed in both grade 0 and grade 1 femoral head cartilage using a standardised image size (1024 × 1024).
Chondrocyte morphology was quantified by manually counting the number of chondrocytes with "normal" morphology, chondrocytes forming clusters and chondrocytes with cytoplasmic processes, using previously published criteria (Karim et al., 2018). The data are expressed as a percentage of the total number of chondrocytes viewed on the standardised image. Chondrocytes with "normal" morphology were considered to be round/elliptical, with no observable cytoplasmic processes. "Abnormal" chondrocyte morphology was defined as cells exhibiting ≥1 cytoplasmic process(es) of at least 2 µm in length. A cluster was considered to be 5+ chondrocytes within one lacuna (Karim et al., 2018). An estimate of the lacunar region surrounding chondrocytes could be visualised because of the secondary antibody labelling in immunofluorescence images as a darker area around chondrocytes.
Fluorescent-labelling of collagen type I was determined on a per-cell basis. This allowed the categorisation of the cells as being collagen type I "positive" (+) or "negative" (−) and the cells were considered to be collagen type I "positive" if the fluorescence was cell-associated, that is, strong specific staining was visible in the direct vicinity and surrounding the cell (see Figure 2). Cells were classified into six morphological and were divided into the superficial zone (SZ) = 15%, middle zone = 50%, deep zone (DZ) = 35% of the total cartilage thickness (Venn, 1978). At low magnification (Figure 2a,b), the heterogeneity of collagen type I labelling was clearly evident with some cells labelling strongly around the cell body (cell-associated), whereas others did not demonstrate labelling.

| Statistical analysis
There was also collagen type I labelling in the ECM of the mildly-  Figure 3b. Of the cells with normal morphology, 78 ± 11% labelled negatively for collagen type I which was greater (p < 0.001) than for the percentage of cells labelling positively (6 ± 5%).
For the cells in the population with one or more cytoplasmic process, the majority (81 ± 14%) labelled negatively for collagen type I which was also greater than for the percentage of cells labelling positively (19 ± 3%; p < 0.001; data from N(n)=11 (3) Note examples of collagen type I labelling associated with chondrocytes (arrows with short solid shafts) or labelling which did not appear to be cell-associated (arrows with long solid shafts). Also shown are examples of chondrocytes that did not appear to be associated with collagen type I labelling (arrows with dotted shafts). The scale bar was 25 µm. (c-h) shows high magnification images of the range of chondrocyte morphologies labelling with or without collagen type I. (c, d) Represent normal, rounded chondrocytes either labelling with or without cell-associated collagen I respectively; (e, f) chondrocytes with cytoplasmic processes either with or without cell-associated collagen type I labelling respectivelyalso note strong collagen type I labelling distant from chondrocytes in this cartilage grade 1 explant; panels (g, h) shows clusters of chondrocytes either with or without collagen type I labelling also in a grade 1-2 cartilage sample. The scale bar for (c-h) was 10 µm STYCZYNSKA-SOCZKA ET AL.

| 7675
small, but noticeable proportion (~10%) of the cells labelled positively for collagen type I (Figure 2a). The observation that for some cells, cytoplasmic processes were evident but there was no collagen type I identified, whereas there was labelling around some of normally-shaped chondrocytes, suggests that the morphological change for these cells was not a necessary prerequisite for the presence of collagen type I.
3.2 | Human chondrocyte morphology and cellassociated collagen type I labelling in mildlydegenerate (grade 1-2) femoral head cartilage With mild cartilage degeneration to grade 1 or 1-2 (see Section 2), the morphological characteristics of the fluorescently-labelled in situ superficial zone chondrocytes were markedly different (Figure 3a). Although there were still more chondrocytes with normal morphology compared to cells with processes (p < 0.05), a large proportion (~40%) of the cells were present in clusters and the percentage was greater than for abnormal chondrocytes (p < 0.05). Compared to grade 0 cartilage, there was a significant decrease (to 39 ± 23%) in cells of normal morphology (p < 0.001) but there were significant increases (to 25 ± 21%) in the percentage of the cell population with one or more cytoplasmic process (p = 0.036) and the percentage of cells in clusters increased markedly (to 35 ± 28%; p = 0.003; data are means ± SD from N(n)=11 (3) The immunofluorescent labelling of collagen I was present around almost all of the cells (Figure 4b) with 91 ± 20% being collagen positive, and 8.7 ± 3.3% collagen negative (data are means ± SD from N(n)=9 (3)). Of the cells with normal morphology, or with one or more cytoplasmic process, the majority (85 ± 10% and 93 ± 10%, respectively) now labelled positively for collagen type I which was significantly greater than the percentage of cells

| Distribution of cell-associated collagen type I with cartilage depth and degeneration
Full thickness coronal images of femoral head cartilage with chondrocytes labelled fluorescently and collagen I identified by immunofluorescence were studied by CLSM ( Figure 5). Cartilage thickness was divided into zones as described (see Section 2) and in grade 0 cartilage there was no evidence of cell-associated collagen type I labelling in any zones. In mildly degenerate cartilage, there was significant (p < 0.001) collagen type I F I G U R E 3 Chondrocyte morphology and collagen type I labelling in grade 0 femoral head cartilage. (a) Shows the distribution of chondrocytes into the classified categories of morphology as described (see Section 2). (b) Summarises the percentage of cells labelled with or without cell-associated collagen type I labelling in each of the three categories. Each symbol represents the averaged data from one femoral head (N = 13 and 11 for [a] and [b], respectively) with at least 3 images analysed for each femoral head. Data were compared between the categories using ANOVA followed by Tukey's post hoc multiple comparison test with ns indicating no significant difference (p > 0.05). ANOVA, analysis of variance labelling identified around chondrocytes in the SZ but not in the MZ or DZ (Figure 5c). High magnification images indicated that there was labelling around SZ chondrocytes but there was also labelling more distant from the immediate vicinity of the cells (Figure 5b). and increased those with one or more cytoplasmic processes. There was also a large increase in the presence of cell-associated collagen type I for both chondrocytes of normal and abnormal morphology, F I G U R E 4 Chondrocyte morphology and collagen type I labelling in grade 1-2 femoral head cartilage. (a) Shows the distribution of chondrocytes into the classified categories of morphology as described (see Section 2). (b) Summarises the percentage of cells labelled with or without cell-associated collagen type I labelling in each of the three categories. Each symbol represents the averaged data from one femoral head (N = 9 and 7 for [a] and [b], respectively) with at least three images analysed for each femoral head). Data were compared between the categories using ANOVA followed by Tukey's post hoc multiple comparison test with ns indicating no significant difference (p > 0.05). For statistical comparisons between the data shown here and that for Grade 0 cartilage see text. ANOVA, analysis of variance F I G U R E 5 Collagen type I labelling in (a) grade 0 and (b) grade 1-2 femoral head cartilage visualised in coronal sections. Low magnification images showed the absence of cell-associated collagen type I labelling in grade 0 cartilage, but extensive labelling only in the SZ of mildly degenerate cartilage. Areas (labelled 1, 2) in the SZ of grade 1-2 cartilage are shown at high magnification. (c) Shows the percentage of cells exhibiting cell-associated collagen type I labelling. Pooled data were obtained from (N(n)=5(3)) with each symbol representing averaged data from separate femoral heads. The difference between number of collagen type I positive cells in SZ in grade 0 and grade 1-2 was significant (p = 0.001). DZ, deep zone; MZ, mid zone; SZ, superficial zone STYCZYNSKA-SOCZKA ET AL. | 7677 particularly in the superficial zone. Although no chondrocyte clusters were observed in grade 0 cartilage, mild cartilage degeneration clusters were evident which labelled strongly for type I collagen.

| DISCUSSION
These observations are important for identifying microscopic properties of individual chondrocytes within otherwise normal, healthy human cartilage, and some of the changes which occur with cartilage degeneration as occurs in osteoarthritis.
A particular strength of the current study was the routine, ready and rapid access to fresh living ex vivo human femoral head cartilage obtained from femoral neck fractures. This required effective coordination between orthopaedic surgeons, theatre staff, and the research laboratory, and the collaboration allowed joints to be delivered quickly from the operating theatre to the research labs.
Preventing injury during removal of femoral heads and maintaining joint hydration throughout were essential as articular cartilage chondrocytes are very sensitive to even mild mechanical (Howard et al., 2020) and drying trauma (Paterson et al., 2015). Many previous studies have utilised tibial plateau or femoral condyle cartilage removed from joints of cadavers or during joint replacement surgery for osteoarthritis (e.g., Bush & Hall, 2003). However, as osteoarthritis is recognised as a whole joint disorder (Loeser et al., 2012) there is a concern that the small remaining areas of apparently "normal" cartilage may have been exposed to inflammatory mediators and degradative enzymes in the synovial fluid for some considerable time before study.
A further problem is that while full-depth osteochondral samples might be obtained and classified as grade 0, the characteristics of the chondrocytes in the SZ which represents a relatively small component of the full cartilage thickness, might be preferentially compro- is recognised that as some of the earliest changes in osteoarthritis are evident at the cartilage surface, the chondrocytes in the superficial zone could either be directly involved in the initial stages of the degenerative process or play a crucial role in the vicious cycle of its progression (Hollander et al., 1995;Robinson et al., 2016). Thus, minor changes in a small population of SZ chondrocytes could be highly significant. It was noteworthy that we did not observe any chondrocytes with cell-associated collagen type I labelling in coronal sections of grade 0 cartilage (Figure 5a). This has similarities to the study by Roberts et al. (2009) who reported that in coronal sections of hyaline cartilage described as "normal", a tiny (1.7 ± 0.8%) proportion of the cartilage area stained immuno-positively for collagen type I. Based on our axial data (Figure 3b) we would have anticipated that 10%-15% of the SZ cells would have labelled for collagen type I, however, the number of cells visualised in coronal sections of the SZ was low and so the chances of observing these labelled cells would be even less. The study of axial images and large numbers of cells ( Figure 2) was preferable as it permitted a relatively large area and a number of cells of the unperturbed cartilage surface distant from a cut edge to be studied. This reveals features that might be missed in relatively thin chemically-prepared and dehydrated coronal histological sections (Loqman et al., 2010) and, therefore, be more representative of the cartilage surface.
Although chondrocytes comprise the vast majority of cells within the SZ of cartilage, a tiny but potentially important population of progenitor cells has been identified. Using cell surface markers, Dowthwaite et al. (2004) reported that in normal calf cartilage 1%-2% of the cell population were progenitor-like cells and could differentiate into chondrocytes. The group also identified chondrogenic progenitor cells (CPCs) in normal human femoral condylar cartilage, which accounted for 0.7% of the total cell population (Williams et al., 2010). Of particular interest in the present context is that the cartilage progenitor cell population expresses chondrogenic markers and matrix components including collagen type I and II.
Although the synthesis of type I collagen is usually considered an indicator of fibrocartilage, the cartilage progenitors appear to follow a developmental process in matrix synthesis initially with collagen type I production followed by collagen type II production as the cells mature (Craig et al., 1987). Indeed, evidence of the presence of both collagen types at the cartilage surface has been described during cartilage development in the chick diarthrodial joint (Archer et al., 1994). This raises the interesting possibility that some of the cells we observed which have cytoplasmic processes and label positively for collagen type I, might in fact be cartilage progenitor cells.
However, the proportion of cells with one or more cytoplasmic process was approx. 10% (Figures 2 and 3a) and so considerably higher than that expected solely for the progenitor cell population.
A potential explanation is that there are two cell types comprising It was perhaps unexpected that there was not a closer relationship between levels of cell-associated collagen type I and chondrocyte morphology. The hypothesis was that chondrocytes with abnormal morphology would demonstrate increased labelling, whereas those of normal morphology would not, but this was clearly not the case (Figures 2 and 3). This suggests an "uncoupling" between these two processes as a significant proportion of chondrocytes possessed "normal" morphology (i.e., no cytoplasmic processes were observed) however, they exhibited strong collagen type I labelling ( Figure 2). Thus, while changes to chondrocyte shape might suggest a shift to a fibroblastic tissue "repair" phenotype, it is unlikely that cell shape, per se, controls chondrogenesis, but rather the organisational status of the chondrocyte cytoskeleton and its interaction with second messenger pathways. Although modification of the actin cytoskeleton does not appear to control the full chondrogenic phenotype (Benya et al., 1988;Watt & Dudhia, 1988;Mallein-Gerin et al., 1991;Parreno et al., 2017), the manipulation of other cytoskeletal proteins which bring about the morphological changes can influence matrix metabolism. For example, tubulin polymerisation reduced the production of interleukin (IL)-1β and protease gene expression in primary chondrocytes (Hui et al., 1998) whereas disruption of vimentin intermediate filaments (reportedly in the absence of changes to F-actin labelling) has been observed in chondrocytes from OA cartilage possibly leading to downstream regulation of protease activity (Lambrecht et al., 2008). Although there might not be a strict relationship between chondrocyte morphology and levels of collagen type I, there is a correlation between the number and length of cytoplasmic processes per chondrocyte and levels of cell-associated IL-1β (Murray et al., 2010). The presence of this potent proinflammatory cytokine will likely have profound effects on the local synthesis/release of degradative enzymes and matrix constituents. We should also be alert to the fact that the present results are only a single "snap shot" in time of the events occurring in cartilage. For example, it is possible that increases in collagen type I levels precede the development of cytoplasmic processes possibly as a result of changes to the lacunar/pericellular structure surrounding chondrocytes (Guilak et al., 2018). In any event, it seems probable that dedifferentiation and any phenotypic changes leading to alterations to matrix metabolism are progressive and complex time-dependent processes. These appear to occur initially in only a relatively small subset of SZ chondrocytes and deserve further study to better understand early degenerative changes in cartilage.
Although there seems to be general agreement that levels of type I collagen are increased as cartilage degenerates (e.g., Fukui et al. 2008;Brew et al. 2010), this begs the question of whether there is evidence of any collagen type I in normal nondegenerate cartilage. This is not a trivial point because as it might be assumed that there was none, the method of analysis needs to be carefully considered in addition to whether the human cartilage was truly nondegenerate. For example, the apparent absence of a mRNA signal for type I collagen in chondrocytes in normal cartilage (e.g., Aigner et al., 1997) does not necessarily rule out the presence of this collagen type in a small subset of cells in the superficial zone. On the other hand, although mRNA levels might be elevated, there are posttranscriptional, posttranslational and protein degradation events which will control final steady-state levels of the protein. Although there is a correlation between cellular concentrations of proteins and relevant mRNA abundance, it is not particularly strong and it has been estimated that in some situations only about 40% of the protein content can be accounted for by the mRNA abundance (Vogel & Marcotte, 2012).
Immunofluorescence with high resolution imaging (e.g., CLSM), is, however, a powerful tool for the localisation of ECM proteins (Hayes et al., 2008). We acknowledge that while we have been able to determine cell-associated levels of collagen type I, these are semiquantitative based on fluorescence determinations and their physiological significance in terms of mechanically relevant parameters, for example, loadbearing, are difficult to assess without specific techniques such as atomic force microscopy. Furthermore, while the present data suggest that there might be a small population of cells with cell-associated type I collagen (Figure 2) in nondegenerate cartilage, there is a concern about whether the grading system would be sufficiently sensitive to determine if the cartilage was truly grade 0 or, for example, grade 0.5 where a more detailed microscopical analysis of the cartilage might be required. This could potentially also include chondrocyte morphology, the presence of collagen type I, and the cartilage surface ( Figure 5). Thus, it is conceivable that while macroscopically consistent with a grade 0 assessment, the relatively aged cartilage used in the present study