ACAN: Aggrecan; ICA: Autologous Chondrocyte Implantation; CRM: Confocal Raman Microscopy; Cir: Circularity; 2D: Two-dimensional; ECM: Extracellular Matrix; GAGs: Glycosaminoglycans; hCSMs: Human Mesenchymal Stem Cells; IRMB: Institute of Medicine Regenerative and Biotherapies of Montpellier; LDA: Linear Discriminant Analysis; PBS: Phosphate-buffered Saline; PCs: Principal Components; PCA: Principal Component Analysis; 3D: Three-dimensional; COL: Type II Collagen.

The avascular nature of articular cartilage makes it a tissue that regenerates little and heals with difficulty spontaneously, particularly in adulthood [1]. This situation has prompted the development of new therapeutic approaches to replace damaged cartilage. Autologous chondrocyte implantation (ICA) is one of the most widely used cellular repair strategies for articular cartilage. The idea is to fill the cartilage defect with autologous chondrocytes harvested arthroscopically from a low-lying or non-bearing site in the diseased joint. The in vitro-expanded chondrocytes are then re-implanted at the site of the cartilage defect directly as a cell suspension. However, the long expansion period and multiple passages generally required to obtain usable quantities of chondrocytes lead to "dedifferentiation" of the isolated chondrocytes [2]. Chondrocyte dedifferentiation is defined as the progressive loss of the molecular markers that define a differentiated chondrocyte.

In general, chondrocytes are surrounded by a cartilage-like extracellular matrix (ECM), a complex network rich in glycosaminoglycans, proteoglycans and collagen which, together with a multitude of intracellular signaling molecules, triggers chondrogenesis and enables the chondroprogenitor to acquire the spherical morphology of chondrocytes [3]. Isolating chondrocytes from their natural three-dimensional (3D) environment, then expanding them under two-dimensional (2D) monolayer culture conditions in vitro, inevitably leads chondrocytes to adopt a different morphology that enables them to adhere to plastic in order to survive. Conventional 2D monolayer cultures are considered unsuitable, as the monolayer expansion of articular chondrocytes does not allow the chondrogenic phenotype to be maintained during passage. Indeed, as the dedifferentiation process persists, the synthesis of ECM molecules such as type II collagen (COL II) and glycosaminoglycans (GAGs) tends to disappear in favor of the production of different extracellular proteins (e.g. type I and X collagens (COL I and COL X)) that incorporate into the matrix, thus offering inferior mechanical properties [4,5].

Chondrocyte dedifferentiation is emerging as a major concern for the functional reconstruction of articular hyaline cartilage. Clearly, there is a crucial need for in-depth assessment of the molecular characteristics of articular cartilage to understand biochemical changes and tailor tissue engineering constructs.

Numerous studies have been carried out either to improve culture conditions in order to optimize the mode of expansion and number of chondrocyte passages [6,7], or to determine a type of culture, notably 3D cultures, in order to obtain a differentiated chondrocyte phenotype [2]. In the context of tissue engineering, monitoring the biological properties of cells and/or their associated modifications provides important information concerning cell status, such as viability and functional state [8].

Raman spectroscopy is emerging as a promising tool for monitoring intracellular processes in vivo and at the molecular level. It is a well-established analytical technique, capable of determining chemical composition and molecular interactions in micrometric samples [9]. One of the most important features of Raman spectroscopy is its ability to measure molecular changes over time in living cells, revealing biochemical processes impossible to obtain with other imaging methods [10-12]. Recently, the application of RCM has enabled us to characterize the ECM [13], and probe the difference between the status of undifferentiated and differentiated cells [14] and the ability to detect proteoglycans and GAGs in differentiated chondrocytes. Due to its non-destructive and label-free characteristics, this technique seems suitable for in situ measurements and would enable rapid detection of biochemical changes, thus significantly improving investigative possibilities [15].

The aim of this work was to characterize the changes occurring in the phenotype of chondrocytes isolated from their natural 3D environment, and to monitor the underlying biochemical changes in real time using confocal Raman microscopy.

Cell source

The cell source was a cartilage sample taken from a healthy site on the knee of a patient at a distance from the lesion site during knee surgery after informed consent. Chondrocytes were isolated by enzymatic digestion with collagenase, then formed into different groups expanded in 2D monolayer cultures by a series of passages at different times. For each group, cells underwent several passages up to passage four (P4) at a rate of one week per passage, during which the monolayer growth rate was deemed insufficient to maintain cell numbers sufficient for further production. Chondrocytes were directly expanded onto CaF₂ slides pre-treated with Cell-Tak®, a widely used polyphenolic protein-based tissue adhesive (Corning Cell-Tak), known to bind strongly to virtually all inorganic and organic surfaces in aqueous environments [16], to facilitate their adhesion and thus avoid further additional manipulations that could alter cell integrity. All cell cultures were produced and supplied by the Daniele team at the Institute of Medicine Regenerative and Biotherapies of Montpellier (IRMB). These experiments were carried out in accordance with the guidelines and regulations of the Ethics Committee of the Montpellier University Hospital of Languedoc-Roussillon, with the approval of the French Ministry of Education, Higher Education and Research. All slides and cells were immersed in Petri dishes containing phosphate-buffered saline (PBS) for Raman measurements.

To corroborate the results of the Raman measurements, quantification analyses of certain genes, including collagen A2 type I (COL1A2), collagen A1 type II (COL2A1), aggrecan (ACAN) and gene 9 (SOX9), involved in chondrogenesis were evaluated, in parallel using real-time quantitative PCR (RT-qPCR), by the Daniele team at the IRMB. Quantitative real-time PCR (qPCR) was performed with supermix IQ SYBR Green (Bio RAD) using a RT-qPCR amplifier (Bio-RAD). The expression level of each target gene was calculated using a formula 2^∧(-ddCt) [17].

Morphological analysis of dedifferentiated chondrocytes

To observe the phenomenon of dedifferentiation, several optical images were taken for each passage using a wide-field optical microscope with a CP-ACHROMAT 10x/0.25 Ph1 objective. Where necessary, image calibration was carried out using the optical and digital magnifications used during scanning. The optical images obtained were then processed using Image J® software, which provides a set of calculation tools [17]. Users expect a set of numerical results characterizing the shape, size and nature of the structuring elements [18], both to characterize and to typify cellular structures and their arrangement, as shown in (Table 1). All images were processed at a scale ranging from "500 pixels to infinity" for all samples, while various parameters were recorded on the Image J platform. Cell morphology was assessed on the basis of the value of circularity, a standardized index classically expressed as the weighted ratio between cell area and perimeter [19]. This circularity can also be determined by calculating the shape factor (S) according to the formula below:

Where A corresponds to the surface (or area) calculated from the sum of the pixels making up the cell. The measurement is given in pixels squared. p corresponds to the perimeter defined as an approximation of the actual length of the cell edge and π = 3.14159. Circularity is therefore given by a value that varies between 0 and 1, 1 being the circularity of the perfect circle. Thus, for S= 1, the cells are considered to have a circular shape and for S< 1 or close to 0, those have a rather elongated shape (characterizing the fibroblastic phenotype).

Analysis and processing of Raman spectral data

For each group and each run, Raman spectral measurements and Raman optical images were taken. A total of 100 spectra were collected for each run, taken at different locations and on different cells. Before subjecting the spectral data to statistical analysis, all collected Raman spectra were pre-processed using well-established and previously described techniques [20], in order to obtain spectra with the same scale and thus enable consistent comparisons where intensity variations may be relative to the intensity of each spectrum [21]. The wavelength range (or "fingerprint") between 600 – 1800 cm^-1 was considered for spectral analysis, due to its higher molecular specificity and based on the study by Pudlas et al. [22]. On the basis of normalized Raman band area ratios, the monitoring and characterization of the various samples was effective by relying on certain Raman bands, notably those of DNA, collagen, proteoglycans and lipids. This made it possible to assess the relative concentration of biomolecules within each sample [23].

Multivariate analyses such as Principal Component Analysis (PCA) and Linear Discriminant Analysis (LDA) were also carried out. PCA is a multivariate analysis method that reduces spectra into a defined number of principal components (PCs) that take spectral variance into account. It takes into account all the pixels in the spectra to describe the variation in the image using a small number of basic contributions (principal components) linked to spectral and spatial behavior [24]. It therefore solves the aforementioned problems of band-intensity heatmaps and is therefore better suited to handling biological variations. In the case of our study, PCA was used to highlight the variability existing in the recording of spectral data sets during different experiments. As an unsupervised method, PCA was used for classification and prediction purposes. In addition, LDA was used, because unlike PCA, LDA explicitly attempts to model the difference between data classes. The various PCA and LDA analyses were performed on the raw data matrix using R software version 3.6.1 (2019-07-05).

Statistical analysis

Given that the distribution of our data did not follow a normal distribution, all our statistical analyses were carried out using a non-parametric test, namely the Kruskal-Wallis One Way Analysis of Variance on Ranks statistical test with p<0.001 for the significance of our results. For comparisons between different zones, the Student-Newman-Keuls statistical test for p-value < 0.05 was applied. All our data were processed with SigmaPlot for Windows software version 11.0 Build 11.0.0.77.

Morphology of dedifferentiated chondrocytes

Morphological changes in chondrocytes were highlighted by determining cell circularity (Cir) on the basis of optical images collected at each passage. Looking at the table generated by the software, we can see that circularity decreases progressively as the number of passages increases. From P1 to P4, the form factor decreases significantly from 0.072 to 0.032 (Table 1). Optical Raman images also show major morphological changes with each passage. Cells at P4 show a larger morphology with a more elongated appearance (Figure 1). Cells at P2 and P3 have approximately the same cell density, but also the same morphology and a slightly less elongated appearance compared to P4.

Spectral data for dedifferentiated chondrocytes

The relative concentration of biomolecules was characterized on the basis of certain Raman bands chosen for their molecular specificity, and also for their ability to provide important biochemical and structural information about the cells. Calculation of the mean and standard deviation of the intensity ratios of each Raman band (n= 100 spectra/passages) made it possible to discriminate between different passages and to appreciate the molecular differences within the chondrocyte spectra over the passages. In (Figure 2), the different passages are clearly classified into clusters thanks to the application of PCA-LDA, a multivariate analysis method for discriminating different samples on the basis of differences in their biochemical content and identifying spectral characteristics.

Observations: Groups A, B, C, and D represent the different passages respectively

The dedifferentiation phenomenon can be followed from the Raman bands located at 788 cm^-1 (DNA/RNA expression), 1063 cm^-1 (GAG expression), 1255 cm^-1 (collagen II or Col II expression), 1449 cm^-1 (lipid expression) and 1659 cm^-1 (collagen I or Col I expression) (Table 2). These bands show more or less the same characteristics in both groups, with nucleic acid content tending to increase with each passage. GAG content decreases progressively after a peak at P2, as does lipid content. Col II decreases progressively with passage, while Col I increase with passage (Figure 3A and 3B). Parallel qPCR analyses on the same samples showed the expression of type IIA and IIB collagen and ACAN to be negatively regulated, while that of type I collagen (Col A1) and the SOX9 gene were positively regulated (Figure 3C).

The introduction of tissue engineering into the biomedical field has raised great hopes for the development of functional cartilage tissue for the repair of damaged articular cartilage. Approaches incorporate the implantation of autologous chondrocytes (ICAs) or the use of human mesenchymal stem cells (hCSMs) to induce biomechanically and functionally equivalent cartilage tissue. However, efforts to date have failed to meet patient expectations, as these cells tend to adopt a fibroblastic phenotype once implanted [25–27]. Chondrocyte dedifferentiation is a troubling and unresolved issue in clinical applications and tissue engineering, in which chondrocytes undergo changes in morphology and molecular markers [25]. In this study, using CRM, we successively assessed the morphological and biochemical changes associated with the phenomenon of chondrocyte dedifferentiation and identified characteristic Raman markers.

Morphological changes

In (Figure 1), the morphological changes in the cells are perceptible as the number of passages increases. Thus, from passage P1, where the cells appeared more or less round and without flagella, they adopt an increasingly elongated shape with more developed flagella. This observation was correlated with cell circularity (S) values at each passage (Table 1, row Cir). It is interesting to note that from P1 to P4, circularity decreases progressively. This clearly shows a change in cell morphology, from a more or less round shape (S= 0.072) to a more elongated, fibroblastic shape (S= 0.032). Our results support the study by Schuh et al. [28], in which they showed that monolayer expansion of chondrocytes was associated with changes in the organization of the F-actin cytoskeleton, forming well-defined stress fibers and exhibiting more of a fibroblastic phenotype. In general, chondrocytes are isolated from cartilage tissue by enzymatic digestion using collagenase and then expanded in monolayer culture to amplify the total number of cells [26,29,30]. These cells are then used to fill cartilage defects and produce a new matrix. However, under in vitro monolayer culture conditions, chondrocytes are forced to abandon their round shape in order to adhere to the plastic in order to survive, resulting in their dedifferentiation [29,31,32].

In Figure 2, the different passages are clearly classified into clusters thanks to the application of PCA-LDA, a multivariate analysis method for discriminating different samples on the basis of differences in their biochemical content and identifying spectral characteristics. Each spectrum is represented on the score graph by a point. Similar spectra are grouped together in certain regions, while very different spectra are widely separated. This allows spectra to be classified according to their similarity, making it easier to appreciate the differences between different samples over time. Remarkably, the P2 and P3 clusters appear to be virtually merged, no doubt due to the strong spectral similarities between these two passages (Figure 2B). Otherwise, this arrangement could lead us to deduce that, from P2 to P3, there would really be no great changes. Indeed, the distribution of spectral signals is the result of the relative concentrations of the sources in each sample, according to their biochemical specificities. The great similarities between P2 and P3 passages could lead us to deduce that P2 or even P3 passages would be largely reasonable to obtain a sufficient number of cells capable of being redifferentiated in 3D cultures. Indeed, it is now known that a 3D chondrogenic culture system can effectively induce dedifferentiated chondrocytes to become functional, redifferentiated chondrocytes, enabling them to regenerate mature cartilage [33,34].

Biochemical changes

To identify the Raman markers associated with chondrocyte dedifferentiation, several spectra were collected for each passage. Whatever the passage, the Raman spectra collected appeared virtually identical at first glance. However, by analyzing the intensity ratios of certain Raman bands, notably the Raman bands at 788 cm^-1, 853 cm^-1, 942 cm^-1, 1004 cm^-1, 1063 cm^-1, 1125 cm^-1, 1255 cm^-1, 1449 cm^-1 and 1659 cm^-1, differences were observed between the different passages. Although these bands have more or less the same characteristics, their relative intensities and ratios vary in both samples, indicating differences in the relative distribution of biomolecules. The variation in band intensity ratios allows us to follow the dedifferentiation of chondrocytes. In particular, we focused on the bands at 1063 cm^-1, 1255 cm^-1 and 1659 cm^-1, attributed to GAGs, type II collagen and type I collagen, respectively. Statistical analysis shows a significant decrease (p<0.001) in the GAG and type II collagen bands, particularly between P1 and P3, while the type I collagen band increases significantly (p<0.001). This reflects a reduction in the main markers of hyaline cartilage, in favor of cartilage rich in type I collagen and therefore fibroblastic. As highlighted above, these changes are due to the 2D monolayer culture conditions and the number of cell passages, which predispose chondrocytes to a phenotypic change, leading them to abandon their intrinsic characteristics and adapt to culture conditions by adopting a fibroblastic phenotype. These results were compared with qPCR quantification analyses to assess the expression of key chondrogenic markers. The qPCR analyses showed that, over the course of passages, the expression of chondrogenic genes, particularly type II collagen (in its two isoforms A and B) and ACAN, was negatively regulated, in contrast to type I collagen, which was more positively regulated (Figure 3C). These same observations have been made in studies showing similar phenotypic changes when chondrocytes expand in monolayer culture in vitro [24,34]. In any case, chondrocytes in prolonged monolayer culture, instead of producing the specific components of cartilage (type II collagen and specific proteoglycans), lead instead to the production of non-specific proteoglycans and type I collagen [35].

Monitoring tissue growth in cell culture remains a major challenge, despite current analytical techniques that probe numerous factors such as RNA expression, intracellular/extracellular protein expression and exogenous components. CRM has been proposed as an analytical method capable of providing a "fingerprint" representing the entire molecular composition of the sample under study, without a priori and without labeling.

In our study, the evaluation of chondrocyte dedifferentiation processes enabled us to monitor intracellular and extracellular biochemical changes in real time. The sensitivity and specificity of the method for single-cell investigations make CRM a promising tool for biomedical applications.

No conflicts of interest are reported.

N'DRE NJ would like to thank the Ivorian Ministry of Higher Education and Scientific Research for its financial support in the form of a scholarship, which enabled him to carry out his research work.

We would like to thank all the authors for their contributions to this manuscript, and our partners, in particular the Bioengineering and Nanosciences Laboratory (LBN) and the Institute of Regenerative Medicine and Biotherapies (IRMB), for enabling us to carry out our work.

N'DRE NJ was responsible for carrying out the experiments, analyzing and interpreting the results, and preparing the manuscript. LECOR PA, ASSOUMOU AA and BLOHOUA MJJ edited the manuscript. All authors contributed comments and suggestions prior to publication.