A Nondestructive Raman Spectral Method for Temporal Tracking of Articular Cartilage Maturation

Animals and tissue preparation

In this study, no live animals were used, and all tissues were obtained from deceased animals culled for other purposes. Fresh-frozen whole fetal and neonatal pigs (Sus scrofa domesticus, Yorkshire cross, female and male) were acquired from Nebraska Scientific. The fetuses of length 17.8 cm–22.9 cm were estimated to be approximately 80 days of gestational age (“Fetal”), and pigs over 35.6 cm that were stillborn were considered “Neonatal.” Knees from juvenile (5–6-month-old) and mature (2–3-year-old) pigs (Sus scrofa domesticus, Yorkshire cross, female and castrated male) were purchased from Corona Cattle, Inc. Prior to sample collection, juvenile and mature knees were fresh-frozen en bloc and subsequently thawed to ensure consistency with fetal and neonatal groups. For fetal and neonatal pigs, unilateral (only the right) knee joints were used, and for juvenile and mature pigs, bilateral knee joints were used.

Knee joint capsules were dissected and opened to reveal femoral condyles with smooth, pristine AC. A total of 20 knees were used, as follows: 5 knees from fetal, 5 knees from neonatal, 5 knees from juvenile, and 5 knees from mature pigs. A full-thickness slice of cartilage, approximately 3 mg in wet weight, was taken from the center of the lateral condyle for biochemical analysis, a smaller full-thickness cartilage slice of approximately 1 mg wet weight from the lateral condyle was used for crosslink analysis, and a 3 mm diameter biopsy punch was taken from the medial condyle for Raman spectroscopy.

Biochemical characterization by Raman spectroscopy

Raman spectra were collected from the surface of cartilage samples. To prevent tissue dehydration and shrinkage during measurement, which can alter spectral features and absolute intensities, the following precautions were taken. The samples were kept hydrated in phosphate-buffered saline, blotted dry, and centered under the objective. Spectra were acquired within 2–3 min of transfer to minimize any potential fluid loss using a modified Renishaw InVia confocal Raman microscope (Renishaw, Inc.), equipped with a near-infrared laser (λ = 785 nm) with incident laser radiation of 109 mW (Laser Spectroscopy Laboratories fee-for-service core, University of California, Irvine). A 20× long-working-distance objective (NA = 0.75) was used for excitation and signal collection. Five scans per spot and five spots per sample were acquired at a resolution of 1 cm⁻¹ with an acquisition time of 5 s to achieve a good signal-to-noise ratio. Spectra were acquired, smoothed, linear baseline corrected per Lorentz fitting, and normalized using WiRE 4.4 (Renishaw, Inc.). Multiple characteristic Raman shifts are analyzed and summed because some peaks become distinguishable earlier in tissue development or molecular organization, enabling early detection of structural changes, where peak shape reflects the degree of (dis)organization, and amplitude indicates quantity. For quantitative analyses, the sum of relative intensity of the main peak(s) was plotted per sample, as shown in the Results section.

The confocal Raman probe (with a lateral resolution of ∼1 μm and a depth of field of ∼2 μm) samples a microscopic volume within the tissue. To avoid bias from local cellular or matrix heterogeneity, we acquired five spectra per sample, each from a spatially distinct spot separated by at least 50 μm across the cartilage surface. This approach ensures the measurement integrates signal from multiple chondrocytes (where present in the sampling volume) and the surrounding extracellular matrix (ECM), providing a tissue-averaged biochemical profile.

The Raman signal originates from the entire focal volume, which includes both chondrocytes and the pericellular/territorial matrix. In immature, cell-dense cartilage (fetal/neonatal), the sampling volume is highly likely to include one or more chondrocytes. In mature cartilage, with lower cellularity, the signal becomes increasingly dominated by the ECM, but chondrocytes within the sampled volume still contribute. This is directly analogous to the destructive biochemical assay, where a homogenized tissue fragment includes DNA from all chondrocytes within it.

Raman shifts of interest were selected based on historical and current literature and can be found in Table 1. The main Raman shifts used in the current work are as follows: 1096 cm⁻¹, corresponding to the phosphate mode of DNA^15,20; 880 cm⁻¹ to 1640 cm⁻¹, corresponding to “very strong” peaks for GAGs including chondroitin-sulfate 4, chondroitin-sulfate 6, and hyaluronic acid (HA), ¹⁶ where “total GAG content” is the sum of all peak intensities; 810 cm⁻¹ to 1275 cm⁻¹, corresponding to collagen-specific proteins including the C–C bond stretching of proline. ¹⁷

OPEN IN VIEWER

Table 1.

Raman Spectroscopy Characterization of Articular Cartilage Biochemistry

Measurand	Method	Raman shift(s)	Description	Refs.
DNA Content	Peak intensity	1096 cm−1	Symmetric stretching of -PO2	15
Chondroitin-sulfate 4 (CS4)*	Peak intensity	Sum of 1635 cm−1, 1411 cm−1, 1376 cm−1, 1341 cm−1, 1050 cm−1, 1079 cm−1	“Very strong” Raman shifts	16
Chondroitin-sulfate 6 (CS6)*	Peak intensity	Sum of 1615 cm−1, 1413 cm−1, 1377 cm−1, 1340 cm−1, 1062 cm−1, 995 cm−1, 937 cm−1, 884 cm−1
Hyaluronic acid (HA)	Peak intensity	Sum of 1412 cm−1, 1375 cm−1, 1124 cm−1, 949 cm−1, 899 cm−1
Total glycosaminoglycan (GAG) content	Peak intensity	Sum of CS4, CS6, and HA
Collagenous protein	Peak intensity	Sum of 1271 cm−1, 1246 cm−1, 920 cm−1, 857 cm−1, 816 cm−1	Raman shifts associated with collagen-specific amino acids	17
Pyridinoline (Pyr)	Peak intensity	1660 cm−1	Raman shifts associated with mature crosslinks (trivalent)	18,19

*

CS4 and CS6 are sulfated GAGs.

Due to the spectral complexity of cartilage, where collagen and GAG signals extensively overlap, we did not assign collagen content to a single Raman band. Instead, the “total collagen-associated signal” was defined as the sum of baseline-corrected peak intensity from several prominent peaks assigned to collagenous protein backbone and side chain vibrations. These include, but are not limited to: the Amide I band (∼1660 cm⁻¹, C=O stretching), the Amide III band (∼1240 cm⁻¹ to1280 cm⁻¹, C–N stretching and N–H bending), and peaks associated with the amino acids proline and hydroxyproline (∼850 cm⁻¹ and ∼875 cm⁻¹, C–C stretching). This summed intensity provides a robust proxy for total collagenous protein content, as it integrates signal from multiple vibration modes of the collagen molecule, reducing bias from any single, potentially overlapping band.

Lastly, the Raman shift at 1660 cm⁻¹, which corresponds to mature collagen trivalent crosslinks via Pyr, has been evaluated as a potential marker of tissue maturation and tissue degeneration.^18,19,21 But, unlike previous work evaluating collagen crosslinking, where peak area measurements were used for pyridinoline analysis, the current study employed relative intensity measurements as a simplified approach for qualitative and quantitative characterization of collagen crosslinking and maturity.

Biochemical quantification by biochemical assay

To validate the Raman spectroscopy measurements, parallel analyses were performed using established, destructive biochemical assays on the same sample. All biochemical samples were processed using standardized protocols to ensure consistency and comparability. Briefly, biochemistry samples were weighed wet, then frozen and lyophilized to acquire dry weights. Lyophilized samples were digested in papain (Sigma-Aldrich, St. Louis, MO) at a concentration of 125 μg/mL for 18 h at 60°C. Total protein was calculated via a modified Bradford assay (Bio-Rad, Hercules, CA). DNA content was calculated with the use of a PicoGreen dsDNA reagent (Invitrogen, Carlsbad, CA). The micro assay protocol was followed, which called for a standard curve composed of 5 × 1 mL solutions containing 0 μg/mL, 10 μg/mL, 20 μg/mL, 30 μg/mL, 40 μg/mL, and 50 μg/mL of bovine serum albumin (BSA), as BSA is the most abundant globular protein and commonly used as a protein standard. Collagen content was measured with the use of a modified chloramine-T colorimetric hydroxyproline assay as described previously. ²² For comparative analysis, sulfated GAG (sGAG) content was quantified using the dimethylmethylene blue (DMMB) assay-based Blyscan assay kit (BioColor, Belfast). All quantification measurements for DNA, sGAG, and collagen content were performed with a GENios spectrophotometer/spectrofluorometer (TECAN) as previously described. ²³

Liquid chromatography-mass spectrometry (LC-MS) was performed to quantify pyridinoline crosslinks within cartilage. Lyophilized samples were hydrolyzed in 6 N HCl at 105°C overnight, HCl was evaporated, and hydrolysates were resuspended in 500 μL of a solution of 25% (v/v) acetonitrile, 0.1% (v/v) formic acid, and 1 μg/mL pyridoxine (internal standard) in water. Resuspended samples were injected into a Cogent Diamond Hydride 2.o HPLC column and eluted using a gradient of two solvents (A: 0.1% formic acid in water and B: 100% acetonitrile) with a flow rate of 400 μL/min for aqueous normal phase separation, coupled to a Waters ACQUITY QDa mass spectrometer. ²³ Pyridinoline crosslinks and hydroxyproline were quantified in MassLynx v4.1 by quantifying the area under the curve of the extracted ion chromatograms of the second charge state of pyridinoline ([M + 2H]²⁺ = 215.1 m/z) and the first charge state of hydroxyproline ([M + H]⁺ = 132.1 m/z), and interpolating into a standard curve of serial dilutions of pyridinoline and hydroxyproline. ²⁴ Pyridinoline content was normalized to dry weight and hydroxyproline concentration.

Statistical analysis

For each biochemical (Raman spectroscopy, liquid chromatography) and spectroscopic (Bradford, PicoGreen dsDNA, Sircol collagen, chloramine-T hydroxyproline, and Blyscan GAG assays) test, n = 5 samples were used. All statistical analyses were performed using Prism 8 (GraphPad Software). Match-paired quantitative data were assessed using a one-way analysis of variance (ANOVA) with a post hoc Tukey’s HSD (honestly significant difference) test at a significance level of α = 0.05. Correlation and correlation probabilities of cartilage photometric and spectroscopic datasets were analyzed by JMP^® Pro 15.2.1 (SAS). All data are presented as means ± standard error and can be found in Supplementary Data. For all figures, statistical significance is indicated by groups not sharing the same letters. Correlation coefficients of R² ≥ 0.5000 and p values of < 0.05 are highlighted in the figures below.

Spectral analysis of cartilage development

The Raman spectra obtained from porcine AC across four developmental stages (Fig. 1) revealed distinct biochemical changes associated with tissue maturation. Specifically, the fingerprint region most informative for biological molecules (720 cm⁻¹ to 1820 cm⁻¹) was baseline-corrected and normalized.

OPEN IN VIEWER

FIG. 1.

Baseline corrected and normalized Raman spectra of farm pig articular cartilage at different developmental stages obtained at a λ = 785 nm with incident laser radiation of 109 mW, 5 scans, 1 cm⁻¹ resolution and 5 s acquisition time. Red arrows point to collagenous protein, and the green arrows point to noncollagenous protein.

Several clear spectral trends were observed. Fetal cartilage spectra showed broader, less defined peaks, whereas mature cartilage spectra exhibited sharp, well-resolved peaks. Specific baseline-corrected relative peak intensity changes were observed that correlated with the developmental stage. The amide I band (approximately 1660 cm⁻¹), associated with C=O stretching vibrations in peptide bonds, ²⁵ increased in intensity and sharpened with maturation. Peaks in the 1000 cm⁻¹ to 1100 cm⁻¹ region, corresponding to carbohydrate vibrations in GAGs, ²⁶ also changed in relative peak intensity and profile shape across stages.

The ratio of the 1450 cm⁻¹ peak (CH₂ bending ²⁷ ) to the 1240 cm⁻¹ peak (amide III^28,29) decreased with maturation. In addition to intensity differences, several bands showed subtle shifts in position. For instance, the amide I band shifted from approximately 1655 cm⁻¹ in fetal tissue to 1665 cm⁻¹ in mature tissue, the complex region between 1000 cm⁻¹ and 1100 cm⁻¹ showed changes in relative peak ratios across developmental stages.

These observations confirmed that Raman spectroscopy could detect meaningful biochemical changes during cartilage development. The next step was to extract quantitative information from these spectra by focusing on specific biomarker peaks associated with key matrix components.

DNA content

Cellularity, as assessed by DNA content, showed characteristic changes during cartilage development. Both Raman spectroscopy and the destructive PicoGreen assay detected significantly higher DNA content in fetal and neonatal cartilage compared to juvenile and mature tissue (Fig. 2).

OPEN IN VIEWER

FIG. 2.

PicoGreen photometric and Raman spectroscopic analysis of DNA content of cartilage with test methods separated by dashed line (|). Correlation (R²) and correlation probability (p value) of matched pair data with R² > 0.5 and p value < 0.1 highlighted (in red). DNA content was compared between Raman spectra and PicoGreen assay values normalized to dry weight, since cartilage has a variable extracellular matrix composition, and normalization ensures that DNA reflects cell content rather than total tissue mass. All data are presented as means ± standard error mean, and statistical significance is indicated by groups not sharing the same letters.

The Raman measurements focused on the peak at 1096 cm⁻¹, corresponding to the symmetric stretching vibration of phosphate groups in the DNA backbone.^6,7 The relative intensity of this peak was highest in fetal cartilage, decreased in neonatal tissue, and was lowest in juvenile and mature cartilage. Statistical analysis confirmed that all pairwise comparisons between groups were significant (p < 0.001) except between juvenile and mature tissue.

Glycosaminoglycan content

GAG content, a key indicator of cartilage functional capacity, showed complex changes during development. Both Raman spectroscopy and the destructive DMMB photometric assays revealed developmental differences in sGAG content of native cartilage (Fig. 3). Fetal and neonatal tissues showed the highest sGAG levels, while juvenile and mature tissues showed significantly lower levels. A strong correlation (R² = 0.6401) was observed between the cumulative relative intensity of Raman peaks (corresponding to CS4 and CS6) and the photometric quantification of GAGs in cartilage samples as a function of developmental stage (Fig. 3). The correlation between Raman and DMMB measurements was strong (R² = 0.6401, p < 0.010) and suggests that both methods capture the same biological trends.

OPEN IN VIEWER

FIG. 3.

DMMB photometric assay and Raman spectroscopic analysis of sGAG content analysis of cartilage with test groups separated by dashed line (|). Correlation and correlation probability of matched pair data with highlighted R² > 0.5 and p value < 0.1 (in red). DMMB, dimethylmethylene blue; sGAG, sulfated glycosaminoglycan. All data are presented as means ± standard error mean, and statistical significance is indicated by groups not sharing the same letters.

Pyridinoline crosslinks

Pyridinoline crosslinks, which stabilize the collagen fibrils and contribute to tissue mechanical properties, showed dramatic increases with cartilage maturation (Fig. 5). Both Raman spectroscopy and LC-MS mass spectrometry detected this pattern, though the absolute values differed between methods.

OPEN IN VIEWER

FIG. 4.

Percent collagen as measured by hydroxyproline photometric assay and Raman spectroscopic analysis of total collagen content of cartilage with Raman (left) and photometric (right) test groups separated by a dashed line (|). Correlation and correlation probability of matched pair data with highlighted R² > 0.5 (yellow) and p value < 0.1 (red). All data are presented as means ± standard error mean, and statistical significance is indicated by groups not sharing the same letters.

OPEN IN VIEWER

FIG. 5.

Mass spectrometric and Raman spectroscopic analysis of pyridinoline content analysis of cartilage with test groups separated by dashed line (|). Correlation and correlation probability of matched pair data with highlighted R² > 0.5 (yellow) and p value < 0.1 (red). All data are presented as means ± standard error mean, and statistical significance is indicated by groups not sharing the same letters.

The Raman measurements focused on the peak at 1660 cm⁻¹, which is associated with the mature trivalent pyridinoline crosslinks.^18,19,21 The relative intensity was very low in fetal and neonatal cartilage, increased substantially in juvenile tissue, and was highest in mature cartilage. All pairwise comparisons were statistically significant (p < 0.001). The LC-MS measurements quantified pyridinoline content normalized to both dry weight and hydroxyproline content. The correlation between Raman and LC-MS measurements was excellent. For Raman intensity versus pyridinoline/dry weight: R² = 0.8403, p < 0.001. For Raman intensity versus pyridinoline/hydroxyproline: R² = 0.8327, p < 0.001. The slopes were 0.89 and 0.91, respectively.

Photometric and spectroscopic Maturity Index

Based on the individual biochemical parameters, we developed a composite MI that integrated information about GAG content, collagen content, and pyridinoline crosslinking into a single quantitative metric (Fig. 6). The MI was designed to approach a value of 1 for fully mature tissue, with lower values indicating less mature tissue. Values > 1 are possible due to variance in the system.

OPEN IN VIEWER

FIG. 6.

Maturity Index comparison of Fetal–Neonatal cartilage as determined by photometric (destructive) and spectroscopic (nondestructive) methods with annotated R² linearity (dotted line). All data are presented as means ± standard error mean.

Raman-based MI (Fig. 6) was modeled after the previously established FI,^14,30,31 where overall tissue quality was quantified by comparing the neocartilage’s properties to those of natural cartilage. Maturity indices were calculated based on correlative photometric and spectroscopic assays and have been developed based on biochemical parameters currently used for cartilage FI measurements, specifically stiffness, tensile stiffness, GAG content, and collagen content parameters. Here, we extend our previous work on cartilage scoring by verifying the following photometric MI (MI_photometric):

M I photometric = [ 1 2 | ( 1 - GAG Mature / Col Mature − GAG Stage / Col Stage GAG Mature / Col Mature ) + ( 1 − Pyr Mature / Col Mature − Pyr Stage / Col Stage Pyr Mature / Col Mature ) | ] − 1

(1)

where, GAG, collagen (Col), and pyridinoline (Pyr) are normalized to dry weight for mature (Mature) and samples at various stages of development (stage) with a new spectroscopic MI (MI_{spectroscopic}) as measured by Raman, as detailed in equation 2:

M I spectroscopic = [ 1 2 | ( 1 − CS Mature / Col Mature − CS Stage / Col Stage CS Mature / Col Mature ) + ( 1 − Pyr Mature / Col Mature − Pyr Stage / Col Stage Pyr Mature / Col Mature ) | ] − 1

(2)

where CS is the sum of chondroitin sulfate-associated Raman shifts, Col is the sum of collagenous protein Raman shifts, and Pyr is the Raman shift associated with mature crosslinks found in Table 1 for mature (Mature) and samples at a particular developmental stage (stage). A linear correlation of the two methods with y-intercept = 0 for tissue samples has been included as a metric for cartilage development and maturity.

The correlation between photometric and spectroscopic MI values was excellent (R² = 0.9653, p < 0.001) with a slope of 0.98 and intercept of 0.02, indicating nearly perfect agreement between the two methods. This strong correlation suggests that the Raman-based MI can serve as a reliable nondestructive surrogate for the destructively measured MI.

Spectral analysis of cartilage development

The observed spectral changes reflect the biochemical and structural maturation of cartilage. The progressive improvement of Raman-based spectral quality with developmental stage suggests an evolution from a more disordered to a more organized molecular structure. This evolution from amorphous to ordered structure is consistent with the known process of cartilage maturation, where initially disorganized matrix components become increasingly structured and aligned. The increase of “Total Organic Matrix” (Supplementary Fig. S1 indicates the increasing collagen content and improved fibrillar organization that occurs in developing cartilage ³² among others. The peaks in the 1000 cm⁻¹ to 1100 cm⁻¹ region showed changes in relative intensity and profile shape that suggested compositional changes in the PG population by age, including possible alterations in GAG sulfation patterns or relative proportions of different GAG types. To quantify the observed shift in matrix-to-cellular dominance, we calculated the lipid-to-protein matrix ratio for each developmental stage. This ratio was derived from the sum of baseline-corrected peak intensities related to the ester carbonyl (∼1745 cm⁻¹) and CH₂ twist (∼1300 cm⁻¹) associated with lipid acyl chains, normalized to the sum of organic matrix components (GAGs, collagenous, and noncollagenous protein) representing the total protein matrix.

This analysis revealed a significant decrease in the lipid-to-protein ratio from fetal (0.29 ± 0.10) to mature (0.20 ± 0.02) stages (p < 0.02, see Supplementary Fig. S2), quantitatively supporting the interpretation that the extracellular matrix (predominantly proteinaceous) becomes progressively more dominant relative to cellular components (rich in membrane lipids) during cartilage maturation. Additionally, the changes in peak shape and position may reflect changes in collagen secondary structure or crosslinking pattern. Together, these systematic changes suggested that multivariate analysis of spectral features could provide quantitative information about tissue development.

DNA content

It is known that the relative cellularity of cartilage tissue decreases as a function of developmental stage, where nascent and immature tissue exhibit a higher degree of cellularity when compared to mature tissue. ³³ Early in development, cartilage contains a high density of chondrocytes that actively produce extracellular matrix. As the tissue matures, the matrix expands dramatically while the cell population remains relatively constant or decreases slightly through apoptosis, resulting in lower cellular density. Our results in Figure 2 are consistent with this expectation, demonstrating higher DNA content in fetal and neonatal cartilage, indicative of greater cell density at early stages of development. The lower DNA levels in juvenile and mature cartilage are consistent with tissue maturation, ²³ where ECM expands relative to the number of cells, leading to reduced cell density per dry weight.

The strong correlation (R² = 0.6046) between Raman and PicoGreen measurements demonstrates that confocal Raman spectroscopy can serve as a reliable, nondestructive surrogate for assessing tissue cellularity, unlike PicoGreen, which requires destructive tissue processing. This correlation is mechanistically grounded: our Raman sampling protocol (multiple spatially averaged spots) captures a tissue-averaged signal that includes contributions from both chondrocytes and the ECM, analogous to the bulk tissue homogenate used in the destructive assay. The higher DNA signal in fetal and neonatal cartilage corresponds to greater chondrocyte density and metabolic activity during active growth and matrix synthesis. As the tissue matures, ECM expansion dilutes the cellular fraction, leading to the observed decrease in both Raman and biochemical DNA measures. Crucially, the Raman method achieves this without destroying the sample, enabling longitudinal tracking of cellularity in the same specimen—an impossible feat with destructive assays. This capability is particularly valuable for monitoring cell proliferation, death, or metabolic changes in developing tissue-engineered constructs or in studies of cartilage degeneration over time.

Glycosaminoglycan content

It is well established that sGAG content is inversely correlated with age, and our experimental data in Figure 3 are consistent with this expectation. ³⁴ Figure 3 findings show that sGAG content is highest in fetal/neonatal cartilage and decreases in juvenile/mature tissue, consistent with matrix remodeling during maturation. High content in fetal and neonatal cartilage reflects the early establishment of the PG-rich matrix, which provides the osmotic properties necessary for cartilage function. The decrease in GAG content with maturation may seem counterintuitive but is consistent with some literature reports showing that immature cartilage sometimes has a higher GAG concentration that decreases slightly in proportion to collagen, whose content increases with age as the tissue becomes more organized.

An important consideration in the GAG analysis is the specificity of the methods. The DMMB assay primarily detects sGAGs (chondroitin sulfate, keratan sulfate) with limited sensitivity for HA, which is nonsulfated. The Raman approach included peaks associated with both sGAGs and HA, which might explain the slight differences in absolute values between methods.

Most physicochemical and biochemical analyses of GAGs from complex biological systems like tissues and cells have involved ion pairing chromatography and mass spectrometry applied to enzyme digest solutions.^35,36 These methods are time-consuming and require specific reagents and enzymes. Therefore, novel and alternative methods of GAG analysis have been developed. The rapid, non-invasive, and nondestructive approach based on Raman spectroscopy presented here appears to be promising because it gives a global molecular fingerprint that can be used for GAG characterization, exhibits good correlation to standard photochemical assays, and requires less time to perform.

Collagen content

Collagen content (Fig. 4) increased from fetal to mature cartilage, reflecting the progressive accumulation and organization of ECM during tissue maturation. This trend matches known cartilage development, where collagen-rich fibrils gradually replace the initially cell-dense, matrix-poor early tissue. ³³ This pattern contrasts with the decreasing trends observed for DNA and GAGs, highlighting the dynamic nature of matrix remodeling during development. Both Raman collagen signals and biochemical assays showed consistent age-dependent increases. The progressive increase in collagen content is biologically significant. Early in development, cartilage matrix contains relatively more PGs and less collagen. As the tissue matures, the collagen network becomes more extensive and organized, providing the tensile strength needed for mechanical function.

The excellent correlation between Raman and destructive methods demonstrates that Raman spectroscopy can reliably quantify these important developmental changes. Thus, Raman may be used as an alternative collagen quantification assay to the hydroxyproline assay. The non-invasive and nondestructive nature of Raman collagen quantification offers significant advantages over the hydroxyproline assay, such as the ability to track samples over time throughout culture periods or to analyze tissues prior to destructive mechanical testing. Furthermore, Raman does not require dangerous chemical reagents that are needed to perform the hydroxyproline assay. ²² Therefore, researchers performing tissue engineering and tissue characterization experiments may consider Raman spectroscopy for collagen analyses.

Pyridinoline crosslinks

The increasing pyridinoline content with maturation is biologically significant. Pyridinoline crosslinks form through enzymatic processes that mature over time, stabilizing the collagen fibrils and increasing tissue mechanical strength. The very low levels in fetal and neonatal tissue indicate immaturity of the collagen network, while the high levels in mature tissue reflect a fully developed, mechanically competent matrix. Collagen crosslinking is an important parameter when assessing tissue maturity and can be measured as changes in the amide I region. This was first illustrated by Fourier-transform infrared spectroscopy,^18,37 where biochemical analysis of collagen model peptides showed that mature trivalent crosslinks (PYR), Figure 5, crosslinks resulted in a band at ∼1660 cm⁻¹ and immature divalent dehydrodihydroxylysinonorleucine (de-DHLNL) crosslinks in a band at ∼1690 cm⁻¹. It is purported that the content of de-DHLNL crosslinks decreases with collagen maturity, while PYR crosslink content increases, as seen in bone ³⁸ and cartilage, ³⁹ likely due to the maturation of the former to the latter. ⁴⁰ Most of the published work has focused on determining tissue maturity and, conversely, tissue degeneration based on relative intensity and peak area ratiometry of the 1660 cm⁻¹/1690 cm⁻¹ bands. Even though this metric was developed for infrared spectroscopy, ¹⁸ it has been used successfully in Raman spectroscopy.^21,41 Unlike the work previously discussed, the current work qualitatively and semiquantitatively measures collagen PYR crosslinks by temporally tracking changes in the relative intensity of 1660 cm⁻¹ Raman shifts to quantitative measurements of PYR as determined by mass spectrometry. The increase in pyridinoline content with development is well established, as collagen fiber bundles mature. ³³ Our results corroborate this finding, showing a corresponding increase in pyridinoline content with cartilage maturation using both Raman spectroscopy and mass spectrometry (Fig. 5). The strong correlation between Raman and mass spectrometric measurements validates Raman spectroscopy as a reliable method for assessing collagen crosslinking noninvasively.

The Maturity Index: Relevance, meaning, and critique

The development of the Raman-based MI is a significant conceptual and practical advancement for tissue engineering. This parameter reflects the extent to which the extracellular matrix is fully formed or mature. Rather than monitoring multiple biochemical parameters separately, researchers and manufacturers could use a single integrated metric to assess tissue quality. MI introduced here is focused on temporal compositional changes as measured nondestructively and verified through destructive assay(s). As a direct measure of tissue maturation and TEMP development, MI can be integrated within biomanufacturing workflows for dynamic in situ evaluation of processes for QA/QC. In addition, process optimization may be evaluated with minimal cost. The Raman-based approach allows this assessment to be performed nondestructively, enabling longitudinal monitoring of the same sample throughout development.

The MI integrates three critical, interdependent CQAs: •

GAGs (CS_Stage): Represent the fixed charge density, responsible for hydration and swelling pressure.

The index’s form of M I = Pyr Stage / CS Stage Pyr Mature × CS Mature Col Stage × Col Mature is physiologically grounded. It essentially calculates the product of the maturing components (Pyr and CS) normalized to the underlying collagen scaffold. An MI of 1 indicates a tissue that is not only biochemically mature but also structurally integrated, implying functional competence.

One potential alternative approach is to employ multivariate statistical techniques, such as Partial Least Squares Regression (PLSR), to develop a model that predicts a ‘maturity score’ based on the full spectral data, rather than relying on preselected peaks. However, while PLSR models can be highly effective, they can also be opaque and challenging to transfer across different instruments or laboratories, and may not provide clear insights into the underlying biological mechanisms. ¹⁵ The proposed MI, based on known biochemical peaks, is transparent, easily interpretable, and directly links the metric to biological processes, which is a significant advantage for regulatory approval and fundamental understanding.

The excellent correlation (R² = 0.9653) between the destructive and spectroscopic MIs is paramount. Comparing the two is not redundant; it is the core of the validation process. The excellent correlation proves that the relationship between these key components is preserved in the Raman data. It demonstrates that the simplified, nondestructive method captures the same complex biological narrative as the combined effort of three separate, destructive gold-standard assays. This relevance is crucial for granting confidence to use the Raman MI as a surrogate endpoint in manufacturing.

Quantitative performance and validation of Raman spectroscopy

This study establishes Raman spectroscopy as a robust quantitative method for tracking cartilage development, moving beyond its traditional role as a qualitative tool. The high linear correlations (R² > 0.96 for Raman vs. destructive assays) across all major biochemical constituents—DNA, sGAG, collagen, and pyridinoline—provide compelling validation. This performance is particularly significant when placed in a broader context. Similar Raman-based quantification efforts in other tissues, such as bone (mineral-to-matrix ratio) or skin (collagen hydration), typically report strong qualitative trends but often achieve lower quantitative R² values (often in the range of 0.7–0.85) when correlated against destructive standards.^7,12 The exceptional correlations (R² > 0.96) we observed for cartilage can be attributed to the tissue’s relatively simplified extracellular matrix (dominantly Type II collagen and specific GAGs) compared to more compositionally complex tissues, making it an ideal model for developing precise spectroscopic biomarkers.

A key question remains: how does the sensitivity of Raman spectroscopy compare to that of destructive assays? Our data suggests its sensitivity is context-dependent and complementary. For high-abundance components like total collagen, traditional assays like the hydroxyproline assay may have a lower absolute detection limit. However, Raman’s strength lies in its multiplexed, nondestructive sensitivity. For instance, it can detect the pyridinoline crosslink signal in nascent tissue where its concentration is too low for easy quantification by LC-MS without extensive sample pooling. ¹¹ Furthermore, while a Picogreen assay might be more sensitive for isolated DNA quantification, it requires tissue destruction. Raman provides a statistically significant (p < 0.05) measure of relative cellularity in situ without damaging the sample, allowing for longitudinal tracking that is impossible with destructive methods. ¹³ While Raman spectroscopy may not necessarily offer a lower detection limit than traditional methods, it provides a significant advantage in terms of informational efficiency, extracting multiple quantitative data points from a single, rapid, and nondestructive measurement.

Methodological considerations: GAGs and limitations

The choice to sum only sGAG (CS4/CS6) peaks, excluding nonsulfated species like HA, is justified for this context. HA is more prevalent in the pericellular matrix and synovial fluid, while the sGAGs (chondroitin sulfate, keratan sulfate) are the primary contributors to the fixed charge density of the cartilage ECM proper. ²⁰ Furthermore, the DMMB assay used for validation is specific for sGAGs. ¹⁶ Therefore, using CS4/CS6 peaks ensures a direct and valid comparison to the destructive benchmark. For studies specifically focused on the pericellular matrix or synovial joint biology, incorporating HA peaks would be essential.

Several limitations and potential areas for error must be acknowledged: (1)

Spectral overlap: The greatest potential for error is in crowded spectral regions, particularly the Amide I region (∼1660 cm⁻¹) where the pyridinoline signal resides. Changes in collagen secondary structure or water content could theoretically influence this peak. ¹⁷ Our validation against LC-MS mitigates this concern for this study, but it remains a consideration for unknown samples.

Generalizability to other tissues

The concept of a Raman-based MI is highly relevant and can be readily adapted to other tissues. While the specific biomarkers may vary, the underlying methodology provides a versatile template that can be applied to different tissue types: •

Bone: An MI could combine the mineral phosphate peak (∼960 cm⁻¹) with the collagen crosslink peak (1660 cm⁻¹) and a matrix ratio.^12,18

In each case, the index would need to be calibrated and validated against destructive standards, as was done here, for that specific tissue. However, the framework established here provides a powerful blueprint for developing nondestructive QA/QC metrics across regenerative medicine.