In this study, cryo-preserved historical skin samples (30 Caucasian female donors, 18–75 years, cosmetic surgery procedures) collected under informed consent were analyzed as part of this explorative study. All the samples were anonymized, and the donors’ age, biological sex, and ethnicity were made available for this study. The sample cohort was split into three groups according to both the donors’ age and anatomical site: Breast Young Skin (BYS, N = 11, age range: 18–29 year-old, median: 22.0 ± 4.2 years), Breast Old Skin (BOS, N = 9, age range: 49–78 year-old, median: 69.0 ± 13.2 years), and Cheek Old Skin (COS, N = 10, age range: 56–80 year-old, median: 65.0 ± 7.7 years). Samples of skin obtained from these donors were split in two for processing. For histology, the skin samples were fixed in neutral formalin and then embedded in paraffin before being sectioned (5 um). The sections were stained using Hematoxylin and Eosin Stain (HES), Orcein, and Sirius Red (SR). For quantitative nanohistology (AFM), the skin samples were not fixed but cryo-embedded in optimum cutting temperature media before being cryo-sectioned (8–10 µm thickness). All sections were physisorbed directly on individual glass slides.

A Leica (Wetzlar, Germany) light microscope (LM) was used for histology imaging of the fixed and stained sections. This microscope was equipped with two crossed-light polarizers (90°) to allow for polarization (darkfield) LM and with an 8-megapixel digital camera (EOS Rebel 100, Canon). Complete histological sections were digitized using the auto-stitch function in Image-Pro Plus software (Meyer Instr. Inc., Houston, United States).

Topological images (10 × 10 μm²) of the tissue were acquired on the unfixed sections by Atomic Force Microscopes (AFM Nanowizard I and III, Bruker-JPK, Berlin–Germany) operated in contact mode (MSNL cantilevers (Nom. tip radius: 2 nm, triangular geometry, Bruker, Santa Barbara) operated in ambient conditions at a scanning rate of 1.0 Hz or above. To avoid bias in selecting the area to be imaged, we used a random walk approach to land the probe in each dermal layer by using the AFM sample holders’ x-y translation screws to move the sample (within each dermal layer). Images were then acquired and optimized where the probe came into contact with the sample regardless of the topology seen in the image acquired. A total of 14 images were acquired for each donor (papillary dermis 7; reticular dermis: 7 on a minimum of 2 sections per donor), leading to a dataset of 420 (10 × 10 μm²) images. Following their acquisition, each AFM image was plane-fitted before being segmented into 100 x (1 × 1 μm²) images using ImageJ, creating a final topology dataset of 42,000 (1 × 1 μm²) images to be analyzed.

The mechanical properties of collagen fibrils of the histological sections were acquired by a Nanowizard I AFM (Bruker-JPK, Berlin–Germany) operated in the force-distance mode in ambient conditions. For these measurements, RFESPA cantilevers (Nom Tip Radius: 8nm, rectangular geometry, Bruker, Santa Barbara) with a spring constant k = 3 N/m were employed. The selection of the sample area to be indented was also performed using the same random-walk approach as previously mentioned. First, a low-resolution image (128 × 128 px over 10 × 10 μm²) was performed to ensure that collagen fibrils with defined D-banding periodicity could be observed. Then, indentation sites were manually selected directly on distinct collagen fibrils’ D-banding (overlap region). All indentations were carried out at 1Hz, with a maximum indentation load not exceeding 350 nN (yielding an average indentation depth of d = 25 ± 5 nm). For each histological section, a minimum of 500 individual collagen fibril measurements were carried out in the papillary and reticular dermis, resulting in a mechanical dataset of 30,000 indentations for the 30 samples. The diameter of each indented fibrils was not measured. All force-distance curves were processed by the JPK Data Processing Software v.5.1.8 using the Sneddon/Hertz model (Harding and Sneddon, 1945). The resultant distribution of Elastic (Young’s) Moduli was plotted as histograms with a fixed bin size (B = 500 MPa) to calculate the median Elastic (Young’s) Moduli for the respective dermal layers.

All data were processed using Origin Pro (Origin Lab Corporation, Northampton, United States). In this study, each histological block was treated as an individual variable, and data obtained from serial histological sections from the same block were pooled together. Four calibrated analysts were used to quantify the prevalence of the four structural biomarkers in each 42,000 1 × 1 μm² image. Analysts’ calibration was performed by agreeing on the assignment of marker 1-2-3-4 on 100 AFM 1 × 1 μm² images. For the remainder of the analysis, all analysts were blinded and did not know the sample groups donors’ assignment until the study was analyzed statistically. The Young’s modulus of fibrils was compared using a two-sided hypothesis Mann-Whitney U test with a significance level α = 0.001. In addition, 2D Principal component analyses (Jolliffe and Cadima, 2016) (PCA) were performed on the dataset using Origin Pro (Origin Lab Corporation, Northampton, United States). The relative position of the variables on the PCA score plot, with their coordinates equal to correlations with the first two principal components, was used to assess the level of discrimination between the variables. Finally, a Partial least Squares Discriminant Analysis (PLS-DA) was performed using SIMCA^® 16.0 multivariate Data Analysis Software (Sartorius Stedim Data Analytics AB, Sweden) with the binary age class as the response variable.

Histological analysis is routinely used to assess structural changes in tissues, including skin, to prognose or diagnose pathologies such as aging. Figure 1 presents the histology of representative sections from our three groups. The young group (Breast Young Skin–BYS) represented by the photo-protected breast skin specimen presents distinct epidermis (Ep), papillary (PD), and reticular (RD) dermal layers (Figure 1A). The dermo-epidermal junction is non-uniform and presents well-defined rete ridges (arrow). Structurally, the collagen in both dermal layers is very dense, and there is little evidence of elastin (Figures 1B-i, iv, vii). In the intrinsically aged skin (Breast Old Skin: BOS) specimen, one observes a thinning of the epidermis and papillary dermis (Figures 1B-ii). The dermo-epidermal junction presents a flattening aspect resulting from the disappearance of the rete ridges. The dermis has an atrophic aspect with a loss of cells and extracellular matrix. As a result, the dermal collagen becomes sparser (Figure 1B-v). Finally, the BOS histology shows evidence of loss of oxytalan fibers but not of solar elastosis/elastin deposition (Figure 1B-viii).

In contrast, the primary histological fingerprint for extrinsically aged skin (Cheek Old Skin: COS) is the abundance of abnormally structured elastin accumulated as elastotic material. This accumulation can be found in the reticular dermis, as shown in Figure 1B-ix. In addition, the extrinsically aged skin specimen presents a flattening of the dermis and a loss of rete ridges found for the intrinsically aged skin (Figure 1B-iii). The ratio of elastin to collagen has also increased (Figures 1B-vi, ix), suggesting that the collagen is being replaced by elastin, which explains the altered functional and biomechanical properties of this skin (Uitto, 2008). The histological presentation of these groups: young photo-protected skin and intrinsically and extrinsically aged skin, are well known and are routinely used to define the pathological age of skin (Khavkin and Ellis, 2011).

Type I collagen fibril topology is one of the most recognized protein structures due to its highly conserved annular-banding periodicity along the long axis of the fibrils, namely, the D-banding periodicity (Petruska and Hodge, 1964; Chernoff and Chernoff, 1992; Stylianou, 2022). In our quantitative nanohistology approach, we have used AFM to image collagen structure on the skin groups’ histological sections. Figure 2 presents our approach to correlating AFM images site with a polarized image of SR-stained skin sections. We observe the presence of long fibrils with a defined D-banding periodicity in all the images recorded. Fibrils can appear as thick bundles, homogeneous sheets, or interwoven scaffolds. Our original skin study defined several potential structural biomarkers for aging in collagen (Ahmed et al., 2017). Four were present across the three skin groups from a list of 7 candidates’ structural (or topological) collagen biomarkers, as presented in Figure 2B. Marker 1 is defined as the presence of interfibrillar gaps, suggesting a loosening of the dense collagen sheet structure, leading to the loss in fibril registration with one another (Figure 2B-i). Marker 2 characterizes areas where the fibrillar collagen structure is not readily observable and could be associated with cell processes or other extracellular matrix components (Figure 2B-ii). Marker 3 characterizes a well-defined collagen matrix, with the fibrillar D-banding apparent (Figure 2B-iii). However, the collagen fibrils in those areas are disorganized and lack registration between them. In other studies, we found collagen fibrils morphology prevalent in the fibrotic area (Strange et al., 2017). Finally, Marker 4 characterizes a well-defined collagen matrix, with the fibrillar D-banding apparent and the fibrils aligned, forming dense collagen sheets (Figure 2B-iv). The prevalence of each of these structural (or topological) collagen biomarkers in the papillary and reticular dermis by segmenting each 10 × 10 μm² AFM image into a 100 sub-image 1 × 1 μm² as presented in Figure 3A. Each sub-image was evaluated independently (1 chosen marker per 1 × 1 μm² image). In addition to these structural biomarkers, we added the median value of the collagen fibrils Young’s modulus for both the papillary and reticular dermis, as presented in Figure 3B. We have already demonstrated the variation of Young’s modulus of collagen fibrils (dry) as a function of the aging process (Wenger et al., 2007; Ahmed et al., 2017).

Dermal collagen phenotype varies among individuals based on biological sex, ethnic origins, and lifestyles (Fenske and Lober, 1986), (Farage et al., 2008). Therefore, to analyze collagen fibril phenotype, one cannot assume that the properties of the dermal collagen for all the individual donors within a given group are similar; thus, markers cannot be represented through a common median value per dermal layer and donor group.

We explored whether the relative structural variations in the collagen matrices between the papillary and reticular dermis could be used to differentiate between the three groups. To do so, we referenced the variations in individual markers in the reticular dermis to those found for the papillary dermis and normalized these variations using Eq. 1 in Figure 4A. The results are plotted in Figure 4A, presenting the %variation of each marker becoming more prevalent towards either the papillary or reticular dermis. Both markers 1 and 2 show an increased prevalence in the papillary dermis of the BYS group. While marker 2 does not readily inform us of the collagen phenotype, marker 1 suggests that the collagen matrix in the papillary dermis of young photo-protected skin exhibits structural loosening due to the increased presence of interfibrillar gaps (or holes) when compared to the reticular dermis. This finding would also suggest that the collagen matrix aging would occur in the papillary dermis before progressing to the reticular dermis. This outcome supports the recent finding by Lynch et al., who established that the mechanical property of the papillary dermis decreases before that of the reticular dermis with age (Lynch et al., 2022). Their study suggested that with aging, the earliest microstructural and mechanical changes occur in the topmost layers of the dermis/skin and then propagate deeper, providing an opportunity for topical preventive treatments acting at the level of the papillary dermis. An earlier study focusing solely on phenotyping the reticular dermis collagen suggested marker 3 is the hallmark of collagen aging. In our present study, marker 3 does not appear more prevalent in either dermal layer, regardless of age. This result does not contradict our previous study but implies that at the nanoscale, one cannot differentiate the two dermal layers solely based on marker 3. Marker 4 shows an increased prevalence in the reticular dermis of the BOS group, suggesting that intrinsic aging promotes the formation of localized, well-aligned collagen fibril bundles. Using our nanohistology approach, we can assert that the reticular collagen is more aligned than the papillary collagen in the sparser region of dermal collagen. This increase in fibril alignment can be directly associated with interfibrillar crosslinks, such as those mediated by the formation of advanced glycation end-products (Perrone et al., 2020). Finally, no clear trends exist in the % variation of any of the markers associated with the COS group. The lack of a specific structural collagen marker to describe the matrix associated with extrinsically aged skin is unsurprising due to the reduced collagen content favoring elastin fiber formation.

In a similar approach for the morphological assessment, we calculated the %variation of Young’s moduli between the papillary and reticular dermis, as presented in Figure 4B. This result shows intrinsic aging promotes reticular collagen stiffening over papillary collagen within the intrinsically aged group (BYS compared to BOS). On the other hand, the extrinsically aged group (COS) did not show any trend toward stiffening of the papillary or reticular dermis. This heterogeneity in the mechanical response corroborates our finding in the nanohistology analysis, for which we could not use our pre-defined histomorphological markers. Numerous studies have explored the variations in the collagen mechanical properties at various scales as a function of aging and focused on calculating the nominal values for the elastic modulus of dermal collagen fibrils or scaffolds. (Crichton et al., 2011; Achterberg et al., 2014; Ahmed et al., 2017; Strange et al., 2017; Penuela et al., 2018). Unfortunately, the range of published elastic modulus values for dermal collagen fibrils varies across several orders ranging from a few 100 kPa to a few GPa (Ahmed et al., 2017). This wide range in the elastic modulus values is directly linked to the histological section conditioning, indentation size, inconsistent indentation load, limited sample size, and collagen hydration level. Unfortunately, comparing study outcomes regarding collagen elasticity measurement is challenging due to the wide variations in the techniques used.

To explore from a multivariate perspective the correlation between the structural (marker 1-2-3-4) and mechanical (elasticity) markers regarding the chronological age of the donors, both unsupervised (regardless of the initially assigned group) PCA and supervised (with the age group as response variable) PLS-DA analyses have been carried out. Figure 5A presents the PCA score plot associated with the %variation of the markers across all donors. The size of the vectors associated with the % variation in markers 3, 4, and elasticity confirm that these three markers are the more potent contributor to the principal components, thus, more prominent differentiators of the entire dataset. On the other hand, the size of the vector associated with marker 1 confirms marker 1 has a limited impact on the dataset. Therefore, we supplemented our PCA analysis with a PLS-DA using the donors’ age binary class as the response variable rather than the markers (Figure 5B). The Biplot chart allows individuals and descriptors (% variation of all markers) to be simultaneously represented and allows us to interpret the individuals in terms of descriptors.

In this biplot, we can observe that the distribution of donors tends to form 2 clusters: all the young donors (<30y.o) are clustered on the left-hand side of the biplot (with two exceptions), whereas the older donors (>50 y.o) tend to cluster on the right-hand side of the biplot. Significant overlap exists between the two sub-populations around the central axis of the PLS-DA. This overlap can be associated with the well-known disparity between the chronological age of the donors and their biological age (Rittie and Fisher, 2015). Although a 20-year gap exists between our younger and old groups, one cannot assume that the chronological age of the donor is the same as their biological age. This is proven by a young donor case on the aged group’s outer edge (red circle). Figure 5C-ii presents the histological images of this young donor (22y.o). Despite the dermo-epidermal junction being non-uniform and presenting well-defined rete ridges, the underlying dermal collagen presents significant signs of aging-induced damage, as found in older donors (Figure 5C-iii, iv). The dermal collagen is heavily atrophic and sparse for this young donor, likely associated with significant Sun damage. The quantitative nanohistological assessment performed in this study could classify this donor’s collagen phenotype amongst the old group. This 22-year-old donor has the dermal fingerprint of a chronologically old donor, demonstrating the disparity between chronological and biological age again.